UNITED STATES DEPARTMENT OF AGRICULTURE

 

 

 

 

 

 

 

 

 

FOOD AND DRUG ADMINISTRATION

 

FOOD ADVISORY COMMITTEE

 

 

 

MEETING ON INFANT FORMULAS

 

 

 

 

 

 

 

 

 

 

 

 

 

Monday, November 18, 2002

 

 

8:20 a.m.

 

 

 

 

 

 

 

 

U.S. Department of Agriculture

Animal and Plant Health Inspection Service Building

4700 River Road

Riverdale, Maryland

Temporary Voting Members Present

 

James Anderson, Ph.D.

Robert D. Baker, M.D., Ph.D.

Margaret E. Briley, Ph.D., R.D., L.D.

Scott Denne, M.D.

Cutberto Garza, M.D., Ph.D., Chairman

James E. Heubi, M.D.

Laurie J. Moyer-Mileur, Ph.D., R.D., C.D.

Virginia A. Stallings, M.D.

Patti Thureen, M.D.

 

Participating Food Advisory Committee

Members Present

 

Annette Dickinson, Ph.D.

Goulda Angella Downer, Ph.D.

Lawrence N. Kuzminski, Ph.D.

Madeleine J. Sigman-Grant, Ph.D.

 

Acting Industry Representative

 

Roger A. Clemens, Dr.P.H. CNS FACN

FOOD ADVISORY COMMITTEE ON INFANT FORMULAS

 

AGENDA

 

Welcome, Introduction and Charges

  Christine J. Taylor, FDA    4

 

Administrative Issues -- Cathy DeRoever,

  Mary Ann Killian and Jeanne Latham, FDA    7

 

Regulatory Background

  Christine J. Taylor, FDA    23

 

Remarks by Chairperson, Cutberto Garza    32

 

Presentations by Invited Speakers

  Physical Growth Measurements

    W. Cameron Chumlea    43

 

  Body Composition, Kenneth J. Ellis    72

 

  World Health Organization Growth

    Reference, Edward A. Frongillo    103

 

  National Center for Health Statistics/

    Centers for Disease Control and

    Prevention Growth Charts, Lawrence M.

    Grummer-Strawn    127

 

  Iowa and Iowa/Fels Growth Data

    Samuel J. Fomon    152

 

  Growth Data for Preterm Infants

    Jon Tyson    169

 

  Analytical Issues, Edward A. Frongillo    202

 

Questions to Previous Speakers

 

  Clinical Studies -- Product Composition

  Considerations, Duane Benton    228

 

  Clinical Studies -- Clinical

    Consideration, Dennis M. Bier    259

 

Questions to Speakers    283

Preliminary Discussion of Issues by Committee    313

Concluding Remarks    383

P R O C E E D I N G S

    DR. TAYLOR:  I'm Christine Taylor, and I'm director of Office of Nutritional Products Labeling and Dietary Supplements at FDA's Center for Food Safety.

    We will be this morning going through a series of conversations, and what you've got right now is our overview.  So given the fact that we started just a few minutes late, we'll go ahead and try to shortcut this overview.

    In the next few minutes, we'll do a brief overview for this Food Advisory Committee meeting.  We'll review a couple of the administrative issues, which will be focused on ethics and conflicts of interest.  We'll spend some time on regulatory context for this meeting, and then we will begin the meeting per se with Dr. Bert Garza serving as chair.

    During the meeting, there will be a series of presentations and white papers, which we have provided as background information for the committee.  There will be some discussion, and then beginning tomorrow there will be public comments, more discussion, and response.

    We're going to spend just a very few minutes this morning going over the role and expertise of the task force members, and I think the key point to be recognized is that there are several different kinds of members sitting with us today.

    FDA considers a number of factors in selecting individuals to serve on the Food Advisory Committee, including their scientific expertise, as well as issues related to conflict of interest.

    We also have sitting with us a consumer representative.  This person is a voting member of the Committee and represents the consumer perspective on issues and actions that come before the Committee.

    We also have an industry representative sitting with the committee.  This is a nonvoting member, and they're responsible for representing all members of the industry, and not any particular association, company or product.

    Basically, the kinds of members we have on this committee are temporary voting members, as well as some members of our larger Food Advisory Committee.  As I mentioned, we also have consumer reps and industry reps.

    What I'd like to do, at this point, starting with Dr. Baker, if you will, Dr. Baker, just so we can get used to using the microphones, if you would let us know your name, where you are, and if you're too modest, I have a listing of your expertise.

    DR. BAKER:  Robert Baker from Buffalo, New York.  I'm a pediatric gastroenterologist, and I have a Ph.D. in biochemistry and in nutrition.

    DR. STALLINGS:  I'm Virginia Stallings, from Children's Hospital in Philadelphia.  I'm the head of the Nutrition Section there, and I do work in healthy children and children with chronic disease related to nutrition.

    DR. HEUBI:  I'm Jim Heubi.  I'm a pediatric gastrologist, as well.  I'm the program director for the GCRC, the General Clinical Research Center at the Children's Hospital in Cincinnati, and I have a longstanding interest in nutrition relating to infant nutrition bone disease, cholesterol, metabolism, you name it, there's a variety of things.

    DR. ANDERSON:  I'm Jim Anderson.  I'm at the University of Nebraska Medical Center in Omaha, Nebraska.  I'm chairman of the Department of Preventive and Societal Medicine, and I'm a biostatistician by training.

    DR. DOWNER:  I'm Goulda Downer, a doctorate in Human Nutrition, with a residency in pediatrics at Georgetown.  Currently, I'm a clinical nutritionist with my own practice, and I'm also on faculty at George Washington University.

    MS. SIGMAN-GRANT:  I'm Madeleine Sigman-Grant.  I'm a maternal and child nutrition specialist at the University of Nevada Cooperative Extension.

    DR. MOYER-MILEUR:  I'm Laurie Moyer-Mileur, from the University of Utah.  I'm a registered dietician with a doctorate in exercise physiology, and I have over 20 years of neonatal nutrition experience.

    DR. GARZA:  I'm Bert Garza.  I'm a professor of nutrition at Cornell University.  I'm both an M.D. and have a Ph.D. in nutritional biochemistry and metabolism, and my primary interests have been in maternal-child health, with interests in growth, and protein and energy metabolism.

    DR. KUZMINSKI:  I'm Larry Kuzminski.  I'm from Duxbury, Massachusetts.  I'm retired from the food processing industry, having R&D responsibilities and operations responsibilities with the Kellogg Company and with Ocean Spray Cranberries.

    DR. DENNE:  I'm Scott Denne.  I'm from Indiana University.  I'm a pediatric neonatologist.  I have a longstanding interest in neonatal nutrition, specifically, and protein and energy metabolism.

    DR. THUREEN:  I'm Patti Thureen, a neonatologist from the University of Colorado in Denver, and my particular interest is in protein and energy metabolism in the extremely low-birth-weight neonate.

    DR. BRILEY:  I'm Margaret Briley from the University of Texas at Austin, and my expertise has been in nutrition of children and child care.

    DR. TAYLOR:  If we could just stop right there.  Margaret is our consumer rep, and on our right we have Dr. Roger Clemens, who is substituting for Annette Dickinson, who is our industry rep.

    I'll go to the next slide and just give a minute or two about the staff you have sitting at the table with you.  As I've mentioned, I'm with the Office of Nutritional Products Labeling and Dietary Supplements.

    We also have Dr. Susan Walker, who is our associate director for Clinical Affairs, as well as Dr. Beth Yetley, who's the lead scientist for nutrition.

    Jeanne Latham, who is sitting next to Dr. Bert Garza, is our executive secretary, and we're being joined today by Ms. Mary Ann Killian, who is program integrity adviser at the Office of Human Resources at FDA.

    Let me just spend a very quick minute, and then we will return with a regulatory context.  I think in terms of mechanics, we need to understand kind of where we are in the process.  Currently, we are operating as an ad hoc task force to the Food Advisory Committee.  In the very near future, we will constitute an Infant Formula Subcommittee of the Food Advisory, but currently we are still in the ad hoc mode.

    The current focus of the Infant Formula Advisory Meetings is to obtain scientific input for evaluating whether new infant formula supports normal physical growth of infants.  This comes under Section 412 of the Food, Drug, and Cosmetic Act, which in a few moments we'll come back to in more detail.  We're looking basically for scientific input, which eventually will inform the Agency relative to regulatory efforts.

    What we're undergoing currently is a series of meetings, and I'm sure most of you remember that last April we held our first meeting on this issue of normal physical growth.  It included a somewhat general discussion, an effort to understand the regulatory context, as well as a few specific questions about extrapolation and attrition in the study.

    This is the second of this series, and the general scientific topics for today fall into three categories: Growth Measures and Methodologies, the Role of Such Measures and Methodologies in  Demonstrating Normal Physical Growth, and then, finally, Principles and Criteria to Determine the Need for a Clinical Study to Provide the Agency an Assurance of Normal Physical Growth.

    It's always helpful to be clear about what's not on the table.  There are so many issues in the area of infant formula, normal physical growth, other issues related to the Agency's regulatory purview that sometimes it's important to realize there are things that are of great interest, but are not on the table for discussion.

    This lists a few, probably the ones that our discussions will most likely tend to gear toward.  The design and conduct of studies is not on the agenda today, other endpoints of clinical studies is not on the agenda.  What constitutes major and minor changes is not on the agenda.  That's, of course, for those of you that are intimately involved in the regulatory, you understand that that has regulatory meaning.  The nutritional impact or efficacy of formulas, the safety of individual ingredients and specific regulatory decisions are not topics for today.

    We have provided specific background for the committee in the form of white papers.  We have a total of nine white paper which, as Dr. Garza will explain in a few moments, we'll go through this morning.  Each of the papers will be introduced by an expert, and then of course discussed by the committee as appropriate.  Those related to the assessment of normal physical growth are listed here, and then for our second topic, changes warranting a clinical study, we have two white papers.  Those should be in your notebooks and available for further discussion.

    Now, the Agency's role is to give you specific charges that are to be accomplished by the end of the meeting on Tuesday, and those charges are in your notebook in the form of seven questions, and I won't go through them now.  I think Dr. Garza will take the time to do that with you later on, but they fall into basically four categories: Metrics for evaluation of growth, which is Questions 1, 2, 3A and 3B; questions about comparators, Questions 4 and 5; controlled feeding parameters, Question 6; and then changes in composition, Question 7.

    Just for the group of us here, the summary of the charges fall into two categories: The criteria for adequate evaluation of normal physical growth during the first six months, and here are several substantive ones.  Again, they are specifically it in your questions; and then, secondly, the type of changes in infant formula that should warrant a clinical study.  Again, those are the remaining questions in your notebook.

    Just in terms of the mechanics, this morning, next, we'll cover the administrative issues, Jeanne Latham, as assisted by Mary Ann Killian, will go through that with you.

    I'll return, and with the help of Dr. Walker and Dr. Yetley, give you some regulatory context and then the actual task force meeting will begin.

    What we'll do is hold questions until after the administrative component, and then again after the regulatory context, and then we should be on our way.

    So, Jeanne, I'll turn the meeting over to you, and dutifully return for the next part.   Thank you.

    MS. LATHAM:  Good morning.  I'm Jeanne Latham and, first of all, in terms of administrative issues, we wanted to have Cathy DeRoever's statement read into the record, and Dr. Garza will take care of that.

    Thank you.

    DR. GARZA:  Catherine DeRoever, the executive secretary of the Food Advisory Committee, was asked to take a few minutes to refresh everyone's memory about a few of the rules of the road, in terms of Advisory Committee operations, so I'm going to be reading her statement.

    It is my understanding that all committee members have been provided with a copy of a Committee Member Guide to FDA Advisory Committees and a video.  The video's title is "A Panel Member's Responsibility."  I believe there are copies of the Member Guide available at the registration desk for anyone who may be interested.  The Committee Member Guide is in need of updating but, by and large, it provides a good operational overview.

    FDA relies on its Advisory Committees to provide the best-possible scientific advice available to assist us in making complex decisions.  Our goal is to do this in as open and transparent a manner as possible.  Part of that openness carries with it a request that the members try to avoid even the appearance that issues are being decided or conclusions are being reached outside the actual meeting.

    We understand that issues raised during the meeting may well lead to conversations over breaks or during the meal.  In fact, we hope the discussions are thought-provoking.  We have had instances where the members have come back from a break and said, "You know, we were talking over break, and we would like to request that FDA provide us some additional information so we can better understand thus and such."  This is perfectly acceptable.

    What we don't want is to have a situation where after the break the members come back and say, "We were talking over break, and we decided that the answer to Question 1 is..."  From our perspective, that would be particularly troublesome because neither the Agency, nor the public, would have had the benefit of listening to the entire discussion, the questions raised, the responses, et cetera.

    In fact, FDA has recently adopted a policy that only matters that can be decided by a show of hands are procedure matters, for example, break times.  I'm not sure I understand that.

    [Laughter.]

    DR. GARZA:  All other votes and comments must be placed on the record, attributed to the member making the statement.  The policy goes even further.  If a member has to leave the meeting early, that member waives the right to vote.  You may wonder why would the person lose their right to vote, but the answer is fairly simple.  FDA believes all parts of the meeting and the discussions are important.  Consequently, voting on issues without having the benefit of all of the discussion would be premature.

    The issue of openness is larger than what transpires during the course of the meeting.  I would like to call your attention to the section in the Members' Guide, titled, "Member Interaction Before, During and After a Meeting."  In essence, this section underscores the fact that all communication with the members should be routed through the Committee's executive secretary.  No one, not even FDA staff, with the exception of the executive secretary, should be contacting the members about upcoming meetings, topics, et cetera.

    This same guidance applies to consultations between members prior to a meeting.  If a member receives an inappropriate contact, the members should feel free to notify the executive secretary and/or refer the person making the contact to the executive secretary.  Our goal in having all contacts routed through the executive secretary is to minimize any situation that could be misinterpreted.

    Appearance issues are always difficult because, as is true of many things, appearances can be deceiving.  We ask that our members, guest speakers, and everyone attending the meeting be mindful of how an interaction between a member and anyone, for that matter, might be perceived.

    Please let me be clear it is not my intention to question anyone's motives or integrity, but I am very sensitive to the issue because I have, and imagine so have you, seen highly respected individuals become the object of negative attention based on a misperception, and I certainly wouldn't want anyone in this room to become such a target.

    I am confident that everyone here is sensitive to these issues and can appreciate that my comments are intended as a gentle reminder.

    Thank you.

    Any questions?  Which I will refer to Ms. Latham.

    [Laughter.]

    DR. GARZA:  From any of the committee members?  Is all of that clear?

    [No response.]

    DR. GARZA:  Thank you.

    MS. LATHAM:  Good morning.  I am Jeanne Latham, the executive secretary for the FDA's Food Advisory Committee on Infant Formula.  I want to welcome everyone, and I'd like to read the conflict of interest statement for the record.

    The following announcement addresses the issue of conflict of interest with respect to this meeting and is made a part of the record to preclude even the appearance of such at this meeting.

    By the authority granted under the Food Advisory Committee Charter of July 2002, the following individuals have been appointed as temporary voting members by Joseph A. Levitt, director, Center for Food Safety and Applied Nutrition:

    James Anderson, Ph.D; Margaret Briley, Ph.D.; Robert Baker, M.D., Ph.D.; Scott Denne, M.D.; Cutberto Garza, M.D., Ph.D.; James Heubi, M.D.; Laurie Moyer-Mileur, Ph.D.; Virginia Stallings, M.D.; Patti Thureen M.D.

    The issues to be discussed at this meeting are issues of broad applicability.  Unlike issues in which a particular sponsor's product is discussed, the matters at issue do not have a unique impact on any particular product or manufacturer, but rather may have widespread implications with respect to all infant formulas and their manufacturers.

    To determine if any conflicts of interest exist, the committee participants have been screened for interest in companies that make infant formula.  As a result of this review, in accordance with 18 United States Code, Section 208(b)(3), Dr. Cutberto Garza has been granted a particular matter of general applicability waiver that permits him to participate fully in the matters at issue.  A copy of the waiver statement may be obtained by submitting a written request to the Agency's Freedom of Information Office, Room 12A30 of the Parklawn Building.

    With respect to FDA's invited guest speakers, there are reported interests that we believe should be made public to allow the participants to objectively evaluate their comments.

    Dr. W. Cameron Chumlea has a grant from Nestle to serve as a coordinating center for a nutritional study of Chinese elderly.

    Dr. Samuel Fomon previously consulted with firms that make infant formula and is likely to do so in the future.

    Dr. Duane Benton owns stock in Abbott Laboratories, and he receives retirement benefits from Abbott.

    Dr. Dennis Bier's employer, the ARS Children's Nutritional Research Center, recently received the Bristol-Myers Squibb-Mead Johnson nutritional 2002 unrestricted nutritional research grant.  As Center director, Dr. Bier is named as the principal investigator, although no funds come to him personally or for his personal research.

    We would also like to note for the record that Dr. Roger Clemens is participating in this meeting as the acting industry representative and a nonvoting member of the Committee.

    In the event that the discussions involve any other issues not already on the agenda, for which FDA participants have a financial interest, the participant's involvement and their exclusion will be noted for the record.

    With respect to all other participants, we ask, in the interest of fairness, that they address any current or previous financial involvement with any firm that makes infant formula.

    Thank you.

    With that, I will turn the program back over to Dr. Taylor.

    DR. TAYLOR:  Thank you very much, Jeanne.

    Our goal for the next 15 or 20 minutes is to set the regulatory context for the discussions we're having today.  For those of you that remember the spring meeting, we did spend some time on that, and hopefully most of this is a review, and all we have to add is an additional focus relative to the topic for today.  As I've mentioned earlier, this is an ad hoc task force of the Food Advisory Committee, addressing infant formula issues.

    Obviously, we have statutory authority relative to infant formula, and its long history goes back to 1980, at which time Congress passed special legislation that amended the Food, Drug, and Cosmetic Act.  We, in the Agency, try to avoid throwing numbers, and clauses and phrases around, but it's almost impossible not to, and the key phrase is that it provided Section 412 to the Food, Drug and Cosmetic Act.

    In 1986, Congress had an interest in adding to this, providing more statutory authority, and so there were some additional amendments in 1986.

    I think what we have to keep in mind is that the infant formula legislation happened for a very specific reason.  Infant formula is unique from other foods.  It is the sole source of nutrition for a vulnerable population.  In Congress's mind, it therefore warranted a special set of provisions for regulation.

    It's clear that the intent and outcome of this action was the following statement from Congress.  It should not only be safe, which I would point out is handled in separate sets of provisions than what we're addressing today and contain all of the necessary nutrients, which again is the separate set of provisions which are not on the table for today, but also should provide those nutrients in a bio-available form to ensure that the infant formula were to support optimal infant growth and health.  That's what we're about today in some respects.

