U.S. DEPARTMENT OF HEALTH AND HUMAN SERVICES
FOOD AND DRUG ADMINISTRATION
CENTER FOR BIOLOGICS EVALUATION AND RESEARCH
+ + + + +
VACCINES AND RELATED BIOLOGICAL PRODUCTS
+ + + + +
+ + + + +
FEBRUARY 18, 2004
+ + + + +
The Advisory Committee met at 8:30 a.m. in
the Embassy Room of the Sheraton Four‑Points Hotel,
8400 Wisconsin Avenue, Bethesda, Maryland, DR. GARY
OVERTURF, Chairman, presiding.
This transcript has not been edited or corrected, but appears as received from the commercial transcribing service. Accordingly the Food & Drug Admin. Makes no representation as to its accuracy.
GARY D. OVERTURF, M.D.
MICHAEL D. DECKER, M.D.
MONICA M. FARLEY, M.D.
RUTH A. KARRON, M.D.
PHILIP S. LaRUSSA, Ph.D.
DAVID MARKOVITZ, M.D.
PETER PALESE, Ph.D.
CINDY LYN PROVINCE
WILLIAM FREAS, Ph.D.
AGENDA ITEM PAGE
Call to Order, Dr. Gary Overturf, Chair 9
Administrative Matters, Dr. Bill Freas 3
Presentation of Plaques to Retiring 9
Members, Dr. Karen Midthun
OPEN Session 11
Strain Selection for Influenza Virus 11
Vaccine for the 2004‑2005 Season
Introduction, Dr. Roland Levandowski, FDA 11
Vaccine Effectiveness, Dr. Carolyn 61
Col. James Neville, DOD 90
Dr. Antoine Flahault, INSERM, Paris 104
U.S. Surveillance, Ann Moen, CDC 118
Open Public Hearing
World Surveillance and Strain
Characterization, Dr. Nancy Cox, CDC 139
Additional Reports, Linda Canas, DOD
Dr. Maria Zambon, HPA, UK 176
Dr. Roland Levandowski, FDA 195
Availability of Strains and Reagents, 216
Dr. Zhiping Ye, FDA
Comments from Manufacturers 220
Update on H5, Dr. Nancy Cox, CDC 241
Dr. Phil Minor, NIBSC 262
NIH Speaker (TBA)
MEMBER FREAS: Mr. Chairman, members of
the Committee, invited speakers, and members of the
public, I would like to welcome all of you to this,
our 97th meeting of the Vaccines and Related
Biological Products Advisory Committee meeting.
I am Bill Freas. I am the Acting
Executive Secretary for today. At this time, before
the meeting, begins, I would like to go around and
introduce to the public the members seated at the head
We will start on the right side of the
table. That is the audience's right‑hand side. And
I will read the names of the people at the table.
Would the members please raise their hands as I call
David Markovitz, Professor, Division of
Infectious Diseases, University of Michigan Medical
Next is Dr. Walter Dowdle, Senior Public
Health Consultant, The Task Force for Child Survival
Next is Dr. Judith Goldberg, Director,
Division of Biostatistics, New York University School
Next is Dr. Ruth Karron, Associate
Professor, Johns Hopkins School of Hygiene and Public
Next is Dr. Walter Royal, Associate
Professor of Medicine, Morehouse School of Medicine.
Next is Dr. Monica Farley, Professor of
Medicine, Emory University School of Medicine.
Next is Dr. Pamela McInnes, Deputy
Director, Division of Microbiology and Infectious
Next is Ms. Cindy Lyn Province, Associate
Director, Bioethics Center of St. Louis.
Next is Dr. Bruce Gellin, Director,
National Vaccine Program.
In the empty chair, we will soon be joined
by Dr. Stephen Phillips, Director, Deployment Medicine
and Surveillance, Office of Assistant Secretary of
Next I would like to introduce you to the
Chair of this Committee, Dr. Gary Overturf, Professor
of Medicine, University of New Mexico School of
Coming around the table, we have Dr.
Philip LaRussa, Professor of Clinical Pediatrics,
Next we have Dr. Martin Myers,
Co‑Director, Public Health Policy and Education,
University of Texas Medical Branch.
Next we have Dr. Bonnie Word, Assistant
Professor of Pediatrics, Baylor College of Medicine.
Next we have Dr. Peter Palese, Chairman
and Professor, Department of Microbiology, Mt. Sinai
School of Medicine.
Next we have Dr. Arnold Monto, Professor,
The University of Michigan.
Next we have Dr. Ted Eickhoff, Professor
of Medicine, University of Colorado Health Sciences
Next we have our nonvoting industry
representative, Dr. Michael Decker, Vice President,
Scientific and Medical Affairs of Aventis Pasteur.
Next we have a nonvoting participant, Dr.
Nancy Cox, Chief of the Influenza Branch, Centers for
Disease Control and Prevention.
Next we have Dr. Roland Levandowski from
Dr. Richard Whitley, University of
Alabama, member of this Committee, has recused himself
from today's participation.
I would like to thank the members for
attending. There is one other person I would like to
introduce at this time. Many people have asked me,
"When are you going to get a real executive secretary
for this Advisory Committee?" I would like to
introduce this morning Christine Walsh, who will be
the next Executive Secretary for this Committee at the
next meeting, which will be announced later. It will
be either in March or May. We have a teleconference
scheduled for March the 17th at this time.
I would now like to read the conflict of
interest statement into the record. Before I do that,
I would like to ask the members of the public if they
could put their cell phones on silence mode, it would
"The following announcement addresses the
conflict of interest issues associated with the
Vaccines and Related Biological Products Advisory
Committee meeting on February 18th and 19th, 2004.
The Director of the Center of Biologics Evaluation and
Research has appointed Drs. Walter Dowdle, Ted
Eickhoff, Bruce Gellin, Judith Goldberg, Pamela
McInnes, Arnold Monto, Martin Myers, and Stephen
Phillips as temporary voting members for this meeting.
"Based on the agenda, it has been
determined that there are no specific products being
approved at this meeting. The Committee participants
have been screened for their financial interests. To
determine if any conflicts of interest existed, the
agency reviewed the agenda and all relevant financial
interests reported by the meeting participants.
"The Food and Drug Administration prepared
general matters waivers for participants who required
a waiver under 18 U.S. Code 208. Because general
topics impact on many entities, it is not prudent to
recite all potential conflicts of interest as they
apply to each member. FDA acknowledges that there may
be potential conflicts of interest, but because of the
general nature of the discussions before the
Committee, these potential conflicts of interest are
"We would like to note for the record that
Dr. Michael Decker is a nonvoting industry
representative for this Committee acting on behalf of
a regulated industry. Dr. Decker's appointment is not
subject to 18 U.S. Code 208. He is employed by
Aventis and, thus, has a financial interest in his
employer. In addition, in the interest of fairness,
FDA is disclosing that his employer, Aventis, is a
manufacturer of a product that could be affected by
"With regards to FDA's invited guest
speakers, the agency has determined that the services
of these speakers are essential. The following
interests are being made public to allow meeting
participants to objectively evaluate any presentation
and/or comment made by the speakers.
"Dr. Antoine Flahault is employed by the
World Health Organization Collaborating Center for
Electronic Disease Surveillance in France. He has
associations with firms that could be affected by the
"Dr. Maria Zambon is employed by the
Respiratory Virus Unit, Health Protection Agency in
England. Her agency's laboratory conducts tests on
licensed influenza vaccines. Her employers
collaborates with firms that could be affected by the
"In addition, there are vaccine
manufacturers making industry presentations. These
speakers have financial interests associated with
their employer and with other regulated firms. They
were not screened for conflict of interest.
"Members and consultants are aware of the
need to exclude themselves from the discussions
involving specific products or firms for which they
have not been screened for conflict of interest.
Their exclusion will be noted in the public record.
"With respect to all other meeting
participants, we ask in the interest of fairness that
you address any current or previous financial
involvement with any firm or product you wish to
comment upon. Waivers are available by written
request under the Freedom of Information Act."
So ends the reading of the conflicts of
interest statement. Dr. Overturf, I turn the meeting
over to you.
CALL TO ORDER
CHAIRMAN OVERTURF: Good morning. I would
like to welcome everybody to this meeting of the
VRBPAC Advisory Committee February 18th and 19th.
Dr. Midthun, would you like to make
presentations to retiring members?
PRESENTATION OF PLAQUES TO RETIRING MEMBERS
DR. MIDTHUN: Good morning. I would like
to ask Dr. Judith Goldberg to please come up to the
podium. I would like to thank her for her many years
of service on this Advisory Committee. She has always
been here with great dedication, always has prepared
extremely well, and provided excellent input to us.
We are really going to miss her and really appreciate
all that she has given to us.
I think I am supposed to stand over here
so we can get with the picture of the plaque. Thank
you. And we also have a letter here for you from Mr.
Peter Pitts, who is our Associate Commissioner for
External Relations. So thank you so much.
DR. GOLDBERG: I just want to thank all of
you because it has really been a privilege to serve on
this Committee. I have enjoyed every minute of it,
and I have learned a tremendous amount.
DR. MIDTHUN: Thank you so much.
CHAIRMAN OVERTURF: I think we will
proceed. As you know, there has been a great deal of
interest in influenza, for those of you who forgot
about last winter. So this year I think will be a
striking update of last year. So that I think we will
begin with the data that is going to be presented by
Dr. Roland Levandowski and his associates. Thank you.
DR. LEVANDOWSKI: Great. Thanks very
much, Dr. Overturf.
STRAIN SELECTION FOR INFLUENZA VIRUS VACCINE
FOR THE 2004‑2005 SEASON
DR. LEVANDOWSKI: I am going to try to
give a somewhat extended review of what has been
happening this last season. Generally I do give a
fairly brief review, but today I am going to be going
into a little bit more detail.
There is an awful lot that is going on.
What we would like to try to cover, just as a
reminder, the real business for today is what is first
on this list of topics for us to take a look at. We
are really here today to make the recommendations for
the strains that should be used in next year's
trivalent vaccine, for the 2004‑2005 trivalent vaccine
for the H1N1, H3N2, and Influenza B viruses.
We do also on this program, as you will
see from the agenda, have some other items that we
wanted to bring to the attention of the Committee. In
particular, there are some items that we have that we
would like to bring to the attention of the Committee
just mainly for information for things that are
There has been a lot of interest this year
about the effectiveness of vaccines. This has been a
discussion point at this Committee on many occasions
in the past. In fact, I can't remember one when it
hasn't been discussed in recent memory of mine.
So we have several speakers: Dr. Carolyn
Bridges from CDC, Col. James Neville from the
Department of Defense, and Dr. Antoine Flahault from
the Institut National de la Sante et de la Recherche
Medicale, who will be talking about some studies that
they have ongoing to look at vaccine effectiveness.
Some of these are still in progress, but we will at
least get to hear what is happening to try to look at
this in an ongoing manner.
I also wanted to bring to the attention of
the Committee what is happening with H5 avian
influenza in Asia. At this time last year, you might
remember we were talking about what is happening with
SARS. So we know that the Committee will be very
interested to hear this information. We also want to
bring it to the attention of the Committee because
there may need to be some activities that go on at a
later date. We just wanted to have them informed as
much as we can at this point.
Finally, on the agenda, we have something
that we would like to have some discussion with the
Committee. This relates to a point that was brought
up last year about use of tissue culture isolates from
field laboratories for preparation of influenza
You will see on the agenda that tomorrow
toward the conclusion of the meeting we have a couple
of presentations by one of my colleagues, Zhiping Ye,
Center for Biologics, and also Dr. Phil Minor from
NIBSC, to talk about our perspectives on what the
issues might be with issue of tissue culture isolates.
So this is the main business. Again, this
is the question that we are asking the Committee to
give us recommendations today. We are asking for a
vote on this. This is the abbreviated version of the
question, what strain should be recommended for the
antigenic composition of the 2004‑2005 influenza virus
vaccine for use in the United States?
Just by way of review, this was the
recommendation that was made by the Committee last
year at this time. It was for a trivalent vaccine
that would contain an A/New Caledonia/20/99 H1N1‑like
strain. Actually, it was A/New Caledonia/20/99.
It would also contain a B/Hong
Kong/330/01‑like strain. And in our case, the strains
that were used for vaccine preparation were the actual
strain, B/Hong Kong/330/2001 or B/Hong Kong/1434/2002.
The H3N2 component that was recommended based on all
of the information was for an A/Panama/2007/99‑like
Why do we change strains for influenza
vaccines? Well, we do it because the vaccine efficacy
is really related to two things. One is how much
antigen is present in the vaccine and then, very
importantly, what the match of the vaccine,
hemagglutinin and neuraminidase, are with the wild
type circulating strains. This has been very clear to
us since the earliest days of influenza vaccine use.
You might remember that influenza viruses
were first isolated in the mid '30s, human influenza
viruses, mid '30s to early '40s. And it was very
quickly that some vaccines were prepared as whole
The first vaccine was licensed in the
United States in 1945. And within two years, by 1947,
it was pretty clear that when there were antigenic
changes occurring by way of mutation in the viruses,
that there was reduced vaccine effectiveness. That
led to setting up the global surveillance system.
What we know from that period of time
onward is that there have been continuous antigenic
changes in the hemagglutinin and neuraminidase of both
influenza A and influenza B viruses.
These are the questions that are asked in
order to answer the question for recommendations. I
will just go over these a little bit with you. First
of all, we want to know, are there new either drifted
or shifted antigen influenza viruses present? Drift
is point mutation occurring in the viruses. And shift
would be exchange of an entire gene segment.
I guess I should remind you that influenza
viruses have a segmented genome. There are eight gene
segments for either influenza A or influenza B. These
can reassort in nature to put new hemagglutinin and
neuraminidases into human influenza viruses. That
usually results in a pandemic. But, anyway, the
question is, are there new influenza viruses present?
This is the purpose that surveillance
serves. It provides us with that information as to
whether there are new viruses that are occurring.
Mainly we are interested in, are they new in terms of
their antigenic properties, mainly for the
hemagglutinin but also for the neuraminidase?
It is also from surveillance that we get
the viruses that are used for vaccine preparation. So
without that underpinning, there really isn't anything
that we would be able to accomplish.
The question to be answered if there are
new viruses, ‑‑ and they almost always are new viruses
that are being identified because of the continuity of
evolution of the viruses ‑‑ are they spreading in
It is not unusual to see that there are
influenza viruses that are really wildly different,
but it turns out they are one off. So that occurs
from time to time. And it takes a while, in fact, to
have an understanding as to whether these new viruses
really have any significance or a potential impact
that we need to take into consideration for vaccine
If there are new viruses spreading, then
we also want to know whether our current vaccines are
going to have any likelihood of having effectiveness
against these new strains. And, for that purpose, we
look at responses from people who have been immunized
with the current vaccines. Often the case is that
although there are some new viruses that are
spreading, the current vaccines actually make
antibodies that cross‑react fairly well.
And while the differences you will see for
these two different activities are somewhat
complementary, there are thousands of influenza
viruses that are examined with a relatively small
number ‑‑ it is not entirely small, but it is a
relatively small number of sera that are used to
categorize them. With the human serologies, we are
looking at the reverse, where we have several hundred
different sera from people who have been infected or
immunized and looking at a relatively small, select
group of these viruses that have been identified in
And then, last but not least, if it is
true that there are new viruses, they are spreading,
the current vaccines don't look like they produce very
good antibody responses to those new strains, then we
still need to know, "Can we do something about it?
And are there any strains that are suitable for use in
And so, to answer the questions, last
year, just to review what we did, were there new
influenza A, H1N1, viruses? No. The answer was no.
