With the historical and regulatory background clarified, the Gender Workshop's three panels - Pharmacokinetics and Pharmacodynamics; Hormonal Influences; and Study Design and Analysis - were convened. During each panel, speakers briefly summarized pertinent research, issues and conclusions. A moderator-led panel discussion followed each set of speaker presentations, allowing relevant subjects to be developed further. The discussions of the three panels and their recommendations for assessing gender differences during medical product development are summarized below.
Panel 1: Pharmacokinetics and Pharmacodynamics
Moderator: Alan Sedman, M.D., Ph.D.
Panel I speakers discussed the following issues: Review of Published Data Regarding PK/PD by Dr. Janice Schwartz, Regulatory Aspects of Gender Based PK/PD Differences by Dr. Mei-Ling Chen, Examples of Gender Based PK/PD Differences by Dr. Raymond Woosley and A New Approach to Drug Development by Dr. Lewis Sheiner.
The pharmacokinetics and pharmacodynamics panel discussion focused on possible clinical development strategies to assess gender effects, using nonclinical data for two hypothetical drugs, termed A and B.
Drug A: Drug A is an inhibitor of plasma renin activity in development to treat hypertension. In hypertensive rats and dogs, this compound demonstrated gradually increasing activity over a broad range of doses. Toxicity was not evident in animals at doses several orders of magnitude greater than those expected to be therapeutic. Oral drug absorption was highly variable and absolute bioavailability averaged five to 10 percent in animals. Binding to human plasma proteins was negligible. Renal elimination of unchanged drug predominated but approximately 30 percent of absorbed drug was eliminated in urine as metabolites. Human liver preparations suggested that CYP2D6 is a potentially important isozyme in the metabolic elimination of the drug. Gender differences in animal toxicokinetic studies were not observed.
Drug B: Drug B is a selective Ml agonist being developed to prevent or provide symptomatic treatment of Alzheimer's disease. In preclinical tests in rats and monkeys considered predictive of cognition in humans, the compound markedly enhanced performance over a narrow range of doses. Intractable seizures were evident in animals at doses six to eight times above those expected to be therapeutic. Oral drug absorption was very reproducible and absolute bioavailability averaged 80 percent in animals. The drug was extensively bound to human plasma alpha,-acid glycoprotein (97%) and was metabolized to numerous moieties, with less than 0.5% of the dose recovered unchanged in urine. Human liver preparations suggested that metabolism occurs via CYP1A2 and CYP3A4 isozymes. In rats, marked gender differences were observed in toxicokinetic studies. The plasma concentrations of unchanged drug in females were six times greater than those in males. Approximately two-fold differences in the same directions were observed in monkeys.
Panelists discussed the preclinical studies for both Drug A and Drug B, noting that Drug A studies suggested little potential for gender-related clinical differences while the preclinical data for Drug B suggested that responses in men and women were likely to differ. Both cases demonstrate the value of preclinical studies to assist the efficient design of further clinical investigations In vitro plasma protein binding and hepatic tissue metabolic profiles can be utilized to determine the likelihood of significant gender-related differences in human Pharmacokinetics. Pharmacokinetics, pharmacodynamics, pharmacology, and toxicology information in both male and female animals can also suggest potential gender-related differences in concentration response, safety, and/or efficacy.
The panel recommended use of animal models with metabolic pathways similar to humans when possible. Attempts should be made to develop preclinical models to understand the gender-related pharmacokinetic and pharmacodynamic differences observed in humans. Acceptable preclinical models that assess the effects of hormones on drug metabolism, pharmacokinetics, or Pharmacodynamics are still lacking. Preclinical and in vitro information regarding drug metabolism, such as the cytochrome P450 isozymes responsible for biotransformation, should be especially useful in determining whether drug-drug interaction studies between a new drug and hormone replacement therapy or oral contraceptives are Rely to demonstrate significant potential for metabolic inhibition. For drugs to be used by both men and women, the panel recommended that both genders be included in early clinical trials to screen for potential differences in pharmacokinetics, pharmacodynamics, tolerance and/or efficacy. If preclinical data suggest little potential for gender effects and no differences are observed in initial human trials, additional trials specifically to explore gender effects may not be indicated. If preclinical or early clinical information suggests a high likelihood of gender differences, a further series of carefully designed studies should be considered to determine whether the drug needs to be used differently in men and women. Such studies would usually be performed prior to large pivotal efficacy studies and would carefully compare the pharmacokinetics, Pharmacodynamics, safety, and/or efficacy of the compound in men and women, possibly distinguishing between pre- and postmenopausal women. For almost all drugs, the panel believed it would be useful for phase 2 and 3 clinical pharmacology databases to be analyzed using population approaches to look for potential gender-related differences of clinical significance, as well as for age, race, and disease state, on response, unless discrete studies of these factors have been already conducted and analyzed.
