Teratology Society

Dockets Management Branch
Food and Drug Administration
HFA-305, Room 1-23
12420 Parklawn Drive
Rockville, MD 20857

On November 11, 1997, Lilly filed comments to the docket for the Pregnancy Labeling Hearing. It has now come to our attention that an inadvertent mistake was included in one section of those comments and Lilly wishes to correct that mistake with this letter. In Attachment 1, on page 9 of our comments, there is a subsection entitled "Human Data". In this subsection, there is a sentence that begins, "Three additional pregnancy surveys (195 total pregnancies)...". This sentence should read, "One additional pregnancy survey (81 total pregnancies)...". The evidence provided by the two additional surveys is not as strong as that provided by the prospective studies, the one survey being included, and the pregnancy registry. Thus, in keeping with the intent of Lillys comments, these two additional surveys should not be included in labeling. For your convenience, a revised copy of our comments including this correction is included with this letter.

Dockets Management Branch
Food and Drug Administration
HFA-305, Room 1-23
12420 Parklawn Drive
Rockville, MD 20857

Eli Lilly and Company (Lilly) is a global research-based pharmaceutical corporation dedicated to creating and delivering innovative pharmaceutical-based health care solutions that enable people to live longer, healthier, and more active lives.

As part of our commitment to pharmaceutical-based health care solutions, we are committed to providing informative and understandable labeling for our products. We believe that information regarding safety of medication that might be used during pregnancy is of great importance.

The issue of pregnancy labeling for prescription drug products involves several perspectives: womens health, medical, toxicology, regulatory, and legal. Two representatives of Lilly, Ms. Judy Buelke-Sam and Dr. Freda Lewis-Hall, presented Lillys views at the September 12, 1997 FDA hearing on content and format of pregnancy labeling for human prescription drug products. Lillys testimony identified four major issues as important to an overall summary of potential reproductive and developmental risk: 1) both animal and human data should be provided; 2) in addition to a narrative text, the labeling should include comparison of replicable or consistent data across reproductive and developmental stages between animal and human findings; 3) a means of evaluating the "quality" of the data should be provided in the labeling; 4) a means for more frequent updating of the labeling as new data, especially human data, become available should be determined. Lilly is submitting these comments to supplement that testimony.

The current labeling regulations provide for a narrative summary on reproductive and developmental risks in several separate sections under "Precautions": Carcinogenesis, Mutagenesis, and Impairment of Fertility; Pregnancy Category, A, B, C, D, or X; Labor and Delivery; Nursing Mothers. The current pregnancy precautions were intended to provide only a brief summary of certain available data on the safety and effectiveness of the drug product in pregnant women and were intended to encourage prevention of pregnancy while taking medications. The Commissioner concluded that each required pregnancy precaution statement, together with a description of the data upon which it is based, is a reasonable expression of the limitations of the data. The pregnancy categories were felt to be necessary to provide consistency in prescription drug product labeling through clear and concise guidance statements concerning the safe and effective use of prescription drug products in pregnant patients. The pregnancy conclusion statements on the use of a drug products are required in labeling when they help to describe the basis under which a drug product is placed in a particular category. However, review of current practice has led to the conclusion that the pregnancy categories often have been assigned inconsistently, misinterpreted both by health care providers and patients, and are based on limited information. It is often difficult to determine what aspects of reproductive or developmental outcomes have contributed to the assigned category for any given drug product. Further, it is particularly difficult to assess both the quality of the data resulting in concern and the severity of the observed animal and/or human outcomes that may lead to concern.

Lilly supports the Agencys interest in improving the current labeling scheme and submits the following comments to assist in identifying and implementing improvements which will optimize physician and patient use of labeling in the decision-making process. Two of Lillys marketed compounds, Prozacâ (fluoxetine hydrochloride) and Humalogâ (insulin lispro), are used to illustrate the range of information available for possible inclusion in product labeling. Throughout the process of updating these regulations, we suggest maintaining flexibility in providing information as reflected in these examples, since the most relevant means of expressing or comparing exposure or effect may be compound-dependent. We believe our examples provide an appropriate balance of uniformity and flexibility.