    This chart is a little complicated at first, but I think it sets the context for what we call to be regulatory boxes, and really what's on the table today is this particular component, but all of this is the regulation of infant formula.  The safety of the individual ingredients, the classic safety considerations are handled under a separate set of provisions, the so-called Section 409.  So ingredients for intended use, that's where most of your classic safety reviews come in.

    Section 412, as provided for by Congress in 1980, is really a statutory check on a particular formulated product.  In providing those assurances, companies consider the required nutrients that have to be in the formula, the good manufacturing practices or GNPs and quality controls, and then quality factors.

    Today, we are focusing on quality factors.  These assurances are provided prior to marketing, and once marketing occurs, in the world of infant formula, the claims then come in, as far as efficacy, truthful, and not misleading, again, a separate set of provisions.  I'll come back to this in a moment, but the key component is that we're here taking a look at that.

    Now, as mentioned just a second ago, in order to provide these assurances, manufacturers submit a notification to FDA 90 days prior to their intention to market that particular infant formula.  Again, as I mentioned, it's specific to a finished product.  The Agency reviews it, again, as I just mentioned, for those three components, and it's here, hopefully, highlighted in red that our questions today will focus.

    The definition of quality factors is not precise.  They certainly do offer the opportunity of expanding, as needed.  There is language from a 1980 discussion in the House Committee, and their references to quality factors focus on things such as pertain to the bioavailability of a nutrient and the maintenance of levels or potency.  They discuss at great length the growth of infants during the first few months of life, and they discuss the concept of healthy growth, the idea being that once you've formulated a product, it needs to support healthy growth.  So, in its simplest form, quality factors are a check on the concern that once you get the entire product put together it works appropriately.

    Now the types of quality factors could be many.  At this point, we basically have two.  In the realm of nutrient-specific, we have provisions for protein efficiency ratios, protein per se, but over time, others could be put in place.  In the world of the formulation itself, the totality of the formulation the quality factor we address is normal physical growth, and, again, others could be put in place over time as needed.

    So, for today, normal physical growth as  quality factor is what's on the table.

    The scientific questions that will come through as you read the charges are basically twofold.  How do you measure and affirm normal physical growth and how and when should assurances of normal physical growth be appropriately provided?

    Going back to that again, this particular slide, quality factors, normal physical growth, assurances for a specific product, along with other components of these assurances.

    Now, just so that we're sure how it works from a regulatory perspective, we've put in this slide, but I think it's redundant to what we've said before.  In order to provide assurances, vis-a-vis Section 412, the manufacturer submits a notification 90 days prior to marketing.  FDA reviews the notification package, taking into all of the components, nutrients, GNPs, quality control and quality factors, and if assurances are adequately provided, FDA does not object to the marketing of the formula.

    If, in the Agency's opinion, assurances are not adequately provided, FDA does let the company know that it objects to the marketing of this particular formulation.  It's important to note, from a regulatory perspective, that this is not a premarket approval process, so manufacturers do have the right to go to market over FDA's objections.

    The scientific input we get from you folks today and tomorrow will certainly guide our thinking about the evaluation of normal physical growth when infants are fed a new formula.  We'd like to point out that it certainly helps us, but it's also helpful to stakeholders in that what is expected becomes clear to them.  It's not as much of a black box if it's quite clear how FDA's scientific considerations are handled.

    It's also going to guide our thinking about when clinical studies should accompany formulation of processing changes in infant formulas, and again it's helpful to us, but it's also helpful to our stakeholders.

    The outcome of today's discussions can be used to inform our ongoing reviews, but we do need to talk a little bit about current rulemaking, in that discussions today have the opportunity or the possibility of impacting on current rulemaking activities.  The current state of our rulemaking, as probably many of you know, is that in 1996, we proposed a rule to implement parts of Section 412, and in that was included the implementation of quality factors.

    That final rule has not been issued, so we are still in the process of what's known as rulemaking.  If input from this Committee is relevant, and it may or may not be, but if input from this Committee is relevant, there would need to be an opportunity to comment on that, and we would, of course, reopen the comment period on this rule for that purpose.  So we retain the option of reopening the comment period.

    So, again, just to review, you've seen this before, today's discussions, vis-a-vis the charges, our growth measures and methodologies, the role of such measures relative to normal physical growth, and the general principles and criteria to determine the need for a clinical study to provide assurances of normal physical growth.

    I think we've gone over topics not under discussion, so I'll mention these only in passing, and then again remind you that the specific charges from the Agency to the Committee are the seven questions in your notebook, and I'm sure Dr. Garza will go over those with you in some detail.

    I do want to introduce Dr. Susan Walker, who's at the table here, our associate director for Clinical Affairs, and Dr. Beth Yetley, who's our lead scientist, and the agreement we have is if you have questions, I will go join them, and we will answer them as a troika.

    Thank you very much.

    MS. LATHAM:  Are there questions?

    DR. GARZA:  Are there any questions to Dr. Taylor?

    [No response.]

    DR. GARZA:  Thank you very much.  That is clear, judging from the lack of questions from the Committee.  I do want to take this opportunity to welcome the Committee members, and guests, and staff that have joined us, and to thank Dr. Taylor because we've made up the lost time.  I was concerned that we would be running late, but we're doing all right in terms of time.

    We have a very full agenda and would like to begin by asking the Committee members if, in fact, they have any questions about the agenda.

    [No response.]

    DR. GARZA:  Very quickly, just to review the procedure, we're going to be launching into one of the major segments of the Committee meeting in just a few minutes, and that is nine presentations, based on the nine background papers which were sent to each of us several weeks ago.  Only Committee members have the privilege of asking questions to any of these presenters.  We will try to hold presentations to about 15 minutes.  Someone will be helpful in alerting the speakers when I think there is about three minutes left in their presentations to help them allocate their time appropriately, and then we will have about 10 minutes of questions from Committee members to each of their presenters.

    You also will have the opportunity to ask questions of those that make comments in the public comment period tomorrow, and we will have blocks of time then to come to some consensus on the seven questions that you have been sent as well.

    It's going to be very important that we address each of those questions carefully, and therefore I'd like to make sure that each of you has a chance to review those questions because I will be proposing time limits to assure ourselves sufficient time to deal with each of them in a way that doesn't shortchange any of them, and so we'll be trying to deal with that time allocation later this afternoon.

    All experts, I am told, will be able to stay throughout today and tomorrow for those questions and answers to that if, in fact, in those blocks of time when we're dealing with any of those seven questions, any of the committee members would like to address any questions to any of the presenters, and then obviously that's going to be possible as well.

    You have seven questions.  They have been divided for us in four sections.  One of those sections is on metrics for the evaluation of normal physical growth, a second section deals with comparators for the evaluation of normal physical growth, and a third is on controlled feeding comparators, and the fourth is on changes in infant formula composition.

    Rather than reading each of the seven questions, I'd like to take just a few minutes to ask committee members if you have any questions about the issues that we've been asked to consider under each of these sections so that, in fact, we can be clear what we're being asked to do, and you can have those clearly in mind during the presentations.

    So let's begin with the first in terms of metrics for the evaluation of normal physical growth.  There are two questions under that section.  Do any of the committee members have any questions about points that you're being asked to address?  I'll give you a few minutes to review those, and we can ask the troika to clarify those for us.  I think that's the way you were described.  That was not my word.

    Having had the pleasure to work with them, you'll get informative responses to your questions, I'm sure.

    MS. LATHAM:  At the end of last week--this is Jeanne Latham, the exec sec--we e-mailed to everyone the updated questions, and I just wanted to make sure that you all have those, and if you don't, we will get them to you.  You've got them. . Anybody that doesn't have them?

    [No response.]

    DR. GARZA:  I think they were in the packets today, again, in case you didn't bring them with you.

    Would any member of the staff want to address any questions in this section?

    [No response.]

    DR. GARZA:  I take it, then, that they're clear--Dr. Thureen?

    DR. THUREEN:  Yes, I have one question.  These are metrics for the evaluation of growth between birth and six months of age.  We will be dealing with both term and preterm infants, I presume.  Should they be handled separately?  Because the preterm infants we are discussing I believe growth after post-conceptional age birth to six months, so should they be handled separately or should we just do a general assessment of these evaluations, presumably for term infants, with maybe later adjustments for preterm infants?

    DR. GARZA:  No, I would assume that we will take those separately, but let me ask the staff if they would object, if there's any reason why we shouldn't take them separately?

    Committee members?  So we'll probably do A and B.  Thank you.  That's a good clarification, with preterm being all preterms, low birth weight, very low birth weight, and extreme low birth weight.

    Any other questions, then, on this first section?

    [No response.]

    DR. GARZA:  Then, on the second, on comparators for the evaluation of normal physical growth, let me give you a few minutes to review those two questions and see if there are any issues that need to be clarified.

    [Pause.]

    DR. GARZA:  Any questions on either of those?

    We have one question on the control  feeding comparators.  Let's take a look at that and see if that's clear.

    DR. ANDERSON:  This is Jim Anderson.

    I wonder if I could get a clarification of the difference between the current infant formula plus new ingredient that's listed on the first bullet and the infant formula plus new ingredient with the asterisks on the last of the bullets.

    DR. GARZA:  On the last bullet, right?  So it's listed below are examples of controlled feeding clinical comparators, and I believe the question is can you clarify the distinction between the first and the last bullets.

    DR. TAYLOR:  We're working on it.

    DR. GARZA:  I gather what it meant was that it was a study in which the new ingredient would be used with some infant formula, but that the intention was to market the new ingredient as a component of some other infant formula.

    DR. GARZA:  That was mine or a generic infant formula, where that new ingredient might be added to any formula, so that it would be a generic comparison was the way I read that.  Am I not clear of that?  If my interpretation is correct, so it's a generic--

    DR. WALKER:  Your interpretation is correct.

    DR. GARZA:  So it's a generic endorsement of the ingredient.

    DR. WALKER:  Right.

    DR. GARZA:  Does that clarify it?

    Dr. Walker, would you--

    DR. WALKER:  The instance in which the generic ingredient is added to the new infant formula and the way that you discussed there, Dr. Garza, is a correct interpretation.

    DR. GARZA:  So it's more an endorsement of the ingredient itself, with a test formulation, but as opposed to a specific formula that had been marketed in the past, where a new ingredient would be added.  I understood it to mean a more generalized evaluation, rather than a specific one.

    DR. WALKER:  I think after we have some of the discussions, some specifics of these will be made much more clear.  I think the speakers will address a lot of these issues in detail, and then we can have more questions.

    DR. GARZA:  And I've been reminded that we must each identify ourselves before we ask questions.  I will try to remember, but I'm probably the guiltiest of all.  I would hope that they would recognize my voice before this meeting is over.

    The statement I was supposed to read said, "I am Cathy DeRoever."  I thought I better not say that.

    [Laughter.]

    DR. GARZA:  It would confuse people, and obviously embarrass Dr. DeRoever as well.

    Are there any other questions on this third section?

    [No response.]

    DR. GARZA:  If not, then, the fourth one is the last.  It is also one question with two parts, A and B.  The table that is attached, obviously, is quite informative, so I would ask you to take a few minutes to look at Question 7, along with Table 1, see if there are any questions.

    [Pause.]

    DR. GARZA:  Are there any questions related to this section?

    [No response.]

    DR. GARZA:  If not, what I propose is the following; that you think about 30 minutes for each of the first six questions, approximately 120 minutes for the seventh question.  If we don't need the entire two hours for the last question, we can always come back and address issues that perhaps we might have felt were not completely resolved.

    Obviously, if it's clear that we need more time with any specific question, as the discussions evolve, then we can always go back and try to reallocate them, but we will come to some agreement on how much time to spend with each before we start the discussion this afternoon.  But as you've had a chance to review them, see if, in fact, you feel comfortable beginning with that type of allocation for the various questions that we're being asked to review.

    I don't think that all first six would necessarily take 30.  Some may take a bit longer, and some will take less, but I want to make sure that we don't shortchange, as I said, any specific question, and so we can get agreement on that as a group, going through the discussion and developing some information that would be useful to the FDA I think is more likely.

    Are there any other questions regarding procedures or the charge to the group that either staff or I have failed to clarify for you?

    [Pause.]

    DR. GARZA:  If not, then why don't we begin with the presentations.  I know that Dr. Chumlea was on the bus, so I assume he's here.  We're starting a bit early, but I think that's fine.  I will just introduce each speaker as they come forward.

    Dr. Cameron Chumlea is a Fels Professor at Wright State University School of Medicine in Ohio.  And for those of you in the field of anthropometry, I don't think Cameron needs any introduction.  For those that may not be familiar with his work, we don't have time.

    [Laughter.]

    DR. GARZA:  It has been quite extensive, and he's certainly recognized throughout the world for his work in this area.

    Thank you very much for the white paper and for joining us this morning, Cameron.

    DR. CHUMLEA:  Thank you very much for the introduction.  It's a pleasure to be here this morning, and I hope I can provide the committee with some information that's appropriate.

    First, I'd like to just simply recognize my co-author, Dr. Shumei Sun, who I know is familiar to many of you and point out that she has just recently become Wright State's Brage Golding Distinguished Professor of Research.

    So, first of all, I'd just like to point out that growth is relative, as you can see from the slide here, it says, "I keep track of my son's growth, which is going up the vertical scale, and my husband's growth.  Frank is age 30, 33, 35, 40."  So growth goes in various directions for all of us here.

    The second thing here is that we're dealing with infants.  Of course, there's our perspective of what infants are, but there's also a public perspective, and I saw this checking out the groceries and decided it really deserved a slide because clearly this, to some degree, is maybe the public's viewpoint, which you can clearly see.  It's amazing what you can do with PowerPoint these days and some slides that are available.

    So what I'm going to do this morning is just basically cover some brief information that's probably familiar to everybody, so that we can just kind of all come up to speed.  Clearly, infancy is a period of rapid growth, and to some extent, this is probably the most difficult group of individuals to measure.  The only other group that's equally difficult is at the opposite end of the age range, but this is generally a very difficult group, but it's also easy in the sense that there's also very few actual measurements that can be collected from them that are really going to be useful.  Weight, recumbent weight and head circumference are the three that are the most important and the ones that should be taken.

    Just to review, weight clearly measures the growth of all body tissues, recumbent length describes the amount of linear growth because we're dealing with both increase in mass and increase in size, and then head circumference reflects brain growth because this is the period of time, the first few years of life, when the brain actually does the majority of its growth.

    This is a period of time when body dimensions increase at a greater rate than in any other period in life.  Weight increases between birth and six months about 115 percent, length increases about 34 percent, head circumference increases 22 percent on average.  The rate of growth in weight ranges from about 1.1/1.2 kilos for boys or girls at one month of age, but then we have to remember that we're on a growth curve here, and then it, of course, starts to slow down, and by six months it's running around a half a kilo a month for boys and girls at six months of age.

    The rate of growth in length is about 3.5 to 4 centimeters per month for boys and girls, and it slows to about 1.5 to 2 centimeters a month for girls and boys at six months.  Just as a reference, the adolescent growth spurt between, say, you know, 12 to 16 years of age, the maximum amount there is only somewhere in the neighborhood of 5 to 8 centimeters a year.  So here we're looking at 3 to 4 centimeters a month.  So they can just put that in comparison because everybody kind of focuses and remembers how much growth their kids did when they were adolescents.  They frequently forget that that rate was a fraction of what they were really doing in the first few years of life.

    The assessment of status, once we've collected measurements, we really need the measurements to be accurate and reliable, and this is really a very critical point, particularly in this particular age range.

    The measurements are really not difficult to take, and there are a variety of mediums in which they're now described.  NCHS produced a video at the end of NHANES III that describes these and all of the measurements that were used in NHANES III on that videotape.  WHO has an in-house video that describes the measurements that are being used in their multi-center growth reference study.  I don't think that one has really actually been distributed yet, but it is available if you can talk to the right people.

    These are also all very similar techniques.  They are also probably being what's currently being done in the current NHANES, and they're all the ones that have come out of the Anthropometric Standardization Reference Manual from 1987, by Lohman.

    When we collect the measurements, this is a point where frequently things get skipped over, and the reason things get skipped over is because there's lack of time, money, personnel, et cetera, but it's the part that's really very important, in terms of collecting the information because, one, we're going to either plot those on a growth chart or refer to status, and the other aspect is what's very important here is we're going to calculate increments of rate of growth, and there you're compounding your measurement errors.

    It really takes two people to measure an infant appropriately.  I'm sorry, folks, but that's really the correct way to do it.  It can be done with one, but that's going to add to the errors that are going to be involved.  You need an examiner who's going to position and take the measurements, and you need a recorder who's going to be writing down the measurements because one person is holding the infant and trying to do the measurements, and they really don't have time to write anything down, and then they need to switch roles because we need to take double measurements here so that we can get as much information as possible.

    Also, what we'd like to do is have the technicians compare their values.  One, this just simply catches transposition errors that occur frequently because people write down numbers in different ways sometimes.  All of the studies that I've described to you have allowable differences between what the measurement values can be between the technicians, again, to control for errors and just to control for variabilities that can occur.

    Did I skip one?  How do I go back on this?

    [Pause.]

    DR. CHUMLEA:  Weight.  An infant can be weighed alone or they can be weighed while the mother is holding them.  It depends a little bit on the situation, but, frankly, I would prefer that the mother hold the infant.  You can weigh the mother, weigh the mother holding the infant, take the subtraction and you're going to get the weight.  The reason I like the idea is it keeps the baby calm, and it provides for a very stable piece of information.

    The infant can be weighed alone, and there are a variety of electronic scales, but pretty much once you take the baby out of the mother's arms and you place it on something else, it starts moving around.  Fortunately, the manufacturers of several of the scales now can compensate for this weight so the stability of the measurement is much better than it used to be.

    It's best that they be weighed nude.  Blankets, et cetera, are available.  However, if they are going to be in undergarments, I think NCHS subtracts about a tenth of a kilo from that for the readings, and spring-type scales and beam balance scales are simply not appropriate for use any more.  The electronic scales are much more available.

    There's a company called Seca that makes some very good scales.  We've used those.  WHO has a really nice platform scale that actually pares the mother's weight.  I don't know exactly what the manufacturer is, but it's a really excellent device.  I think it's been specially made for them, but there are a variety of scales that are available for use, but my preference is that the infant should be held with the mother or the caregiver, whoever is there, and then subtract the weight, if that's possible.

    This is the one nobody really likes.  It's recumbent length.  It takes two people to do it, again.  A variety of different pieces of equipment that are useful for doing this.