In summary, the panel concluded that the potential for clinically significant, gender- related differences in pharmacokinetics, Pharmacodynamics, safety, and/or efficacy should be considered early in drug development. Preclinical and early clinical information is critical for optimal design of early and late phase clinical trials and to determine whether special studies in humans are needed to define gender-related differences in clinical response. All studies in all phases of drug development should be carefully designed to answer specific sets of questions, rather than based on standard protocols and test methods.
Panel I Recommendations
- Acceptable preclinical methods to assess the effects of gender and gender-related factors, such as hormonal influences on drug metabolism, pharmacokinetic, or pharmacodynamics, are still lacking and should be developed.
- Clinical pharmacology databases should be examined to identify potential gender-related PK/PD differences of clinical significance.
- Where appropriate, phase 2 and 3 databases available to FDA should be analyzed to look for gender effects, as well as the effects of age, race and disease state, on safety and efficacy, unless discrete studies of these factors have already been conducted and conclusions drawn.
- Pre-clinical and early pharmacokinetic and/or pharmacodynamic information should be provided to allow optimal design of subsequent clinical trials and to determine whether special studies are needed to define gender-related differences in clinical response.
- Studies in all phases of drug development should be carefully designed to answer specific sets of questions based on prior knowledge of the intended use and therapeutic range of the drug under investigation.
Panel Discussion II: Hormonal Influences
Moderator: Jean P. Rowan, M.D.
Panel II speakers discussed the following issues: General Physiology of Female Hormones over the Lifespan by Dr. Edward Wallach, Important Hormonal Influences on Drug Metabolism by Dr. David Flockhart, Hormonal Influences on Metabolism of Cardiovascular Drugs by Dr. Paresh Dandona, Hormonal Influences on Metabolism of Psychotropic Drugs by Dr. Jean Hamilton and Methodologies for Studying Hormonal Influences by Dr. Scott Lukas.
There is substantial data to show that endogenous and erogenous steroid hormones affect the kinetics and dynamics of drugs. The panel on hormonal influences structured its discussion around four questions as it tried to determine the clinical relevance of these PK/PD changes:
- What mechanism of drug action should prompt an investigation of hormonal influences?
- What enzyme systems are apt to be influenced by changes in hormonal environment, and thus impact the safety and efficacy of a drug?
- In clinical trials, what findings suggest hormonal influences may be active?
- Should studies on endogenous and erogenous hormonal influences be conducted as phase I studies, or confined to studies in women under treatment, in phase II and III trials?
The panel concluded that pharmacodynamics, rather than pharmacokinetics, more accurately predicts clinically important safety and efficacy effects. It also agreed that the mechanism of action of the gonadal steroids is not well understood, and that side effects from drug therapies may often be key indicators of an underlying hormonal influence. The panel noted that therapeutic agents can have a variety of unanticipated effects on estrogen-responsive tissues. Often, such tissues are not traditionally considered "target tissues" of either the gonadal steroid hormones or the agents themselves; blood vessels are an example. The panel also discussed the effect of the various isoenzymes in the CYP 450 system - for example, the 1A and 3A systems. While there are relatively small changes (approximately 20 to 30 percent) for the cytochromes under discussion, these cytochromes have a broad range of metabolic activity, e.g., the 1A and 3A system, at the extreme end, the difference between women and men is relatively large and might well be clinically relevant. This is an important area for development of preclinical screens of drugs using the CYP 450s of steroidogenic enzymes, similar to those used for hepatic CYP 450 isozymes, to study how drugs can interfere with the endocrine system.
Side Effects: The panel noted that certain drug side effects may indicate an underlying hormonal interaction. A variety of mechanisms need to be examined carefully, and compartmentalized by possible site of action to unravel the relationships that occur and to determine the differing effects that may be produced by the same drug. Drugs that have a small margin of safety, or serious adverse effects, warrant special scrutiny.