Current labeling format does not adequately address the issues that contribute to decisions concerning drug product therapy in women of child-bearing potential or in pregnant or nursing women.

The reproductive and developmental process is an interrelated and interdependent series of stages. This process raises many issues of concern to clinicians, including potential effects on fertility, spontaneous abortion, premature delivery, perinatal death/stillbirth, major malformations, birth weight/perinatal complications, postnatal development, and potential neonatal risk of exposure through breast milk.

Limited information concerning the effects of drug products on outcomes of the reproductive and developmental process appears in many separate sections of the labeling. Additionally, the summary of risks and benefits of treatment does not always take into account information on the entire reproductive and developmental process. The current pregnancy categorization deals primarily with fetal risk and traditionally has emphasized the adverse outcomes of death and malformations. In light of the limitations described above, we suggest a broader scope for the summary of potential reproductive and developmental risk. We would also include quality publicly available information regarding the risk of adverse gestational outcomes in both unexposed women with no medical condition and in unexposed women diagnosed with the medical condition which the drug product is intended to treat.

For example, mood disturbances during pregnancy may themselves result in adverse consequences for the offspring (Cooper et al. 1996). Women with mood disturbances are more likely to engage in activities that are also risk factors for adverse pregnancy outcome, such as smoking, drinking, and abusing drugs. Consequences of untreated depression during pregnancy may include higher rates of premature delivery, reduced birth weight, and increased incidence of neonatal complications (Steer et al. 1992). Increased fetal risks also have been demonstrated in diabetic pregnancies. The risk of major malformations is known to be increased in infants of diabetic mothers, with reported rates as high as 5.2% to 16.8% compared with 1.2% to 3.7% in infants of non-diabetic mothers (Kitzmiller et al. 1996). In one study, the risk for cardiovascular anomalies was increased 18-fold in infants of women with type 1 diabetes (Becera et al. 1990). These risks are associated with poor glycemic control. Although optimizing glycemic control prior to conception and throughout pregnancy reduces the risk of congenital malformations, it does not eliminate that risk.

In conclusion, we encourage alternative means of summarizing risk/benefits which provide quality information on the entire reproductive and developmental process, and we urge that both maternal and fetal risk associated with the medical condition itself be included as part of this overall summation, where publicly available. Physicians must have the data to weigh the alternative of not treating a medical condition, and the associated risk to the mother and fetus, against the potential risks of drug product exposure as part of their overall treatment recommendation.

Additional information for inclusion in labeling to address issues that contribute to decisions by clinicians and their patients.

The definitions of the current pregnancy categories refer to "adequate and well-controlled" studies in pregnant women. Medical and ethical considerations restrict the conduct of placebo-controlled, randomized clinical trials in pregnant women. Thus, for many medications, information is not available on risk to the mother and the fetus associated with treatment during pregnancy. In the absence of the traditional double-blind, placebo-controlled studies, data are available for many drug products in the form of published case reports, retrospective surveys, case-controlled surveys, prospective surveys, or cohort-controlled studies. Lilly recognizes as does FDA that there may be methodological or other flaws and biases in the available published data which limit their extrapolation to the treated population, quality published data are nonetheless useful.

While published case reports are useful for generating hypotheses concerning potential risk, they should not be used to imply a cause-and-effect relationship between a particular drug product treatment and adverse outcomes. Case reports, by definition, are based upon single observations without appropriate comparator or control observations. Consequently, any association between treatment and outcome may be due to chance, especially when the baseline rate of fetal and newborn complications in healthy mothers is considered. Use of case reports is particularly problematic when the underlying condition being treated (eg, diabetes) is associated with a high background rate of complications. Viewed in isolation, a case report may suggest an association where none actually exists or, conversely, obscure a true relationship between treatment and outcome. Nevertheless, such case reports are often referenced in the product labeling, potentially misleading the physician and patient concerning risk.