    It requires one person holding the infant's head.  It requires the mother, the caregiver standing there beside the infant reassuring them that nothing is going to happen, and it takes a third person then to position to footboard up against the soles of the foot of the infant.  You're holding the head so the child is looking straight up in a vertical Frankfort plane, and it takes then another person to hold the legs, both legs, for the infant, if it's very small, and get the length, and by the time they're six months, generally, the best you can sometimes do is grab one leg and try to hold the other one with your little finger and get it.  It's not easy to do, and it's one that can be particularly prone to error, and it's also important that the kid stays straight down the table.  I think I've covered what's there.

    This simply gives you a description.  I think this child is about two, the age there, but again positioning the head up against the headboard, keeping the legs straight and keeping the feet straight up in terms of taking the measurement, but clearly it takes two people to do.

    Head circumference, it should be measured with an inelastic tape of fiberglass, metal, something like that that's good.  It's really best that the infant is seated in the mother's lap.  I don't like it being done with the infant.  I've seen that done.  I like the kid up in the mother's lap, which requires then the person getting, who's taking the measurement, to get down beside the mother.  This allows the mother to cuddle the infant, keep him quiet, and you can slip the tape over their head and get the measurement before they pretty much know what happened to them.

    It's placed right across the front of the skull, and it can be quickly moved up and down the back of the skull.  The insertion tape is a nice piece of equipment that's useful for doing this until you find the greatest circumference, and then you pull the tape tight, and this is something we have to coach people in because it doesn't hurt the infant.  There's no pain involved with it.  They feel a little pressure, and they want to kind of shake it off, but it does need to be tight.

    And you can see here it's just anchored really right over the kid's eyebrows, worked up and down.  Most kids at this age aren't going to have quite this much hair, and so it's generally pretty easy to get this one from them, particularly, again, with the child being comforted by the caregiver.

    Now, what are some other measurements that could be taken?  Well, there's really a bunch of them, but I don't think those are really going to be really appropriate in the instance here.  Crown-rump length is sitting height.  Crown-rump length, I think that was pretty only used in children with special cases.  Chest circumference, limb lengths, one that's potentially possible is skinfold thicknesses.

    The problem with all of these measurements is they really kind of have a restricted utility in terms of describing normal or healthy growth.  They're frequently prone to high measurement errors, and there's really a limited amount of reference data available for all of these measurements pretty much within the age range that we're looking at, six months.

    If you're going to do something like a skinfold.  The skinfolds you're going to take are going to probably be triceps and subscapular.  Now, again, you're bringing two people into the program, if not three, and you've got to go to landmarks, and to do a triceps, you've got to find the midpoint of the arm, which means you've got to measure it, and make that determination, and you can see clearly it's requiring one person is holding the child and the other person is taking the measurement, and then you're going to have to go and take the measurement.

    The question comes up there in terms of equipment.  Skinfolds are dependent upon the type of caliper you're using.  There's two major brands, really.  There's a Lang and a Holtain, which you see here.  NCHS and WHO are both using the Holtain caliper.  I think the Lang is still used out there, to some degree, but I'm not that familiar with it any more.  There are differences there.

    Really, if you're wanting to do skinfolds, the question is what do you want to get out of that?  And probably what you wanted to get is total body fat, and there's probably, I think Dr. Ellis is probably going to talk about better ways of doing that today than taking skinfold measurements.  I don't really feel comfortable in doing it on anybody until they're about two years of age.  It's just difficult to do.

    Now, there's indices that can be used from the information that's collected.  BMI is the one that we all get informed, and just for your information, I'm 28, so you can kind of put that in reference.  I always think everybody should, when we have meetings like this, they should always walk around with their BMI on it, so we'll all be honest about this folks.

    [Laughter.]

    DR. CHUMLEA:  The problem with BMI is that, in infants, you've got 25-percent body length is composed of the head, so that throws off the proportionality aspect.  The relation of BMI with direct measures of body composition in infants hasn't really been established.  Weight for length is probably a better descriptive indices of relative leanness adiposity within children.

    Measurement error is very important and needs to be paid attention to.  The catch here is that the error which may be small is actually going to be very large because of the small size of the child that you're measuring.  So you really need to pay a tremendous amount of attention to error in measuring infants.  Of course, they can have a tremendous impact on the interpretation, if you're going to go growth increments.

    We need to get good-quality equipment.  Measurements should be taken.  If they're taken on a daily basis, the equipment needs to be calibrated.  That includes scales.  People forget that scales can go out of calibration, and then particularly the technicians need to be trained in a standardized way of taking the measurements.

    We need to collect inter- and intra-observer reliability.  Quality control is really important, particularly if there's going to be more than one center used to collect information because we need to control for inter-site differences.

    Measurement schemes.  You need baseline, an interim and a final.  I really like something that's going to be getting a measurement at 1, 2, 4 and about 6 months of age is my preference for collecting things, generally, starting after about 10 to 14 days.  With measurements at 1, 2, 4 and 6 months, you're going to get a good accountability of weight measurements over that period of time.  Clearly, if you can collect more measurements, the more measurements the better.  I'd be very happy to have those.

    I'm going to cover just very briefly growth increments, which are going to be calculated from the repeated measures of growth, and there are charts that these can be plotted on from birth to 12 and 3 to 6 months of age, which are examples here.

    These are from Fels data, and just contrary to popular opinion, the majority of Fels infants were breast fed for at least three months, exclusively, so that has been reported.  We tend to not get a good press on that.  I just want to kind of correct that.

    In terms of growth velocity data, there's the Fels data.  Also, I'd like to point out that WHO is collecting longitudinal data from its multi-center growth reference study, but this data and report from that study has not been available, and that'll probably be given later.

    So recommendations, from what I've just described to you, weights should be measured I think at 1, 2, 4 and 6 months.  I'd like to see recombinant length and head circumference at the beginning and end because it just gives you additional information on the quality of the size of the infant.  Close attention needs to be given to methodology and errors.  Two technicians are really important and reliability data needs to be collected, and use of existing increment charts until the WHO charts are available.

    So thank you very much.  I'd just like to say, personally, the last time I had to give a paper in front of Dr. Briley, she gave me a B--

    [Laughter.]

    DR. CHUMLEA:  --which is about 25 years ago.  So I hope I did at least that good this time.

    Thank you.

    DR. GARZA:  We'll take that up at the break, I guess.

    DR. CHUMLEA:  Okay.

    [Laughter.]

    DR. GARZA:  Thank you very much, Dr. Chumlea.

    Are there any questions?

    DR. SIGMAN-GRANT:  Sigman-Grant.  I have a question.  You recommend weight starting at one month, and you talked about the regain from the loss from birth weight.  So much is happening in that first month.  Why don't you measure it before--

    DR. CHUMLEA:  I said that, really, as early as, say, 10 to 14 days.  I would like to see it done that way, but within no later than one month of age.  So let me kind of restructure that between--

    DR. SIGMAN-GRANT:  Why not between--why not the first week, instead of 14 days?

    DR. CHUMLEA:  Well, there's a shift in weight, as far as I'm familiar with, that's supposed to occur after birth, and so I think, just my understanding is, that there's a period of time within the first week or State or local that the infant basically kind of stabilizes after the birth experience.  Now, Dr. Fomon, I'm sure, could give you more information on that, if I'm incorrect on that.

    The more measurements you can get out of this, the better.  I was being, trying to give you what I think is the very minimum that you have to collect there.  You can measure them every week.  That would be fine with me.

    DR. DOWNER:  Goulda Downer.  I understand, when you talked about examiner variability and possibly downright error, but can you talk a little bit more about why you don't think that subscapular skinfolds and triceps skinfold are important--

    DR. CHUMLEA:  Useful information?

    DR. DOWNER:  Yes, at this juncture, because I think it is.

    DR. CHUMLEA:  The reason I don't like them is this.  First of all, they're extremely difficult to collect in children at this age, so the amount of error that's in the measurement is extremely high.  The question I would have is that what information are you going to get out of this particular measurement that you're not going to be getting by bodyweight alone?  Because if weight is going up, the skinfolds are going to go up; if the weight is going down, the skinfolds are going to go down.

    So the question that you're really interested in is total body fat, and, yes, 90-some-odd percent of total body fat in a child is principally subcutaneous; that if you want to go total body fat, there are now better ways of doing that, such as DXA, that I think are going to give you the information that you really want.

    If you do go and collect the skinfolds, then you're faced with some reference values that are useful.  There's only two that are out there that are fairly good--what's available from NCHS and NHANES III, and then Dr. Fomon's data on skinfolds.  But outside of that, there's really little other reference data that's available.

    So I guess if I'm going to, what I want to know is total body fat, and if I want to measure total fat, I'll go measure total body fat with something that's going to give me I think better, and more accurate, and reliable information about the child than I'd get from the skinfold.

    DR. HEUBI:  Jim Heubi.  I don't want to misunderstand what you're saying, but you're not recommending that people weigh infants in garments and subtracting--

    DR. CHUMLEA:  I'm sorry, what?

    DR. HEUBI:  You're not recommending that people weigh infants in garments and then subtracting one-tenth of a kilogram for the--

    DR. CHUMLEA:  I would prefer they be weighed nude, yes.

    DR. STALLINGS:  Stallings.  To follow up a bit on the other question.  We're beginning to think I think about looking at infants who are growing too slowly as the historical way of the failure to thrive related to this, but the issues of growing too fast are also of concern.

    So, to go back to the question, could we get most of that information with weights and heights, and weights for heights, rather than looking for data related to adiposity, or if we were looking for excess growth, whatever that concept means, what would you recommend, derived values from the anthropometry or DXA?

    DR. CHUMLEA:  I think--we're still sticking between this birth and 6-month range, and I think if you're getting excess growth, you're going to, you may need to take more frequent measurements so that you can plot and get a better description of the curve as what's going on there.  And if you have more information, then you can discriminate between the children who have excess growth and those who don't.  So that's I think something that's important to consider.

    I think you will get everything you want from weight and length.  If you go to DXA with a child at this age, and Ken will address this more I think in his talk, you're going to get, you know, fat, lean and bone, and the fat is probably the most important aspect here that you'd be concerned about for excess growth, but it's going to be described in weight, also.

    So unless you're wanting to tease a tissue out and say, okay, we're really concerned about the increase in fat here, in addition to the increase in weight, then, yeah, then something like DXA I think would be important.

    DR. STALLINGS:  Follow-up.  The velocity, then, would be what we would be looking at, more than just attained weight?

    DR. CHUMLEA:  I think you have to do both of them.

    DR. STALLINGS:  We, historically, are always looking I think at the attained weight.

    DR. CHUMLEA:  Yes, and you'd have to include the velocity in there because the velocity would, these children should, I think, potentially have much higher velocities.

    DR. STALLINGS:  And that might be a way of discriminating between the concept of normal growth and excess growth?

    DR. CHUMLEA:  Yes, right.  You could have children, let's put it this way, who have, say, after, say, three or four months, when their velocity should be declining, these children might not be declining as what the average is, so they're still obtaining a rather high velocity of growth at that point.

    DR. THUREEN:  Thureen.  I would argue that DXA is not a very useful body composition measurement for most studies because it's not that readily available, especially as a field tool, and I think that a lot of people are now starting to do more caliper measurements for assessment of body fat, even in very tiny infants.  And certain people, like Suda Kashyap, have gotten very reliable measurements over time.

    Do you think that the data from the NHANES study on body composition, using anthropometric measurements of body fat, was not useful or do you think it is useful?  And if you want to do large population studies, do you think that there is a future for caliper measurements?

    DR. CHUMLEA:  The DXA thing you can talk to Ken about.  I'll let him address that when he gets up here.

    There's two issues here, whether we're talking about small studies or large studies.  If you're wanting to do large-scale studies, population studies, like NHANES has done with NCHS, then collecting caliper information is going to be what you can do because there can be limits to what you can collect for DXA, particularly because of the issue of radiation exposure, although it's very minimal, even within the current NHANES, where they have DXA machines in all of the trailers, I think the limited age there is Age 8, from what they're collecting, although they technically have the availability of doing it in those particular studies.  Other people clearly don't have access to such expensive pieces of equipment.

    If you don't have access to that, then I'm not opposed to collecting the skinfold data.  The issue comes up that it is extremely difficult to collect accurately and reliably, and so, I guess, I kind of am in favor sometimes of no data is better than bad data, and I know that, in collecting it, it is something that people have to pay very close attention to, the technicians have to be very careful, and this is frequently something that in the course of studies, we pay lip service to it, and there's good attention, but these things do tend to fall out.

    Now, in smaller scale studies, this is something that can be done, and the information can be collected.  Overall, I'm just not that happy with the information.  Yes, there are studies where it's been done very well, and so I'm not putting those studies down at all.  I'm just talking about, in general, my experience has been, in collecting from infants in this age range, that this is really hard to do, and when it's something that's hard to do, it doesn't sometimes always get done the best way.

    DR. THUREEN:  Thureen, one more question.

    In your opinion, if you're looking at a growth outcome study, do you think incremental data are the gold standard--growth data are the gold standard, attained growth are both critical to an outcome study?

    DR. CHUMLEA:  I'm assuming, when you say a "growth study," you're going to be collecting repeated measurements from the same children, so you're going to have both pieces of information available there.

    The status value simply describes where the children are in reference to whatever reference values you're using for peers at that age.  That simply tells you that they're at certain percentile levels, but at the same time children also grow at different rates, and so there's a distribution of the rates at which they grow.

    So children who may appear to be at one percentile level, their rates of growth can be a different percentile level, so it gives a much clearer picture upon what's available.  And since any study where you're going to collect repeated measurements, you're going to have all of that information available to you.  So I would take advantage of it.  Again, the errors are difficult to control and need to be paid attention to for collecting it.

    DR. GARZA:  Cameron, I have two questions.  Given the fact that we are going to be providing advice to the FDA on the approval of specific formulas, how many measures do you recommend be taken if, in fact, one has an interest in the pattern of growth?

    DR. CHUMLEA:  If I was going to design the study, and you're not going to restrict me to what I want, okay.

    DR. GARZA:  From your perspective, if you're going to be protecting the public health and infants' health, what should the American public ask?

    DR. CHUMLEA:  I'd want a birth weight.

    DR. GARZA:  Birth weight.

    DR. CHUMLEA:  Clearly.  I would like it at, say, two weeks, one month--I'd like it again at two, and then at four, five, and six maybe, something like that.  The more measurements I could get out of the thing the better.

    DR. GARZA:  But you think that with seven measurements, one would be able to assess both the pattern of growth, as well as velocity of growth, at those specific time periods.

    DR. CHUMLEA:  Yes.

    DR. GARZA:  And you mentioned there were various sources of error.  Is there any consensus that FDA could rely on that deals with the nature of the equipment, the type of calibration that should be insisted upon, the training that obtaining the measures should be able to obtain and document, and the--those would be the three: equipment, calibration and the training of the technicians.

    DR. CHUMLEA:  There's a little bit of information about equipment errors and a little bit of information about the inter- or intra-observer errors for collection of measurements in the Anthropometric Standardization Reference Manual that's collated in one location.

    There's other pieces of information that are clearly scattered around the literature that are available.  From NCHS, there's really a limited amount of information.  There was really limited error data that was collected in NHANES III.  What's available from NCHS is principally from the earlier NHANES studies, and NHES.  So there's not much there.

    The techniques are described in a variety of locations, but there's not really anything that I know of that's really written down that says, okay, you can refer to here, and this is what you should do, in terms of training, collecting the measurements, et cetera, in one central location, no.

    DR. GARZA:  Any other questions or comments?

    [No response.]

    DR. GARZA:  Thank you very much.

    DR. CHUMLEA:  Thank you very much.

    DR. GARZA:  We'll move on, then, to the next presentation.  It's a topic that has already come up, body composition assessment in early infancy.  Dr. Ken Ellis, from the USDA/ARS Children's Nutrition Research Center, with Baylor College of Medicine.

    Again, Dr. Ellis, thank you very much for joining us.

    DR. ELLIS:  Thank you.  What I'm going to present today is a probably a little different from what most people have had experience seeing.  Some of this--in fact, all of this is going to be body composition beyond simple weight.  If one was interested in what the composition of weight is--let's see.  Which button do I...this is the laser, okay?  And which one of these is--can you hear me now?

    [Pause.]

    DR. ELLIS:  This is supposed to move the slides, she said.  I'm so used to pointing to the screen these days, so it's my fault.

    As you already heard, most people--or most of the work, at least in infants, has been with weights and heights, all kind of weights and heights charts.  But body composition, at least the first models that were attempted, the basic classic 2-compartment model is to measure--is divided in two compartments, fat and non-fat.  The direct measurement of body fat is really very difficult to do.  It's not an easy process to do that.  And so for many years what we did was we said that if we could measure precisely some parameter of the body that would represent the fat-free mass, then subtraction of the fat-free mass from the total weight would give us a measure of the fat mass.

    Part of the problem with that is that all the years in the  (?)  scale for the fat-free mass translate directly to the fat mass, and you'll see what I mean by that in a few minutes.

    Three classic methods that have been used for 50 years, or maybe even longer:  underwater weighing in adults--you can't do this in babies and infants.  Practitioners as well as parents tend to object to holding babies underwater.  Hydration, this is probably the more common thing you'll find in the literature that is done.  It's the dilution technique.  You give a tracer, collect the blood sample or some fluid sample several hours after that, then do some manipulations on that,  (?)  space to get water, and make some assumption about how much water there is in the fat-free mass.  And as we all know, hydration content of the fat-free mass in children at very early ages changes dramatically.  So depending upon what you assume, you can then quickly be off in your estimate of the fat.

    Whole-body counting is another method used to measure potassium content, primarily the body cell mass.  And, again, how much that relates to the--how much of that is a constant fraction or not of fat-free mass at these ages has also been questioned.

    So, again, like I said, the difficult underwater weighings, difficult to do the infants.  For the water measurements, they must swallow all the tracer, collect some kind of fluid sample.  Plasma is your best choice.

    The problem with this is you can't repeat it.  If you want to do the trial a week or so later, you can't.  You need to leave a sufficient amount of time for the previous tracer to clear, or you start increasing the doses of the tracer to compensate for that.  And, again, in the past, the most accurate assays required one to have a mass spec or availability of a mass spec, which is not in everyone's laboratory or garage, as I usually say.

    Whole-body counters, the problem with those have been over the years really they haven't been designed for infants.  They're really designed for adults.  There are a few of us who have done this, but in general they simply don't exist.  And even if you do have a whole-body counter available to you, most of the time it's not in a clinical setting.  You're going somewhere else to get the measurement done.  Again, the reason that tends to exclude it, at least for infants and children.  But one nice feature is you can repeat this as often as you wish, and so you could do it on a daily basis if one chose to do that.  So if you have access to a counter, one can count these infants as frequently as one chooses.