The HA of all of the strains was very similar to the
For H3N2, the answer, however, was yes.
There were quite a few strains that were identified.
Although most of these strains were originally very
much like the current vaccine strain, there were some
strains that were identified early in 2003 that were
antigenically distinguishable. And it was a
relatively small proportion to begin with, but that is
not unusual either that it starts out small and
However, after collecting information and
analyzing, it wasn't really until February of last
year that it was clear that there was a cluster that
were antigenically and genetically related that seemed
to be the ones that were most likely to spread
For influenza B viruses, again, the answer
was really no. There weren't really any new viruses
that were found. The majority of the strains were
very similar to what was in the vaccine, but there
were a small number of strains that are different.
There actually have been two hemagglutinin
lineages for influenza B viruses co‑circulating for at
least the past 15 years. One or the other of these
hemagglutinin lineages has tended to be the
predominant strain. We had just left a period of time
where for about ten years, the strains in the
so‑called B/Yamagata/16/88 lineage were the ones that
were predominant, particularly outside of Asia.
For the last two years, however, the
strains that have been predominating are on the other
HA lineage. They are in the B/Victoria/2/87
hemagglutinin lineage. And that is what we have
currently in our vaccine. But there was a small
proportion of viruses last year that were identified
that were in the B/Yamagata/16/88 lineage. That was
being paid attention to, but it was not sufficient to
think that there was something really happening there.
Were these viruses spreading? For the
H3N2, as I mentioned, the answer was yes. By the time
the Committee was meeting in February or March, it was
pretty clear that there were some of these viruses
found on several continents, including Asia, Europe,
and North America. Were these viruses inhibited by
the current vaccines? And the answer to that was
There were a number of strains that were
very much like the Panama/2007/99 vaccine strain.
Those were very well‑inhibited. But for the group
that we are now calling A/Fujian‑like strains, some of
these were relatively well‑inhibited by current
vaccines and some were not. It was not a very
Then, to answer the question, were strains
suitable for manufacturing available? The answer was
actually no. And it related to the fact that all of
this information was coming out just at the time that
decisions need to be made in order to prepare a
vaccine. I will give a little bit of explanation
shortly about why that is true, why the timing was
Sort of in a nutshell here, the
manufacturing does depend on having an egg‑adapted
strain that will grow well. It could be either
wild‑type or a high growth reassortant for the
influenza A viruses. Generally it needs to be a high
The fact that these first Fujian‑like
strains were first identified in February made this
difficult. The first egg isolate of an A Fujian‑like
strain wasn't until April. The first high growth
reassortant wasn't prepared until toward the end of
June. That timing also is fairly typical for dealing
with new influenza viruses as they are appearing.
So the implications of the strain
selection from last year were that the preparation of
vaccines this current year was very much on schedule.
I will provide some information about that. And the
supply of vaccine matched the demand that was expected
by the previous year's experiences.
There were some other implications. One,
this year there was an early widespread appearance of
drift variant of A Fujian‑like viruses in the United
States. There were reports of mortality in children.
Now, that significantly increased vaccine demand.
And although there were several million
doses of vaccine, both of inactivated and live
vaccine, that were still available in mid November.
And it appeared that we were headed toward a situation
where a lot of vaccine would not be used again, which
has frequently been the case in the past, not just the
year before but for many years running. The amount of
vaccine that was available was not sufficient to avoid
some spot shortages that occurred after the
Thanksgiving holiday in the United States.
And then again, the effectiveness of the
vaccines against this drift variant has been
questioned. So there are some studies that are
ongoing. We are going to be hearing about those.
So for the United States, we have three
licensed influenza vaccine manufacturers. Two of them
produce inactivated vaccine: Aventis Pasteur and
Evans Vaccines. Evans is now part of Chiron. So I
have to be careful. It is hard for me to keep up with
the changes that occur business‑wise, but these two
companies have been licensed in the United States for
quite some time, as you can see here.
And last year, between the two companies,
there was production of about 83 million doses of
vaccine. Put that into a little bit of perspective.
Inactivated vaccines around 1990, there were
approximately 25 to 30 million doses produced per
year. So over the decade of the '90s, vaccine
production had increased substantially.
And our license manufacturer for live
attenuated vaccine is MedImmune. They were licensed,
as you might recall, in June of 2003. And they
produced about four million doses of vaccine for use.
The timelines for vaccine production are
shown here. And it is a little bit of a pyramid
scheme. What most everybody is interested in or what
gets the most visibility is vaccine use, which occurs
in the fall through the early winter. But supporting
that, underneath that, is all the work that the
manufacturers have to do to prepare the vaccine. And,
even before that, surveillance and other activities
are required. We are right here in February to March.
So we are right down here in this period of time. It
is early days for vaccine preparation, and it is
months away from vaccine use.
As I mentioned, without surveillance, we
would not have strains for use in vaccine production
in the first place. And there is a lot of work that
goes on between surveillance and trying to develop new
strains throughout the year, although there may be
periods of time when there is more activity than
others. But there is some activity going on pretty
much all the time.
Recommendations are generally made by the
WHO for the Northern Hemisphere and for the winter
months here and for the Southern Hemisphere and for
the winter months there. But these recommendations
are important so that the manufacturers know what they
Too, as there are reference strains that
are getting worked, the manufacturers throughout the
year are working on their seed viruses, which were
proprietary to them. They worked with the virus to
make sure that it is going to be appropriate for the
manufacturing conditions. And although there may be
some early seed viruses that are used in production,
there is some continuous work that goes on to try to
make that better so that manufacturing can be smoothed
out. I will show a little bit of information about
Production of the monovalent components of
the vaccine takes many, many months. And, really, it
starts maybe earlier than January. Manufacturers may
be working at risk before recommendations are made to
produce monovalent components. They don't do this
without some education. They are paying close
attention to the surveillance that is being reported
throughout the year by WHO and our colleagues at CDC.
Once all three of the components are
present, trivalent vaccine can be produced, but you
will see that there is still overlap. There is still
for quite a long time, actually, many years, work
being done with the monovalent vaccines. And so there
is some vaccine that starts to come out, but it
doesn't all come out at one time. And that vaccine
uses the desired goal at the top again.
So to try to also give some understanding
about how long it takes, when there is a new strain
that is recommended, when there is a new reference
virus that is identified, for the point of time that
that new reference is identified to the time that that
is available for sending out to manufacturers to
develop their seed viruses, for the period of several
weeks, part of this is trying to understand, is this
the best strain that is available or are these the
best strains that are available for producing the
vaccine? It takes some analysis. It takes some
collaboration between the WHO centers to come to that
Part of the time for influenza A viruses
and probably for influenza B viruses in the future is
preparing the high growth reassortants that make it
more expeditious for producing the vaccine. At the
same time that that is happening, reference reagents
for standardization of the vaccine need to be
prepared. And this is true not only for inactivated
vaccines, but it may be necessary for the live
Potency testing for these is dependent
upon having antisera that can be used for the tests
that are done to try to standardize in terms of
Once the manufacturer has a seed virus
prepared, then they can start to manufacture. As I
mentioned, there may be manufacturing at risk when the
strains are not changed. Those strains can be
prepared in advance of this meeting if the
manufacturers so desire, but they really can't do
anything until they have in their hand something that
is appropriate for making a seed virus. And they
can't manufacture all the strains without having those
So this just shows manufacturing three
different strains. And I am showing down here vaccine
release. There are activities that go on between the
manufacturer and regulatory authorities to try to make
sure that these seed viruses: first, are appropriate
for use, that they maintain their antigenic
characteristics; and, second, that other kinds of
qualities are maintained.
Now, this bar is about three weeks long
for each of the monovalents. That is an
approximation. Most of that time I think you will
hear from our colleagues from industry relates to
quality control, not necessarily interaction with
regulatory authorities but just needing to meet their
own good manufacturing practices and be sure that the
vaccines are going to meet all of the specifications
that are set for them.
Once the three strains are produced, then
it is possible to go ahead and formulate vaccine and
to fill it and to send it out. You see, each of these
points have bars that are about three weeks long as
well. And there are some release activities that go
on for the trivalent vaccines.
This duration of time here again is not so
much the actual physical manipulation of the vaccine.
It has to do with the quality control measures that
need to be met and some very important ones, like
sterility for inactivated vaccines, for example.
So once that has all happened, then, of
course, the vaccine can be distributed. But, you see,
this time line up here, for a new strain, I am
indicating about 20 weeks. I think that is a
reasonable estimate. If you put that in months, that
is about five months from the time of the first
appearance of the new strain until there is really the
possibility of having a product in hand that can be
That is just the first. Once the first
comes out, then, of course, it just keeps rolling.
Again, to try to put this into some more perspective,
it seems to be cut off a little bit on the boundaries.
That is okay.
I went back and reviewed monovalent
vaccines that were produced for inactivated vaccine
over the last several years. What I am trying to show
you here is that when strains are changed and those
are shown across the bottom, the relative proportion
‑‑ this should add up to 100 percent for all 3 strains
‑‑ of the strains that are produced really tends to
favor the new strains that are added. So you can see
between 1998 and 1999, there was a new influenza B
virus. And although 38 percent of all of the
monovalent concentrates produced in 1998 were
influenza B, over half of them, about 55 percent, were
You see that with other changes. In 2000,
we added a new H1N1 and H3N2. And you can see in both
instances, the amount of effort, the relative amount
of effort, in terms of the number of monovalent
concentrates that had to be produced was mostly
devoted toward the new strains. You can see that all
the way across here, that when new strains were added,
that there needs to be an adjustment that it is the
early work that has to be done by the manufacturers to
figure out how best to get things growing.
Once they get it organized, you can see
that it is possible that things may even out a little
bit more between the three strains, but the strain
that has changed, the strains that are changed are the
ones that are the most difficult in terms of overall
production, at least for the first year.
In terms of timing of these things, this
is what we see in terms of submission to the Center
for Biologics for Release. I am showing the number of
lots here against the month for both the monovalent
vaccine and the trivalent.
What I really want to point out to you on
this slide, the numbers aren't so important. It is
the overall pattern. You see that there is kind of
this buildup of more and more monovalent concentrates
coming in up until about August‑September. And then
it starts to wane.
And this relates to the point at which
manufacturers when they are trying to meet the need
for vaccine in October and November have already
planned out how they are going to be putting together
how they are going to be manufacturing the vaccine
components, when they need to have them on tap and so
on. And so they come to a decision point about August
or September where they have to decide whether it is
worthwhile for them to continue manufacturing or not.
There is a lot of effort and money that
goes into that continuation. And it is possible for
them to do so. They could keep going if they knew
that there would be demand, for example, within our
current system. It is possible to make more vaccine.
It doesn't have to stop right here, but it does
because there is a target that has been developed from
sales and demand. And it is really kind of a
You see that there still is overlap
between the preparation of the trivalent vaccine and
preparation of the monovalent right on out to the end
of the overall campaign for the year.
For this year, because the strains were
the same as the previous year and the demand was
fairly well‑understood at the beginning of the year,
it was possible to get everything ready.
This curve shows cumulative percent of all
the lots that are submitted to us for release from
June to December. Two thousand was the year that we
had the shortages or delays that were concerning. And
this was an atypical year in that the point at which
50 percent of the vaccine that was available was
shifted out substantially from where it normally is.
These curves over here are more typical of
what we would be seeing. And generally 100 percent of
the vaccine in the past has been out by about October.
And that is where we are here. This is the red color
here, the diamonds. The red diamonds are this year,
So you can see that vaccine was being
produced fairly consistently throughout this period of
time and very expeditiously met the overall goal for
this production campaign without any delays.
So why are influenza vaccines important?
Well, they are important because influenza has a lot
of economic consequences, the lost work, school time,
and so on. We know that morbidity is high,
particularly in the very young.
Pneumonia and Influenza, that is the only
category that is in the top ten causes of death in the
United States, the only infectious diseases category
that is in the top ten causes of death in the United
States. And this is for ages overall. It is not for
a specific age group but for the ages overall.
We know from other statistics that we can
expect somewhere between 20,000 and 40,000 deaths in
a typical year related to influenza. That is
generally in the elderly. And we know that pandemics
cause even more.
I just wanted to read a couple of things
that were from some publications that sort of put this
into perspective. So I am quoting here. It says,
"Early apprehension was increased by the fact that
when the first indications of the outbreak were
observed in the country, the influenza had already
attained epidemic proportions in England.
The sharp rise in influenza deaths,
however, was found not to be due to virulence of the
causative organism but to a high case incidence. The
term "lightening influenza" was used in newspaper
Also, the epidemic caused by influenza A
viruses was unusually severe for the inner pandemic
period. The attack rate in children was much higher
than for adults. At least 30 percent of children
under 5 years of age were ill. And most were taken to
medical care facilities.
Over 320 children per day crowded into the
pediatric outpatient clinic at Ben Taub Hospital. So
you might think that was this year, but the first one
is from 1943 and the second one is from 1975.
I just wanted to try to remind everybody
that what we are dealing with here is something that
maybe has been a little bit forgotten but that we
should remember that influenza is a very serious
And to try to highlight that more, I have
got some other slides here that I have taken from some
of the older literature. This is data from
door‑to‑door surveillance activities in Baltimore that
were undertaken during and after the pandemic of 1918
to try to get some information about what was
happening. Unfortunately, my legend is cut off over
here. The red one is 1918. The purplish one down
here is 1919. The green one is 1928 to '29. And the
black one down here is '40 to '41. This is 1943 to
That quote that I was reading partly
related to this. There was what was seen as a
relatively large incidence of case attack rate in
children predominantly during that period of time.
There was a large fear that this was the return of the
1918 pandemic strain. So you can see, by comparison,
it wasn't quite as high an attack rate, but it was
much higher than what had been seen in some of these
intervening years. So there was a lot of concern
I think what it indicates to us is that
attack rates can be higher or lower. It is sort of
interesting that in the case of both of these years,
there is sort of a relative disproportionality in
terms of younger children and then sort of young
adults. It caught my attention because I think that
may relate a little bit to what we are seeing this
year as well. I think we are seeing more activity,
and I think we will hear more about that.
So in terms of pneumonia ‑‑ and these were
cases per 10,000 shown on the other slide, and it is
the same scale here, but the numbers are drastically
different. So this is the pneumonia cases in those
You can see from 1918 to 1919, this very
much parallels the mortality curve, where there was a
kind of instead of a U‑shape, where it is very high at
both ends in the very young and the very old, there
was this extra peak occurring in young otherwise
healthy adults. And there is a small echo of that in
the year following 1918, during 1919 and 1920.
What I am showing here is that this blue
down here again is 1943, where there is a huge number
of cases occurring, but the amount of pneumonia that
was being identified in Baltimore was relatively low,
particularly in young healthy adults. There was a
little bit of an increase more in elderly than in
children, but the young children were affected as
These are some data from 1975‑1976 in
Houston during an A/Victoria/3/75 epidemic of
influenza. I am showing this. Again, I am trying to
use that same scale. This is hospitalizations per
10,000. These are the ages of the individuals.
I guess what I should have said in my
previous two slides is that a lot of this, of course,
1918, some of this could have been something other
than flu. We didn't really have virologic
capabilities until later, but it is based on the
sharpness of the peak of the epidemic. And it is
probably true a little bit here for these data as
These are not all virus isolates. These
are clinical studies that were done to try to define
what was happening in the epidemic. But this is
during an epidemic in Houston. Again, you see this
U‑shaped curve, where the hospitalizations are most
marked for the very young and the very old. In fact,
the number of hospitalizations in this particular
instance appear to be probably more than in the
elderly. But you see some level of hospitalization
during this relatively severe influenza season in all
Different from that, however, is what has
been seen for mortality in some of these epidemics.