Similarly, published retrospective surveys and case-controlled surveys may have ascertainment bias and are better for generating hypotheses than for drawing conclusions. Published cohort-controlled studies usually include a control group and a larger sample size. However, they rarely include the comparator group required to meet the definition of an "adequate and well-controlled" study, ie, the placebo control group in which the medical condition has been diagnosed, and thus are rarely referenced in labeling.

An additional source of human data is the prospective survey, and in particular, prospective and retrospective pregnancy registry data. In this regard, Lilly has established a pregnancy registry to collect information on developmental outcomes of women treated with our products during pregnancy. For example, data from over 2000 women who have taken fluoxetine during pregnancy have been entered into this registry. As of April 1996, pregnancy outcomes from 759 fluoxetine-treated patients had been collected. Although there are limitations to the conclusions that may be drawn from information collected in this manner, it is an excellent complement to published human and preclinical data contained in the labeling. Unfortunately, the current categorization scheme and interpretation of the requirement for "adequate and well-controlled" studies has prohibited including this information in the fluoxetine labeling. Since the case-controlled studies and registry data are not included in labeling, regulations do not allow Lilly to disseminate this information directly to health care providers. While some of these data have been published in the medical literature, they are not readily available for use in the individual physicians and patients risk-benefit assessment.

Therefore, we strongly support a regulatory standard that would permit inclusion of quality published human data, in particular published cohort-controlled data and registry data, within the labeling summary of reproductive and developmental product information. We recognize that the quantity of available animal and human outcome data at each developmental stage may become extensive. We propose that for animal studies, precedence be given to regulatory quality (Good Laboratory Practice, GLP) data. In addition to GLP studies, the regulations should permit the drug product manufacturer to include in labeling quality publicly available outcome data relating to each developmental stage. Further, we propose that for human studies, published cohort-controlled data, when available, take precedence over prospective survey data. We recommend that published retrospective survey or case-controlled data be used only when no other data are available and that they be replaced as more robust, prospective data become available, as long as proper balance of outcome information is maintained. We further recommend that published case reports not be included in labeling.

An option for communicating risk and an alternative means of summarizing overall reproductive and developmental risk/benefit information.

There are two basic perspectives of risk communication relevant to the issue of pregnancy labeling: 1) treatment of non-pregnant women of reproductive potential (prospective) and 2) treatment during pregnancy and/or lactation (retrospective). Issues associated with the former include how actively a pregnancy should be avoided during treatment and the risks for a potential planned pregnancy during treatment. Issues arising in relation to the latter include interpretation of risk for a pregnancy which occurs while on treatment and interpretation of risk for treatment of a medical condition which develops during pregnancy or lactation.

Prevention of pregnancy in women of reproductive potential is a more conservative focus and the one given stronger consideration when the current pregnancy category regulations were adopted. However, treatment during pregnancy or lactation is of immediate concern for the patient who becomes pregnant during treatment and, from this perspective, the risk-benefit assessment changes for both the physician and the patient. Others who testified at the September 12 hearing provided vivid reports of the consequences of withdrawing needed pharmaceutical treatment once pregnancy was confirmed, as well as of termination of wanted pregnancies due to unsubstantiated fear of harm to the fetus. Providing more systematically-collected information in the labeling will allow more informed decision-making, whether assessing risk for a potential or actual pregnancy.

In addition to a narrative summary, we suggest using a table (such as those appended to these comments) to summarize comparative animal and human reproductive and developmental data. If data were not available for any portion of the reproductive and developmental process, an entry of "insufficient data" could be included in the table. We believe a table makes the absence of data and the weight of evidence apparent. The table might be preceded by a brief assessment of the known pregnancy risk of the medical condition itself, observed risk of drug product exposure in animals, and whether those or other risks have been identified in humans. A means of obtaining upon request a more detailed summary of the relevant literature could be provided.