    And because of the way in which the counting procedure works, there doesn't have to be any really significant constraints.  They can move around and be--it will not really affect the results that much.

    Now, I looked at the various things that we received in reference to this meeting, and there's the document in there from the American Academy of Pediatrics where there's one little paragraph on body composition.  This was in June of '88.  And it says, "Normal growth implies appropriate composition of the increment in body weight.  Sequential measurement of various aspects of body composition"--such as water, fat and bone--"have the potential for defining changes in body composition."  However, at that time the opinion of the Task Force was that such measurements have not yet reached the stage of precision, non-invasiveness enough, and they're not that very convenient, and which I've just showed you that sort of in the three previous procedures, which I would agree in 1988, I would agree with that statement.

    That gives us a quick summary of that.  It just says you want to be able to--that the weight is appropriate composition, you will be able to do longitudinal measurements.  Again, for clinical testing, the precision, noninvasiveness, and convenience are the issues that they put out.

    Again, the 2-compartment model, I quickly talked about the limitations.  The density is not constant.  The hydration is not a constant.  Extracellular and water ratio is not constant.  Bone accretion is not constant.  Basically babies aren't constant.  We know that, right?  Because we know that when the baby comes back a month later, it's not the same baby we saw a month before.  It's a whole different child totally, at least from a body composition point of view, unlike adults, which really change very slowly over time.

    Now, so what's happened in pediatric body composition research since 1988?  There are sort of three general areas where there have been advances made.  One is bioelectrical techniques.  In pediatrics, I want to take a broad sense of pediatrics.  That's anybody under the age of 18.  And there's been quite a bit of work used with this technique in older children.  When it has been tried or used in infants, it has not been very successful at all.  In fact, in general, most conclusions with this technique--I should say the first two techniques, the bioelectrical impedance and the bioelectrical spectroscopy, which is the same as this but at two frequencies, have been that the information gained from those two techniques really hasn't been much more than you already knew when you had just simply weights and heights with these individuals.

    TOBEC, on the other hand, has been more successful, but the problem with TOBEC is that these machines are very--they're not common, there are not many out there.  There's probably not more than a dozen--a half a dozen, in fact, for infants.  And so, again, this is a technique that holds promise or has held promise, but, again, it's not a technique that is widely available at all.

    Absorptiometric techniques, DXA.  Remember that Academy of Science report?  It was in 1988.  Well, in 1993, the world of X-ray absorptiometry changed because at that time it went to what's called DXA.  It went to X-ray sources, different detectors.  The whole technology advanced substantially such that one could now consider this technology for whole-body measurements in infants, and one can do a localized region, such as the spine, if that's specifically what one would want to do.

    Another area, body volume measurements.  Remember, underwater weighing doesn't work.  This is an air displacement plethysmograph technique now, which is just actually started this year, so maybe in another few years, when this committee reconvenes, we'll be able to talk more about this.

    I'm going to basically focus on the absorptiometry or DXA or DEXA methodology because that right now holds the best promise for this kind of broad application.

    So the basic model, again, in 1988--I mean, there are more models, but the very basic model, again, we're talking about fat and fat-free mass.  And today, when one talks about body composition, one is almost forced pretty much to really address this kind of a model over here, 4-compartment model:  fat, bone mineral, mineral composition, ash, the water, and protein content.  And, in fact, it is these compartments that we're interested in when we look at body composition at any age and look at change in body composition at any age.

    So that 4-compartment model is now shown on your left, and what does DXA provide us?  DXA provides us--the only method we have that provides us a 3-compartment model for a single measurement.  We have a measurement of fat, which is directly the fat.  We get a measurement of BMC, which is bone mineral content, which is for the mineral compartment.  Eighty percent of this is this; there's another 20 percent which is distributed in the non-mineral, non-osseous compartments.  Then a third large compartment called the lean tissue mass, which is the non-bone, non-fat compartments.

    So when we do a DXA measurement of anyone at any age, we get this basic model, which has been a huge advance.

    This is what an image looks like in a child.  Most of the time people show images of adults.  Adults are very nice.  They hold their arms right.  They put their legs down straight.  Children, amazingly, 14 years of age, haven't learned how to do that yet, have they?  So they tend to lay the way they want to.  We do make measurements in children at our place.  We've done probably, I don't know, 600, 700 DXAs, at least.  Maybe a thousand.  I really don't know the number these days.  Quite a few studies.

    Let me show you, again, if one is interested in just the spine, this is not an infant spine.  It happens to be an adult image there.  But one can localize and make a measurement just at the spine for bone.  But for body composition work, we do total body measurements, which is shown here.  I think you can get an idea of the skeleton seen.  I think you can see the soft tissue parts that are obviously not the bone, and we can then get that information.

    Now, as you heard earlier, the BMI in terms of the height of heads, if one chooses to, one can actually decapitate the image and just worry about this part if you're concerned about how does the head contribute to all this information.  So it is possible to do that kind of stuff.

    By the way, the time it takes us today--in 1988, if you attempted an infant, which you would never be successful at, it would be 25 minutes.  Today we do an infant in less than 3 minutes with the newer scanners.

    So DXA, what are the advantages and some of the disadvantages?  First, DXA has almost achieved a reference status within the body composition field.  It still has some improvements to be made, but, again, in terms of everything else, it is the better technique that we have.

    Advantages, it does give us good precision and accuracy.  It is the only technique for a single assay that gives us basically a 3-compartment model:  bone, fat, and lean.  As you can see from that image, we can get some regional information if we choose to do that.  It has a very low exposure risk.  There's a very minimal amount of that.  And there are more the reference populations out there for adults, for children, and they are being developed, and several for infants, if you know the references to look for.

    The disadvantages, very low exposure risk, the same thing.  One could argue whether it's an advantage or disadvantage.  One of the ways I talk about this--in fact, I just thought about it coming over here yesterday on the plane--was on the flight I was on, there was at least five children under the age of 2 on that flight with me, and the radiation dose they got on that flight exceeds what you get from the DXA.  In fact, it's two to three times higher.  So there's an idea what the risks are involved.

    Scanners are not optimized for infants.  If you get a scanner, it's adult size.  They have not--the industry has tended to resist this, primarily because of the market that they are focused at, which is osteoporosis in older women.

    It doesn't give us a 3-D image.  You saw that 2-dimensional image.  It gives us 2-D not 3-D imaging.  That's what I call boot-strapped 3-C model.  It's not a perfect 3-C model, but it's not bad.  And the one different problem has been that the results differ between manufacturers, so that if you do a study--a multi-site center study, you want to stay with the same instrument, same software.

    Okay.  Precisions and accuracy of different body composition measurements and the minimal detectable change in an infant.  For this I chose basically a full-term infant and made the assumption it's 15 percent body fat.

    If you look at these methods, the water, the dilution method, or the bioelectrical impedance method or TOBEC, this is the potassium one here.  DXA, the bottom three, the fat, fat-free, and the bone, precisions, these are optimistic.  Precisions tend always to be a little bit better than--whenever you do a precision measurement study, they always do much better than they do random.  I'll guarantee that.

    This shows you the precision measurements here.  This is generally the accuracy.  Precisions are, let's say, in the 1 to 5 percent range or 2 to 5 percent range.  Accuracies tend to be in a 3 to 5 percent range.

    If you take these, this information, and take this size of an infant and you translate those into one of the minimum detectable changes for that infant, the values are shown here, the last column on the right.  And the percents are those percents of what that person had in terms of the composition at that age.

    So one can measure water, changes at 5 will start to show up.  If you use (?)-ium, they can get worse, TOBEC or BIAs.  Fat-free mass, 125 gram changes, only about 5 percent of the total fat-free mass.  Forty grams of fat, if that starts to change, it's 8 percent of that to implement this weight and composition.  You can start to see changes relatively quick with the single input.

    This shows the relationship between precision of the methods and what kind of a change, minimal change is required for that to become statistically significant, at the 5 percent level and a power of 0.8.

    I call this the clinical application in individuals, what I consider to be a clinical application, what's changed in that individual.  And if you look, for example, this is the relationship that the minimal detectable change--this is approximately 3 times the precision.  And so that if our precisions for BMC are somewhere around 2.5 percent, 6 percent change would occur, lean tissues at about 10 percent, and fat-free mass at around--if the precision is 4.5 to 5 percent, we'd have to see changes in the range of about 14 percent to be significant for that individual.

    Now, this I show you because this is the difference between 1988 and today.  I only got 30 seconds left?  Oh, well.  Well, okay.  Very quickly, this shows you the methods.  This shows you what the precisions are for FFM, fat-free mass.  If you translate those into fat-free mass, this is the tail here.  The top three were in 1988.  That's why you couldn't do it.  DXA is hugely improved since then.  This shows you the kind of weight gains that would have to occur in infants at, I guess, again, the standard term infant here, very small weight change with DXA, large weight changes for--I'm out of time.  She says zero.  Anyway, the--I'll keep on going anyway.  And the number of weeks would have to be changed in that individual.

    I can't believe I've taken up all the time already.

    This just shows you the rates of change that occur with age.  You can find this from several different sources.

    This shows you the first six months where you have to figure out what the rates of change are in the composition, and here's a series of papers that are on infants.  There's one on TOBEC.  This one used a series of methods which we can do.  All the rest you'll notice, with the exception of one here and dilution, were done with DXA.  The weight ranges are shown here.  The number of infants are shown over here.  We are now doing a meta-analysis to bring this together into one common reference database.

    And I want to just quickly go through these, again, precisions, 2 percent, 3 percent, 6 percent reported here.  And if you did a calculation--this is an interesting study in twins.  They looked at the weight difference, which is about 14 percent, and they calculated again with these alpha-5, power of 0.8, 40 to 45 infants would be needed to detect a 15 percent difference in one or more of any of these three compartments.

    Another paper here, let's see, this particular paper does give percentile curves for each of these values as a function of weight.  And, again, what I'll point out here is that even though you may have a 3.5 kilogram weight infant, the fat range can range from 10 to 26 percent.  By the time they're 10.5 kilograms of weight, it can range from 22 to 23 percent.  So weight does not represent fat.

    This paper is another one that has percentile curves of each of these compartments versus weight.  Again, comparable to the other one, actually these are lower fats.  Interesting, this is an European study.  The previous one was a U.S. study.

    And I want to show you--this is the last one, slide here, and I'm a minus five I think now.  But here this shows the changes--total body DXA.  These were preterm infants.  Initial weights were about 17--under 1750.  This was fortified human milk formula, and this was a preterm infant formula.  And the baselines are measured at 3 weeks of age, repeated again at approximately 3 to 4 weeks later.  And this shows the statistical ability to measure changes in body weight, for example, and in the fortified human milk you can make--you can change--see the difference at this level.  And the preterm formula, it's about 19.9 grams per kilogram per day.  Differences were four, and this is statistically different.  You can see these--in other words, you can compare in groups of 20 versus 30, you can see differences in weight, you can see differences in lean mass and fat mass and bone mass by DXA at three weeks between these two groups of children.  So it is possible with DXA to measure not only weight but the composition of that weight change in relatively small sample sizes.

    And if you want to convert those to growth kind of numbers, this represents about 2.3 grams per kilogram per day in terms of growth, and that's about 12 percent of the mean weight gain in terms of composition.  This again is 2.1 grams of lean mass per kilogram per day, 15 percent, about 1.2 grams per kilogram for fat mass and 76 milligrams per kilogram per day for the BMC.

    So the point here is that we can measure--I think there was a question, could we measure changes of composition that would be comparable to 3 grams per day?  The answer is--from this study the answer would be yes, we could do that.  Again, relatively small sample sizes of 25 to 30 children.

    I'll end there.  Thank you.

    DR. GARZA:  Thank you very much.

    Any questions or comments?

    DR. MOYER-MILEUR:  I have a question.  When you do your measurements in your babies, are they sedated?  Because we find that we require sometimes more than two technicians to keep a baby quiet to minimize the movement artifact.

    And my other question is, with preterm babies, we found it somewhat difficult to do early measurements because of equipment artifact, that they have leads and monitors on that make it very difficult to get a true assessment using DXA.

    DR. ELLIS:  Yes, two things.  One, none of the infants that we measured and none of the infants that any of these studied were sedated.  These are--again, they're all healthy children.

    Our experience has been if you feed them right before you want to do the measurements, they tend to be rocked in the chair by the mother or someone, they go out, and then you can make the measurement pretty easily.

    I don't know which machine you were using.  Was it a 4500A or 2000 or--

    DR. MOYER-MILEUR:  We have a 4500A, and we also have--

    DR. ELLIS:  You should do it in three minutes or less.  You do have to work at it.  I mean, normally with older children who will cooperate, that could be anywhere from age 5 to 18, depending upon what you're looking for.  But those children can get on a bed and will cooperate and can be--and they'll do it.

    Here you can do the whole procedure in ten minutes or less.  Here sometimes you have to spend as much as an hour to get the one measurement done.  You have to work at it.  But none of these children were sedated.  We don't sedate any children in any of our studies.

    DR. MOYER-MILEUR:  Yes, and I just, you know, would caution DXA in the infants in that it requires people with specialized training so that you can't just--

    DR. ELLIS:  Yes, yes.

    DR. MOYER-MILEUR:  --go to a community hospital and get their--

    DR. ELLIS:  Yes, you can't--you cannot send these children to a radiology department even with the hospital because they simply are not experienced with measuring children.  They just don't like it when children show up.  They don't have--infants, they basically will send them back.  They will not--they will not take that hour, hour and a half to do it in.  It takes effort sometimes.  Sometimes they go right on the bed and out.  It's always the ones that show up at 4:30 that take the hour and a half to two hours, though.

    As far as the artifacts, you're right.  You have to be careful about artifacts.  You can delete those off the images, though, pretty well.  If you take leads out, for example, out to the side, you can delete those right off the images.  And so that's a minor effect if you deal with it right.

    DR. GARZA:  Dr. Denne?

    DR. DENNE:  I was wondering if there are any direct comparisons in infants between skinfold thicknesses and DXA for fat mass.

    DR. ELLIS:  There may be a few, but, again, the issues have been that skinfolds are probably more difficult to get than the DXA.  We have skinfolds in some of our kids, but we just don't rely upon them for anything.

    DR. DENNE:  It would be an interesting comparison to make.  You know, relative difficulty depends on what you're actually used to doing.

    The other question is:  How is DXA validated in infants?  I mean, most of this body composition, you know, was validated against the other techniques which all have their own sets of issues.

    DR. ELLIS:  Yes.  The validation of DXA are done two ways.  One is with animals, small animals have been done.  We have done 73 piglets under the weight of 10 kilograms.  Other people have done comparable size piglets.  And probably if you add everything up, it's probably about 200 pigs have been done over the years at different centers with different machines.

    The pig is not the best of models because, for example, its bone is more mineralized than infant's.  Weight-wise, composition-wise, soft tissue is not that bad.

    The other way we've done it is we actually built phantoms.  We've actually fabricated mock-ups of the human body with parts made from polyester resin, doped with calcium and phosphate--phosphorous compounds to simulate that.  But that's how it's done.

    I have also done cadaver work.  The problem--not the problem.  The situation is that, unlike Elsie Widdenson, today's environment would not allow one to chemically digest the infant body, so we have done that in about 30--more or less 30--these are all preterm infants, and we did that by a technique called neutron activation analysis where we do a nuclear, chemical--nuclear chemistry technique where you measure calcium, phosphorous, sodium, chlorine, phosphorous, manganese and magnesium and potassium.

    And so if I look at the BMC bone versus fat, and if I look at the other ones and make some model--I have to make some modeling assumptions now about how much sodium is in the water, extracellular water and so forth, but they come out pretty well, with the 5 percent kind of accuracies.

    DR. THUREEN:  In the past several years, it's been recommended that at different centers, even if you have the same type of machine, you should do your own phantoms.  Do you think there's enough phantom data out there now that that doesn't need to be done?  Or if you're going to do a multi-center study, do you think that needs to be done?

    DR. ELLIS:  Well, for the multi-center study, there should be at least a common phantom that is going around to all those sites.  One, to do the initial calibration to be certain everybody is within reason of the numbers, and then continue on throughout going for the study.  That's typically what we do in all studies at all ages, whether it's infants, children, or adults.  That's what we do these days.  For multi-site studies, there's a common set of phantoms that go around all the time.

    DR. GARZA:  Dr. Stallings?

    DR. STALLINGS:  I'm one of those other six people in the world that's got the TOBEC, so I agree those, you know, have tremendous advantages.  But I don't think that we would be able to use them.  So I think, you know, bringing us to issue with DXA and how we could use it is an important question.

    Laurie asked one of the big questions that I'm always asked, which is about sedation, and I would agree that, you know, natural sleep and that sort of thing and working in the research setting.

    The other question sometimes is:  How many images do you really have to take to get the one right?  I even noticed on your slide the hand is--

    DR. ELLIS:  The hand was a little off.

    DR. STALLINGS:  And I just spent last week working with DXA and trying to figure out which one had all the body parts there and minimal movement.

    But would you share with the group, you know, how frequently do you need to do two scans or you get halfway through a scan and then you do it again to get a good research quality measurement?

    DR. ELLIS:  I would say it's definitely less than 10 percent that we have to repeat the scans.  It's like you say, there's a technician there.  The image is being acquired while the scan is being done, and you can stop it immediately to start again, as you well know.

    We have some--we have looked at some scans where the infant has moved, but we finish the scan and then repeat the scan again and looked at those.  It has a lot to do with what kind of movement you have.  As you well know, you can--if the child's arm is here--or say here when it starts and here when it ends, you have a three-armed child in the image you end up with, because it was here the first time you scanned through and caught on the second, on the lower case.

    We have found that if we have motion in this direction, there tends to be a minimal effect because you're not changing anything.  You're just moving the slice over a little bit here.  But it's when there's movement like this, a flapping of the arms or kicking of the legs, if they're doing that, we don't scan them.  We stop.  But that's usually what happens, they wake up or something.  Less than 10 percent.

    DR. GARZA:  Any other questions?

    [No response.]

    DR. GARZA:  I have two, Ken.  How well described are the specs of equipment that one would need to be able to measure infants reliably?  Is there pretty much a consensus on the quality of the equipment, the DXA equipment that would be needed?

    DR. ELLIS:  You'd have to have something that's equivalent to what's called--there's basically two manufacturers in this country.  One is Lunar, the other one's Hologic.  You have to have at least the DPXL for the Lunar at least the 4500A or DelphiA for the Hologic.

    We always use Hologic's, and, in fact, we, you know, are constantly trying to improve those machines.  I'm not going to tell you they're perfect, but they're the best thing we have.  I think they could make them better.