This is a different age 3 and 2 epidemic in Houston,
encompassing Harris County, Texas, the statistics,
health statistics, from there. Again, these are
deaths per 10,000 at different ages.
Here you can see that, really, it is the
elderly who are most markedly affected. They are the
ones who die when they become ill and develop
pneumonia. But there are deaths that are reported in
all ages. It is kind of a small number here in the
young adults. And there are quite a few more seen in
young children but not nearly what we see in the
So from this information, the effects
here, we know that influenza attack rates are often
highest in children who are less than ten years ago.
There is serious illness in all of the age brackets,
with the young and the old most affected. And the
mortality is generally highest in the elderly,
although it is also seen in young children. And in
some instances, it seems to parallel the incidence of
pneumonia during the period of time that the influenza
epidemic is occurring.
So a brief history of influenza vaccine
efficacy. In 1941, there was a request to license the
first inactivated vaccine in the United States, but
the regulatory authorities at the time, who were part
of NIH under the Public Health Service Act, thought
that it was best to get efficacy. That wasn't really
required, I believe. It was mainly safety data that
were needed. But there was a request to show that the
vaccine would actually be efficacious.
They were set to do the study. They had
all of the materials in place and the desire to do the
study in 1942, but this often happens to those people
who are trying to study influenza. There was no
epidemic that year. So it was not possible to do the
studies. Instead, there were some challenge studies
that were done at the time, which demonstrated that
these vaccines were effective against influenza A and
Those studies were published as well as
the information from the studies that were done later
from large‑scale field studies. Those were done in
1943 through 1945. And they were done with bivalent
vaccine using influenza A and B viruses.
The first vaccine was licensed in the
United States in 1945. And, as I mentioned before, it
was very shortly after that that it was recognized
that antigenic drift could reduce the effectiveness of
vaccines and the Global Surveillance System was
inaugurated to try to identify changes that were
occurring and to be able to make alterations in
influenza vaccine as necessary.
So the first studies that were done by
Tommy Francis and Jonas Salk and others with the armed
forces and a special commission that was set up to
investigate influenza, the studies that I am going to
be talking about were done as randomized
placebo‑controlled field efficacy studies between 1943
and 1945. The vaccines that were used at that time
were whole virus, formal and inactivated. They were
highly formalin‑inactivated reactogenic.
A large percentage of the people who got
the vaccines felt ill for a couple of days. Some of
them went to infirmary. The antigens that were
contained in the vaccine, it was actually trivalent
vaccine. It had two H1N1 components: A/Puerto
Rico/8/34 and A/Weiss/43. There was an influenza B
component. It was B/Lee/40.
The studies were done at the Army
specialized training program centers around the
country at the time. These were located in a number
of universities and medical schools. In these
particular studies, there were more than 10,000
What they were looking at, mainly they
were looking at the clinical endpoint. It was
influenza illness. It was most important. They did
have the capability. And they were using it during
the studies to identify infection by culture. They
also could look for serologies. But, really, the
endpoint here was the illness.
Illness was supposed to be characterized
by symptoms that included abrupt onset fever,
myalgias, cough, sore throat, and nasal symptoms. And
the cases were further categorized by illness
severity. Those who had a temperature over 100 by
whatever the going criteria were, they were sent to
the infirmary for hospitalization to get them away
from the rest of the men in the barracks where they
were staying. They also, of course, did X‑rays when
they wanted to look for pneumonia.
There were some differences between the
multiple centers in terms of the way the actual study
was run. So that in some instances, hospitalization
could have been for lesser fever, but generally this
is what was followed.
I am showing the two different studies
that were done here looking at influenza A and
influenza B. Again, this was a clinical measurement.
It was clinical influenza that resulted in febrile
illness that needed hospitalization. There were
approximately 12,000 individuals who were randomized
to get either the vaccine, the trivalent vaccine, or
somewhat identical placebo.
The number of cases that occurred was
substantially higher in those who got the placebo than
those who got the vaccine. If you calculate
protective effectiveness from that, it works out to be
about 69 percent protective effectiveness against the
clinical febrile illness requiring hospitalization.
A similar study was done for influenza B.
The numbers here are approximations because the people
involved in these programs at the time were sort of
going in and out. There was a lot of personnel
transfer in motion. So they did the best they could
to try to determine what the denominators were here,
but it really is kind of an estimate.
The number of cases, it is firmer. Again,
you can see that for the vaccine, there were
substantially fewer cases than in the placebo group.
Translated, it would be a protective effectiveness,
around 88 percent.
There were some subanalyses that were done
in this set of studies that have been published also.
In one subset at the University of Michigan, they
tried to look at the effect on illness, different
levels of severity. They looked at people who had any
kind of respiratory illness. This included the common
cold or what they called the common cold. It was
illness that was obvious, but it wasn't severe enough
to result in hospitalization. And it didn't have
other symptoms that they thought would be more typical
of the syndrome that we call influenza with all of
those symptoms that I listed early on.
We know very clearly that influenza
infection can cause what seems to be a common cold in
some people, and we know just as well that other viral
infections can cause what seems to be influenza by its
clinical manifestations with fever, myalgias, and so
on. So it is a very nonspecific indicator.
For those that they thought were more
likely to have influenza based on the symptomatology,
again, that is based on the clinical symptoms but
being more typical for influenza syndrome. These
inpatients had fever. Then, of course, they were also
looking at pneumonia.
What you see across the bottom here if you
try to figure out a protective effectiveness, you can
see that there is increasing effectiveness of the
vaccine against the more severe forms of illness.
It is very difficult to show vaccine
effectiveness when it is diluted by many different
types of respiratory viruses, none of which were known
at the time. They were identified specifically at the
time these studies were being done, but they were, the
giants whose shoulders we stand on were, very much
aware of the fact that there were other etiologic
agents out there that needed to be categorized. You
can see that it is very difficult to show that.
If you go to the most severe forms of
illness, it is a lot easier to try to show that there
is some effect. They commented that throughout the
study, there were no cases of pneumonia in anybody who
got the vaccine. There were ‑‑ this is a relatively
small number ‑‑ only four cases of pneumonia in the
recruits who got the placebo. That does fit, however.
These studies were done in 1943 and 1944, when I
showed that there was very low incidence of pneumonia,
even though there was a high attack rate for
So there were some other observations they
made from this. One of them was that in these kinds
of studies, the placebo group was actually diluted by
having an immunized cohort that may have been able to
reduce transmission in the placebo group. This was in
one of the first thoughts about herd immunity.
The differences in the attack rates
between vaccine and placebo were really greatest at
the peak of the epidemic. And as the epidemic
receded, it was harder and harder to be able to show
One corollary to this part was that there
was at least one center where there was an attempt to
immunize in the face of what seemed to be the
developing epidemic. And they noted that it was very
clear‑cut that during the first week, they couldn't
really show any difference in the attack rate in the
placebo or the vaccinated individuals, but after one
week, it was very clear who had been immunized. There
seemed to be a big difference, even after that
one‑week period of time.
I mention this just because we often talk
about needing two weeks after immunization, somebody
who is immunologically primed. Of course, these
individuals all were immunologically prime by previous
exposures. But there may be protective effects that
are kicking in, even in an earlier period of time.
So, to add a little bit of information
about some of the other studies that had been done, I
wanted to concentrate on a few studies. These are
selected by me to make some points about the effect of
vaccine when there is antigenic drift that is
This first study that I want to talk a
little bit about was done in Texas in 1976. It was
the Houston family study for this particular
publication. There were 37 families who had 155
members of different ages, ranging from infants up to
about mid '40s.
The A/Port Chalmers/1/73‑like viruses had
caused an epidemic in 1975. And so these individuals
who were not immunized specifically had antibodies
that were directed against or should have had some
possibility of having antibodies directed against Port
Chalmers, but then the following year, A
Victoria/3/75‑like viruses caused an epidemic. These
viruses, the A Victoria/3/75 viruses, are really drift
variants of the previous strains.
At the time, it was noted that this was a
very dramatic difference in terms of antigenic
characteristics between these two viruses. I don't
know whether it is fair to say so, but it probably was
at least as different as what we are seeing this past
year with the Panama‑like strains and the Fujian‑like
strains and possibly more because, actually, it was
remarkable and there was a lot of comment about how
different those strains were.
They were able to use virus isolation and
serologies to try to document infection. As I
mentioned, there was no vaccine used. Now, what I am
showing here are the preexposure hemagglutination
inhibition titers from the people who were in the
study. They were able to get blood from 154. They
tested them for antibodies to both Port Chalmers 73
and Victoria 75 strains.
What you can see here, I think, is that
there was some proportion who had, really, very low
antibodies in both of these groups. These are the
same people, of course. So you see that some higher
frequency of those who they tested for antibodies to
Victoria were more likely. They were more likely not
to have antibodies is what I am trying to say.
If you look at the distribution from low
antibody to high antibody, you can see that in
comparison to Port Chalmers, there is a shift toward
lower antibody titers for Victoria strains. And that
is what we are usually dealing with when we are
looking at our serology. So this is very similar.
They were able to do some other things to
look at infection and illness. I think that what you
can see is that there is a relation between antibody
presence and protection. As you get higher antibody
titers, there are fewer and fewer people who are
infected. The same thing is true if you look at
clinical illness. Those who have higher antibody
titers are less likely to be clinically ill.
What this also says is that the number of
people who are infected who are relatively
asymptomatic is fairly high compared to the numbers
that we would recognize as having had influenza. So
it is something else to keep in mind in terms of
trying to make sense out of what is there.
So another study done by the same group in
1986. Again, it was the family study, Houston family
study. They had 98 families enrolled with 192
children who are between 3 and 18 years old.
These children were randomized. I am not
sure that is quite the right word. They were groups
that got either placebo or inactivated trivalent
vaccine or a live attenuated bivalent vaccine.
Each of those vaccines contained an
A/Chile/1/83‑like H1N1 virus. These children all got
a single dose of vaccine. That particular year, the
H1N1 virus was a new one that had appeared only in
March, was first identified in March.
And you might remember that that was the
year that there was a supplemental vaccine that was
produced for A/Taiwan/1/86. It was not used in this
study, but it was recognized that the Taiwan/1/86
virus was substantially different from Chile so much
so that it was thought that for younger people, that
an A/Taiwan vaccine would be a good idea.
Again, they had virus isolation and
serology to try to document infection. What they
showed was both of these vaccines had protective
effect against infection with A/Taiwan/1/86. And this
is infection and not illness that we are looking at
now. So it is the measurement that would pick up more
Anyway, you see this is fairly similar.
It was about 52 percent protective effect on this
drift variant from the live attenuated vaccine in use
at that time and about 61 percent for the inactivated
The authors commented that they thought,
actually, in the younger age group, the live
attenuated vaccine performed better than the
inactivated vaccine and vice versa for the inactivated
vaccine, which you can see here. Nevertheless, both
of these vaccines were substantially better than no
vaccine in terms of what happened to the placebo
So, finally, one more drift variant story,
a nursing home in Colorado in 1987 during an outbreak.
There was an outbreak that was caused by an H3N2 drift
The vaccine strain at the time was
A/Leningrad/6/86. The viruses that were being
isolated have names. They are Colorado, of course.
They were all similar to this reference strain, the
Sichuan/2/87 strain. And that strain was different
enough that the following year, it was included in the
vaccine for use. So that there was some difference.
I am not sure that the difference between
Sichuan/2/87 and Leningrad/6/86 is the same as the
Victoria/Port Chalmers difference, but there was still
enough that it was thought a good idea to change the
vaccine after those had appeared.
Not everybody in this nursing home was
immunized, but they immunized a very high percentage
of them after the outbreak started. The outbreak
itself had a peak that occurred about two weeks after
the immunization campaign.
This analysis was done retrospectively,
but they were able to get pretty good documentation
from the nursing home records about who had fever; as
measured by thermometer, who had illnesses, what kind
of illnesses they were. Of course, pneumonia and
death were pretty obvious because the residents needed
for the treatments. In a subset, they were able to
confirm that infection had occurred because of H3N2.
But generally this again was a clinical observation.
This just shows the epidemic, how it
occurred. When it was first recognized, it was a
fairly sharp epidemic. There were about five of the
residents of the nursing home who were infected. The
following week, vaccine was given to all who wanted it
or could receive it. There were a number of
individuals who refused the vaccine, and there were a
number of individuals who had other ongoing illnesses
that were thought to be contraindications to getting
So the numbers peaked around week four
here on the epidemic. And then it kind of quickly
tapered off afterward. What you can see is that
although there are quite a few cases in both the
immunized and the unimmunized populations, looking at
vaccine effectiveness, as calculated by the authors,
they mainly were looking at febrile upper respiratory
They excluded a number of individuals who
had been immunized in that two‑week interval. From
the time they immunized until two weeks, they excluded
those from their true analysis. And that was true for
both vaccinees and non‑vaccinees. There were some
other exclusions as well.
Looking at the incidence, these are
numbers and not percents here. Looking at the number
of febrile upper respiratory illnesses that occurred,
the proportion was significantly less in those who
received vaccine. And protective effectiveness
against febrile upper respiratory illness in that
group was calculated to be about 65 percent.
There were no pneumonias in those who got
vaccine and there were no deaths in those who got
vaccine; whereas, there were pneumonias and deaths in
those who did not. This isn't really randomized
prospectively. So there may be some other reasons for
that. But if you look at all of the residents all
together, you can see that all together, the residents
who got vaccine, there were no pneumonias and no
deaths; whereas, there was a substantial number of
both incidence of pneumonia and death in the residents
who were not immunized.
Some facts that I guess we could take from
those studies are that the vaccine protective effect
is a lot more obvious for severe forms of illness and
for complications that are related to influenza and
The vaccine shifts the spectrum of disease
toward the less severe consequences and milder
illness. Whatever you are looking at, they have to
keep that in mind. And higher antibody titers are
more likely to result in protection from clinical
Also, for infection, this is not an
absolute. There is in my own mind not an absolute
number to use for this, but it is pretty clear that
the more antibody you have, the better.
The vaccine administered in an ongoing
epidemic still may reduce illness, pneumonia, and
death, even when there is antigenic drift that has
Turning away from that at the moment,
these are the recommendations that were made by the
World Health Organization for influenza vaccine
composition for the Northern Hemisphere for 2004‑2005.
The recommendations from there that are
based on the information that was available to WHO
last week were to continue to use an A/New
Caledonia/20/99 H1N1‑like virus, to use an
A/Fujian/411/2002 H3N2‑like virus, and to use a
B/Shanghai/361/2002‑like virus for the B strain.
Again, the question for the Committee,
"What strain should be recommended for the antigenic
composition of the 2004‑2005 influenza virus vaccine?
Should it be based on the epidemiology and antigenic
characteristics of the viruses, the serologic
responses, and availability of candidate strains?"
All the information that we are going to
be presenting apart from the vaccine effectiveness
studies this morning will relate directly to answering
this set of questions by the Committee. And I think
I can stop there and ask if there are any questions.
CHAIRMAN OVERTURF: Were there any
questions for Dr. Levandowski? Yes? Please identify
DR. MARKOVITZ: David Markovitz. I am
speaking from the somewhat claustrophobic right
corridor of the table, where it is difficult to see or
breathe. So if you will excuse me, I am not
My question is this. You showed a graph
that implied that the vaccine manufacturing process
was quite effective this year but, yet, also alluded
to some early gaps, early in the season, which we all
noticed just in our hospitals or reading the
newspaper. What is your overall assessment of how the
manufacturing process kicked in this year?