In Attachments 1 and 2 are draft examples of such tables, narratives, and references for two Lilly drug products, Prozac and Humalog, respectively. These examples demonstrate a range of data that may be available for inclusion in such a table. Extensive human data exist for Prozac. At this time for the newly available insulin product, Humalog a hormone replacement therapy not expected to increase pregnancy or developmental risks, insufficient human data are available regarding treatment during pregnancy.

At the September 12, 1997 hearing, FDA requested comments on liability issues related to the content and format of pregnancy labeling for human prescription drug products. Lilly believes that use of both a narrative and a summary table to display published and registry data of appropriate quality, as suggested by Lillys testimony and these comments, would provide physicians the information they need to weigh the benefits and risks of use of drug products in pregnancy as well as allow a drug product manufacturer to better meet the legal duty to provide adequate information to the physician.

Lilly appreciates the opportunity to comment on this subject and looks forward to working with the Agency to help develop informative labeling regarding reproductive and developmental risk. Our comments summarize the issues related to this topic. We have provided a specific proposal as a beginning to a discussion process. We hope these comments will provide the basis upon which clearer labeling information on the reproductive process can be provided to physicians and patients for decisions on drug product treatment during pregnancy.

Overall risk: Low for fertility, pregnancy (all trimesters), and lactation. Fluoxetine should be used during pregnancy and lactation only if clearly needed.

Summary: Animal data indicate increased stillbirths, neonatal mortality, and decreased birth weights at high doses during gestation in one species. The relevance of these findings to humans is uncertain. Human data demonstrate that untreated depression may increase the risk of an adverse pregnancy outcome, most likely resulting in prematurity, low birth weight, and perinatal complications. No consistent effects on premature delivery, birth weight, or perinatal effects have been found in infants of women treated with fluoxetine during the third trimester. Fluoxetine does not appear to increase the risk for major malformations or spontaneous abortion when compared with the general population. Preliminary information in humans on lactation and postnatal development of offspring has shown no increased risk.

Evaluations	Animal Findings	Human Findings
Fertility	No effects ^a,b,c	Insufficient data
Spontaneous abortion	No increase^a,b,c,d,e	No increase^f,g,h,i
Labor and delivery	No increase in premature deliveries. No effects on labor.^a,b,c,d,e	No increase in premature delivery ^f,h,iIncrease in premature delivery for early vs late exposure^g
Perinatal death and stillbirth	Increased in some rat strains at ³ 7.4 mg/kg ^a,b,c,e	No increase ^f,g,h,i,j
Major malformations	No increase in rats, rabbits ^a,b,c,d	No increase^f,g,h,i,j
Birth weight	Reduced in some rat strains at >7.4 mg/kg^a,b,c,e	No effect ^f,hDecrease for late vs early exposure^g
Perinatal complications	No increase ^a,b,c,e	No effect on APGARs; no increase in complication rate ^f,h,kIncreased complication rate in late vs early exposure^g
Postnatal development	No effects on growth, behavior, or reproduction^a,c,e	Neurobehavioral development unaffected ^h
Lactation	No effects^a,b,c	No effects ^l

GLP studies: ^aWold 1980, ^bBrophy 1982, ^cHoyt 1989, ^dByrd 1994.
Other: ^eVorhees 1994.

Prospective survey: ⁱ Goldstein 1997, ^j McElhatton 1996,^kGoldstein 1995, ^l Taddio 1993.

Animal Data: Two fertility studies conducted in rats at doses of up to 7.5 and 12.5 mg/kg/day (approximately 0.9 and 1.5 times the maximum recommended human dose [MRHD] on a mg/m² basis) indicated that fluoxetine had no adverse effects on fertility. In embryo-fetal development studies in rats and rabbits, there was no evidence of teratogenicity following administration of up to 12.5 and 15 mg/kg/day, respectively (1.5 and 3.6 times, respectively, the MRHD of 80 mg on a mg/m² basis) throughout organogenesis. However, in rat reproduction studies, an increase in stillborn pups, a decrease in pup weight, and an increase in pup deaths during the first 7 days postpartum occurred following maternal exposure to 12 mg/kg/day (1.5 times the MRHD on a mg/m² basis) during gestation or 7.5 mg/kg/day (0.9 times the MRHD on a mg/m² basis) during gestation and lactation. There was no evidence of developmental neurotoxicity in the surviving offspring of rats treated with 12 mg/kg/day during gestation. The no-effect dose for rat pup mortality was 5 mg/kg/day (0.6 times the MRHD on a mg/m² basis).