The tendency to ignore what patients say, particularly in terms of side effects, was noted as troubling. As the data presented at the workshop demonstrate, women have vascular systems that are, in many ways, different from men and that may result in such side effects as light-headedness, headaches, and nausea. Another side effect, and one that is often overlooked when drug effect is considered, is change in the menstrual cycle, such as amenorrhea or excessive bleeding. Mood swings and depression may also suggest a hormonal response that merits a closer look.
The data also suggest that women are more likely to have side effects if the vasodilatory effect associated with the drug is very great. In that context, it is important to ask whether there are any data to show that the flushing, dizziness, throbbing headaches, and ankle swelling associated with certain drugs, such as calcium-channel blockers, are more common in women. This may prove relevant to a finding several years ago that grapefruit, which contains a flavonoid called naringin, was shown to inhibit the metabolism of nifedipine, increasing the drug's bioavailability. This interaction may produce more marked side effects in women. The possible role of estrogen in this reaction needs to be examined.
Studies with the dihydropyridine calcium channel blocker, amlopidine, found three drug-related side effects - flushing, edema, and palpitations - were considerably more common to women than to men. Used at a dose of 15 milligrams, amlopidine gives a 30 percent rate of edema, which is high By contrast, a 10 milligram dose has a low edema rate. Possibly, smaller body size puts women a little further out on the dose response curve, but other basic physiological or metabolic factors may also be involved.
Unanticipated Drug Effects on Estrogen-Responsive Tissues: A variety of mechanisms need to be examined carefully and compartmentalized by the site of action. Many agents exist with unexpected side effects on the reproductive system. One important realm of scrutiny is the central nervous system, and drugs acting at the level of the central nervous system to influence hypothalamic function. Signals governing hypothalamic synthesis, storage, and secretion of GnRH involve adrenergic, serotonergic, and dopaminergic mechanisms that influence GnRH output and, ultimately, ovulation. One possible example are the psychopharmacologic agents that give rise to hyperprolactinemia at this level, and ultimately suppress ovulation, causing amenorrhea and bone loss.
Over-the-counter anti-inflammatory agents provide another example of unexpected side effects on the reproductive system. The mechanism of ovulation has been Uened to an inflammatory process by a number of researchers, with release of histamine, prostaglandins, especially F2 alpha, and oxygen-free radicals. Nitric oxide has also been implicated. Cyclo-oxygenase inhibitors and anti-prostaglandins are commonly used over-the- counter agents that may, in certain dosages, influence ovulation by acting locally at the follicular level. Another unexpected cross-reactivity between drugs and hormones may occur when drugs and hormones share, and compete for, the same receptors. For example, spironolactone taken by pregnant women competes for, and blocks, androgen receptors, profoundly influencing intrauterine development of male external genitalia. In a related example, tamoxifen blocks the estrogen receptor when given to treat breast cancer but also acts on estrogen receptors in the endometrium to increase the risk of endometrial cancer. The manufacturer claims tamoxifen also supports non-destruction of the bone.
The mode by which estrogen acts at the bones is not precisely understood. Mononuclear cells from post-menopausal women generate more cytokines, such as TNF-alpha, IL-1 and IL-6, than from pre-menopausal women. When post-menopausal women are treated with estrogens, their cytokine levels return to normal. This is interesting because the circulating monocyte macrophage is a cousin of the osteoclast, which is a phagocytic cell that breaks down bone. Moreover, the so-called osteoclast activating factor has been shown to be TNF-alpha.
The steroidogenic CYP 450 enzymes are certainly affected by drugs, including ketoconazole, which inhibits steroidogenesis. Recent clinical evidence also points to griseofulvin as affecting steroid metabolism. Women treated with griseofulvin for toenail fungal disorders develop dysfunctional anovulatory bleeding, which stops when griseofulvin is discontinued. Post-menopausal women on hormone replacement therapy also experience this side effect.
Finally, drugs that affect vascularization or neurovascularization may interfere with corpus luteum function, since the function of the corpus luteum is to develop a transient neurovascular structure from which progesterone is transported.