DR. LEVANDOWSKI: Well, I am not sure what
gaps you are referring to, but from the point of view
of production and vaccine release this year, it went
about as smoothly as it ever goes, which means that
manufacturers were busy producing monovalent vaccines
and busy producing trivalent vaccines and having those
released and being able to get them into distribution.
I don't know that I have any other
information. From our perspective, things went
DR. MARKOVITZ: I guess at our hospital,
in the fall, for example, people were not able to get
vaccine. I don't know what the cause of that was.
Was that just too early in the year? Did they strike
too soon or what? Is that just an anecdotal
observation from what I have seen? I believe I read
about that nationally, too, I thought.
DR. LEVANDOWSKI: Again, from our
perspective, I don't understand it. I don't have any
information that relates directly to distribution. We
don't get involved with distribution per se at FDA.
It is more understanding that the vaccines meet their
specifications and that they are okay to go out.
What we saw was a steady stream of vaccine
preparation release. And maybe this is a question for
the manufacturers and not for me because I don't know
what issues there might have been for them in terms of
DR. MARKOVITZ: I wasn't really thinking
of distribution. I was thinking more of manufacture.
DR. LEVANDOWSKI: Well, that is an
important part of manufacturing: getting the vaccine
out to where it is supposed to be used. But, again,
that is not a part that FDA really interacts with
directly as the vaccine is being prepared. It is more
of the release specifications being met for the
vaccine and that all of the good manufacturing
practices have been met to make sure that the vaccine
is ready to go.
Again, I would have to say that this year
‑‑ and the graph I showed I think is an indication of
that ‑‑ all the vaccine that came through FDA came
through at a very early time point.
CHAIRMAN OVERTURF: Dr. Myers?
DR. MYERS: Roland, the data sometimes is
presented for children less than five and sometimes
broken down by a year or less than five. The data you
showed is striking for the morbidity being for
children less than a year of age. Is there any
protective effectiveness data for that population?
DR. LEVANDOWSKI: There is relatively
little direct information. Again, I would have to ask
some of our other colleagues out there what they know.
I don't know that there has been a specific study to
look at vaccine effectiveness in children who are less
than one year of age.
I think we do have some understanding that
immunogenicity may be decreased in the very young
children as well. I think that is partly reflected in
the setting, the cutoff at six months for use of
vaccines, if there is an understanding that maybe the
vaccines won't be so immunogenic in those children and
maybe the reactogenicity is a bigger concern than any
clinical benefit that they might get.
There is, however, relatively little for
inactivated vaccines. There is more recently for live
attenuated vaccines. In children 15 to 72 months of
age, the studies that were done by MedImmune show a
very high level of vaccine effectiveness, efficacy
actually, in those children.
CHAIRMAN OVERTURF: Dr. Gellin?
DR. GELLIN: Well, two questions. The
first is you made a comment in 1986 that there was a
supplemental vaccine produced. Could you give us a
little more insight into what that was?
DR. LEVANDOWSKI: Right. The vaccine was
made for the A/Taiwan/1/86 strain. That virus was
first identified, I believe, in March of that year.
It was at a very late point in time. There was a
recommendation that a supplemental vaccine be
prepared. And the manufacturers did that, but the
timing for it was not available until late November
And because of the way it came out, there
was a lot of confusion, part of the confusion because
I was in clinical practice at the time trying to
figure out what to do with the vaccine late November
and early December. There was a lot of confusion on
the part of practitioners about what to do with it.
Not much of it got used. Most of it was subsequently
discarded. It had to be thrown away, basically. That
strain was what was used in the vaccine, then, the
following year for the trivalent vaccine.
DR. GELLIN: So it was a monovalent
DR. LEVANDOWSKI: Yes. Sorry. It was a
monovalent supplemental vaccine, right, that one year.
DR. GELLIN: The second question was, in
your chart about the efforts, intensive efforts, that
go into when they are changed each year and then the
subsequent graph about the delay in production.
In 1992 to 2000, there were two changes.
And that was the year that there was a delay in the
release of the vaccine that was used. I was wondering
if they were related.
DR. LEVANDOWSKI: It is partly related.
Some manufacturers had some difficulty with
replication of the H3 strain early on, but that was
worked out, as it usually is. Manufacturers are
actually quite resourceful at making things work.
If you will look at that chart that you
are talking about ‑‑ I am not sure I can get to it
because of the touch pad here on the computer. It is
not that friendly. If you look at that graph that you
are referring to, you can see that most of the effort
actually goes into producing the H1 strain. And
although the H3 is the one that got all the notoriety
for being difficult to work with, at least initially,
the H1 strain took up more manufacturing time in terms
of number of monovalents.
The way I presented that information, you
can't directly equate that with the overall amount of
vaccine that is being produced because their
variability, lot sizes are variable from manufacturer
to manufacturer and even within a manufacturer. So it
is not like you have one box that has 100 units in it.
You have a box that might have ten units. You have a
box that might have 150 units. It is not that direct.
But I just tried to give some impression
as to the overall effort. It is actually the H1
strains that have been more difficult the last few
years in terms of overall manufacturing effort.
CHAIRMAN OVERTURF: Dr. Monto?
DR. MONTO: I would comment about the
nursing home and how to interpret drift in terms of
nursing home outbreaks. We have had a surveillance
going on in a number of Michigan nursing homes for a
number of years now. Two years ago, with a rather
wimpy A/H3N2 outbreak with a non‑drifted variant, we
had confirmed transmission in 26 percent of our homes.
This year it is going to be higher. And
we had an outbreak in December. It is going to be in
the 30 percent range, we think, once we finish the
What I am saying is that you really have
to look at what happens in nursing homes, even in a
non‑drifted year, in terms of putting things into
context because the vaccine really does not protect
all that well against just influenza‑like illness,
even laboratory‑confirmed, even in a non‑drifted year,
in this population. Our nursing homes were typically
80 percent and many of them 90 percent vaccinated.
CHAIRMAN OVERTURF: Dr. Farley?
MEMBER FARLEY: To me, one of the most
striking features of this year's influenza profile was
the early onset of disease. Given the manufacturing
timetable that you presented, if there were any
indication to attempt to begin immunizations earlier
in the fall or late summer even, is it even possible
within the constraints of the timetable?
DR. LEVANDOWSKI: I think the answer, is
it possible, yes, I think it is. I mean, I think that
this year the vaccine was prepared at a very early
point. If you have that graph? Again, I am not sure
I can get to it easily here because of the computer
There was a substantial amount of vaccine
in trivalent form. That graph that I was showing was
for trivalent vaccine. There was a substantial amount
that had been released for distribution, even in the
summer months. So it is possible that it could have
This year, although there was an early
epidemic of influenza, although it started early,
there was a substantial amount of vaccine that could
have been available around the country at that point
based on the manufacturing timelines for this past
If we had made a strain change this last
year, I don't know that that would have been true. I
am not sure that there would have been further delays.
Given the timelines for preparation of seed viruses
and so on this past year, I would assume that we would
not have been seeing vaccine early in the summer but
probably the first vaccines might have been available
September, rather than in July.
So, again, from my perspective, I think
manufacturing went very efficiently and was early and
on time with a total amount that was intended for
production for this year based on what the demand
parameters were that the manufacturers understood for
all of the vaccines.
CHAIRMAN OVERTURF: Dr. LaRussa?
MEMBER LaRUSSA: Just to carry that a
little further, can you just sort of estimate?
According to one of your slides, the high growth
reassortants were available in June for the Fujian
strain. If you had decided to go ahead and make a
monovalent vaccine, what would be the earliest that
would have been available?
DR. LEVANDOWSKI: Well, that is what I was
trying to get to with that other slide that I had
about manufacturing timelines, the 20 weeks. If you
take about six weeks for development of the reference
virus at that point, what would that be? That is
about three and a half months.
So from June until sometime late September
probably by the time we would be seeing any vaccine
produced that was trivalent; whereas, with the
manufacturing system as it was, the first trivalent
vaccines were actually coming in end of June,
beginning of July.
CHAIRMAN OVERTURF: We have time for about
one more question. Any other questions?
CHAIRMAN OVERTURF: I think we will go
ahead and proceed, then, with the discussion regarding
vaccine effectiveness. I guess Dr. Carolyn Bridges is
going to make that presentation.
DR. BRIDGES: Good morning. Today I am
going to be discussing some studies that have been
done on vaccine effectiveness of the inactivated
influenza vaccine this year that CDC collaborated on.
I will be discussing a little bit of
background about the flu season, most of which Roland
has already covered. Then I will be describing two of
those studies, where we have preliminary results; and
then listing some other studies that are currently in
progress; and then end with some final remarks.
As Roland stated, influenza activity
started earlier than usual this year. And children
appeared to be disproportionately affected compared
with recent years. There were widely publicized
reports of pediatric deaths that received quite a bit
There is also unprecedented demand for
vaccine more than in some recent years. And there was
discussion at the HHS level about additional vaccine
purchase. Although, as Roland says, the amount of
vaccine that was purchased or was manufactured was
equal to the demand from the previous year.
In addition to these pediatric deaths,
there is also a drifted variant of influenza H3N2,
which predominated, which was different from the
vaccine strain. And influenza vaccine effectiveness
In order to assess the effectiveness of
this year's inactivated vaccine, several studies were
initiated simultaneously using various age groups and
looking at different outcomes.
Preliminary results are available from two
of those studies, which were conducted in the State of
Colorado. One of those studies was a retrospective
cohort study among health care workers. The principal
investigator for that study is Dr. Nidhi Jain.
The other one is a case cohort and
subsequent case control study among persons aged 50 to
64 years who have laboratory‑confirmed influenza. And
the co‑PIs for that study are Drs. Marika Iwane and
This is some surveillance data from the
Children's Hospital in Denver, Colorado. And, as you
will look at the left scale, it goes from zero to 400.
This is a scale for influenza A viruses. On the
right‑hand of the graph, it is a scale for RSV and
influenza B, which goes from zero to 40. So there is
a tenfold difference in the scales.
As you can see, influenza A activity began
early in November. It peaked towards the end of
November and then started on this decline. There were
very few influenza B viruses and very few cases of RSV
that were identified at the Children's Hospital among
hospitalized as well as outpatients.
This is the remainder of their
surveillance data. Again notice the scale, zero to
20. So there also was very little in the way of
paraflu, adeno, rhinovirus, or pertussis that was
Incidentally, of the respiratory specimens
that were tested in the hospital laboratory, the
percent of specimens that tested positive for
influenza at the peak was around 60 to 70 percent. So
this is a very high percent positive rate from
We decided to conduct a study among health
care workers at the Children's Hospital because the
staff provided a large cohort for rapid analysis. And
we knew that a large number was needed if we were
going to look at nonspecific outcome.
Also, this cohort has substantial
opportunities for exposure to influenza as they had
many hospitalized patients who were influenza‑positive
and they had conducted the bulk of their inactivated
influenza vaccine campaign in the month of October.
Influenza‑like illness was the outcome.
This has been used, as Roland has described, in many
prior studies, although we clearly understand that
this underestimates the vaccine effectiveness that
could be seen looking at more specific outcomes, such
as laboratory‑confirmed influenza.
We thought that this may provide us with
a reasonable estimate or reasonable chance of finding
vaccine effectiveness because influenza was so
predominant as the cause of flu‑like illness in the
population based on the Children's Hospital
surveillance data. Preliminary results of this study
were published in the January 16th MMWR.
The objective of this study was to
estimate the effectiveness of the 2003‑04 inactivated
influenza vaccine in preventing influenza‑like
illness, or ILI, among adults working at the
Children's Hospital in Denver, Colorado.
This is a retrospective cohort study. A
questionnaire was distributed via e‑mail and also
paper surveys to approximately 3,100 employees. This
is an anonymous survey, and very limited demographic
information could be collected. Information was
collected on age group, sex, whether they had patient
contact, whether they had one or more high‑risk
conditions, whether they are vaccinated, and the
timing of their vaccination, illness onset and
symptoms, and whether they had physician visits or
were influenza‑tested. We also asked about missed
workdays from flu‑like illness. The questionnaire was
distributed from December 11th through December 17th.
The ILI definition used was self‑reported
fever plus either cough or sore throat, which is
similar to the CDC surveillance case definition.
Illnesses were counted if they began on or after
November 1st or through the date of survey completion.
We conducted two different types of
analysis, one a categorical analysis and the other a
person‑time analysis. For the categorical analysis,
we estimated vaccine effectiveness against ILI only
among persons who were vaccinated before November 1st
and compared those with those who were never
We looked at two different vaccination
definitions for persons who became ill less than two
weeks after being vaccinated. In one instance, we
categorized those persons as being unvaccinated. In
the second analysis, we excluded them from the
For the person‑time analysis, again, the
person‑time began November 1st and ended on the date
of survey completion. We did not exclude persons who
were vaccinated during the illness period for this
analysis, but an individual could contribute both
vaccinated and unvaccinated time if they got
vaccinated during the period of interest of November
1 through survey completion date.
The outcome for this study was ILI
incidence density rate. And, similar to the first
study for those who became ill one to 13 days after
vaccination, we either counted that time as being
unvaccinated time in one analysis or we excluded those
person‑days from the analysis.
This graph shows the number of
influenza‑like illness cases among staff. Those are
in the short light blue bars; the number of
laboratory‑confirmed influenza cases among patients at
the Children's Hospital, which are the tall purple
bars; and the line graph shows the percentage of the
cohort that was included in the study, the percent of
the cohort that was vaccinated by time.
So as of November 1st, 54 percent of the
persons who answered the questionnaire were
vaccinated. An additional 24 percent were vaccinated
during the outbreak period of November 1 and later.
So of 3,100 persons to whom the
questionnaire was distributed, 1,886, or 61 percent,
completed the survey. Half of those completed it
online and half of those completed it by paper.
Persons were excluded if they did not
report their vaccination status as being yes or no or
if they did not report date of vaccination. We also
excluded persons who did not report whether or not
they had an illness or if they did not report date of
This is the demographic information from
the persons who completed the questionnaire. It
includes all persons vaccinated or unvaccinated,
regardless of the timing of their vaccination.
Persons who were vaccinated tended to be
in the older age group. And a higher proportion of
them, though not statistically significant, were
In addition, physicians and nurses were
more likely to be vaccinated than persons in neither
of those occupation categories. And people who had
patient contact also were more likely to be
vaccinated. There is a considerable correlation
between these two variables.
As I mentioned, we asked people if they
had been tested for influenza and also what the
results of that testing were. Twenty‑eight people who
had influenza‑like illness reported being tested for
influenza. And of those, 13, or 46 percent, reported
that they were positive.
Now let's skip to the results. This is
for the categorical analysis, where persons were
counted as unvaccinated if they became ill in the less
than 14 days prior to vaccination.
So of the vaccinated persons, about 15
percent of them developed influenza‑like illness
compared to almost 17 percent among the unvaccinated.
The vaccine effectiveness adjusted for age group,
high‑risk condition, and patient contact was
approximately 14 percent with confidence intervals
that included zero.
If we looked at the additional vaccination
definition, where we excluded persons who became ill
in the 14 days after vaccination, that eliminated 9
people from the analysis. This analysis, then, again,
15 percent of vaccinated were ill and approximately 15
percent of unvaccinated were ill, with an adjusted
vaccine effectiveness estimate of approximately 3
percent, again, with confidence intervals that
In addition to the categorical analysis,
we conducted a person‑time or survival analysis. In
this analysis, again, people were not excluded if they
got vaccinated during the outbreak period.
