Comments on FDA Proposal on Labeling Requirements for Pregnancy and Lactation Drugs

August 27, 2008

Re: Content and Format of Labeling for Human Prescription Drug and Biological Products: Requirements for Pregnancy and Lactation Labeling (Docket No. FDA-2006-N-0515)

Division of Dockets Management (HFA-305)
Food and Drug Administration
5630 Fishers Lane
Room 1061
Rockville, MD 20852

To whom it may concern:

Public Citizen submits the following comments in response to the Food and Drug Administration’s (FDA’s) Proposed Rule on pregnancy and lactation labeling for drugs and biologics. While the Proposed Rule contains many improvements over the current haphazard approach to risk disclosure, we are concerned that the heightened emphasis on human studies, many of which are poorly designed and/or underpowered, might provide false reassurance regarding the safety of these products. We provide comments only on the pregnancy section of the Proposed Rule.

The current system of pregnancy labeling consists of five letter-categories (A, B, C, D, and X, in order of increasing risk) based primarily on animal studies, but also on human studies and theoretical considerations. The FDA now proposes replacing this system with “Fetal Risk Summaries” that include a one-sentence fetal risk conclusion, followed by a separate narrative explanation of the human and animal data underlying the Summary. The new label would give primary importance to human studies, if they exist, while downgrading the emphasis on animal data.

Limitations of Human Data

Of course, if the human data are of good quality, such data would be central to the evaluation of the risk posed by these products. But the essential problem is that, for a large proportion of drugs, the quality and statistical power of the human data fall well short of the desirable. Consequently, human data must be accompanied by clear acknowledgment of their deficiencies. It is important to remind readers that, unlike animal studies, these are not randomized, controlled trials, that they usually include only small numbers of women, and have limited follow-up. Very often the available studies conflate different doses and durations of use of the drug and do not clearly state when exposure occurred, even though these are critical to evaluating risk. Moreover, while prospective studies typically do an adequate job of determining the presence or absence of particular teratogenic effects, registries and retrospective cohort studies are much less certain to have done so. Very few studies include follow-up long enough to rule out late-manifesting effects such as behavioral abnormalities. Thus, while the human studies have obvious advantages (and attractiveness to those intimidated by animal studies), they bring with them deficiencies not associated with animal studies.

Of the problems with human reproductive studies mentioned above, the most concerning is that such studies are typically underpowered. Most studies have fewer than 300 subjects, yet with a baseline birth defect incidence of 3% (and this is for all defects, not particular ones), such a study has only 85% power to detect a relative risk of 2 (alpha = 0.05, one-sided). Most patients will find little solace in knowing that all that has been excluded is a more than doubling of their risk. It is worth recalling that cigarette smoking, an accepted risk factor in pregnancy, is associated with “only” a 1.34-fold increase in the risk of birth defects.[1] (Of course, smoking has other adverse effects on the fetus.) Societal and medical admonitions against smoking during pregnancy are very strong.

The solution to this problem is to make clear the limitations of any human studies cited. These should include mention that the study was not randomized, that there may have been problems of ascertainment (especially for registries), how long the follow-up was and whether the study combined exposures of different doses, durations and weeks of gestation. Most important, for studies that fail to find an elevated risk associated with the drug, a sentence with the following formulation should be included: “While a study of XXX human exposures to the drug did not show an elevated risk of birth defects compared to background rates, this study only excludes a YYY-fold increase in the rate of birth defects due to the drug.”

The FDA also seeks comment on the degree of detail to be included the animal studies section of the data component of the label. Details on numbers of species, consistency of the findings and the extent of the elevated risk for the most important defects should all be provided. The level of detail should be similar to that provided for the human studies. To not do so prevents those with sufficient expertise from properly weighing the animal data, particularly when in many cases human data will be missing or deficient.

Finally, companies are unlikely on their own to seek better human data.  It is therefore incumbent on the FDA to make the collection of human data on fetal damage a more common element of pre- or post-approval studies.

Clinical Aspects

The FDA proposes boilerplate language that asserts “All pregnancies have a background risk of birth defect, loss, or other adverse outcome regardless of drug exposure.”   Under Clinical Considerations, it provides fairly detailed reasons why the use of the drug may be justified in pregnancy (e.g., “Asthma complicates approximately 1 percent of all pregnancies resulting in higher perinatal mortality, low birth weight infants, preterm births, and pregnancy-induced hypertension compared to outcomes for nonasthmatic women. Because of the risks of even mild maternal hypoxia to the developing fetus, asthma should be clinically well-controlled during pregnancy.”)   The inclusion of such statements is deeply ironic, for the FDA has long held to the claim that it does not engage in the practice of medicine. Yet here, in the pregnancy section of the label, it makes a spirited case for the usefulness of the drug. The agency has repeatedly presented itself as simply a provider of data clinicians can use to make informed decisions; we recommend that the agency adhere to that role in this case.