Clinical Endpoints in Pharmacodynamics: The discussion regarding the nature of clinical trial findings, which might suggest that hormonal influences are active, reflected a broad understanding of the role, and mechanism of action, of the gonadal steroids. The knowledge that erogenous estrogens in the postmenopausal woman affect many different systems is apparent in the treatment of vasomotor symptoms, vaginal atrophy and dryness, and osteoporosis. Strong epidemiological and clinical information, based on surrogate endpoints, demonstrates that estrogens are active in cardiovascular health and in preventing Alzheimer's-type dementia. If exogenous estrogens in the postmenopausal period are active in all these systems, then endogenous estrogens almost certainly are active in those systems prior to menopause. As such, endpoints become dynamic indicators of hormonal influences. Certainly, when studying drug mechanism of action, and drug activity in disease areas such as those just mentioned, both endogenous and erogenous sources of hormone must be considered.
The menstrual cycle is such a sensitive mechanism during reproductive life that menstrual irregularity may be an early indicator of a drug effect. Often, information about a subject's menstrual cycle is sparse or unavailable but it is a potentially useful data source that should perhaps be collected more rigorously. In a woman, the clinical evidence to examine are menstrual irregularities, amenorrhea, failure to conceive despite attempts to do so, and conception and early pregnancy loss, especially if that loss is associated with a structural or chromosomal abnormality. Because males have no comparable indicator, it may take longer to know whether a drug has a damaging impact on the hypothalamus and pituitary.
One panelist emphasized the need to look for the source when results differ between men and women. Likewise, any time a statistician finds considerable variability, gender effects and underlying hormonal mechanisms should be considered. Although pharmacokinetic differences, per se, are not a reason to pursue gender studies, a disassociation between the pharmacodynamics and the pharmacokinetics suggests there may be an underlying hormonal difference, or an interaction of endogenous or exogenous hormones, of interest. Any changes in drug efficacy, or anything with a marked effect on dose ranging or the side effect profile, should immediately lead to an exploration of hormonal influences.
As clinicians, there is a great need to look at new ways to optimize treatment. In an emerging area such as gender effects, it is important to keep an open mind, and pursue investigation, even when some of the effects initially seem modest. Dosing changes of anti-depressants during menstrual cycle illustrate the need to define clinically sensitive sub-populations, and optimize therapies for them. It is also important to look at variations within each gender group and across genders. In many dose response trials, it's not easy to identify patterns, either in measures of effectiveness, or in side effects. They are probably most significant when the dose response curve is reasonably steep within the therapeutic range. Where titrated doses are used, each patient becomes his or her own dose-response trial and the requirements for eliciting gender differences become very different.
Timing of Gender Studies: The panelists discussed whether studies of endogenous and exogenous hormonal influences should be conducted as phase I studies, or confined to phase 2 and 3 trials. They agreed that case-by-ease decisions are needed and must be driven by the science. As the knowledge base expands, and the in vitro and in vivo areas in which generalization is, and is not possible becomes clearer, studies can be expected to be more focused. While more definitive generalizations may be possible in five years, a few guidelines were proposed by the panel:
- If there is primary metabolism through CYP 450 la or 3a, kinetic effects are more likely, and gender and hormonal influences should perhaps be considered in phase I. Other factors also come into play, including high first pass metabolism through the liver, and great variability in bioavailability.
- If there is a very strong suspicion that hormonal effects exist, from metabolic and enzymatic standpoints, as well as from prior experience in that class of drugs, it may be highly desirable to establish those effects early in drug development.
- Because the diverse side effects that indicate hormonal influence are so much greater in women than in men, they can be very difficult to predict and use in phase I testing. From that perspective, phase 2 and 3 may be more appropriate places for testing hormonal influences.
- There is a need to study drug interactions, particularly for drugs that have predictable effects, such as the oral contraceptives, hormone replacement therapy, danazol for endometriosis, and estrogen therapy for prostate cancer.
- Menstrual cycle data should be collected routinely from female subjects.
Panel II Recommendations
- Metabolic pathways should be described during the preclinical phase of development.
- If metabolism of an investigational agent occurs primarily through CYP 450 IA2 or 3 A4, pharmacokinetic differences due to hormonal influences are more likely; this influence should be examined in early drug testing.
- Although blood vessels have not traditionally been considered target organs for estrogen action, a number of studies support the finding of intrinsic vascular regulatory mechanisms that are responsive to estrogens. Further research in this area should be conducted.
- The mechanism of action of the gonadal steroids is not well understood. Side effects from drug therapies may often indicate an underlying hormonal influence. Drugs can have a variety of unanticipated effects on estrogen responsive tissues. Therefore, pharmacodynamic differences between genders, rather than pharmacokinetic differences alone, should be studied because the former predict clinically important safety and efficacy effects more accurately.