Human Data: Three independent, controlled, prospective evaluations of fluoxetine-exposed pregnancies have been reported. One study of 128 first-trimester fluoxetine exposures did not demonstrate an increased risk of major malformations, stillbirths, or neonatal deaths in offspring exposed to fluoxetine during pregnancy. A second study of 37 first-trimester and 18 first-, second-, and third-trimester fluoxetine exposures found no differences in neurobehavioral measures of activity, temperament, or behavior in children up to 4 years of age of fluoxetine-treated women. A third study of 228 women taking fluoxetine during pregnancy found no increase in risk of spontaneous pregnancy loss or major fetal anomalies. However, third-trimester exposure, complicated by other known risk factors, was associated with increased risk for perinatal complications. One additional pregnancy survey (81 total pregnancies) and a Lilly pregnancy registry (796 pregnancies) also did not demonstrate an increased risk of abnormalities in fluoxetine-exposed pregnancies. An evaluation of third-trimester exposures through delivery (112 pregnancies) from the Lilly registry did not demonstrate an increased risk or pattern of perinatal complications. A case series of 11 infants of 10 women showed no adverse events associated with breastfeeding. Fluoxetine and norfluoxetine were present in the breast milk.

Overall risk: Low for fertility, pregnancy (all trimesters), and lactation. Humalog should be used during pregnancy and lactation only if clearly needed.

Summary: Data from non-diabetic animals treated with Humalog indicate a slight increase in gestation length and offspring startle changes. The relevance of these findings to humans is uncertain. Human cohort-controlled or prospective survey data are inadequate to draw firm conclusions. Human data of pregnancy outcome in diabetic women indicate that poor glycemic control is associated with a 3- to 5-fold increased relative risk of major malformations and an elevated risk of perinatal complications.

Evaluations	Animal Findings	Human Findings
Fertility	No effects^a,b	Insufficient data
Spontaneous abortion	No effects^a	Insufficient data
Premature delivery	Gestation length increased by <1 day ^a	Insufficient data
Perinatal death and stillbirth	No effects^a	Insufficient data
Major malformations	No increase in rats, rabbits^a,c	Insufficient data
Birth weight	No effects ^a	Insufficient data
Perinatal complications	No increase ^a	Insufficient data
Postnatal development	Behavioral effects at 20 U/kg/day; no effects on growth or reproduction^a	Insufficient data
Lactation	No effects ^a	Insufficient data

Animal Data: There is no evidence from animal studies of Humalog-induced impairment of fertility. Reproduction studies have been performed in pregnant rats and rabbits at parenteral doses up to 4 and 0.3 times, respectively, the average human dose (40 units/day) based on body surface area. The results have revealed no evidence of impaired fertility or harm to the fetus due to Humalog.

Human Data: Although there are no clinical studies of the use of Humalog in pregnancy, published studies with human insulins suggest that optimizing overall glycemic control, including postprandial control, before conception and during pregnancy improves fetal outcome. Although the fetal complications of maternal hyperglycemia have been well documented, fetal toxicity also has been reported with maternal hypoglycemia. Insulin requirements usually fall during the first trimester and increase during the second and third trimesters. Careful monitoring of the patient is required throughout pregnancy. During the perinatal period, careful monitoring of infants born to mothers with diabetes is warranted. It is unknown whether Humalog is excreted in human milk. For this reason, caution should be exercised when Humalog is administered to a nursing woman. Patients with diabetes who are lactating may require adjustments in Humalog dose, meal plan, or both.