    DR. GARZA:  You also indicated that individuals or personnel had to be specialized or had to be highly trained.  How much training do individuals need to be able to use this equipment reliably, or was the training in reference to just training and dealing with pediatric populations?

    DR. ELLIS:  It's more dealing with the pediatric populations.  It's more of that than it is simply for this, because once they understand--basically what you want to do is you want to minimize motion and have them in the right position and things like this.  But it's more dealing with the pediatric population, dealing with a child that may want to cry for 20 minutes or something, or a mother that could be apprehensive when she hears the child crying.  It's more that issue than it is anything else.

    DR. GARZA:  Dr. Stallings?

    DR. STALLINGS:  I just want to ask a little bit what Bert was doing.  If we were doing such a study, a multi-center study, what would your advice be about centralized reading of the scans, the technician at the instrument site?

    DR. ELLIS:  It is a good point.  These days, again, it is common practice now to send all the scans to a common central reading site because at least what happens there--well, I'm thinking more of the adults.  If there is any kind of bias--in adults you set regions of interest.  In the infants, it's a total body scan.  There's no region of interest set.  So it's less of an issue there.

    But, again, the judgment about good scans or bad scans would come from one source and not from different sources.  So it would be a reasonable thing to do.

    DR. GARZA:  Any other questions?

    [No response.]

    DR. GARZA:  Thank you very much, Ken.

    DR. ELLIS:  Sure.

    DR. GARZA:  Committee members and guest speakers are invited next door for coffee.  We're going to break right now, instead of at 10:35 as on your schedule, so that we don't break up the following three presentations.  But I will ask everyone to try to get back here at about 10:25, 15 minutes from now, so we can assure that we don't eat into any discussion time with either the speakers or anyone else.

    Everyone else that is not a speaker or on the committee is invited to the cafeteria.  These are federal rules.  I didn't make them.

    [Laughter.]

    [Recess.]

    DR. GARZA:  The committee is seated at the table.  If I can have our guests please take your seats, we're ready to start.

    Our next speaker is Dr. Frongillo.  Dr. Frongillo is an associate professor in the Division of Nutritional Sciences at Cornell University, and he's going to give us an overview of the World Health Organization Growth Reference Study that was referred to a bit earlier by Dr. Chumlea.

    DR. FRONGILLO:  Good morning.  I'm going to stand here in the middle and use this archaic technology.  It might help if we could dim the lights up at the front here a little bit.

    These are some growth data from a single child, and you can tell what country they're from if you look at the units of measurement in pounds.  And if we plot the data, it's actually more interesting to look at.  We see a trend like we expect.  But to try to really discern anything about what the pattern is, it's helpful to compare it to something.  And so this is a graph--this is the old U.S. reference, the 1978 reference, and this is the same child.  And you can see this graph on the left, the child started off at the bottom of the distribution and then seemed to climb into the chart a bit, and then at about four months or so started to really--three to four months in there, started to really--its trajectory is now falling well below the chart.  And eventually by--this is about two years where you see it comes back onto about the same percentile where it originally started.

    This discrepancy that we see is either telling us something about this particular child or it's telling us something about the reference.

    Well, it turns out it's telling us something about the reference because that's a breast-fed child.  And if we look at a data set, this is a comparison that the WHO infant growth--an analysis that was done in the early 1990s, and what it shows is that, if you look on the left here, this is for boys, this is weight in kilograms and age going up to 12 months.  And the dotted line to the 1978 U.S. reference which was adopted about that time, just after that by WHO, is the international reference.  And the breast-fed data set, these are infants who were exclusively breast-fed for four months and then continued breast feeding through the first year.

    And you can see again that about four months you start to see the solid curves deviating from the dotted lines, which is the same pattern that we just saw.  And so this is showing that in a sample of about 426, I think it was, infants epidemiologically what we just saw in that individual child.

    And a similar pattern was seen for girls.  We can amplify this in a way by--what I've done is just simply take the current reference, the current international reference, the 1978 reference, and that's what would be at zero.  So if these children were growing exactly like the reference, there'd be a horizontal line right at zero here.  But what we can see is that this breast-fed set seemed to grow, if anything, a little bit faster at first, and then by the end of the first year had increased their weight substantially less than the U.S. set.

    In fact, if we calculated the difference in rates for these two groups from zero to 12 months, it's about 2.7 grams per day.  So this is something to keep in mind for later when we're thinking about how big our meaningful difference is.

    The difference from about one month, which is at the top there, from the maximum to the minimum at 11 months is about five grams per day.

    So that was for weight.  If we calculate a z-score--and we'll have--I'll put this up here because we're going to hear about z-scores at various points.  A z-score is where we take a particular measurement for a child and compare it to a reference median, whatever the reference is, and then divide it by a reference standard deviation.  So the graph I just showed you just showed the numerator there, but for a z-score we also divide by the standard deviation.  And the reason we do that is because then it's easy to imagine that the growth of a, quote, normal population would fall between about minus two and plus two z-scores.  About 95 percent of the distribution would fall there.

    So if we look at z-scores for this breast-fed set, then what we saw was that, regardless of the index that was used, whether it was length-for-age, which are the triangles, which is the curve near the bottom, whether it was weight-for-age, which is the circles, or weight-for-length, which is the squares, we saw a very similar pattern with this breast-fed set in comparison to what was then the U.S. and international reference.

    So this discrepancy, along with other information that was obtained during the review that the Infant Subcommittee made during the early 1990s leading up to the WHO publication in 1995 of the uses and interpretation of anthropometry, the recommendation was made that consideration should be given to making a new international growth reference.

    The justification for having an international reference is, first of all, that it allows cross-national comparisons to be made that otherwise couldn't be made; and since there's been an international reference since the late 1970s, it's allowed us to do some things in a comparative way globally that we weren't able to do before.  For example, this is a graph from a WHO publication that was in the bulletin of the WHO showing that the trends that have occurred from 1980 to about now so that we could actually look at the progress in Africa on the left, the very rapid progress that occurred in Asia.  This is in percentage--the percent of the population that's stunted--and the rapid progress that was made in Latin America and the Caribbean region.

    This kind of comparison has been made possible because there is a common reference being used throughout the world.

    In addition, we know that it's very expensive to make local references, and also that in developing countries where there's still a very strong cyclical trend in growth, if a local reference were made, it would have to be revised very quickly because of changes that are occurring.

    The justification for having an international reference goes back to work that was done in the early 1970s.  This is a well-known graph from a well-known paper, in '74 I think it was, showing that these curves right here were all curves of high SES children in well-off countries, whereas the ones that were down here were children in developing countries that were not so well off.  And so the fact that these were all so close together meant that growth roughly from one place to another where children are growing in conditions that are favorable to growth tends to be roughly about the same.

    Some work that the WHO has done with a cross-national data set collected by the Human Reproductive Program recently shows that--this is for girls--across a number of different countries, these are children who were reasonably well off SES, not necessarily the highest SES, showed basically that, with the exception of this lower curve, which is in China, that these other curves all pretty much are very close together.  Again, giving more recent evidence of the idea that it was reasonable to make an international reference where data from multiple countries could be combined.

    So let me tell you, then, a little about the effort that's underway in the Multi-center Growth Reference Study.

    First of all, I wanted to point out that a reference--the idea of a reference is that it's a tool for providing a common basis for the purposes of comparison.  So we're interested in references because it allows us to compare as opposed to a standard which then involves a judgment.  So here we're talking about making a reference, and during the early 1990s, both the U.S., in preparation for the revision of the U.S. reference, and also WHO examined the current reference which was being used in both the U.S. and internationally.  And the sample that had been used for the early infancy especially was from Fels, which was one particular place in Ohio.  The measurements were taken every three months, and in the very early period we might wish for more than that.

    There were very few infants that were breast-fed for an extended period of time, and at the time that this reference was made, there simply wasn't the technology to do curve-fitting that we now have.

    So those reasons, plus the main factor that the breast-fed infants seemed to grow differently than infants who were not necessarily breast-fed according to feeding recommendations, drove the decision to make a new reference.

    At the time the WHO feeding recommendation was that infants should be breast-fed exclusively from birth up to about four to six months, and then after that they should continue to be breast-fed for up to two years or beyond.  You may know that WHO recently--I guess about a year and a half ago--revised this to be from birth to about six months for exclusive breast feeding.

    So the objective of the Multi-center Growth Reference Study that WHO is doing is to build a set of growth curves for all children under age 5 years to be adopted as a new international reference for assessing the growth and nutritional status at both the population and individual level.

    When this effort was started, it was clear that there were a couple of conceptual issues that needed to be thought through.  One was that some references have been constructed, especially, for example, in the U.S. or in England, to take two examples, have been constructed to be descriptive references, meaning that they were intended to describe the growth of the population at a particular time.

    This is different than what's going on in the Growth Reference Study, the WHO Multi-center Growth Reference Study, which we can think of as perhaps a prescriptive reference, meaning that it's meant to be a reference that depicts the growth of infants who were fed according to current recommendations for how children should be cared for during infancy.

    The other issue had to do with maximal growth versus optimal growth.  In the past, we've had a tendency to think that maximal growth and optimal growth are the same thing.  The graphs I just showed you indicate that when infants are breast-fed, at least during the first year, and perhaps into part of the second year, they are not the maximum size they would be if they were breast-fed, but we think that because they're fed following feeding recommendations that that corresponds to optimal growth.

    So the design of the study involves multiple geographically diverse sites.  There's a longitudinal component which goes from zero to 24 months.  And each site was asked to recruit about 300 infants per site in the hopes that at least 70 would be available for inclusion in the final reference.  We've actually done better than that because the compliance with the feeding recommendations by the mothers and infants has been much higher than the 25 percent that we feared might be there.  So we've actually ended up with quite a bit more than 70 percent.  And then very frequent measurements, I'll show in a minute, and then there's a cross-sectional component which overlaps the longitudinal component.  It starts at 18 months, goes up to 71 months, past 5 years, to make sure we have enough data on the right-hand side to be able to characterize growth well up to at least 5 years.  And sites were asked to recruit about 1,400 per site, which, again, would give a minimum at each age of about 70.

    In the longitudinal component, which is a very demanding part of this study, measures of weight, length, and head circumference are collected frequently during the time.  At birth, there's one visit, of course, and then in months 1 to 2 they're biweekly.  So there's four visits there.  In months 3 to 12, measurements are monthly, so there's 10 visits for that.  And then in roughly the second year, they're bimonthly, which is six visits during that time.   And then arm circumference and skinfold measurements are also taken in the same schedule, starting at 3 months, as in the other measurements.

    Now, the way the study was constructed, there were a set of criteria at the population level and then a set of criteria at the individual level.  So at the population level, the idea was to find populations of infants who did not have socioeconomic constraints on growth, where mobility would be low so that they could be followed, where at least 20 percent were willing and able to follow the WHO feedings recommendations, with support, so there had to be existence of or at least the ability to build breast-feeding support systems; and then there had to be local presence of collaborative institutions who were capable of carrying out this kind of exacting work.

    Then at the individual level, individual criteria were set that there was an absence of health, environmental, economic constraints on growth; the mother was willing to follow the WHO feeding recommendations; that the mother was a non-smoker; that gestational age would be at term, which we defined to be 37 to 42 weeks; and that the infants wouldn't have any severe illnesses that would be expected to affect growth.

    The protocol for the study site selection then applied to subpopulations the fact that socioeconomic status did not constrain growth, it was low altitude, low mobility, the minimum of the 20 percent, existence of breast-feeding support systems, the local institutions.  We looked at the rate of hospital deliveries because we had to know that there were enough infants being produced quickly enough that they could be enrolled in the study so we could get the study done sometime in our lifetime; that there would be sufficient numbers of eligible birth; and that it was feasible within those locations.

    In some places, for example, in a really huge city of 10 million, it's just not feasible to do a study like this.  The logistics are too difficult.

    The Steering Committee also considered some other factors in its thinking about in looking at mean birth weights, maternal heights, complementary feeding practices, health-related behaviors, and the existence of environmental hazards.  The Steering Committee looked at geographic distribution.  It's a global reference, so WHO tried very hard to have geographic representation throughout the world, and funding issues, because it's expensive to do this kind of study and we had to think about where the funding would come from and how that could be arranged.

    The protocol was developed by this set of characters here.  The main reason I put it up here is to show you that it's a multidisciplinary set of people representing a diversity of backgrounds and institutional relationships.  So this group put together the protocol roughly in the '95, '96 time frame.  And then the study is being run now with an advisory group:  Cameron Chumlea, Tim Cole, myself; Ray Martorell is the Chair of this group; John Van den Broeck, who recently moved to South Africa; senior scientists representing CDC and UNICEF, previously was Roger Shrimpton in UNICEF; and then WHO, the day-to-day work gets done at the sites and at WHO Secretariat.  Mercedes de Onez coordinates all of this with her staff there.  A very dedicated group of people.

    The sites that have been selected are the following:  in Pelotas, Brazil, in the south part of Brazil near the coast; Victoria is the PI with Cora Post; in Oslo, Norway; in the U.S. at Davis; in Muscat, Oman; in Accra, Ghana, the capital of Ghana; and in New Delhi, India.  These are the six sites.  Each of these sites represents very differing and very large challenges to carrying out this study.  And it couldn't be done without the commitment of the teams there who are doing the work on the ground.

    The Steering Committee is chaired by Cutberto Garza representing UNU.  Data management is done by local data entry and checking at the local, each site, and then the data are shipped to the WHO Human Reproductive Program.  They have extensive experience in handling large, multi-country data sets, and they've done a fantastic job in coordinating all of this and in helping to ensure data quality.

    The decisions and information about the study, there was a working group on the growth reference protocol.  We have Steering Committee and Advisory Committee meetings periodically, other meetings.  Particular tasks are handled by other meetings.  We do a lot through electronic mail.  Various site visits are made, were made before the study started in preparatory work and are made throughout the study, and rapid surveys were done at the beginning to get information that was needed to actually do the planning for the data collection in the particular sites.

    To give you an idea of what's involved, I just made a list here of sort of the documentation that's been produced, which will give you a feel for what was necessary to carry this off at the level of scientific quality that was being strived for.  The protocol was developed, a measurement of standardization protocol was developed.  A manual of operations, a generic manual of operations was produced, and each site had to adapt that for its particular location.

    There was a protocol for the 12-month visit.  A special effort was made throughout this study on the epidemiological quality of the study to follow all infants.  Even if mothers wanted to drop out or weren't complying, an effort was made to follow them as much as possible so we could keep measurements for every infant even if they weren't exactly following the feeding recommendations.

    In particular, we had an effort made that at 12 months we could go out and get at least some measurements on all those who were not willing to continue.

    There were guidelines for complementary feeding.  A protocol was developed for assessing diet, for the cross-sectional study, for data management, and then questionnaires were produced for both the longitudinal and the cross-sectional study.  And if any of you have been involved in questionnaire production, you can imagine what those meetings were like as people argued about the exact wording of every question and every answer.

    So that's an overview of the study.  Basically where we are now is that the last site--the sites have been selected over time.  Some were able to start earlier than others.  Some had technical challenges that took longer than others.  So in the next few months we'll be finished with data collection.  There's a meeting coming up to look at and try to decide on the final methodology for analysis.  Some preliminary work is underway.  So this will be proceeding, and we're talking about having a reference be available in the 2005 year.

    There's a lot of work that needs to be done in order to prepare for not just producing the reference but preparing for how it will be depicted and how it will be used and testing that will have to be done with the reference under the auspices of WHO.

    DR. GARZA:  We have about five minutes left for questions.  We might be able to go over.

    Dr. Stallings?

    DR. STALLINGS:  Well, one, to compliment the group.  This is an extraordinary effort and an extraordinary study.  But to cut to the chase, do you think that we will see this used in the U.S. as the growth standard for infancy through 5 years or even infancy through 3 years with what we think of now as our traditional infant chart?  So I'd be interested in your opinion and certainly in relation to what we're here for, which is to start to understand the best comparison group for children in the U.S. who are taking infant formula.

    DR. FRONGILLO:  Okay.  Well, I think I'll probably not try to answer your question so directly.  Let me just say that I think the advantages of this reference will be that it's longitudinal, that the longitudinal data in the first couple of years will have the ability to look at--to have a reference, a velocity reference, and so a judgment will have to be made whether that's better than, for example, the Iowa/Fels data that are available.

    The second thing is that we know that infants who are fed following the breast-feeding recommendations will show a different pattern of growth than formula-fed infants.  So to the extent to which it's seen as desirable to have a reference which fits that growth pattern for infants who are being breast-fed, then I think that would certainly be the advantage of the new reference.

    DR. GARZA:  It may be useful to describe the references that are going to be available.  Is it just weight and length, or are there others?

    DR. FRONGILLO:  Well, those reference data will be available for all of the measures that I showed, so we will have data on weight and length, head circumference, arm circumference, and skinfolds.

    DR. GARZA:  Any other questions?

    DR. THUREEN:  Why did you choose to include skinfold measurements, and what kind of information did you hope to get from that?

    DR. FRONGILLO:  Well, I think that there was a debate about, you know, the importance of this and when it should be started, and I think the lack of reference data on skinfolds was very compelling and it was thought that, given the potential usefulness of that information in the future, that while a study of this effort was--while this large effort was being made, it would be important to have that kind of information available.

    DR. GARZA:  Dr. Baker?

    DR. BAKER:  I have a question about the prescriptive nature of this.  If you're going to do a study like this using a prescription, it assumes the prescription is right.  It also assumes that it also would change, presumably, if the prescription changed.

    Now, the WHO has changed it since this study was done.  Does that make a difference?

    DR. FRONGILLO:  Well, certainly we thought about this a lot during the time in which the study was being planned.  I don't think anyone imagined that the basic recommendation about breast feeding and complementary feeding is going to change appreciably, and at least not for quite some time.

    Now, if it does, one of the things we wanted to do and part of the reason for the intensive follow-up even for infants who didn't exactly follow the current feeding recommendation was that it allows us to have the information available so that if 15 years from now we have new knowledge and decide that some slight revision of the feeding recommendation is made, anybody who's ever thought about changing the feeding recommendation will run after that possibility.

    But, anyway, if anybody gets brave enough to try to do that, we will have the information available from the cohorts in the six sites so that one could conceivably reconstruct the reference to conform with that.

    DR. GARZA:  Yes?

    DR. DOWNER:  Have you decided exactly which tool you will be using to do the skinfold measures?  And my second question is:  Because what we consider SES for different world populations differ so widely, how are you going to decide on what to use in this study?