Panel Discussion III: Study Design and Analysis
Moderator: Murray Lumpkin, MD.
Panel III speakers discussed the following issues: Subsets: Opportunities and Cautions by Dr. Robert Temple, Statistical Aspects of Assessing Gender-Related Treatment Specificity by Dr. Richard Simon, Survey of Gender Analyses in the NDAs for Selected New Molecular Entities in 1993 by Dr. Stella Machado, Methodological Considerations for the Safety and Evaluation of Gender Differences in the Clinical Evaluation of Drugs by Dr. Robert O'Neill, Prospective Studies: Issues and Rationale by Dr. Roger Williams, Pharmacokinetic Screen: What Can Be Accomplished by Dr. Jean Louis Steimer, FDA Gender Guideline: Perspective of an Industry Statistician by Sarah JH Kogut and Special Consideration in Using Cardiovascular Devices in Women by Dr. Julie Swain.
In an effort to learn more about study design and analysis, this panel focused on four key questions:
- What types of results in early phase studies imply the need for more intensive evaluation of potential gender differences in phase 3?
- When and how should trials be designed to determine if different dose regimens are needed in women?
- When are separate, fully powered trials needed to gain safety and efficacy information about gender differences?
- Does it make a difference in planning, if a potential difference in safety, rather than in effectiveness is possible?
Panelists agreed that early indications of gender differences in PK studies are signals that gender effects should be considered in later trials. Any PK non-linearity difference is a cause for concern although it is rare that investigators will study different regimens in men and women to determine this possibility. In general, clinicians assume there will be no efficacy differences by gender unless available data shows an important interaction.
There was broad agreement that very few studies are designed to optimize the dose in individuals, that most studies are analyzed by group data, and that investigators need to pay greater attention to dose-response at all phases. Given the complexity of the related issues, the value of bringing together people from different disciplines and with differing perspectives was noted by several panelists. The dose-response issue becomes most important for drugs with narrow therapeutic indices, especially those accompanied by steep dose-response curves. Such situations generate a lot of overlap in the data and are characterized by many confounding variables that make it hard to determine dose- response differences. Very often, PK and PD screens can help sort out these differences. Such analyses may be performed using sparse data from late phase studies to yield valuable information about gender and other effects that would be otherwise difficult to determine.
The relative merits of determining dose response in phase 2 and phase 3 trials were discussed extensively. An industry spokesperson said that one reason for determining dose-response only late during phase 3 trials was to respond to the pressure both from the public and from private industry to develop drugs as quickly as possible. From industry's perspective, it is important to conduct studies in a certain sequence in order to acquire the information necessary to move onto the next step. If dose estimates are not established until later in drug development, this approach saves time, although it also involves risk. If there is some reason to expect toxicity, and the dosage is not going to be titrated, perhaps because the adverse effects are sudden, securing good dose findings in phase 3 trials becomes especially important. Another speaker noted that while every available pharmacologic index or indicator of potential effectiveness should be used in phase 2, these studies are unlikely to identify the correct dose because the endpoints of interest are not used and the trials usually aren't sufficiently large. For example, the dose of drugs designed to increase cardiac output or reduce symptoms of heart failure are not necessarily the doses that produce the best survival outcome. In addition, safety findings during phase 2 are too infrequent to yield much information about gender-specific differences.
Nonetheless, there are many limitations to doing dose-response studies in phase 3 and some arguments for gathering the information earlier. Some panel members noted that it is more cost-efficient to determine dose-response relationships in phase 1 and phase 2 and that this allows for a fuller understanding of safety and efficacy profiles prior to phase 3. Panelists generally agreed that as much dose-response information should be collected as early as possible. Certainly, the pharmacokinetic screen should not be confined to late phase trials, especially because PK data obtained earlier can provide significant information relevant to decisions about dose-response. On a related topic, panel members urged sponsors to encourage more communication among the clinicians, scientists and statisticians involved in both phase 2 and phase 3 studies. The necessity of enhanced communication was highlighted by two recent trials that resulted in unexpected findings at the end of phase 3. This may have been avoided by a more careful review of phase 2 studies, where the data clearly showed significant blood level differences between study groups.