The agency would also use such vague terms as “low,” “moderate,” and “high” risk of developmental abnormalities (see p. 30842 of Federal Register notice).   We do not believe that these terms are precise or that individual clincians or patients would interpret them similarly. Finally, while it is certainly true that a decision to use a drug during pregnancy should be made in the context of the effectiveness of the drug and the seriousness of the condition being treated, the Proposed Rule does not consider whether an alternative drug may be less likely to be teratogenic.

Procedural Considerations

Because the new label will consist largely of narrative descriptions of studies, a critical part of the process will be who decides which data sources get cited, how critically they are analyzed, and what weight they are given. The majority of published human analyses of safety done by Teratology Information Services (e.g., OTIS and Motherisk) conclude that most drugs are without risk.

The synthesis of the new label will be a time-consuming process. It is not clear who will be responsible for writing/reviewing: industry or the FDA. Who will choose which studies to include and make the judgments as to the seriousness, reversibility, and correctability of an abnormality? Will the FDA provide training for staff to analyze the literature, adverse event databases (U.S. and international), pregnancy registers, and animal studies? Will these staff have the expertise to critically evaluate statistics, medicine, physiology, and toxicology? Will the FDA call on outside experts, and who will identify them?

Problems concerning the sample labels provided by the FDA

Example 2: Drug for which only animal data are available; lack of developmental toxicity findings
The Fetal Risk Summary that the drug “is not predicted to increase the risk of developmental abnormalities” is based on a study in rats and rabbits treated at doses only 1.5 times the maximum adjusted human dose. There is no information to alert the reader that the use of doses this low does not constitute a valid test, and, thus, in addition to no human data, there are also no valid animal data.

Example 3: Drug for which animal and some human (insufficient) data are available
The Fetal Risk Summary states that “the risk of major congenital abnormalities or spontaneous abortions is low,” a conclusion based on less than three weeks of exposure in the first trimester in an unknown number of women. Only if one takes the time to go the data section below does one find the reference and learn the power of the study. Will readers make this effort in practice? Will the reader be able to interpret the conclusions presented? Finally, how does one interpret the statement that, based on animal data, the risk of developmental abnormalities is predicted to be “moderate”? Do physicians and patients have similar interpretations of this term?

Example 4: Drug for which sufficient human data are available
The Fetal Risk Summary states that “Human data do not indicate that DELTAMAN increases the overall risk of major congenital malformations or neural tube defects.” This conclusion is based on a pregnancy exposure registry and a retrospective cohort study of a group of women using this drug and “pharmacologically similar drugs.” Although the label states that the registry has a 90% power to detect a 2.5-fold increase in major malformations and only an 80% power to detect a 10-fold increase in the rate of neural tube defects, we do not know doses, duration of use, or when exactly exposure occurred during the first trimester. Exclusion of a 2.5-fold increased risk does not merit a conclusion that there is no risk, the conclusion most will draw from the Fetal Risk Summary.

Completely ignored in the Summary are the data from three animal species (rats, mice, and monkeys), all of which demonstrated a dose-dependent increase in neural tube and skeletal anomalies. Webster and Freeman state that where there is “an increased incidence of one or more birth defects in two animal species and evidence of a dose-effect … [s]uch findings are treated as though the drug would cause birth defects in the human regardless of the pharmacokinetic data.”[2] The proposed FDA algorithm states that human data outrank animal data, apparently even where there are compelling data from animals that contradict those from humans. Finally, this is a drug for gastro-esophageal reflux disease (GERD), for which changes in diet and behavior might be all that are needed.


Predicting the safety of drug use in pregnancy is a very inexact science, particularly due to the relative absence of well-designed and –conducted human studies. Using cryptic Fetal Risk Summary statements, basing safety on a potentially selected set of human studies, and downplaying the usefulness of animal data, fails to provide pregnant women and their health care providers with the information the FDA is required to provide by law.

Yours sincerely,

Elizabeth Barbehenn, Ph.D.
Research Analyst

Peter Lurie, M.D., M.P.H.
Deputy Director

Sidney M. Wolfe, M.D.
Health Research Group at Public Citizen



[1] Shepard TH, Brent RL, Friedman JM. Update on new developments in the study of human teratogens. Teratology 2002;65:153-161.

[2] Webster WS, Freeman JAD. Is this drug safe in pregnancy? Reproductive Toxicology 2001;15:619-629.