    DR. FRONGILLO:  Okay.  The skinfolds, which tool, do you mean which skinfold caliper?  Basically I think--and Cameron can comment on this because he's the expert here.  But my experience is that it's not the caliper that makes much--any difference at all in the measurement.  It's the quality of the enumerator and their training in using the instrument.  So that's not going to make any difference.

    DR. DOWNER:  What instrument have you planned on using?

    DR. FRONGILLO:  The instrument we are using is--what is the instrument we're using?  Holtain, right.  Okay.

    And the second question?  I'm sorry.

    DR. DOWNER:  The SES.

    DR. FRONGILLO:  Oh, the SES.  In each site, a survey was done before the study began, the main study began, to actually look at the relationship between socioeconomic status and growth so that we could develop in each site exactly what the criteria needed to be from a socioeconomic standpoint in order to ensure that the population of infants selected was at high enough SES to not constrain growth.  So that was done separately in each of the sites, and some of us traveled around to different sites to help them actually carry that out.  And you're right, in each site different criteria were needed because the conditions were different.

    DR. GARZA:  Any other questions?

    [No response.]

    DR. GARZA:  All right.  Thank you very much.

    We'll move on then to Dr. Larry Grummer-Strawn, who is the branch chief of the Maternal and Child Nutrition Branch at CDC, and he will tell us about the NCHS/CDC's growth charts.

    DR. GRUMMER-STRAWN:  Good morning.  If we can figure out how to forward this?  Which one?  Just here?  Okay.  Thank you.

    This morning I want to give kind of an overview of the new growth charts and contrast them to the old NCHS growth charts.  I'm going to start off with kind of a historical perspective.  I'm sure many of you know the history but to kind of just set a context for all of us, do some comparisons of differences and similarities between the old and the new, and then go into some of the differences a little bit more explicitly, and then finally end with some analytic issues that the charts pose for us.

    The original NCHS charts were released in 1977.  Those charts were only percentile curves.  They were published by Hamill, et al., really became the standard of reference for all U.S. infants.

    Subsequent to that, there was a normalization of those curves at CDC.  Those were actually published in 1987, but were actually available for use long before that.  So people who were interested in normalized curves had access to them earlier.  And, finally, the WHO adopted those curves as being the international reference, really referring to the normalized curves.  The adoption by WHO actually came prior to the publication of the normalized curves.

    The reason that I point this out is that those curves never became one and the same.  The percentile curves never matched with the normalized curves, and so someone who was using clinical charts that actually saw the graphs in front of them was not necessarily using the same cut-off points as someone who was using computer software or might be analyzing data sets.  They were very similar to one another.  They were analyzed off of the same data, but were slightly different from each other.  And then, finally, in May of 2000, CDC released a revision to these charts.

    Now, at the time that the original charts were created, NCHS was a separate agency, and so they were referred to as the NCHS Growth Reference.  When people talked about the normalized curves and put them in an international context, they might have referred to the NCHS/CDC/WHO because of the separate role each of those agencies played.

    Subsequent to 1977, NCHS was actually incorporated into CDC, and so the new charts are referred to as CDC charts.  That does not mean that NCHS was not an active player.  They actually were the progenitors that moved the new charts forward.  But it was in a larger context of CDC, and other parts of CDC were also involved.

    So what are some of the similarities?  First of all, both sets of charts are looking at the same indicators.  We have weight-for-age, length-for-age, weight-for-length and head circumference-for age.  Both sets are sex-specific.  In neither case did we have any separation according to the parental anthropometry, race, ethnicity, infant feeding mode, different things that might impact on the growth of the infants.

    Ed just described for us kind of the difference between the idea of a reference and a standard or a descriptive reference and a prescriptive reference.  These clearly are references, not standards.  The only kind of movement toward a standard is that very low birth weight infants, that is, less than 1,500 grams, were not included in the new charts, the CDC 2000 charts.

    These charts reflect attained size, not incremental growth, and in both cases, we have accessibility of percentile scores and z-scores.  In the new charts, the z-scores are one and the same.

    The differences between the old and the new are, first of all, that in the 2000 charts, the data for infants are now nationally representative.  These represent a broader spectrum of race and ethnicity across the United States, a broader spectrum of socioeconomic status, and there's an increased representation of breast feeding in the charts.  That doesn't mean that they are primarily breast-fed children, but there's a mixture of formula-fed and breast-fed children.

    The 2000 charts are based on a pooling of several data sets coming together, whereas the 1977 NCHS charts were all based on the Fels data set.

    There were some minor changes to the smoothing techniques, which I will mention briefly.  As I said, the z-scores now are one and the same as the percentiles.  There's a one-to-one match on those, so it doesn't matter whether you use computer software or you're using printed charts.  You're going to be looking at the exact same cutoffs.

    Another minor difference is that the length now extends down to 45 centimeters rather than 49 centimeters when we're looking at weight-for-length, and when ow have on the clinical charts the accessibility of 3rd and 97th centiles.  Of course, based on normalized curves, you can get any kind of centile that you're interested in, but the difference is that the clinical charts that are produced actually do extend out to the 3rd and 97th percentiles, and the smoothing-out rhythm is intentionally extended out to those centiles to make sure that we're incorporating the original data out that far.  And, finally, as I mentioned before, very low birth weight infants are excluded.

    So what are the data sources?  Well, in 1977, I'm sure you're all familiar with the Fels Research Institute study.  It was done in Yellow Springs, Ohio, primarily represented Caucasian, middle-class families.  And while not exclusively formula-fed, this group is considered to be almost all formula-fed infants.

    There was a longitudinal follow-up study.  The children were followed from birth, 1 month, 3 months, 6 months, and at three-month intervals after that, but we're focusing on the first 6 months today.  The data were collected between 1929 and 1975, and there were a total of 867 infants.

    Now, the data for the CDC 2000 curves, as I said, represents a number of different data sources.  What I've put up here is a graphic showing you how at different ages, different data sets come in.  I'm going to talk about kind of these different data sets at different points.

    The primary data source is the NHANES III data here when we're talking about children birth to 6 months of age.

    Now, this is represented by the long line here in light blue.  The NHANES III was the only data set for which we had nationally representative data prior to 6 months of age.  Starting at 6 months, the NHANES II data were also available, and starting at 12 months the NHANES I data were also available.

    Now, the reason that these are important for us today is that because we're smoothing these curves across age, the influence of those older data sets does come in at 6 months of age and at 12 months of age.  The smoothing is across all ages, and so the curves below 6 months are also affected by those other data.

    However, the NHANES III data started at 2 months of age, and among 2-month-olds, it was a fairly small sample size.  In order to extend these curves down to birth, we had to look to other data sets.  And in each case, whether we're talking about head circumference-for-age, length-for-age, weight-for-age, weight-for-length, in each case we have to turn to different data sources.

    I'm going to start in the middle here to talk about in the weight-for-age.  In this case, it was fairly straightforward to use a birth point coming from the national birth certificates.  We had all of the birth certificates during the years that the children in these NHANES surveys had been born and had the birth weights available on all of those.  So it's a huge sample size, a very precise point that is truly not only nationally representative but a census of all births in the United States.  And so in creating the curves, we were able to connect that particular point, actually anchored the curves to that point, and then smoothed it with the data starting at 2 months of age from the NHANES.

    However, we don't have national data on any of the other indicators, head circumference or length.  With regard to length, we did find that there were two states that routinely collect length data at birth in a representative fashion.  We analyzed the data in those states for their birth weights against the national birth weight distribution and found that they were quite representative--those states were Missouri and Wisconsin--and felt that because the birth weight distribution matched the national distribution, we could expect that they should represent the national birth length distribution, even though we do not have data on the national birth length.

    So when we looked at the weight-for-length curves, those are based on connecting the dots between the Missouri and Wisconsin data on weight-for-length versus the NHANES III data weight-for-length, and those curves were connected together.

    Initially, we intended to do the exact same thing with length-for-age, use only the data from Missouri and Wisconsin to connect these curves across age and with the NHANES III data.  We did that in our first pass but analyzed--as we were evaluating the curves against alternative data sets, we found that we were comparing against the Chicago data set, the WHO pooled data set that Ed just described for us, as well as some of CDC's surveillance data, and found a common pattern in all three of those that the curvature between birth and 6 months did not match what we found in external data sets and felt that this was partly an artifact of the fact that we only had about 35 infants from the NHANES III data that clearly were not matching the normal pattern of growth.  And so the curves were being pulled in the direction of those NHANES III data from a very small sample size.

    So what we opted to do was to choose an additional data set to add in here between just beyond birth--these were not birth points, but at the first visit to a clinic--up through 5 months of age from the CDC's Pediatric Nutrition Surveillance System.

    These are data on low-income infants.  However, we didn't use a representative sample of low-income infants.  Instead, what we did is we chose clinics that matched the national distribution in terms of their mean, standard deviation, and skewness at each age from birth--from 3 months of age through 11 months of age compared to the NHANES III data.  So we were pulling out clinics that the children in that clinic happened to look exactly like the national distribution and chose those clinics and assumed that they would also look like the national distribution would have looked between birth and 3 months, and then added those data to the curves here and connected using the Missouri and Wisconsin data, the CDC nutrition surveillance data, as well as the NHANES III, and continued the curves using that.

    Finally, for head circumference, we had no national data, and we returned once again to the Fels Institute data for the head circumference at birth point and, again, connected that with the NHANES III data.

    So you see the picture here is one of bringing together multiple data sets.  We had a number of comparisons to try and make sure that these were valid comparisons to make, but it certainly leaves us with a difficulty using multiple data.

    So what are some of the other differences?  First, with regard to the smoothing, the old curves were smoothed with cubic splines, with knots at birth, 6 months, and 18 months, for those of you who work on these statistical arenas.  What that meant for us was that there were six independent parameters that characterized growth between birth and 3 years of age.

    In the CDC 2000, a completely different approach to smoothing was applied.  Here we used fractional polynomials that had been used previously in other growth studies, primarily used in the Fels data as well as some Canadian data.  And so they had kind of been proven methodologies for assessing growth during the first three years of life.  However, there were a smaller number of parameters that described growth, really just three independent parameters--roughly three, because there were some other things that were done to the curves to get them to fit.

    Finally, on weight-for-length, there were no set models, and so we used a 5th degree polynomial to maximize the flexibility of the curves there.

    I mentioned before that in 1977 the standard deviations were estimated as a separate path and created a separate set of curves.  In that case, we had two different standard deviations.  There was a standard deviation above the median and a standard deviation below the median.  And so if you think about kind of plotting the standard deviation as it goes across the curve, it was a low standard deviation that instantaneously at the median rose to a higher level.

    We calculated these scores in the normal way, taking the measure minus the median divided by the standard deviation.

    In the CDC 2000, this kind of immediate change in the standard deviation at the median was thought to not be desirable, and so instead we had a more continuous change in the standard deviation.  In this case, we transformed the data with a Box-Cox transformation, a power transformation.  People understand how you take the log transformation of a data set or a square root transformation.  The Box-Cox is a continuous set of transformations that you can then choose a parameter to say how much you want to transform that data to make it symmetrical.  And then once it is symmetrical, you can fit parameters to normalize that curve.

    Some other differences, clearly this group would be interested in the differences between the infant feeding in the groups.  As I mentioned before, the old charts were virtually all formula-fed infants.  The quality of the infant formula across that time, of course, has changed, so it is a mixture of a number of different kinds of feeding across those year '29 through '75.

    In the CDC 2000, we have a mixture of breast feeding and formula feeding, but it still is primarily a formula-fed group.  At 2 months of age, about half were formula-fed, half breast-fed, but by 6 months of age that was down to 28 percent currently being breast-fed.  And we're not talking about exclusive breast feeding.  As you can see, the exclusive breast-feeding rates are much lower than that.  Down to less than 10 percent by 6 months of age were exclusively breast-fed.

    If I can just take a couple of minutes to show you a comparison between the old and the new curves with regard to breast feeding, and we'll stop there, Ed showed a number of things as to how the old curves performed against WHO pooled breast-feeding data set.  We did some additional comparisons seeing whether the new curves have actually improved that situation.  So we've compared the WHO pooled data set that came together from six different studies of exclusively breast-fed children and pooled those data together.  And instead of plotting the means, which is what Ed was showing us, here I'm going to show the percent below the 10th percentile using the old curves versus the new curves.

    When we're looking at weight-for-age, you can see that there really has not been much of a change between the old curves and the new curves.  We have this same problem that as children get older, we're going to diagnose more of them as being underweight.  Just as Ed was showing us that the means go down, the percent that would be low is going to get higher.  And we still have that problem with the new curves, that it is considerably--we would have considerably more older children being considered underweight in this breast-fed data set compared to the younger children.

    However, with regard to height for age, we've improved the situation somewhat.  With the newer curves, there still is a tendency toward increasing the percentage that would be considered low as we get to older infants, but it's not as steep a trend as we had with the older curves.

    And, finally, with regard to weight-for-height, whereas the old curves showed that same pattern of worsening nutritional status over the first year of life, on weight-for-height the new curves have pretty much wiped out that problem.  We see a much flatter distribution across age.

    I'm going to skip over these because Ed is going to come back to many of these points in his description of the analytic issues.

    Conclusion:  The interpretation of the new charts is really not widely different from the old charts.  We're still using the same kind of way of thinking about growth and the way we analyze growth as very similar in the old curves to the new curves.

    There are a number of enhancements that argue for changing over to the new curves.  I'm not arguing that we haven't made enough of a difference to adopt these new curves.  However, I do think that the WHO reference that Ed has described would relate to a more substantive change in our interpretation of growth parameters, and we really need to give much more thought to different ways of thinking about growth than we have so far.

    Thank you.

    DR. GARZA:  Thank you very much, Larry.

    Any questions?  Dr. Stallings?

    DR. STALLINGS:  I actually have a series of questions, lots of things jotted down.  One, also, thank you and your team for doing this.  It was a monumental change to give us this from the clinical point of view.  It's wonderful to have the charts revised.

    In that last set of slides where you were showing the less than 10th percentile, just so I'm thinking about it correctly, the perfect outcome would be 10 percent would be less than the 10th percentile, because if we were looking at a population study, would you by definition expect 10 percent to be less than the 10th percentile?

    DR. GRUMMER-STRAWN:  Yes.  What I'm more concerned about is the pattern of growth there than the actual level.  We're comparing curves that are based on formula-fed infants mixed with some breast-fed infants against a group of breast-fed infants.  And so I wouldn't have been surprised if the level was somewhat different in a group of breast-fed children that might have less malnutrition than a general U.S. population.  But I would expect the pattern should be representative of the pattern of growth.

    DR. STALLINGS:  But, still, when I was looking at it, when there was almost zero less than the 10th percentile, I saw that as unusual as when we have 20 percent less than the 10th percentile, that that's sort of the concept.  If we were getting--whatever right is, but if we were getting it right, the population would go along that.

    While we have all the experts in the room, we've heard three different people say three different things about whether the Fels data included a lot of breast feeding or not.  And before you guys leave today or tomorrow, I'd like that to be readdressed.  You know, you were very helpful in showing in your slide the percentage of children who were breast-fed and then exclusive, and so obviously it's not one number.  Where you are at 2 weeks of age is different from 6 months.  But I think it would be helpful for the committee and for the FDA to have an understanding of what we believe the Fels data represented as exposure to breast feeding and what you believe your 2000 charts represent.  This is--

    DR. GRUMMER-STRAWN:  I'd like to have that, too, because I certainly--

    DR. STALLINGS:  Okay.  Well, then, you guys aren't leaving until we get this right.

    [Laughter.]

    DR. STALLINGS:  The other part of that is I think getting a handle on that will be even more important as the WHO new charts come out and those data are there.

    My last question, which really is a question, is:  If I understand, then breast-fed babies grow faster during the first 4 months of life compared to our usual reference data and more slowly between 4 and 6 months, if we were to look at the zero to 6-month period that we're really supposed to be focusing on.

    DR. GRUMMER-STRAWN:  I think it's more like around 3.  Is that about the peak?  Three months of age.  So faster in the first 3 months of life and slower after.

    DR. STALLINGS:  So that's part of the pattern that we're trying to capture that has made individuals difficult to look at.

    Thank you.

    DR. GARZA:  Any other questions?

    [No response.]

    DR. GARZA:  I have two.  In thinking through the reasons why the WHO took more frequent weight measurements, one of the principal things that drove that was that very often the pattern of growth was used to assess the nutritional management of infants, the first 3 to 6 months.

    To what degree can the present reference between used with that amount of detail, the first 3 to 6 months?  Were the smoothing techniques in your judgment sufficient to capture the differences in growth patterns that Ginanne just described?  Or did the smoothing eliminate much of that?

    DR. GRUMMER-STRAWN:  First of all, you have the whole difficulty of dealing with the cross-sectional data as opposed to longitudinal data.

    DR. GARZA:  I know.  That was Part B to 1.

    [Laughter.]

    DR. GRUMMER-STRAWN:  So all of those issues come in there.  You have not a very large sample size in that age range, and so there is a fair amount of noise.

    In addition to that, you're fitting basically a three-parameter model to the first three years of life.  That doesn't give you a whole lot of degrees of freedom to really let that first few months take on a particular shape.

    That said, we did examine what the curves were doing in that age range, and they looked like they fit fairly well.  But there's a large amount of noise in the cross-sectional data that bounces from month to month, and you look at the curves and say, you know, I think we did the best we can given the data that we have.

    So I don't want to blame the three-parameter model, but I would say that if one had better data and wanted to capture really what is the pattern of growth in the first year of life, I probably would not do the smoothing in the way that it was done.

    DR. GARZA:  And the second, since you answered Part B of 1, we've been asked to look at various control groups, either historical or using specific references as controls in clinical studies.  You also make the distinction between standards and reference.  To what degree can, in fact, one use the present reference in making judgments, value judgments for clinical studies and making comparisons between feeding groups that might be placed on new formulas and the current CDC reference?  Is it sufficiently robust to be used as a standard in making that clinical judgment that control groups normally play in clinical studies?

    DR. GRUMMER-STRAWN:  I think the question is one of what do you want to compare to.  What is right?  In comparing to these curves, you are implicitly saying I am comparing this child's growth or this group of formula-fed infants' growth to the way children have grown in the United States essentially over the last 10 years, maybe 15 years.  Are you comfortable with that kind of a comparison?

    You would say based on formula--in assessing a formula, you are saying this particular formula that we are evaluating generates a pattern of growth similar to the way children grow in the United States, whether they are formula-fed or breast-fed, fed on all kinds of different formulas, mixtures of solid feedings coming in at various ages.  If you are comfortable with that kind of a comparison, this formula produces a pattern of growth like children in the United States, like a cross-section of all the children in the United States, then I say yes, this is a perfectly fine comparison to make.