In reality, most studies are not classical dose-response studies but can best be described as "find me a winner among one of the doses." For example, in studies of nucleoside analogues, the low dose was preferred because of toxicity at high doses. In terms of individualized dosing, dose-response studies might be designed to use two different high doses and two low doses, for males and females. This provides a variation on dose-response for males and females, since one need not look for a smooth curve but rather a curve that maximizes the efficacy/toxicity ratio.
Despite the challenge of determining individual dose response relationships, several concrete strategies were proposed. Dr. Sheiner has written about an approach to dose finding in which individuals are titrated to a therapeutic endpoint. This technique can be applied to conditions that are relatively responsive and reversible. Careful analysis, using NONMEN, is required when trying to relate dose to response, because those receiving higher doses tend to fare less well. This approach offers an opportunity to develop an initial idea of the dose range and may offer advantages over trials in which a single dose is given to all subjects. A specific goal, therefore, is to develop study designs that generate interpretable ways of titrating dosages.
Another proposal for determining the proper dose level is to develop a table showing the percentage of patients who gain the intended benefit, with an acceptable toxicity level, at each dose level or range of dose levels. Such a table can be used for all patients, hopefully including a representative sampling of women and men and some ethnic diversity. If dose levels prove not to matter much, and a fairly high number of patients benefit from the drug, then a single fixed dose level can be chosen. If the dose ranges vary substantially but patients benefit from the drug, the information can be used to establish an appropriate dose level.
Panelists also suggested the savings generated by greater flexibility with dosage should be considered. Greater flexibility in dosing, as defined during studies of gender-based dose differences, may provide savings of benefit to practitioners and consumers.
Another possible benefit is the opportunity to use advertising to highlight dosing that reflects gender-specific differences in efficacy. The need for early recognition of gender differences in device use was also discussed. In general, little attention has been paid to these differences by either the device industry or the FDA. Panelists called for the FDA to take the lead in promoting awareness of this issue, perhaps by issuing guidelines.
Panel III Recommendations
- More complete by-gender tabulations and graphics of safety and efficacy results are needed, as demonstrated by the results of an internal FDA study showing that three of 15 new molecular entities approved in 1993 did not include a gender analysis. In addition, more analyses of combined data sets are needed and new ways should be found to facilitate meta-analyses when combining studies with varying designs.
- Although, suitable study designs allow subset differences in safety and effectiveness to be identified and rational dose labeling provided, strategies for improving the methodology are needed. Early indications of gender differences in PK studies should signal the need to examine gender effects more closely in subsequent trials and offer the prospect of improving late phase clinical trial designs, possibly saving time and money.
- Good population and individual dose response curve information for safety and efficacy effects is critical to establish the need for different dose regimens between genders. Gender-specific dose response information can be determined in phase 2 and phase 3 studies.
- Gender analyses in large-scale clinical drug trials have often been performed as tests of "interaction" in the analysis of variance. Such tests have low power, compared to tests of the main treatment effect, i.e., the study would have to be at least four times its planned size to achieve the same power. More information is obtained when separate statistical tests are performed on mates and females, but even these tests are less powerful than the test for the effect in the whole population. Thus, it will not be unusual to fail to see a significant effect in either subgroup. The panel recommended that confidence intervals for the treatment effect should be calculated for males and females separately, and compared with the confidence interval for the treatment effect for the whole population. Graphs of the confidence integrals are helpful to show consistency between genders and the amount of certainty in the findings are readily apparent. Bayesian approaches may also be useful in this context.
- When combined with phase 3 and other data, population PK/PD methods may be useful to enhance understanding of gender effects, and in particular, how gender, age, race, and other patient factors may be interrelated.
- Generally, designing a larger clinical trial or combining databases can provide more definitive subset information from trials. Although the quality of inferences drawn from a single study may be better than from a combined data base of equivalent size, there are a variety of reasons that enlarging the size of a single study to increase the power of separate analyses may be less attractive. Relevant factors include the drug and the condition being treated and certain ethical constraints may also come into play. For example, the possibility of an early halt to a mortality study may preclude collecting sufficient data on both genders. On the other hand, meta-analyses should be conducted with care, especially when the studies to be combined have different designs. Graphical methods are a helpful adjunct to meta-analyses.
- Where information about potential safety and efficacy gender effects of drugs has been identified, it should be better communicated among sponsors, FDA review- ers, health care providers, and patients.
- Earlier recognition of gender differences during the evaluation of medical devices is needed. FDA should take the lead to help promote awareness of this problem.