Again, looking at the two different
definitions for vaccination, vaccine effectiveness for
both of these, the point estimates were negative, but
the confidence intervals were very wide and were not
statistically significant from zero.
There are a number of limitations to this
study. One, the response rate was only 61 percent.
And there was certainly a possibility for responses by
us. Secondly, vaccination and influenza‑like illness
status were self‑reported. And vaccination was
self‑selected. So there may have been some
confounding by indication for those two who were
In addition, there was a high vaccination
rate among the respondents. And this decreased our
power to detect a lower vaccine effectiveness
estimate. However, the vaccination rate that we found
in this cohort was very similar to what the hospital
estimated for their entire population. They estimated
that they had approximately 75 to 80 percent of their
In addition, because the influenza season
was early, some persons were vaccinated during the
peak influenza activity. In addition, we used a
nonspecific case definition, which likely
underestimated true vaccine effectiveness if one were
to look at laboratory‑confirmed flu.
The conclusions from this study are that
we were unable to demonstrate vaccine effectiveness
against ILI and this study when it drifted, influenza
H3N2 virus predominated. However, vaccine
effectiveness is likely to be higher against
laboratory‑confirmed influenza and against
influenza‑related hospitalizations and deaths.
The recommendations that were published in
the MMWR were to continue vaccination, particularly of
high‑risk persons, their contacts and health care
workers because H1N1 and influenza B may circulate
later in the flu season and because based on
historical information, vaccine is expected to protect
against influenza‑related complications and
The study pointed out that retrospective
analyses and sort of on‑the‑fly analyses are very
difficult to do during the middle of flu season and
that prospective annual vaccine effectiveness studies
against laboratory‑confirmed influenza are needed for
more accurate yearly assessment of vaccine
effectiveness and to assess the impact of the vaccine
I am now going to move on to the case
cohort analysis for persons 50 to 64 years of age.
This study was initiated to estimate the effectiveness
of the inactivated vaccine against
laboratory‑confirmed influenza in persons 50 to 64
years of age. And this was initiated at the same time
as the health care worker study.
For the cases, these all were
laboratory‑confirmed cases that were identified
through the surveillance system in the Colorado
Department of Health. Colorado requires that persons
with laboratory‑confirmed influenza are reported to
the state Health Department. However, they are not
required to report contact information. They report
age, gender, and name but not phone number or address
or county many of the times.
Over 10,000 cases were reported to
Colorado by the end of December. And over 500 of
those cases were in persons 50 to 64 years of age.
However, more serious cases of influenza were more
likely to be reported.
The cases were interviewed by phone. And
they were excluded if they did not recall being tested
for influenza or they denied being ill. In that case,
we may have had the wrong person when we tried to hunt
down a phone number based on name only.
Information was collected on demographics,
illness onset, and duration, vaccination, and timing
of vaccination, health care provider visits, and
For the cohort, the cohort was considered
the Colorado population aged 50 to 64 years. To
estimate the coverage in Colorado among persons 50 to
64 years, we looked at the Colorado behavior risk
factor surveillance survey. This is an annually
conducted survey that is state‑based. And for a
number of years, the question has been asked, "Have
you had a flu shot in the past 12 months?"
We looked at data from 2001 through 2003
for the limited 2003 data that was available. We used
a screening method or case cohort method to estimate
vaccine effectiveness, which was estimate to be
approximately equal to one minus the relative risk.
The confidence intervals calculation included the
variance of the various aspects and coverage
estimates. So the confidence intervals were
Again, similar to the other study, we used
two different classifications of vaccination for
persons who were ill in the two weeks following
vaccination. In one sense, they were counted as
unvaccinated. And then they were also in a second
Vaccine effectiveness was calculated for
the different possible cohort coverage rates in a
sensitivity analysis. Because of the difficulty in
determining the vaccination rate of the cohort,
because the outbreak happened during the typical time
where vaccine is administered, we have also initiated
a case control study.
So controls are being recruited that will
be age group frequency matched. We are attempting to
recruit three controls per case. And they are being
recruited through random digit dialing. Hopefully
this will provide a more accurate assessment of
So among the cases, there were
approximately 574 cases reported to the Colorado
Department of Health. Of those, interviews were
completed on 56 percent. Among those where there was
not an interview completed, for most of those persons,
it was an inability to contract the person or for most
of those inability to identify a telephone number. If
there were 1,200 John Smiths in Colorado, it was
difficult to identify who that name and age may have
belonged to. However, those who were interviewed and
those non‑interviewed did not differ by gender or age.
So this graph shows the number of cases by
week of their illness onset. As you can see, most of
these started again in early November and a cumulative
percentage of these cases that were vaccinated, which
is represented by the red line.
Of the cases, approximately 329 were
included. As you may have noticed from the slide
earlier, there were 330, but the vaccination status on
one of those persons couldn't be confirmed.
Of the 329, approximately 50 percent were
high‑risk and 50 percent were not high‑risk using our
vaccination definition of counting them as
unvaccinated if they became ill less than 2 weeks
after the vaccination date.
Among all persons in this category, 42
percent were vaccinated, 52 percent of those were
high‑risk, and 32 percent were non‑high‑risk who were
vaccinated. A high proportion was hospitalized.
Forty‑eight percent of high‑risk people were
hospitalized, and 17 percent of those classified as
non‑high‑risk were hospitalized.
We found very similar numbers when we did
a second analysis and excluded people if they were
vaccinated one to 13 days before illness onset.
Again, about 50 percent were high‑risk. And
vaccination rate was about 56 percent among
non‑high‑risk and 36 percent among non‑high‑risk
persons. And, again, high rates of hospitalization,
about a third.
I would like to look at the results of the
cohort. So this is the overall Colorado population 50
to 64. We had a number of different estimates. The
middle estimate was using the 2002 BRFSS final
weighted data. And the estimate, vaccine estimate, is
The high estimate may be the 2003
unweighted monthly data from BRFSS. So this is
preliminary data. And the December estimate was 52
A low estimate would be the 2002 Western
regional national health interview survey data for
persons 50 to 64. That data would suggest that the
vaccination for high‑risk be 40 percent and
non‑high‑risk 32 percent.
However, we know that Colorado generally
has substantially higher vaccination rates in all age
groups compared to most of the other Western region,
particularly California, which tends to weight that
estimate down quite a bit.
This is some preliminary results from the
controls that have been recruited from the case
control study. Of 304 persons who have interviews
completed and data entered, 26 percent of those have
high‑risk conditions. And vaccination rate for those
for this flu seasons, for 2003‑04, was 58 percent
among non‑high‑risk and 61 percent among high‑risk, so
considerably higher than the previous behavioral risk
factor surveillance survey estimates. We hope to have
recruitment through about 1,100 persons completed by
So if one considers this range of vaccine
coverage estimates in the cohort, we then use the
screening method to estimate what the vaccine
effectiveness may be in the overall cohort of 50 to
If you look at the non‑high‑risk group,
where among the cases, the vaccination rate was 32.5
percent, assuming the vaccination coverage for the
cohort ranges somewhere between 40 and 60 percent,
then the overall vaccine effectiveness may range
somewhere between 28 percent and 68 percent.
When you look at the high‑risk group,
there is substantially more content by indication. I
think it is very difficult to assess these results.
In general, it looks like vaccine effectiveness is
lower. I think that is very difficult to assess.
Again, looking at the second definition
for vaccination for cases being excluded if they
became ill in that two‑week period after vaccination,
the estimates are somewhat lower, but the
non‑high‑risk, the estimate is somewhere between 16
and 63 percent, and for the high‑risk, again, most of
the estimates are negative. That I think is very,
very difficult to assess.
The limitation of this study is that these
lab‑reportable cases over‑represent sicker and
hospitalized patients and certainly over‑represent
high‑risk persons. The case cohort method provided
for a very wide confidence intervals of the vaccine
effectiveness estimate, which gives us quite a bit of
degree of uncertainty.
However, overall the vaccination rate at
the time of onset of cases of illness is unknown due
to the early season. So we don't really know what the
vaccination rate is, really, for the controls at the
time of the outbreak. So use of the historical
vaccine coverage estimates would overestimate overall
So the case control study will help us in
determining what the vaccination rate was during the
time of the outbreak. So that should give us a little
bit more confidence.
The conclusions from our case cohort study
are that vaccine effectiveness among the non‑high‑risk
is likely somewhere between 16 and 68 percent. It is
a very wide estimate.
The vaccine effectiveness among high‑risk
persons is difficult to interpret because of content
by indication. And further analysis is pending for
the case control study.
I would like to acknowledge our
collaborators at the Children's Hospital, the
University of Colorado, the Colorado Department of
Public Health and Environment, and colleagues at CDC.
I just wanted to show you some other
studies that are currently in progress on which CDC is
collaborating. As I mentioned, the Colorado case
control study, we hope to have some results from that
by the summer.
There is also a cohort study being done
among children 6 to 23 months using Colorado HMO
database, which will look at ILI as an outcome. In
addition, the vaccine safety datalink sites are going
to be looking at vaccine effectiveness among 6 to
23‑month‑olds with outcome of hospitalized
And in Georgia, a study is ongoing to look
at the effectiveness in children 6 months to 4 years
with laboratory‑confirmed influenza as outpatients.
The new vaccine surveillance network in
Rochester, Cincinnati, and Vanderbilt will be using
the case cohort method to look at vaccine
effectiveness for influenza, hospitalization, and ED
and outpatient visits.
There is an ongoing study in Iowa among
college students, which is a cohort study to estimate
effectiveness against influenza in outpatients.
Among all of these studies, one of the
issues has been for persons who are tested for
influenza. What is the bias in terms of who gets
tested? And are people being tested based on their
The new vaccine surveillance network is an
active prospective surveillance system. And children
who come into that system who are enrolled, there is
no bias in terms of their vaccination status. So this
may provide us with one of the more accurate
So our results, we feel, are compatible
with many prior studies and years with the suboptimal
vaccine match, where low or no vaccine effectiveness
against ILI can be demonstrated. However, there is
likely to be vaccine effectiveness against
laboratory‑confirmed influenza as our case cohort
study suggests. These studies also illustrate the
difficulty in estimating vaccine effectiveness in
Our resources should be devoted to annual
assessment of vaccine effectiveness and to the
vaccine's health impact. We hope to have a pediatric
assessment using the new vaccine surveillance network.
And we would like to be able to use that system for
yearly estimates in children.
I will stop there.
CHAIRMAN OVERTURF: Yes, Dr. Myers?
DR. MYERS: Carolyn, I guess my conclusion
is it is hard to do studies with vaccine effectiveness
in the midst of the epidemic. So I think it is great
that the CDC is trying to do that.
On the first study, I am not so troubled
by your using ILI in association with a lot of
clinical isolates of influenza as your endpoint, but
I am troubled by the apples and oranges of the groups
that you evaluated.
You showed a statistical likelihood for
people who are likely to be exposed to get vaccine.
So the doctors and the nurses or people who had
patient contact were more likely to get the vaccine.
So I would have liked to have seen and
maybe you have available of that subgroup looking at
vaccine effectiveness for the people who had patient
contact during that interval.
DR. BRIDGES: You know, we really thought
a lot about asking questions about exposure, more
exposure, variables. But because there was so much
influenza circulating in the community, we thought
that that was extremely difficult to do, particularly
given that many people affected with influenza may
shed virus but are clinically asymptomatic.
We did adjust in our analysis for patient
contact. We adjusted for occupation in one analysis,
which I didn't present. And it didn't make any
We did a subgroup analysis. Also, it
didn't make any difference in terms of being able to
find vaccine effectiveness.
CHAIRMAN OVERTURF: Dr. Palese?
DR. PALESE: In terms of your first study,
where influenza‑like illness is being used as an
endpoint and there was no vaccine effectiveness
demonstrated, is it known what percentage the Fujian
strain made up in terms of influenza activity?
In other words, what is the reason for the
ineffectiveness of the vaccine? Do we know that 80
percent was Fujian there? You said it was
predominantly Fujian, but do we know? Do we have
precise values in terms of how much Fujian was and how
much H3 basically was occurring at the time in that
DR. BRIDGES: Very few isolates were
antigenically characterized from Colorado by CDC.
They obviously characterized a huge number of isolates
but relatively few isolates from every state. Some of
those isolates have overworked Fujian. I think it is
very difficult to assess what the proportion of H3N2
viruses in Colorado were Fujian based on the small
number of isolates that were characterized.
DR. PHILLIPS: Wouldn't they be interested
to know why the vaccine was not effective?
DR. BRIDGES: Nationally I think Ann Moen
will present this. About 80 percent. Is that right?
Seventy‑five or 80 percent of the H3N2 viruses were
CHAIRMAN OVERTURF: Dr. Monto?
DR. MONTO: I want to congratulate you all
for putting together the case cohort study in
Colorado. I know how difficult it is in the midst of
an outbreak trying to identify the places where you
can get useful data, which may have important public
I just want to recall that the multiple
studies that were put together during the period when
HCFA questioned the use of influenza vaccine for older
individuals, the 65‑plus, that there the endpoint was
hospitalization for pneumonia, influenza indications
in the course of an influenza season and not
I would assume that many of your
confirmations here are in the more severe individuals.
And controlling for confounding because we all know
that the people with undermined conditions are both
more likely to get vaccinated and more likely to be
hospitalized, the effectiveness there was from 31 to
maybe 50‑55 percent, fairly wide.
And these were all non‑drifted years. So
this really is not all that incompatible. You would
have a better endpoint here in terms of laboratory
confirmation. So it is all in the same ballpark and,
thus, suggests that there was some degree of vaccine
effectiveness in the past year.
In terms of the other study, I want to ask
you a question and make a comment, namely since the
outbreak was so abrupt and you have a fairly large
population group, could you look at the peak couple of
weeks of the outbreak in terms of the reported illness
and also in terms of the ILI case definition?
We have done some studies which we are
going to be publishing suggesting that sore throat
actually is a negative predictor in a febrile
individual of the isolation of an influenza virus,
which suggests that it may be appropriate to change
the very long used ILI case definition, but this has
been present in two very large studies of antivirals
in which laboratory confirmation was made before
treatment began. Sore throat clearly predicts against
the presence of influenza virus.
DR. BRIDGES: Thanks, Arnold.
We also looked at analysis from just
November 15th on, which was more of the peak week. We
didn't look at a two‑week interval, but we did narrow
it down to about a four‑week interval and still didn't
In addition, we looked at several other
outcomes that I didn't present today. One of those
was looking at a definition where persons had fever
for three days, plus two or more respiratory symptoms.
In addition, we looked at influenza
associated with loss of work, physician visits, and we
also looked at days in bed to try and find a more
severe outcome. And with none of those were we able
to demonstrate effectiveness.
CHAIRMAN OVERTURF: I will take just two
more questions: Dr. Karron and Dr. Goldberg.
MEMBER KARRON: Given information that has
been available in the past about potential increased
efficacy of live attenuated vaccine in children
against drifted strains, do you know of any studies
that are currently being conducted, either
head‑to‑head comparisons to look this year at the
efficacies or just live attenuated vaccine versus
DR. BRIDGES: I don't know if there is
someone from the manufacturers that could address that
question, I think. I see Kathy Coelingh in the back.
MS. COELINGH: I am Kathy Coelingh from
MedImmune Vaccines. There are some ongoing studies
that are being conducted. None of those data will be
available for several months, head‑to‑head comparisons
during the first year.