    If you want more of a prescriptive statement that this formula produces a pattern of growth that is the most healthy, I don't think that you can say that.  I don't think that we can infer that a cross-section of infants from the United States with all of the variety of primarily infant feeding patterns--there are also varieties in terms of socioeconomic status, the kinds of conditions children are running up against.  I think that in terms of the impact of those on these curves, those are not as great.  But the variation in feeding patterns, I don't think that we can say that we have the best pattern of growth here.

    DR. GARZA:  Any other questions?

    [No response.]

    DR. GARZA:  Thank you again.

    We'll move on to the next paper.  I don't think Dr. Fomon needs an introduction.  I can't think of a more senior person in pediatric nutrition than Dr. Fomon, and I'm very pleased that he was able to join us today and was willing to leave lovely Texas for Washington for this purpose, from one native Texan at least.  And he will be addressing the Iowa data and the Iowa/Fels growth data.

    Thank you very much for joining us, Dr. Fomon.

    DR. FOMON:  As the most senior member of the presenters, I'd like to make a statement for the presenters that we were asked to prepare a 20-minutes presentation, and if we run over and get the zero sign at 15 minutes, we're not very apologetic.

    [Laughter.]

    DR. FOMON:  First I want to get out of the way what are the Iowa data and what are the Fels data, and then we can talk about more interesting things.  The Iowa and the Iowa Fels data, Iowa Fels data is published by Guo, et al.  Term infants measurements all made with highly quality controlled efforts.   Caucasian, we were able--and I will show you that--no, but it's in my paper--that the length and weight of the Fels and the Iowa series were very similar at three months, which gave us some encouragement in combining them.

    Iowa data.  There are other Iowa data, but I wanted to first speak about 8 to 112 days.  These were all formula fed infants, 380 males, 340 females, and mostly infants of upper socio status, but not necessarily economic, because they were mostly UI personnel, University of Iowa faculty and students.

    They were measured within 4 days of 8, 14, 28, 42, 56, 84 and 112 days.  No exceptions.  And the age of measurement by interpolation or extrapolation was made to the target age.  So we adjusted if a baby was measured at 16 days.  We used the 28, 16 and 8-day measurements to compute a 14-day measurement.  It was fully longitudinal.  There is no one of those 380 males or 340 females who was not measured at every time.  This is published, so you can read about it in books.

    Then we did also measure a number of infants from 112 to 196 days of age, and they were 165 males, 188 females, and there was a subsample of these that were also in the other group that I talked about.  There were 63 males and 74 females who went from 8 to 196 days.  We have many more now, but it's too expensive to do the analysis.

    The Fels data, there are a lot of Fels data.  The Fels data that I'm talking about are the Fels data that are included in the Guo, et al. paper, and that included 240 males and 236 females, wide range of socioeconomic status, few measurements during the first 3 months.  The target ages were 1, 3, 6, 9, 12, 18 and 24 months.  Most of the children were measured within 3 weeks of the target ages, and there were some missed points, and they arrived at these ages by mathematic curve fitting.  And there are other people who know a lot more about that here than I know.

    So the Iowa Fels data, we ignored the Fels data during the first 3 months, so it doesn't matter whether they were breast fed or formula fed.  There were, for the first 3 months, just Iowa data, 580 males and 562 females.  The reason that's bigger than the numbers I showed you before was what I showed you before was formula fed.  We included breast fed in this, and I'll tell you why.

    Then we used both Iowa and Fels data, having established that the Iowa Fels size at 3 months was very similar.  We used both for 3 to 6 months of age, and then we used only Fels data for 6 to 24 months of age.

    So those are the Iowa data and the Iowa Fels data.

    The Iowa data and the Iowa Fels data and most other referenced data, except those for international comparisons, have been developed to detect abnormalities of infants, of individual infants, and to detect abnormalities of individual infants you have greatest interest in the outlying centiles.  The more individuals you have, the more confidence you have in those outlying centiles. What you want is early detection of growth abnormalities.  Growth will not tell you whether a child is normal or not, but it gives the most important single clue to telling you that that baby is one that needs closer attention than the general garden variety baby.

    And for that reason, weight gain is more important than length gain, and that's because it gives you this clue much earlier than change in length.  I think I have that on the next slide.  And you need data for at least the first 2 years.  Iowa data are only good, up to at best, 196 days.

    I have to go back, but I don't know how to do that, so I'll tell you that--can you go back?  Just go on back.

    The weight gain is more important than length gain because as I said, weight gain gives you the clue earlier.  Length gain is very difficult to measure accurately enough to be useful for determining changes in length the way it is done in hospitals, clinics and doctors' offices.  So length is not really a very feasible way, and moreover, there are very few instances in which length gain will be abnormal and weight gain normal.  So weight is the most important thing.

    Now, when you evaluate an infant formula--we didn't think about this until much after we published most of the Iowa growth data--you have different criteria for what you need as reference data, and the characteristics are it should be longitudinal.  It's difficult for me to agree that you can use cross-sectional data as a sensitive way of analyzing longitudinal data.  If you're doing an infant formula study, you're examining how the infants grow over the period of study.  For that you need longitudinal reference data.  It should be gender specific.  I haven't heard any argument about that.

    The study integral must include at least part of the neonatal growth spurt, should include all of it or most of it.  Neonatal growth spurt is from 8 to 42, maybe 8 to 112 days of age.  After 112 days of age growth rates are substantially less, and we'll come back to that.  And you need length data as well as weight data because it is possible that you would find--and I'll give you an example of this--a situation in which babies would grow normally in weight or maybe super normally in weight, and the weight to length would be outside of what we see with usual infant formula, suggesting that maybe this formula is not fully adequate.

    The reference population should be similar to the study core, and that's always going to be a problem.  It will never be exactly like the study core, but the question is, how close can you get and how close do you need to get?

    I missed the last one.  If I knew how to use this, probably I'd know how to use my camera too.

    [Laughter.]

    DR. FOMON:  The duration of study should be at least 84 days.  That's a new minimum length.  I made up the old 3 months minimum length that's in the AAP report, and I didn't have any good basis for that, but I think that maybe you can agree to 84 days as well as you can agree to some slightly longer figure, and we'll come back to it.

    I said this.  The most sensitive evaluation of the longitudinal growth study of a cohort in the longitudinal growth study requires longitudinal reference data.  I don't say that this is gospel.  It's just what I believe. Gains in weight and length are more rapid in infant males than in infant females.  The formula may be adequate for females but not for males.  Nobody's arguing about the gender anyway.

    The study integral must include at least part of the neonatal growth spurt.  A formula may be adequate for older infants but not for younger infants.  The reason is that during the period of most rapid growth the ratio of specific nutrient  to energy is highest, and if you get beyond that period, the ratio of protein or calcium or whatever to energy may be down at a lower level.  So if you start a study at 4 months and you get 4 or 5 months of additional data, the formula may be fine, but it doesn't tell you that it will be fine starting at 8 days or at birth.

    Here is an example. These are males, and this is weight gain from 8 to 56 days, and we did a study of a relatively low protein diet based on isolated soy protein with or without a methionine supplement.  We were interested in getting a fix on the requirements for sulphur containing amino acids.  And with the methionine supplement the gain was 42.3 grams per day, and with no methionine supplement it was 38.8 grams per day.  Compared to the reference data, 8 to 56 days, this was not significant and this was significant.  From 56 to 112 days there was no difference.

    I wish I had 3 or 4 more studies to demonstrate this, but this suggests at least that you need to have an early portion where you have the maximum postnatal growth included in your evaluation of a formula.  You need data on length as well as weight, and this was the best example.  I think if I spent more time I could find more examples and then it would be more convincing, but I couldn't do that, because I had to work in my yard.

    [Laughter.]

    DR. FOMON:  These are males and this is the BMI, and this is a low-protein formula and this is the reference.  And at 8 days, when we enrolled them, the low-protein, the cohort receiving the low-protein formula had a BMI of 13.  That was significantly less than the reference data.  At 112 days the low-protein cohort had a BMI of 18.6 which was significantly greater than the reference data.  And all our data, including--if you pardon my expression--skinfolds, indicated that these babies were fat.  And what we speculate is that they were--and they took more volume.  We, in all our studies, record how much the babies eat.  We weigh the bottles in and the bottles out, and they took more energy in and they gained more weight, and we speculated that they ate more because of the low-protein content.  They were eating--I don't know how they knew how to do it--but they were eating more to make up for the low-protein concentration in the diet.  They got enough protein.  They grew normally in length.

    The study cohort must be similar to the reference cohort.  They should be healthy, that is they shouldn't include babies with illness.  They should be term if what you want to know is, is your formula going to be adequate for term infants.  You don't want to increase the noise in your experiment by adding preterm infants.  And then the question of ethnicity, terribly important in international studies, may not be so important in the United States where you're comparing what seems to be similar groups, but that's a question that needs to be carefully examined for each study.

    So the duration of study should be at least 84 days.  I think that 8 to 112 days, which is why our data are mostly 8 to 112 days, or 14 to 112 days, almost as good I think.  At 8 days, there are great advantages of 8 days.  Most formula-fed babies have regained their birth weight.  You can get a really good measurement of length at 8 days.  You're not there to get it when the baby is born, and even if you are there, the hardest time to get an accurate measurement on a baby is at birth.  Howard Meredith, many years ago, showed me some publications on how very hard it was to get an adequate length at birth.

    Now, I think that in the current U.S. climate, where you can hardly recruit any formula-fed babies before 42 days of age, which is still fairly easy, but you may be able to recruit them at 28 days.  Later than 28 days you miss too much of the postnatal growth spurt.  Earlier it's too hard to recruit, so maybe 28 days is the most feasible, and 28 to 112 is 84 days, and that's how I came to 84 days.

    Now, if I give you my recommendations to the FDA, straight out, no hedging, I would say that size data are not relevant.  I mean I know half the people in the audience hate me.  Size data are not relevant.  Data over 6 months of age are not relevant.  I'm not even sure that data over 4 months of age are relevant.  Breast-fed babies are not relevant, but that's not so serious because during the period we talk about they gain about the same.  I just don't like to muddy up a study of formula-fed infants with a mixed group that I can control.  And unless you have--you must match the cohort with the reference group.  If your reference group consists of term infants, then you can't muddy it up by including infants, preterm infants in any number that they might be present.

    So those are my messages and I'm willing to take the flak because I've done it before.

    DR. GARZA:  Thank you very much, Dr. Fomon.  I regret the miscommunication between FDA and the speakers.  We'll try to get that resolved for you in terms of the 20 versus 15 minutes.

    Are there any questions?  Dr. Anderson?

    DR. ANDERSON:  Anderson.  I understand that the Iowa Fels data are longitudinal data.

    DR. FOMON:  Absolutely.

    DR. ANDERSON:  And that the recently published CDC standard is largely from cross-sectional data.

    DR. FOMON:  Absolutely.

    DR. ANDERSON:  To what extent do the percentiles generated from the two sources differ in substantive ways?

    DR. FOMON:  I don't know that offhand.  I think from the point of view of infant formula evaluation, that it's not relevant because you can't get good analysis of longitudinal data from a cohort under study by comparing with size data.  In 1976, just to prove my seniority, I had a dialogue with Peter Hamill [ph] over about 9 months, trying to convince him to call the NCHS charts size charts so that people wouldn't be confused by thinking that they're growth charts, but I lost that argument.

    DR. GARZA:  Dr. Stallings?

    DR. STALLINGS:  Dr. Fomon, I'm interested in your opinion of a couple things.  One, the need for a control group, a contemporary control group when you're actually doing a study, you know, particularly for the kind of things we would be looking at, a change in formulation.

    And secondly, when you said that you thought the breast-fed baby didn't have any role in thinking about evaluation of formula-fed babies, I think rather than is breast feeding--the growth pattern of breast-fed babies a pattern that should be strived for with formula-fed babies?

    DR. FOMON:  Well, let's see, question number 1.  Tell me question number 1 again.

    DR. STALLINGS:  About a contemporary control group.

    DR. FOMON:  I think there are circumstances under which a concurrent control is essential, and one was the study that I mentioned to you, where we had a formula, a low-protein formula fortified with methionine or not fortified with methionine.  In that case we had to demonstrate that with methionine it maps the reference group.  But I think in general it's probably not really essential if you have good reference data for comparison, don't think it's really essential to have a concurrent control.  It greatly increase the number.  The number that you need to compare one cohort with the reference data is considerably less than the number that you need if you're going to compare it with a concurrent control.

    So just from the practicality of making it possible at all to study new formulas, I think it's a reasonable compromise not to require a concurrent control unless something about the ingredient change or whatever suggests that a concurrent control would be valuable.

    And on the other question, should a cohort fed a new infant formula be compared to a breast fed control reference group, I think that's more a philosophic matter than a scientific matter, and my own conclusion is that if you want to study a new formula, you should study it in comparison with old formulas, and not with some group that we think might represent ideal growth.  That's just what I think.  Other people think other ways.

    DR. GARZA:  Any other questions?

    [No response.]

    DR. GARZA:  Thank you very much.

    It is 5 minutes to 12:00.  We're going to try to make up the 20 minutes from this afternoon's schedule because we've been running a bit over time in the presentations, so that rather than coming back at 1:35 as the schedule suggests, I'm going to ask people to come back at about 1:10.  So that we can make sure we can started by that time, having everybody here at 1:00 o'clock would be ideal.  So we will ask the Committee to reconvene at 1:00 and we'll get started after that as possible.

    Lunch for the Committee and the speakers is in the room where you had coffee, and the cafeteria is available to everyone else.  1:00 o'clock.

    [Whereupon, at 11:57 a.m., there was a luncheon recess.]

A F T E R N O O N   S E S S I O N

(1:07 p.m.)

    DR. GARZA:  We can get started.  Our next speaker--can I have all our guests please take their seats?  I think the Committee is at the table.

    Our next speaker is Dr. Jon Tyson, who is a professor of pediatrics, obstetrics, internal medicine and epidemiology at the University of Texas Medical School in Houston and the School of Public Health.  Dr. Tyson will not be able to stay with us tomorrow, so I want to make sure that the Committee members ask all questions or clarify any outstanding issues that they might have related to the topic of growth data for preterm infants, because in fact Jon will be leaving soon after his presentation.  So it's important that you try to get your questions to him before he returns to Houston.

    Jon, thank you very much for coming.

    DR. TYSON:  Thank you very much.

    Well, as a neonatologist and epidemiologist, I'm going to try to make the case today that if what you mean by normal growth is desirable or healthy growth, that the evaluation of early growth in preterm infants will necessarily involve evaluation of health and development.  I'm going to try to go through this in a orderly fashion to promote a rational and evidence-based decision making.

    The first question is:  can the growth rate sustained by a new formula be adequately assessed using published growth norms?

    This is a growth curve from the Neonatal Research Network, observed rates of growth, are weight--sizes, Dr. Fomon would say--in babies according to their birth weight.  Throughout their hospital stay there is serial information assessed.  Also in length and head circumference and mid-arm circumference for these babies.  There's no data for babies greater than 1,500 grams and no data beyond discharge.  I would add that there in my little handout, there's a website that you can go to where you can, for an individual baby, enter the measurements at birth, and print out a growth curve for that particular baby.

    For preterm babies beyond discharge, this shows what I think is the best available data from the Infant Health and Development Program which had 985 preterm babies, and it provides data for 3 groups:  2,000 to 2,500 grams, 1,250 to 2,000 grams, and less than 1,250, plotted according to post-conceptual age--I think that's really post-menstrual age--up to 3 years of age.

    Now, do these both norms describe normal growth?  And I think we have to be really careful what we mean when we say normal values, because I think that's a term that often causes confusion.  I actually try to avoid that word.  I see it used sometimes as referring to values that are expected or typical, typical values, values that are not associated with adverse outcomes or low-risk values, values that do not cause adverse outcomes, healthy or optimal values, and values for which intervention has not been demonstrated to be beneficial.  I think it's really important that we keep clear what we're talking about.  At best these growth grids describe what are typical values.

    How should they be used?  I think that they're a useful clinical tool to assess the growth pattern of individual babies.  Whether they are appropriate regulatory standard to evaluating the formula, I'm sure the answer to that is no.  They provide neither the optimal values nor an adequate basis to compare the growth with new formula to conventional formulas.

    As I've spelled out in the handout, if you compare the growth for the new formula to one of these growth grids, what you describe as an effect on growth might be due to any of a large number of factors, including intervening changes in care and outcome since the growth grid was developed, the fact that in most studies you're using selected patients, whereas the growth grids are based on all patients; a myriad of differences between centers, and the opportunity for bias in patient care, selection, care and assessment in evaluating new formulas.  And I think it's particularly important when the sponsor or the investigator has a financial or even a professional interest in the outcome of the studies to attempt to avoid bias.

    I think we also have to ask ourselves whether the statistical tests that are commonly done in evaluating growth studies are misused and then misinterpreted in assessing interventions using historical controls.  The babies in feeding studies and the babies in these norms that we see for preterm babies and for term babies, are clearly not a random or even a representative population of the sample of the same population.  Moreover, a p-value of less than .05 is often taken to mean that the difference is due to the intervention, when it may be due to any of a number of factors a difference in population, differences in the way the populations are assessed, et cetera.

    So I think we should be asking the question, why use historical controls at all to evaluate new formulas for preterm infants or for term infants as well.  For every other intervention that we talk of in medical care, the randomized trial is the gold standard.  The concurrent cohort, carefully done, carefully studies, is a silver standard.  Historical control is a bronze standard.  Why do we want to use the bronze standard?  And these are also issues in concurrent controls as well.

    Should carefully designed randomized trials be required?  I think the answer to that is clearly yes with a number of other features designed to minimize random error or systematic error and increase the signal-to-noise ratio, which would include mass caregivers and evaluators, well-standardized evaluations shown to be reliable by the people who do the assessments in the study, effective procedures to avoid attrition, and intention to treat analysis predefine stopping rules in an adequate sample size.  And I would add to that I think a commitment to publish the data at least on the website, no matter what the data show.

    Now, there has been opposition to the use of clinical trials by formula companies that would be charged with this responsibility, and I think part of that opposition is the expense and the feasibility of such trials, and I think with the progress in organization of research effort, that this is more feasible now with lots of neonatal research networks out there, some that already include follow up evaluations in at least some if not all the centers, and by the recognition that you don't--you need only do simple management trials.

    Most people, when they think about randomized trials are very expensive, are thinking about the usual traditional kind of explanatory trial, which are designed to determine whether therapies work in ideal or restricted circumstances, or that are designed to define the mechanisms of action.  Management trials, on the other hand, or so-called effectiveness trials, are designed to determine whether therapies work under routine clinical circumstances, so all the effort that goes into trying to control all the co-interventions in explanatory trials is inappropriate in a management trial.