MEMBER KARRON: Did you do some studies to
assess the effective coverage probabilities in one of
these studies? Did you look at different models to
try and assess the different levels of
non‑responsiveness in the different groups and whether
you can model this out just to get an idea of whether
that is what is causing the problem?
The definition problem you addressed in
several ways, but there is a whole other piece about
what happened to the non‑responsiveness. You made
different classes of assumptions about them. Have you
MS. COELINGH: We haven't done that, but
that is something that we easily can do.
CHAIRMAN OVERTURF: Carolyn, if you could
just stay there? I am going to have Col. Neville come
forward and go ahead and set up his presentation. We
can take one more.
DR. DOWDLE: Thank you very much. I will
congratulate you for some remarkable analysis in a
very short period of time.
On the first study, in the 28 staff
members, as I recall, that reported being tested for
influenza, which was considerably less than the
percentage among the patients, if I recall correctly,
the question is, did you go further, do any further
analysis, of those who had reported influenza‑like
illness and were tested?
By the way, was this serology? Was this
DR. BRIDGES: These were all rapid tests.
As far as we know, it is anecdotal. We didn't ask
people to specifically to antidote. The hospital
epidemiologists and laboratories were testing some of
the staff in the lab. And also physically they used
the staff for the advantage of testing.
We did look to see if we could find
something very different in terms of the vaccine
effectiveness point estimate among those 20 people.
And in looking at that, we have estimates of zero,
DR. DOWDLE: Thank you.
CHAIRMAN OVERTURF: You get the last
MEMBER DECKER: And a couple of segues, I
think. First, I don't find your presentation and the
handouts. Will it be made available to the Committee?
DR. BRIDGES: The case cohort, the
information on the health care workers study was
published in the MMWR. And that is in the handout.
MEMBER DECKER: Right. The other one.
DR. BRIDGES: The other one is very
preliminary. So at some point, when we have more
analysis, we will be able to provide that.
MEMBER DECKER: All right. And the second
question is, if I understood your table correctly, the
first availability of the data that looks at
hospitalization or mortality is by vaccination status,
which I think is the real outcome of interest because
I think that is why the country hires the vaccines, to
prevent hospitalization and death and not
The first data on that will be available
at the end of this year at the sites? Is that right?
DR. BRIDGES: Right. The NVSN sites, we
want to have some data, hopefully by the end of the
year. For the case control data, because we have such
a high proportion of those persons who were
hospitalized, maybe also we are able to look at
hospitalization outcomes for the 50 to 64‑year‑olds.
CHAIRMAN OVERTURF: I think we will
proceed at this time. Thank you very much.
Dr. Neville is going to make a
presentation from DOD.
COL. NEVILLE: Thank you very much. I
appreciate the invitation.
I am here because the Air Force is the
executive agency in the DOD for influenza
surveillance. And the Air Force has been conducting
systematic influenza surveillance since the late
1970s. In 1999, it became the DOD program.
My organization is the Air Force Institute
for Operational Health, Brooks City, based in San
Antonio. We collaborate with the Naval Health
Research Center you see here in San Diego. In
addition the Army, this here, you will see in a minute
conducts clinical and virology services at the medical
centers. And the DOD oversees labs participating in
febrile respiratory illness projects.
The overall program is conducted under the
auspices and guidance of the DOD global emerging
infections surveillance and response system, or
Later you will hear from Linda Canas this
afternoon, who will present her annual summary of the
data from the DOD lab‑based surveillance program. I
won't present her data here, but, rather, my intention
is to describe other aspects of the DOD's involvement
in influenza surveillance with an emphasis on tests to
meet vaccine effectiveness, which, of course, received
attention this year.
DOD typically does not conduct influenza
vaccine effectiveness research, although this year
several efforts were made, as I will describe.
I will briefly overview the surveillance
programs that exist in the DOD and describe some of
the data that was available for estimating vaccine
effectiveness. Those include a Navy estimate, an Army
estimate, and Air Force estimate. These are the three
that I am aware of in DOD.
All three medical services and also the
Coast Guard contributed to influenza surveillance in
the Department of Defense. The Air Force program is
laboratory‑based, radiology‑focused, its main purpose
being the collection of respiratory pathogens from
military populations around the world that are
considered good sentinel sites. Don't worry about the
map details here. You will see it again this
afternoon from Linda Canas.
Selection factors for several sites
include overseas location, major ports of entry, and
the history of participation in the program. There
are 27 sites from all 3 services and the Coast Guard.
In addition, the overseas laboratory
network supplies specimens from various countries
where those programs are active. Some of the overseas
labs sends isolates to us in San Diego. Some of them
process their own or ship to the regional health
AFIOH in San Antonio has the only
full‑service clinical virology laboratory in the Air
Force. The Navy's febrile respiratory illness
surveillance program is managed by the Naval Health
Research Center. It focuses on eight military
On‑site investigators at FRI visits,
febrile respiratory illness visits, they all take the
denominator data and collect specimens of
systematically health care‑seeking trainees who meet
an FRI case definition, which is a temperature greater
than 100.4 and a cough or a sore throat. we may have
to change that maybe. We will see.
Since these trainee populations are
well‑characterized and monitored, attack rates can be
established with a high degree of confidence. NHRC
has the name "Respiratory Virology Lab."
U.S. Army has six regional medical
centers, each with clinical diagnostic virology
capability. While there is no systematic influenza
surveillance program per se in the Army, clinical
results are characterized as needed. And local and
regional preventative medicine staff track FRI rates
and etiologies as needed.
Special operating investigations also are
conducted as needed, more of which I will describe
briefly here in a minute. I should mention that all
three services have agreements or a plan in place for
sharing laboratory services in situations where demand
outstrips capacity, such as how major febrile
respiratory illness operates.
I will briefly describe some of the
relevant data sources that have been identified from
the DOD. Certainly these are not all unique data
sources, but in the context of an urgent public health
effort to estimate the effectiveness of the vaccine in
the face of this nationally occurring disease, these
are the data sources that we used both to monitor
disease activity and to attempt the evaluation of
The first is the local public health
officer reports, which is a surprising part of the
report. The primary responsibility for response and
monitoring rests at the local level in DOD. Reports
include routine reports of reportable diseases and
evidence of unusual patterns of disease or outbreaks
which might require special responses or assistance.
Our medical surveillance currently takes
advantage of data collection systems, largely not
designed specifically for surveillance activities.
Among this existing data sources is familiar IC‑9
coding for ambulatory visits.
Many of you may be familiar with the
electronic surveillance system for the early
notification of community‑based epidemics, or ESSENCE,
which monitors ambulatory visits at every DOD health
These data can be manipulated or used in
several ways. Here this chart may be hard to see.
This data accumulates all of the Pacific Rim DOD
medical care facilities and the percent of visits for
ILI, influenza‑like illness, which is comprised of
about 30 different codes, IC‑9 codes, that make this
ILI family. So this can be monitored on a daily
Facility‑specific data can also be
monitored. And a similar approach is used to monitor
diseases and non‑battle injuries for deployed
locations. Data collection at those locations is
often more lacking for a variety of reasons.
Laboratory specimens, submissions, and the
results can also be used to identify surges in
activity generally or at specific sites. This is an
example from last year where an outbreak occurred in
the Tidewater region of Virginia. Percent of total
commissions gives a good indication of pathogen
activity as well as illness occurrence. This is
similar to data displayed by CDC.
Officers. This tracks vaccination
delivery. These here just show the Air Force's data
because that is what it was most able to meet.
Annual influenza vaccination is mandatory
for military members, barring contraindications, of
course. High vaccination rates are achieved fairly
Again, this may be a little difficult to
see from the back reading from the handouts. Eighty
percent coverage was achieved by the 2nd of December
in this past year, 2003 and by the 6th of December,
Because influenza vaccination is mandatory
and it is a command program, it is not a medical
service program, it is a commend program, analyzing
vaccine effectiveness among active duty personnel
during this time period would be a fairly complex
Because such a small portion of service
members remain unvaccinated comparing attack rates
between vaccinated and unvaccinated groups is not very
Now to the vaccine effectiveness data.
First, the Navy's. As I described earlier, the Naval
Health Research Center in San Diego monitors FRI
illness, febrile respiratory illness, on basic
trainees of all three services. They have used
existing data here to estimate influenza vaccine
Their calculation is based on data for
four representative centers, training bases: Fort
Leonardwood, Fort Jackson, Marine Corps Recruit Depot
in San Diego, and Lackland Air Force Base in Texas.
For the month of December 2003,
influenza‑specific attack rates were calculated based
on culture‑proven influenza cases identified in the
ongoing surveillance program.
All trainees are routinely vaccinated upon
arrival of basic training, especially during this
time, the early fall. For this analysis, they were
considered unvaccinated or non‑immune for the first
two weeks after vaccination. So for an 8‑week
training course, for example, 75 percent of the
person‑time, they are considered immune or vaccinated.
So just to display the data that NHRC
worked with, these numbers are for their overall
surveillance program, not just for the sites that they
Only influenza A has been identified so
far this season as a ton of data, which was about a
week and a half ago. All of the isolates that they
sequenced were of the Fujian strain.
The dark red segments are influenza
isolates from trainees who had been vaccinated at
least two weeks prior to the onset of illness. As you
can see, the number of influenza cases is dreadfully
small, but there were no isolates segmented away.
This is the size of the calculation.
Because of the specific calculations they are showing,
viewing the vaccine effectiveness estimate of 9.9
percent, it is basic trainees. These data are not yet
published. I don't have other parameters, like age
ranges and gender and so forth, at this time.
The Army's Fort Lee is an Army post,
design post, in southeast of Richmond, Virginia. It
is home of the Army's quartermaster center and school.
It is not a basic training base, but it is a basic.
And that is specialty training.
At any one time, there are a little over
3,000 students present. And courses at that school
are of varying length, from 4 to 12 weeks. Students
arrive weekly for different courses. They are
organized into companies of 200 or so trainees. They
are housed in dormitory rooms of three to four people
each. And so it is a fairly complex model of exposure
in populations at risk.
An influenza outbreak began on the 31st of
October to find out it was in the students and staff
of the 23rd quartermaster brigade, which was a
training. Also, that epidemic occurred here. So the
31st of October, that is when that outbreak began.
Massive vaccination campaign began.
EPICON is the term for epidemiology
consulting team, the Army's Center for Health
Promotion. So the next outside team came and did some
work here. A team came back later in December to get
more information to try to evaluate this vaccine
Fort Lee had just done their routine
annual influenza immunization campaign. And few were
immunized when the outbreak started. So the medical
personnel on the post responded with a rigorous case
isolation protocol and mass immunization campaign that
reportedly reached about 95 percent of the target
population, which in this case is the trainees.
Data for two different cohorts were
analyzed, primarily based on data availability. One
cohort arrived early November, and the other cohort
arrived later in November. These two different case
definitions now that we see, in a minute we will see
four different analyses.
There are two case definitions. The loose
case definition and any ILI‑related IC‑9 code, a
visit, a health care visit, with a diagnosis of an ILI
code, and also a tight case definition that documented
febrile illness with sore throat or cough or
lab‑confirmed influenza infection.
There are few limitations here for us.
One of those is that it occurred right after an
epidemic, an influenza epidemic, in that training
post, which means that the threat of a subsequent
outbreak may have been modified somewhat by the
population on the study, being surrounded by people
who had achieved the natural immunity grasp.
So two different cohorts are listed,
population size for each, 67 and 463 there. The
vaccine effectiveness point estimates range from 40.5
to 100. And here there were no cases among the group
considered to be immune. Obviously those confidence
intervals are pretty well. And we put zero. So it's
hard to reach any conclusions of effectiveness
Now, the Air Force is here. The Services
Branch of AFIOH has a retrospective force underway
using telephone surveys to collect data on household
contacts of influenza‑infected index cases. They
identify influenza‑exposed cohort by identifying index
cases, who are persons whose cultures have tested
positive for influenza in our own lab.
We included only Air Force personnel for
simplicity and excluded persons from the Air Force
Academy and those who are deployed, in deployed
locations, because our real target is household
We contacted the sponsor of each positive
culture, which may be the person with the culture or
in post housing, and sought voluntary participation in
the survey and took some surveillance within two weeks
of the index case as well as vaccination status.
The secondary attack rate among vaccinated
and unvaccinated is used to estimate vaccine
effectiveness or is being used to estimate vaccine
There are almost 2,500 specimens in our
lab this year as of a couple of weeks ago. Of those,
859 were positive for influenza. And of those, 114
were considered eligible for this. Among those
eligible, there should be about 400 family members of
So as of February 12th, last Thursday, we
had collected data on 219 of these family members.
And the preliminary estimate of vaccine effectiveness
is 40.1 percent. We feel this could be a greatly
recurring annual study because it is relatively
efficient to do.
Plan B, we would like to do the same study
with a little more effort to validate some of hte
data, like looking at medical records, linking the Air
Force's immunization tracking, electronic tracking,
system to validate vaccine delivery and so on.
So, to summarize, NHRC had a vaccine
effectiveness of 91 percent. CHPPM's vaccine
effectiveness was 40.1 percent, although our concern
was pretty wide. It included zero. Our vaccine point
estimate in the Air Force was preliminary point
estimate is 40 percent, but that it gave us
collections that were none.
With that, I will be happy to answer any
CHAIRMAN OVERTURF: You didn't mention,
out of those 240 isolates, what the type was.
COL. NEVILLE: Every isolate we got was an
A, an influenza A. And every one that we have
sequenced is consistent with Fujian strain. I don't
know that we sequenced, but it might be something like
CHAIRMAN OVERTURF: Are there any more
DR. MYERS: I missed it, the NHRC Fujian
COL. NEVILLE: Every one that they
sequenced as well as those after the first isolate
that they sequenced ‑‑
PARTICIPANT: They have sequenced about 20
COL. NEVILLE: Pardon me?
PARTICIPANT: They sequenced about 20
COL. NEVILLE: About 20 percent.
DR. GELLIN: Can you comment about the
regularly recurring annual study?
COL. NEVILLE: Well, competing interests,
I suppose, supplies and manpower and stuff like that.
We were able to achieve pretty rapid IRB permission to
do this next year. It's on the chart. We will do it
again. And it is relatively easy and cheap.
But there is another operator, another war
somewhere and competing priorities, then yes. We plan
CHAIRMAN OVERTURF: Dr. LaRussa?
MEMBER LaRUSSA: Would you just remind me?
The Air Force study, was it everyone who had the
COL. NEVILLE: The sampling is designed
primarily for collecting pathogens. So not every
person who shows up at a clinic, even if the symptoms
are right, is sampled. It's just the clinician in the
local public health system.
We try to get them to send us six a week,
but that is the target. So if they have not much
disease, they don't get that many out of these. And
we may get a lot more than that.
So I don't know what portion of people
presenting to a clinic with febrile respiratory
illness is sampled. It is variable.
CHAIRMAN OVERTURF: Any other questions?
CHAIRMAN OVERTURF: We are on time. We
will hear the INSERM presentation.
DR. FLAHAULT: Ladies and gentlemen, first
of all, I would like to thank the adviser very much
for inviting me to present this work. This work is
followed by INSERM in France, which is the national
institute, the French NIH, I will say, and by the
Universite Pierre et Marie Curie, which is a
university of Paris.
Of course, I will not come back too long
to the strain selection for the influenza vaccine,
just to say that in Europe, that was, of course, the
same recommendation. And vaccine which was not
created in Europe also used Panama strain, rather than
the Moscow, and used several other strains of the B
virus for the epidemic current season. Eventually it
is a main circulating strain. As you say, it was also
in Europe, the Fujian strain.
So it is a question of which of those is
during, at the beginning of the epidemic, and not
after the epidemic was, how effective was the vaccine
against clinical disease? So it was nearly impossible
to try to provide some answer due to certain
difficulties but also to measure it for sure during
that time in Europe as well as in the U.S.A.
Also in Europe, we had a very early
epidemic. It was just above the man, the average,
epidemic, but as an early epidemic, it was not so easy
The question you had in the previous talk
was about the particular rates you found in your
revised studies and also in the comprehensive review,
which was done by Roland at the first talk.
So what is an appropriate protective rate
is difficult to answer. It has been proposed by the
weekly epidemiological record of the WHO that against
ILI, clinical disease, at least, that should be
between 15 and 80 vaccines.
How can we measure vaccine effectiveness
in the flu season? We have seen several methods also,
of course, particular efficacy, which is mainly used
for trial vaccines. And the clinical effectiveness is
usually observed by the so‑called screening methods in
the cohorts that are the case controls and the case
cohorts and all of these kinds of models, which, in
fact, try to compare attack rates in vaccinated versus
The vaccine effectiveness with the
so‑called screening methods is this formula, which is
1 minus the ratio of attack rate in vaccinated
population and of the attack rate in unvaccinated
To assess vaccine effectiveness in real
time in France, we used the screening methods because
they are efficient, because cases are drawn from our
sentinel system, which is a very clinic surveillance
system, which provides some figures in real time or
close to a real‑time basis. I say "closely" because
we have all of the consolidated figures each week.
We chose the controls of the cohort from
a regular poll which had been collected by the French
Social Security using a private company specializing
in these kinds of polls for studying and estimating
the vaccine coverage in France for years.
It had been set up in our communities 20
years ago in November '84. It continues on the clinic
surveillance of 12 conditions: ID, of course, but
also acute diarrhea, measles, mumps, chicken pox,
spinal meningitis, viral hepatitis, and also other
infectious conditions, such as asthma.
It can hold a network of 1,200 sentinel
GPs, which are virtually unpaid and all private
practitioners, as they provide information on a
numbering basis. As I told you, we have re‑created a
ton of information which is widely available on the
internet, including the vaccine effectiveness.
So cases are ILI reporting ID cases, which
are only clinical cases with high fever and myalgia,
upper respiratory symptoms, as you mentioned. What is
at work is in each case is individually described with
the age in year, either in months for children;
gender; vaccine status. That is the trademark, only
the vaccine or not, during the preceding four; and
also hospitalization, the case of and with many
details of hospitalization, such as condition dates
and so on, in the case of. It is a mix for France.
Controls. As I told you, we conducted a
poll, which has been done annually for estimating in
terms of vaccine compliance. This is a national
survey. It was a representative sample of the French
population. We obtained data until ten years ago
conducted by a private organization.
As I told you, we have to make
assumptions. These are the figures which were 65
years or less, between 60 persons and 70 persons now.
Of course, we did not have the level of the
cooperation of the current season. So we assume it
was constant, it was the same as the year before.
For the others, the illness in 50 and 64,
there is a rate of conditions being low in France. It
is assumed to be 11 percent for the current season as
it was for the preceding season.
This slide is just to show you some
existing tools for monitoring that we have in France.
We have an early warning system based on accepting
what it takes to detect epidemic in time and to
measure the excess morbidity of influenza.
Also, we have a set of predictors. The
size of influenza epidemics prior to its occurrence
has been presented in our application 5. It predicted
fairly well the level of influenza activity in last
September for the current system. We predicted a
sweep on two million cases. And eventually the size
was Sweden in person‑time, which is a country of 16
We also have set up a forecasting
time‑space dynamic of influenza epidemics broadcasted
on French TV, which was for broadcasting three weeks
ahead the times of influenza. These maps are
broadcast on the French TV each week now during the
epidemic, of course. It was actually this content in
As I told you, we have now set up a field
vaccine effectiveness, which is run for years for
measles. It has been published in 1993 and which is
on the routine basis since 1998 in France for
influenza and clinical disease. We also have some
maps showing the current evolution of the epidemic
week after week in the country.
So the available material for influenza
effectiveness assessment is to have this contingency
table for each age strata and each age group strata.
We need to have the number of vaccinated ID cases,
vaccinated ID cases, and say which coverage in the age
In fact, we did not choose the screening
methods as it was presented because it is too
simplistic. If we want to adjust an age group, we
need to use this for vaccine efficacy which has been
proposed by Greenland in statistics medicine in '86
and which also is used to compute the 95 confidence
interval, which is very easy to meet demands in our
The results of these measures, the ongoing
measurements, are this one. You have here the figure
of this year, which is vaccine efficacy/effectiveness
of 60 percent, which computes along with the score to
Of course, the confidence interval may be
large. And if the point estimate, the confidence
interval, which is very important, which was very
important to our public health authority, is to see
that in 2003‑2004, the level of vaccine effectiveness
was pretty similar to that of the other years with the
exception of the year '97‑'98 where when the H3N2
Sydney strain was circulating and it was very
well‑known that this strain was not included in the
vaccine and the vaccine was clearly not protective
against this strain and the vaccine effectiveness
using the same methods was 26 persons developed, 13 to
We presented this measure on the first
week of the epidemic. And we completed this series
during the whole epidemic using two relative figures,
but it did not vary a lot. Of course, the confidence
interval was much larger at the beginning of the
epidemic, but the vaccine effectiveness was about the
We put all of the data each week on the
Web site, which is now available also available. Many
things are in French. We have set up for WHO since we
have been designated five years or six years ago as a
collaborating center for epidemic disease
surveillance. We have set up for them the FluNet
System, which is still based in Paris, our unit, but
we put the move, really, in 2004 for Geneva. It was
set up in 1997 trying to connect all of the network of
412 centers and also the center between WHO, which was
Because of that, it is possible to have
some very interesting findings. The first thing is
that there are similar patterns in the Northern
Hemisphere in terms of morbidity. It really begins in
the U.S. early. It begins also in Europe very early.
The duration of epidemics was the same in '98 and '99
presenting the figures, but also for the four
The virus circulation with the FluNet,
which is reported in FluNet, you can see that in the
Southern Hemisphere, not for Hong Kong, which is more
a subtropical area, but in the temperate zone of the
Southern Hemisphere, you have a very good coherence of
time occurrence of isolates reported in the
hemisphere. Of course, in the Southern Hemisphere,
there is a lag of six months, which is also very, very
We have tried to study the mortality in
France, in the U.S., and in Australia. We have seen
that the average duration of activity with the same
kind of model organization using a periodic regression
model for estimating the size of epidemics to show
that the average is really very, very similar in
France and in the U.S. It is a little shorter in
Excess mortality is highly correlated
between France. So we can say between Europe and the
U.S. It is not correlated between the South and the
North Hemisphere as in Europe or in the U.S. The
correlation between France and hte U.S. was open to 60
To understand if the peak is happening
here the same date in France and in the U.S. and also
in Australia, we can see this is a way to get the
analysis. We show very, very high coincidence in the
phase of rapidity.
And we can see also a high correlation
between the peak date in the U.S. and the peak date in
France, showing that the peak occurred at the same
time in both countries. So we see a peak between the
U.S. and France in the last 20 years. It was opened
five weeks, so very, very short.
There is a strong coherence in Northern
Hemisphere, the temperate zone at least. It may
suggest the starting date, date of peak, duration of
influenza epidemic are similar in the U.S. and Europe.
Also size and severity of influenza incidence are
proportionally similar in the U.S. and as for Europe.
So we can probably assume that virus circulation and
probably also vaccine effectiveness among clinical
disease as we see in France is probably similar. And
this is going to be around 60 persons in the U.S. as
well as Europe. That was around 60 persons using the
method reviews, the general practitioners, the
clinical nurse, and so on.
So I want to take this opportunity to
thank all of my team, which worked with us on this
thing. Tank you very much for your attention.
CHAIRMAN OVERTURF: Are there any
questions? Yes, Michael?
MEMBER DECKER: I just wanted the same
question I had for Dr. Bridges. Will you make a copy
of your presentation? Will it be available?
DR. FLAHAULT: Yes, of course. I just did
not provide that. In fact, I was a little embarrassed
because we wanted to publish the material. And we
know that FDA will publish it on the Web site. So we
will wait for the presentation to provide all of the
figures and charts with that.
CHAIRMAN OVERTURF: Dr. Farley?
MEMBER FARLEY: Yes. I am wondering what
predicts an early onset of a flu season.
DR. FLAHAULT: In fact, we did not try to
predict. We did not include that in our predictions.
So we only predicted the size in terms of morbidity
and mortality of the epidemic in France.
We did not predict the onset. We only
predict the onset with the other model, which has been
very recently published in the American Journal of
Epidemiology by Cecile Viboud, a nurse, which is
predictions three weeks ahead. So that is providing
a very big warning for the beginning. And after we
predict three weeks.
CHAIRMAN OVERTURF: Yes, Dr. Monto?
DR. MONTO: I think your studies show the
idea of long‑term ability to study influenza vaccine
effectiveness from year to year because then you
really can put what you observe into context.
What I would like to ask you about is the
variation in peak time in influenza incidence in
various countries in Europe. I had better be careful
what I say. France is a relatively small country.
And the peak occurs over a reasonably short time
period in the entire country.
I understand there were different times,
just as there were different times of peak occurrence
in Europe. And we have the same situation in the
United States. How well does your model predict
occurrence of peaks in western and central Europe, for
DR. FLAHAULT: It is not so easy to have
precise data and available data on that. In fact, we
felt that the distribution of the peak was much wider
than it was, effectively, either in France, within
France. We used to say that the peak was moving from
a region to another one. The reasons were so that, in
fact, there is a good coincidence everything is done
within ten weeks.
So ten weeks may be not exactly
coincident, but all activity is done in ten weeks.
And all of the peaks are done within three weeks, so
very, very short. So within Europe, we need to have
the same kind of system or at least the same kind, not
the same system but to know what we know, if anything,
about the peak.
And in terms of mortality data, when we
have them, we can say that the peak of mortality is
really, really coincidental. And so I don't know what
is your experience in the United States between the
states of the United States, but I am not so sure it
is a very wide distribution of peak in your country,
CHAIRMAN OVERTURF: I assume that the
reason for no breakdown in the age group from 15 to 60
or 65 is because there is no recommendation for
vaccine routinely to be given. I assume that the
rates of immunization in 50‑year‑olds is the same as
it is in 15‑year‑olds, which is 10 percent. Is that
true or not?
DR. FLAHAULT: Yes, not completely true
but approximately true. That is right. In France, it
is free of charge to be immunized after 65 and plus,
and it is not reimbursed before. It is not completely
exact because in several companies, you may have free
of charge vaccination. And because of that, in the
same age group, in some particular situation, you may
have some higher figures, but in children at least,
the figure ‑‑ not in children. It is very low, but
between 15 and 30 is very low, too.
CHAIRMAN OVERTURF: There are no European
countries that are considering immunization of
children that you know?
DR. FLAHAULT: All of us are considering
immunization of children, but we are still waiting for
the product, which is still not yet marketed, also to
have some recommendation of regulatory authority with
work on that in the future, we are sure.
CHAIRMAN OVERTURF: Dr. Gellin?
DR. GELLIN: We are enlightened by this
series of vaccine‑effective studies. Do you or does
anybody know if there are similar studies in the
DR. FLAHAULT: I am sorry. Can you repeat
DR. GELLIN: Does anyone know if there are
vaccine‑effectiveness studies in the Southern
DR. FLAHAULT: No, I am not aware of that.
CHAIRMAN OVERTURF: Yes, Dr. LaRussa?
MEMBER LaRUSSA: Do you plan to have any
data on the importance of influenza‑like illness in
DR. FLAHAULT: We collect influenza‑like
illness in children. In France, general practitioners
are taking main charge of the presence of the children
in terms of immunization, in terms of like illness.
So we have these down. They are not immunized at all.
So we did not compute any vaccine effectiveness. So
the measurement of the coverage, vaccine coverage,
only concerns age 15‑plus.
CHAIRMAN OVERTURF: Any further questions?
CHAIRMAN OVERTURF: Well, I would like to
thank all of the presenters thus far. We have a
little extra time for a break. We are to be back here
by 11:30. And we will hear about U.S. surveillance.
(Whereupon, the foregoing matter went off
the record at 11:15 a.m. and went back on
the record at 11:36 a.m.)
CHAIRMAN OVERTURF: Ann Moen is going to
present the data on the U.S. influenza surveillance.
MS. MOEN: Good morning. I am going to
spend about the next 20 minutes giving you an overview
of the U.S. influenza surveillance data for the
current 2003‑2004 season.
This first slide depicts the general
schematic of the U.S. influenza surveillance season
and the major components of that under which CDC
collects data on a weekly basis from October to May
There are four major components that we
collect data on weekly. The first is virologic data,
which we collect from the system of laboratories in
all 50 states. And then there are sentinel providers
that we collect influenza‑like illness on throughout
the 50 states. That is data reported weekly to CDC.
Each week, the state and territorial
epidemiologists in each state report to CDC the level
of influenza activity in their state based on defined
criteria. And then the 122 cities' mortality system
we receive vital statistics, registrars' reports to
CDC on a weekly basis, reporting influenza mortality
due to P and I.
There are some other various forms of data
that come in sporadically, such as institutional
outbreaks and sometimes cruise ships. This year we
had some extra data we collected on pediatric
mortality, which I will talk about later.
All of this information flows into CDC
through the state health departments. And we work
very closely with our partners in the states. Then
the data is analyzed weekly at CDC and then published
in reports that go back out to public health
officials, physicians, the medics, and the public.
This slide shows the virologic data that
is collected through the WHO and national respiratory
and enteric virus surveillance system collaborating
laboratories. There are approximately 120
laboratories throughout the United States that collect
virologic data. They report to us weekly the number
of respiratory specimens that they tested and the
percent that were positive for influenza.
This graph shows just the positive
samples, and the yellow bar shows the influenza A
viruses, which were unsubtyped. And that smaller
subset, which were subtyped, the influenza A(H3N2)
viruses, are shown in red. There is a smattering of
B viruses down here, just a small handful, shown in
green that you may or may not be able to see. And
then the blue line shows the percent of overall
respiratory specimens that tested positive for
To give you an idea of the magnitude of
the data collected through that system, of the
viruses, the total specimens tested by the WHO in
NREVSS laboratories in the U.S. for the week ending
February 7th, there were over 92,000 respiratory
specimens tested for influenza. And of these, 22,419
were positive for flu.
All of the laboratories in this
surveillance system type the influenza viruses. Of
those positive for influenza, 22,286, or 99.4 percent,
were influenza A viruses. And then there was a small
handful of influenza B viruses, just 133.
Now, a smaller subset of these 120 or so
labs also do subtyping of influenza viruses. And of
the 22,000‑some viruses positive for influenza A,
5,862 were subtyped. Of those, 99.9 percent were
influenza A, H3N2, viruses with just one influenza A,
H1 virus, which was detected from a military base in
Virginia, I believe, and was probably associated with
So of all of these viruses collected
through this system, a subset is sent to CDC for
further characterization. This afternoon, you will
hear about the antigenic and genetic characterization
In reference to an earlier question about
the percentage of Fujian‑like viruses, I believe for
the week ending February 7th, approximately 82 percent
of them were antigenically similar to Fujian viruses.
And 18 percent were Panama‑like. You will hear those
So this slide shows the percentage of
estimates testing positive for influenza. The red
line shows this year's data. You can see that it
peaked at about 36 percent positive around week 51‑52.