    Who should be enrolled?  Who should be excluded?  I think you want to enroll representative sample of the babies for whom the formula is intended or at least the highest risk group.  So you want to include any babies who make up an important part of that population, small for gestational age babies, very sick babies, twins, et cetera.  You would want to exclude relatively few infants, say the babies who have major congenital anomalies or overt nonbacterial infections.

    Should infants fed their mothers' milk be included?  I don't think it's absolutely necessary, but I think it's highly desirable to increase the generalizability of the results of the trial because a large proportion of all preterm babies are fed at least some of their mother's milk, and also to help identify limitations of the formula, and areas for potential advances based on the benefits of mother's milk over formula after adjusting as best feasible for other factors.

    What assessments should be performed?  I think, obviously, body composition or biochemical, physiologic or functional variables need be considered, but most of the time those would have been studies in prior explanatory trials.  There might be some need to get some of that in some of the patients though.

    What about health outcomes?  Well, I think one variable that has to be considered is the percent of infants with necrotizing enterocolitis.  This is a serious disorder with a mortality exceeding 50 percent in surgically treated extremely low-birth weight babies, and it may well be related to feeding.  Death is a competing variable for necrotizing enterocolitis.  You have to live long enough to get NEC, so you would need also to look at the composite outcome of death or necrotizing enterocolitis.  We also have to be worried when we feed babies with chronic lung disease that rapid growth may not be attainable or even desirable.  If you have marginal pulmonary sufficiency, how fast do you want that baby to grow?  So you would like to know about the combination of death or prolonged mechanical ventilation.

    Neuro developmental outcomes I think are at least highly desirable if not mandatory, because first growth and development may be differentially affected.  You need to exclude adverse outcomes on development, even in the presence of good growth rates.  I think also we need this information to better define the optimal growth rate and the appropriate goals for growth rates sustained by formulas for infants with or without serious illness.

    While growth assessments, weight, length, head circumference and weight-length ratio I think are essential, there may be others as well.  What minimum period of assessment is needed, I think we should remember that a reliable identification of major neural developmental impairments is probably not possible any earlier than 18 months adjusted age, that is, post term, and this would allow evaluation of potential late effects beneficial or hazardous on time-limited interventions given in the NICU or later.

    What standard should be used in judging the growth of preterm infants fed new formula?  The American Academy of Pediatrics has said that the goal should be to achieve rates in an extrauterine environment like those that would have been achieve in utero had the baby not been delivered early.  But we have to wonder if this is really the right goal.  We have some uncertainty about what this rate is currently and we can discuss why.  If we can only measure gestational age well, we could do this better.  Currently we think that it's about 15 to 17 grams per kilo per day weight gain, about 1.1 centimeters per week in length, and about .7 centimeters increase in head circumference.  We again need to ask is this an appropriate goal for infants with severe lung disease, and we also had this observation of persistent growth deficits after reaching full feedings.  So even though we can get babies to grow rapidly, once they get to full feedings, there's this long period of time when they're growing poorly as they recover from illness after birth, and this shows you data for the neonatal research network for babies of different gestational ages, 24 to 25 weeks, 26 to 27 or 28 to 29 in relationship to a so-called growth grid that Alexander published.  And you can see that the babies don't do that bad once they start growing and taking a full intake, but they end up with most of them smaller than the 10th percentile for babies developing in utero with the same gestational age.

    Whatever goal we try to take we have to think of it as provisional, but we could ask, should the current standard for judging preterm formulas be the formula that sustains the best catch-up growth, and that could be that the weight, length and head circumference and the body proportions would be most like that of term infants of the same adjusted age, providing there were no adverse effects on the health or development through 18 months as identified in a well-designed trial.

    How many infants would you need to study to assess a new formula?  This is a really complex and important generic issue in assessing intervention, any intervention where there may be an uncommon but serious potential hazard like necrotizing enterocolitis, and I'm going to spend some time on this even though this may seem to you like a statistical issue only, I think it's an important practical issue, because the kind of things I've said would make feeding studies so large that formula companies or indeed the NIH may be unwilling to fund these.  So I want to try and see if we can find some way to address this.

    So somebody may say, well, wait a minute.  The old formulas have not actually been tested that well, and I've got a new formula here that has strong a priori evidence and rationale for using it.  Say it has a component that's provided before birth across the placenta and in human milk after birth, not given in prior formulas.  It's not well synthesized from precursors in preterm babies.  And we think it's important for healthy development.

    I think even in that circumstance you still need to rule out the possibility that there are important unrecognized hazards of this formula and I'm going to try to list what I think those are or the most important.

    The first would be an absolute increase of at least 3 to 7 percent or more in major adverse neonatal outcomes, particularly necrotizing enterocolitis.  A 3 percent absolute increase corresponds to what's called the number needed to treat of 33.  That is, for every 33 babies that you this formula to, you would cause one baby to have necrotizing enterocolitis.  I think that would be unacceptable even if all the other babies benefited in growth or perhaps even in development.

    A second would be a reduction in developmental quotient at 18 months of a quarter of the standard deviation or more.  That's the mean developmental quotient of 18 months and reduction of a quarter of a standard deviation or more.  If you observe that, that would substantially increase the number of preterm infants with a deficient or marginal IQ that would be eligible later for educational intervention programs.  In the neonatal network this would correspond to a reduction of almost 5 points on either of the Bayley subscores.

    Third would be a reduction of a quarter of a standard deviation in length or head circumference at 18 months.  And this of course is arbitrary, but at least after recovery from serious illness, there's no apparent benefit of slow growth so that I would think a modest decrease in length or head circumference, if not weight, could be seen as presumptive evidence of harm.  And this would correspond in the neonatal network in 18 months to about 250 grams in weight, 1-1/4 centimeters in length and a half a centimeter in head circumferences.

    Now, you may think that this is too small to look at, but I would reassure you that the sample size needed to assess necrotizing enterocolitis, if you use a sample size that's large enough for that, you can evaluate very small effects on growth.

    Now, if you take a conventional approach to sample size, you would need 315 per group to have 80 percent power to identify a quarter of a standard deviation difference in either development or growth or size at 18 months, and an alpha error of .05, assuming you lose fewer than 20 percent of kids to follow up.  The power to identify an increase in necrotizing enterocolitis would be 78 percent, for a large increase, 7 percent, that would be a doubling of the right of necrotizing enterocolitis in the neonatal research network.  It would fall to only 22 percent for a 3 percent increase, so a really small power to look at a clinically important increase.  If you said you'd like 90 percent power to identify a quarter of a standard deviation difference of 18 months, you'd need 421 per group.  Your power to identify an increase in NEC would still be only 30 percent for a 3 percent increase in NEC.

    What can you do about this?  One potential way to address this is a non-inferiority trial.  For the sake of time I'm not going to talk about that.

    Another, and I think this is, practically speaking, the most attractive option, is to increase the p-value considered statistically significant in evaluating a serious hazard.  As you know, the same p-value, usually p less than .05 is used for benefits and hazards in studies, and this is an arbitrary and not well justified practice.  For a serious hazard like necrotizing enterocolitis, a higher p-value might be justified on multiple grounds.  First is, we know in clinical studies the direction of bias is toward finding benefits rather than looking for harms.  There's a lot more effort put into it in general, and the studies are powered to evaluate benefit rather than harm usually.  But the hazard may be much more important than the benefit.  And as pragmatic evidence, we know that data safety monitoring committees that review the accruing evidence in clinical trials will stop a clinical trial at a much higher value of p for hazard than for benefit.

    The appropriate p value shouldn't depend in part on the cost of drawing the wrong conclusion.  For a serious hazard like NEC, I would contend that we might select a p less than .30.  That would still result in a 70 percent chance or higher that a difference of that magnitude would not occur by chance under the null hypothesis.  If you did this, what you're doing is you're increasing the risk of a false positive conclusion, that is, that you would conclude that the formula causes NEC when in fact it doesn't.  In order to reduce the risk of a false negative conclusion, that is, a conclusion that the formula doesn't cause NEC when in fact it does.

    So if we go back to the numbers we calculated before for benefit, at 315 infants per group, again, that was for 80 percent to look at a .25 SD difference, the power to identify an increase in NEC would be 96 percent very high power for a large increase, 58 percent for a 3 percent increase.  So you're slightly more than 50 percent likely to identify it.  If you use 421 per group, you then get down to a power of about 2 in 3 to identify a 3 percent increase.

    If you found hazards at a p of .30 and benefits at a p less than .05, what would you do?   Well, I think you wouldn't recommend the formula, you'd recommend further study, and that would be a departure from what has been done in usual practice.

    I'm going to skip that one, and just conclude by saying that I hope I've convinced you that the growth of preterm infants should not be assessed in isolation from effects on health and development, that a large trial evaluating growth health and development to 18 months or more is needed to assure that the benefits of any new formula outweigh any hazards in preterm babies and to better define the effects of different growth rates, and the growth rate that we should be looking for in deciding how to design preterm formulas.

    Thank you.

    DR. GARZA:  Thank you very much.  Questions or comments?  Dr. Stallings?

    DR. STALLINGS:  It sounds like you've silenced us pretty well.  You know, we at the beginning, talked about dividing up preterm from term infants completely, and it sounds like, from your presentation, that you really, in the area of growth, that there really isn't anything you learned from term studies that would influence you on preterm.  Would that be a fair--

    DR. TYSON:  Well, I wouldn't say wouldn't influence you, but I don't think you can determine whether a new formula is appropriate for a preterm infant based on observations in term infants.

    DR. STALLINGS:  The other thing I'd like for you to elaborate on a little bit is I think it's often that--you were telling us a bit about who the sample should be, and in essence, the inclusion/exclusion criteria.  Which infants, if you elaborate, which infants should not be in a growth study?  Which preterm infants should not be in a growth study of preterm?

    DR. TYSON:  Well, I think it would be babies with the kind of problem that's very unusual, and that's going to have an overwhelming effect on growth like trisomy 13, growth and mortality, nonbacterial infection.  Other than those things, I think you're talking about 3 percent of babies or something.  The rest of them I would vote to include.

    DR. THUREEN:  Thureen.  Dr. Tyson, I know that in your paper you said you would include growth restricted infants as part of this because they're such a large portion of the population, but that you would substratify those infants for further analysis.

    DR. TYSON:  Yes, right.  You can of course include explanatory evaluations within a management trial, so it might be that that formula has a different effect on those babies.

    DR. THUREEN:  In terms of neuro developmental outcome, do you think it would be fair to exclude infants who had had very high risk factors for significant neuro developmental outcomes, such as intracranial hemorrhage, prolonged asphyxia, evidence of white-matter disease, before the trial even started?

    DR. TYSON:  If the formula is going to be fed to babies with severe asphyxia, then I think you would want to test it in those babies.  For some of those conditions you had mentioned, they would occur after you started the feeding, so like cystic white matter disease you might not identify till 36 weeks post conceptual age or something, and that's really, that's potentially an outcome variable.

    DR. THUREEN:  Would you pair match those infants then at all with other infants who had similar risk factors or known disease that affects neuro developmental outcome, or would you just do a purely prospective randomized trial?

    DR. TYSON:  If you do a large randomized trial, first of all, it gets really cumbersome to try and pair anybody at birth.  As long as you're stratifying by center and maybe a couple of other things like birth weight less than 750, 750 to 1,000, something like that, that you will end up with an approximately equal number of those babies in the two groups, and then you can go back and do an analysis.  If you try to stratify for birth weight, SGA, gender, birth asphyxia, et cetera, you end up with so many huge strata that the study gets really complicated to do.  And I don't feel as strongly about that as most statisticians, but my understanding of the school, most statisticians are towards the minimal prognostic stratification at randomization, and more toward post hoc looking at individual groups who should have been predefined ahead of time which group you were going to look at.  Does that answer your question?

    DR. THUREEN:  That makes sense.  And would you change any of your ideas about how to conduct a study if you are going to look at patients who this is their exclusive formula fed from initial feeding versus studies started when infants really attained full feeding?  Do you think that it makes any difference on how you conduct the study if you're looking at those two issues?  Did that make sense?  Because yours sound like you're referring to infants who may start minimal enteral feedings with the study of formula, rather than waiting until they attain full feeding and then starting from that standpoint?  Do you think it's preferable to do one or the other, or do all of your idea really apply to--

    DR. TYSON:  It depends on when the formula is going to bed.  If it's a formula that's going to be fed in the first week or something like that, I think you want to test it as it's going to be used.  Let's say that it was a formula that was recommended for us from the first feeding.  Let's say for the sake of argument that it cause necrotizing enterocolitis, and you didn't enroll baby, and you didn't start that formula until 3 weeks of age or something, or at a point when the babies were on full feedings, you might miss that effect, so you want to test it as it's going to be used in the real world.  Does that make sense?

    DR. THUREEN:  Yes.  And then lastly, do you believe that there are really no good reference standards for growth in the preterm infant or at least a certain subgroup of preterm infants that may be extremely low birth weight?

    DR. TYSON:  In the Neonatal Research Network, we have research nurses that are doing standardized--that have done standardized evaluations of anthropometry in intervals.  There are huge center differences.  If you try to take the data from any center to apply it to another center, you could easily be mislead just by the center differences.  So I don't see why you would want to use comparisons that would involve center differences or time differences.  This was gathered data 3 years ago when they were using steroids, postnatal steroids more often or less often than they are now.  Why not randomize and get the cleanest--I think the belief that you don't have to use controls, you don't have to use randomized controls, that you can answer the question with fewer patients is an illusion, that at a given number of patients your ability to get a unbiased answer to the question is going to be greater with randomized controls than with historical controls.

    DR. THUREEN:  Thank you.

    DR. GARZA:  Heubi?

    DR. HEUBI:  I think, Jon, this is all very interesting.  I wanted to actually ask you a couple of questions, and you'd have to follow my line of thought here.

    The number of subjects that you would entertain as being appropriate for a study is about 10 times what a typical current formula study would be.

    DR. TYSON:  Right.

    DR. HEUBI:  With that in mind, knowing what you knowledge is of the Neonatal Network, would the Neonatal Network sponsor studies like this because this is specific to preterm infants and it would be a potential rationale to study with partnering between industry and NIH money to do studies like this?

    DR. TYSON:  The Neonatal Network has a protocol review committee and standard procedures for--there's no reason that couldn't be proposed and seriously considered or accepted if it went through all those things.  There are lots of networks out there.  In Texas we've started a Texas network, and there's an Oxford network, and the Canadians have a network, and the Australians have a network, and I'm sure there's networks developing in the United Kingdom if they're not already in place.  So there are a lot of people willing to do this.  The hardest part is going to be the 18-month follow up.  That's a lot harder than studying NEC, but as more and more people say, "If I'm going to take care of babies this size, I need to know how they turn out.  That means I have to have a really well functioning follow up system.  So I think there are going to be people out there that can do it at much lower cost than if you just went to them on day one and said, "We're going to fund your whole follow-up effort in order to answer this one question.

    DR. HEUBI:  But I was looking at it from the standpoint of it being economically more attractive to industry to do studies through the network that exists through the NIH because some of the infrastructure already existed and was already being paid for in part by federal money.

    DR. TYSON:  Right.  Kathleen Kennedy and I proposed to the network a feeding study, and one of the things--and I'm glad I was involved in this effort--the business about the p-values that I presented today, that was aware stumbling block to us when we got to the--the reviewers really liked it and the statisticians said, "Well, you're going to have to study 6,000 babies or something," and I was working under the same mindset I had been before, well, that's if you want to look at a p less than .05.  But why not accept a higher p-value, or as I was going to say on the last slide, predefine what you think is an acceptable ratio of the number of babies who benefit to the number of babies that are going to be harmed, and then test that and say, does the number of babies who are helped by improved growth or development,  relative to the number who are harmed by NEC or worsening BPD, if indeed that's a hazard, is that an acceptable ratio or not?  And try to design studies not to look at one outcome variable but at the relationship of one or two variables or perhaps even more.  And I think that's a cutting edge issue in the experimental design that the time is ripe to do now.

    DR. HEUBI:  And this is a circumstance where clearly DSMB or some monitoring board, during the--

    DR. TYSON:  Yes.

    DR. HEUBI:  --would be pretty accepted.

    DR. TYSON:  Right.

    DR. DENNE:  Jon, you've made an argument about following preterm infants out to 18 months for neuro developmental outcome and I understand the rationale for that argument.  How do you feel about similar studies in terminants?

    DR. TYSON:  Do you mean randomized trials or--

    DR. DENNE:  No.  I mean the necessity for evaluating neuro developmental outcome at 18 months in studies of new term formulas?

    DR. TYSON:  I don't see why not.  I mean everybody in this room thinks nutrition's really important.  You take these formulas and you feed them to millions of babies.  Why test it in only 50 or 60 babies?  Why allow it to go on the market without knowing does it have beneficial or harmful effects as far as you can tell with an appropriate sized study in term babies?

    DR. ANDERSON:  Anderson.  In your discussion of NEC much of the calculations were done based on a baseline rate of 7 percent.  How would you feel about some of the adverse event monitoring being done not necessarily in the context of a randomized clinical trial, but against some fixed standard?  That is an infant formula would be unacceptable if it produced a rate of NEC above 10 percent.

    DR. TYSON:  There are several problems with that.  One is there is a lot of institution variation in NEC and with any institutions, there are periods when the NEC rate really goes up, and when it comes down, nobody quite understands that.

    And finally you have the potential problem of bias.  There have been studies, for example, they took x-rays of kids with NEC and x-rays of kids thought not to have NEC, going to every pediatric radiologist in California, and they found this incredible variability in what was called NEC and what wasn't.  And so when you're in the context of a study like that, I just think the opportunity for bias is there, and that your ability to relate that to some other institution in a different point in time I just don't think is worth the effort.

    DR. STALLINGS:  Stallings.  A slightly different angle.  And we talked this morning a little bit about, or inferred a little bit about term babies growing too fast, and you know, that's of concern, but certainly for my clinical time, the worry over preterm babies growing too fast, and I can remember bedside debates about too fast and it's only fat and it's no brain and it's no muscle and all of that.  I don't think we have nearly as good a handle on the body composition component of the preterm babies.  And then you add to all of that the concerns that we're all reading more and more about, is early postnatal growth a very--I mean we know it's an important time, but a differently very important time in lifelong health.  Would you make a few comments?  I know you made the caveat about babies with really chronic lung disease and concern, and that's really related to CO₂ retention, but put those babies aside, and can we grow preterm babies too fast, and how do we determine when we're approaching that?

    DR. TYSON:  Well, I clearly don't have the answer to those questions.  It seems to me the only way we can get it is to randomize babies to different feeding regimens that produce different growth rates and see who turns out to have the best h