For comparison here, I have shown the data
from the 1999‑2000 season in green, which is the most
recent moderately severe influenza AH2 H3N2 season
that we have had. And then, for further comparison,
the purple line shows the percent of positive
specimens tested through this system of labs for the
What this slide does a very good job of
showing is that the season did come earlier this year
and started very early and peaked earlier than most of
the previous seasons. The 1999‑2000 season was
considered early at the time. And this season was
even earlier than that.
This slide showed the percentage of visits
for influenza‑like illness reported by sentinel
providers through our national system of sentinel
providers. Each week, a series of sentinel providers
in the U.S. report the total number of patients seen
in their practice and then the total number of patient
visits for influenza‑like illness using a case
definition of fever of greater than or equal to 100
and cough or sore throat with no other known cause of
The smooth white line across the center at
2.5 percent is a national baseline. And you can see
that this season that started early peaked around
eight percent. And it has currently come back down
under the baseline. We are currently at the end of
last week around 1.5 percent.
Again, for comparison, the 1999
season‑2000 season is shown in green. And then the
much milder season that we had last year is shown in
purple, where there wasn't that much time spent above
the national baseline.
Just to give you an idea of the numbers of
sentinel physicians in this program, we currently have
1,931 sentinel providers that are enrolled in this
program. And they voluntarily report this information
to CDC on a weekly basis. Of the providers enrolled
in this program this year, 1,141 of them have been
regularly reporting so far, which is considered
reporting more than half of the weeks.
From these sentinel physicians, we
received almost 20,000 reports to the end of last
week, which was for over 5 and a half million patient
visits. And of these, about 170,000 of them were for
reports of influenza‑like illness.
This slide shows the third major component
of the influenza surveillance data collected at CDC.
It represents the pneumonia and influenza mortality
data for the 122 cities. Each week the vital
registrars in the 122 cities report the total number
of death certificates filed and then the number of
death certificates that have either pneumonia or
influenza listed somewhere on them so that we can get
a percentage of deaths due to P and I.
This data represents about a third of the
mortality data for the U.S. The 122 cities represent
about a third of all deaths in the U.S. On this
slide, you can see the bottom smooth line is the
seasonal baseline that is calculated using a
mathematical modeling method. And the upper smooth
line is the epidemic threshold, which is 1.645
standard deviations above the seasonal baseline. Then
the red jagged line shows the influenza mortality due
to pneumonia and influenza.
You can see for the current season that we
have peaked at about 10.3 percent and just for the
week ending 2‑7 came down to about 8.7 percent is
still above the epidemic threshold. Any time the red
line crosses the epidemic threshold, then we consider
that there is excess death due to influenza and
For comparison on this graph, you can also
see last season, which was much milder and barely up
above the epidemic threshold. Then you can also see
the 1999‑2000 H3N2 season, which peaked at 11.2
percent over here. So it would sort of show that this
season wasn't greater in magnitude in terms of
mortality than previous H3N2 seasons.
This slide shows the fourth major
component of our influenza surveillance system, where
we have received weekly influenza activity estimates
reported, as assessed and reported, by state and
The white color shows no report. The
yellow color is no activity. Green represents
sporadic activity. Purple or the light purple
represents local activity. Blue is the regional. And
red represents widespread.
This is the very first week that we have
started reporting data for this season. And you can
see that already in Texas, local activity was being
The next series of slides is going to take
you on a quick trip through the weekly reports as
reported and assessed by the state and territorial
epidemiologists. You will see how the season
By the middle of November, there was quite
a bit of activity reported, especially in the western
half of the United States. And by the beginning of
December, there was widespread, lots of influenza
activity throughout the United States, where we peaked
about the middle of December or towards the end of
Then you can see the activity continued to
decline, and there is still as of the week ending
February 7th some activity or relatively a bit more
activity in the East, where the outbreaks and the
epidemics started just a bit later.
This year we made some changes in the
activity reports from the state and territorial
epidemiologists. In previous years, the activity
levels were assessed at four levels: none, sporadic,
regional, or widespread. These were based on a
percentage of the population and counties. And then
the criteria used was either outbreaks of
culture‑confirmed influenza or influenza‑like illness.
In response to some comments from states,
we worked closely with our partners in the states to
make the criteria by which reporting was done more
defined so that it could be more uniform from state to
So for this current year, five levels of
activity reported: none, sporadic, local, regional,
or widespread. And this was based on state‑defined
regions within the states. The defined criteria used
a combination of influenza‑like illness and outbreaks
and laboratory data. There were specified time frames
for which the lab confirmation needed to take place,
though hopefully we will be able to assess this at the
end of the season and see if states are happier with
this local new definition for providing activity
So, just to take you back to the peak
week, though, for the week ending December 20th, 351,
there were 49 states reporting either regional or
widespread activity in contrast to the previous
1999‑2000 H3N2 season, when there were 45 states
reporting either regional or widespread activity.
You can see that that season peaked about
three and a half weeks later than the current season.
This is in contrast to last year, where at the peak of
the epidemic, there were 34 states reporting regional
or widespread activity. And that peaked at the
beginning of March.
I also want to spend a couple of minutes
talking about influenza‑associated death among
children less than 18 years of age. Because of the
early attention that pediatric mortality received, CDC
requested reports of influenza death in children by
sending out health alert networks, publishing in the
MMWR and Epi‑X.
As of the end of last week, we had 134
influenza‑associated deaths that were lab‑confirmed
reported to CDC. The median age was 3.36 years with
the range of 2 weeks to 17 years.
Of these, 82 children were less than 5
years of age, 36 children were 6 to 23 months of age.
And of the 134 children, 32 of these children had
underlying medical conditions.
Of the available vaccination histories, it
is shown that 76 children were unvaccinated, 38 had
missing or unknown vaccination histories, 20 had some
sort of vaccination, but only 3 of these were
vaccinated according to the recommendations. It was
pretty much equal opportunity with about half and half
male and female.
This slide shows the epidemic curve by day
and week of the influenza impact in children. You can
see or you will see that the peak here correlates with
the peak of the nationwide aggregate data.
So the question is, is the 2003‑2004
different in impact among children? Well, this has
not been a normal part of our regular surveillance
systems. The influenza‑associated deaths are not
reportable conditions in the U.S. So the average
annual number of influenza death is unknown. There is
no baseline data.
There was a study that looked at ten
years, from 1990 to 1999, and estimated the annual
average of 92 deaths, respiratory and circulated
deaths, occurring among children less than 5 years of
age. This estimate is based on mathematical modeling,
and it is not counting laboratory‑confirmed
Studies to determine if hospitalization
increased in children are ongoing. And discussions
are underway with our state partners to consider
making laboratory‑confirmed deaths in children or
This slide, in summary, shows some of the
main components of our influenza surveillance system
overlay so you can see how they correlate. The red
line shows the percent of visits for influenza‑like
illness. The blue line shows the number of states
reporting widespread or regional influenza activity.
The yellow line shows the pediatric deaths. And the
green line shows the percent of positive isolates as
tested by the WHO collaborating and NREVSS
You can see that the ILI, the state
reports, and the influenza mortality, pediatric
mortality all nicely correlate and peak around week
51. The virologic data seems to precede the other
indicators a bit. And it may be due to some early and
very heavy reporting by some of the states who tested
a lot of respiratory specimens in the West and
submitted a lot of specimens that were positive.
I think I will stop there and take any
CHAIRMAN OVERTURF: Are there any
questions? Dr. Myers?
DR. MYERS: Ann, in those pediatric
deaths, particularly the ones in young children, were
there risk factors other than age? And did you look
specifically under one year of age?
MS. MOEN: They have got age data on all
of the children. The data are being analyzed. The
form that they used is to collect as much information
as they could on children.
But some of the information was hard to
collect because we had reporting of a lot of death
before they decided to collect certain information.
They are going back and trying to fill in some of that
information. Hopefully more full information will be
CHAIRMAN OVERTURF: I believe from the
published data already, about 50 percent of the
children did have underlying illness, though. Isn't
MS. MOEN: That was through a few weeks
ago. This is the most current data as of the end of
last week. So the underlying illness for this is a
CHAIRMAN OVERTURF: Dr. Monto?
DR. MONTO: One of the questions we are
always asked is whether the deaths in children that
have been in the newspapers, et cetera, and locally is
something new or is something that has been around for
a while that we haven't recognized.
One thing that was striking to us in
southeastern Michigan is that we had several deaths of
children last year. And some of them were type A H1.
And some of them were type B. And none of them were
A H3N2, which is the subtype we generally associate
My question is, how much is the use of
rapid testing to identify influenza virus associated
with the correlation between children's deaths during
an influenza outbreak and identification of influenza
as the etiologic agent?
MS. MOEN: I am not sure how many of these
children were tested by rapid test or influenza
culture. So I can't answer that question.
CHAIRMAN OVERTURF: There is a question
from the audience. Yes?
DR. RUBEN: Fred Ruben of Aventis Pasteur.
I just wondered, Ann. We would normally
think of deaths during influenza periods as associated
with pneumonia and influenza. My understanding from
a presentation by Tim Uyeki was that these weren't all
pneumonia and influenza‑like type deaths, that there
were other attributes, like neurologic conditions and
so forth. Could you comment on that?
MS. MOEN: I think that is true. Some of
them were sudden. It runs the gamut. There were some
that were quite rapid and not associated with
long‑term pneumonia illness.
And there were definitely some excess or
additional cases of pediatric encephalopathies
associated with some of these cases. And we
additionally have been collecting information on other
severe cases of influenza associated with
encephalopathies that may not have resulted in death.
CHAIRMAN OVERTURF: Dr. Gellin?
DR. GELLIN: Ann, you commented early on
about the surveillance, the virologic surveillance,
system. I don't have a context for how many samples
normally go through that system. I guess the question
is akin to the rapid diagnostics. With the increasing
availability of rapid diagnostics, is there a problem
in the system with getting enough viral cultures?
MS. MOEN: I think right now it is
something we remain concerned about. We know there is
a good useful rapid test on occasion, but we also need
to maintain viral isolation so that we can
characterize the viruses.
I don't know what percentage of these
tests were rapid positives, but those are also
reported. An in some cases, say in a year like this
year, when there weren't very many B's, if we're just
getting mostly rapid test reports of B's, that doesn't
leave us very many isolates.
But so far, I mean, the 92,000 that have
been year to date this year is almost equivalent to
the total number of respiratory specimens that were
looked at last year. So I think we are getting more
isolates so far this year.
CHAIRMAN OVERTURF: Dr. Myers?
DR. MYERS: I guess after the presentation
we had just before the break, the obvious question is
to ask, as the numerator data, the case selection
method that France is using, is available to us
through our sentinel physician system, has CDC
considered the possibility of doing an electronic
surveillance analogous to what is done in France,
which, really, I was impressed by the fact that it
gives real‑time data?
MS. MOEN: Yes, their system is
impressive. Right now our sentinel physicians don't
collect vaccination data or patient‑level data. We
get only age data on the sentinel physicians. So we
can't look at vaccination right now.
And on a wide scale, I don't know how well
we would be able to convince all of these providers to
report that level of detail gratis, as they are doing
But it would be great if the U.S. could
implement some way to have annual estimates of vaccine
effectiveness based on surveillance data.
CHAIRMAN OVERTURF: Yes?
DR. ROYAL: Are you able to comment more
specifically on what is known about some of the
neurologic complications these children have
developed, either pathologic findings or other autopsy
MS. MOEN: No. I wouldn't be the best
person to comment on that.
CHAIRMAN OVERTURF: Is there anyone
present who can comment on that?
CHAIRMAN OVERTURF: Are there any further
questions for Dr. Moen?
CHAIRMAN OVERTURF: We only have three
minutes left before the scheduled break for lunch. I
think Dr. Cox was going to go early, but I don't think
we have time for Dr. Cox. So I think we will just
wait until after that time.
We are scheduled to reconvene at 1:00
o'clock. So I will adjourn the meeting until that
(Whereupon, at 11:59 a.m., the foregoing
matter was recessed for lunch, to
reconvene at 1:00 p.m. the same day.)
DR. FREAS: I think we're ready to resume
CHAIRMAN OVERTURF: At this time of the
meeting, there's time set aside for public comment.
We've heard of nobody who wants to make public
comment, but at this time I would encourage members in
the audience or others who want to make public comment
to step forward to a microphone, identify themselves.
Before you comment, I need to make one
statement which I need to read for the FDA. "Both the
Food and Drug Administration and the public believe in
a transparent process for information gathering and
decision making. To ensure such transparency at the
open public hearing session of the Advisory Committee
Meeting, FDA believes that it's important to
understand the context of an individual's
presentation. For this reason, the FDA encourages
you, the open public hearing speaker at the beginning
of your written or oral statement to advise Committee
of any financial relationship that you may have with
any company or any group that is likely to be impacted
by the topic of this meeting. For example, the
financial information may include the company's or
group's payment of your travel, lodging or other
expenses in connection with your attendance at the
meeting. Likewise, the FDA encourages you at the
beginning of your statement to advise the Committee if
you do not have any such financial relationships. If
you choose not to address this issue of financial
relationships at the beginning of your statement, it
will not preclude you from speaking."
MS. BARR: Thank you. My name is Geeta
Barr. I'm with the National Vaccine Information
Center. And I have no financial conflicts of
And this question is addressed to Ann
Moen. During the question and answer period after
your presentation, something came up regarding flu
related deaths in children, especially in regards to
neurological complications, encephalopathy and you
mentioned encephalopathy has been associated as a
cause of complication with these flu related deaths.
Is there or will there be available data
on how many of these children were vaccinated?
MS. MOEN: I gave the vaccination data for
the kids that were vaccinated. I believe it was 76 or
78 of the children were unvaccinated and then there
was missing data on I believe it was 38 of the
children and vaccinated children were approximately 20
with three of them vaccinated according to the
MS. BARR: Thank you. And I had one other
Dr. Levandowski mentioned in his
introduction that protection from being vaccinated has
been noted as early as one week following vaccination
and considering this point, is it really appropriate
in many of the effectiveness studies since they have
considered subjects as unvaccinated for the two weeks
CHAIRMAN OVERTURF: Dr. Levandowski, would
you like to address that issue?
DR. LEVANDOWSKI: All right, I'll try. I
was just ‑‑ I think I was trying to point out in that
early study in 1943, it was commented that they could
tell the difference between the people who were
vaccinated and unvaccinated as early as one week
Traditionally, we think about protection
as taking two weeks, mainly because it takes at least
that long for peak antibody titers to be achieved
after immunization and somebody who has been
immunologically primed and we believe strongly, at
least for the inactivated influence of vaccines, it's
the antibodies' direction against hemagglutinin and
that they are the most important of the protection.
I'm sorry if that seems a little confusing, but I
don't think we actually know exactly when protection
kicks in for any one person and on a population basis,
we would, I think, normally be expecting it should be
at least within about two weeks, but there may be some
effect earlier or it might be later in some
MS. BARR: Thank you.
DR. FREAS: Mr. Chairman, just for
clarification, I would like to say that the open
public hearing, it really is to address the Committee
and make comments before the Committee. We would
appreciate you holding questions until the end of the
discussion or the end of all presentations just in
case those questions may be answered in subsequent
CHAIRMAN OVERTURF: Are there any
additional public comments?
I think we will proceed then. Dr. Cox
will present the information on the world surveillance
of strain characterization.
DR. COX: Okay, I want to start out my
presentation by reminding you that CDC houses one of
four WHO collaborating centers for influenza and one