Abstract
Objective. To present any congenital anomalies with respect to tumor necrosis factor (TNF) antagonists reported to the US Food and Drug Administration (FDA) to determine if there are common findings.
Methods. A review of the FDA database of reported adverse events with etanercept, infliximab, and adalimumab from 1999 through December of 2005 was performed. Key words for congenital anomalies were employed as search tools. Duplicate reports were eliminated. Any concomitant medicines were recorded.
Results. Our review of > 120,000 adverse events revealed a total of 61 congenital anomalies in 41 children born to mothers taking a TNF antagonist. Of these mothers, 22 took etanercept and 19 took infliximab. There were no reports in women taking adalimumab. The most common reported congenital anomaly was some form of heart defect. Twenty-four of the 41 (59%) children had one or more congenital anomalies that are part of vertebral abnormalities, anal atresia, cardiac defect, tracheoesophageal, renal, and limp abnormalities (VACTERL) association. There were 34 specific types of congenital anomalies in total, and 19 (56%) of those are part of the VACTERL spectrum. Nine of these 19 (47%) types of VACTERL anomalies were observed statistically significantly more than historical controls (p < 0.01); in 4 of these 9 the p value was ≤ 0.0001. Thirteen (32%) of the children had more than one congenital anomaly; 7 of these 13 children had 2 defects that are part of the VACTERL spectrum. However, only 1 child was diagnosed with VACTERL. In 24/41 cases (59%) the mother was taking no other concomitant medications.
Conclusion. A seemingly high number of congenital anomalies that are part of the VACTERL spectrum have been reported. These congenital anomalies are occurring at a rate higher than historical controls. This commonality raises concerns of a possible causative effect of the TNF antagonists.
Autoimmune conditions that may require the use of a tumor necrosis factor-α (TNF-α) antagonist often involve women of childbearing age. TNF-α antagonists are rated “category B” and are presumed to be safe in pregnancy based on animal data. However, animal reproductive studies are not always predictive of human response. For obvious reasons, this has never been formally studied in prospective trials involving humans.
Efforts have been made to form pregnancy registries to assess the safety of TNF-α antagonists and other drugs. However, few patients have been documented to take these medications throughout the entire first trimester and even fewer throughout their entire pregnancy. One such registry, the Organization of Teratology Information Specialists (OTIS), prospectively follows women with autoimmune diseases monitoring the potential effects of treatment on the developing embryo or fetus. The number of patients followed in these registries is still rather small, yet birth defects have been observed1–3.
We previously reported a case of VACTERL (VATER) association in a child of a mother taking etanercept throughout her pregnancy4. Recently, we had another case of a child with tracheal stenosis, bronchomalacia, patent ductus arteriosis, and a skeletal disorder (defects part of VACTERL; possible VACTERL association) born to a mother who took adalimumab through her first trimester. VACTERL is a nonrandom association of birth defects that occurs in ~1.6/10,000 live births5. The acronym VACTERL is used to describe the association of Vertebral defects, Anal atresia or imperforate anus, Cardiac abnormalities (most commonly atrial septal defect, ventricular septal defect, and tetralogy of Fallot), Tracheoesophageal fistula or tracheal atresia/stenosis, Esophageal atresia, Renal and/or Radial abnormalities, and pre-axial Limb abnormalities. In order to be diagnosed with definite VACTERL a child must have at least 3 of the associated congenital anomalies. There are about 300 cases of VACTERL reported in the literature. All of the birth defects that are part of VACTERL are linked, but the exact etiology is unknown.
Interestingly, other drugs that inhibit TNF-α have been linked with similar congenital defects that are part of VACTERL. Thalidomide is known to cause a variety of limb abnormalities6. Thalidomide also works, at least in part, through its ability to inhibit TNF-α7. Valproic acid has been associated with vertebral body defects in humans8,9. It is also known that valproic acid inhibits TNF-α10.
After publication of our proband case with VACTERL4, we decided to further analyze the congenital anomalies associated with the TNF-α antagonists. We requested the complete lists of all reported adverse events for all 3 TNF-α antagonists (etanercept, infliximab, and adalimumab) from the US Food and Drug Administration (FDA). Our aim was to determine if there were any common findings.
MATERIALS AND METHODS
Data collection
A written request for the complete lists of all the reported adverse events for all 3 of the commercially available TNF-α antagonists (infliximab, etanercept, and adalimumab) was submitted to the FDA. These data are available through the Freedom of Information Act. After approximately 4 months, we received 3 compact discs from the FDA with the requested data. The database was complete from the time that TNF antagonists were first commercially available through December of 2005.
In order to search for all of the reported congenital anomalies for each drug, the keywords “congenital anomaly,” “congenital anomalies,” “birth defect,” and “birth defects” were employed as search tools. Each adverse event has a unique report number. Any duplicate reports were eliminated so that each event was counted only once.
Each adverse event listed includes the following information: a unique report number, the date of the report, the adverse event(s), the report source (e.g., health professional, consumer), the medication and all concomitant medications, and the medication that is the primary suspect for the adverse event listed (as determined by the reporter). All of these data were recorded.
Statistical analysis
The specific congenital anomalies in the FDA database were compared to historical controls. In order to create a direct comparison for the FDA database, we had to change the historical incidences per live births to incidences per child born with a birth defect. Knowing that congenital anomalies encompass 3%–5% of live births, we divided the historical incidences per live birth by 0.04. A Poisson probability and cumulative Poisson probability were performed on each congenital anomaly in the database. A Poisson probability refers to the probability of getting exactly n occurrences and a cumulative Poisson probability refers to the probability of getting at most n occurrences. P values were determined from the Z-score using the equation Z = n – lambda/(square root of lambda). Because we analyzed multiple p values, there is an increased chance of Type 1 error. Therefore, we considered p values to be significant only if they are < 0.001. In order to prevent bias of including the congenital anomalies from our proband case, this child’s birth defects were excluded from the statistical analysis.
RESULTS
This database review totaled over 120,000 adverse events. A total of 41 children with 61 congenital anomalies born to 40 mothers taking a TNF antagonist have been reported to the FDA through December 2005 (Table 1). Twenty-two of these mothers took etanercept at some point during their pregnancy and 19 were taking infliximab. There were no reports in women taking adalimumab. In all 41 cases the TNF-α antagonist was considered the “primary suspect” as the cause of the birth defect. Thirteen (32%) of the children had more than 1 congenital anomaly.
Children with congenital anomalies.
The congenital anomalies from our proband case with VACTERL (Table 1; Subject 19) were excluded from the statistical analysis, leaving 54 reported congenital anomalies in 40 children (Table 2). Our second case of a child born with probable VACTERL association occurred after 2005, and was therefore not included in this database either. The most common reported congenital anomaly was some form of heart defect (n = 11) [4 congenital heart disease, 2 ventricular septal defect (VSD), 2 atrial septal defect (ASD), 1 great vessel malformation, 1 tetralogy of Fallot, and 1 ventricular hypokinesia]. Other anomalies reported more than once were cystic kidney (3), pulmonary malformation (3), teratoma (3), tracheal stenosis (2), hypospadias (2), trisomy 21 (2), and hydrocele (2). Five of the congenital anomalies were not specified.
Fifty-four congenital abnormalities were reported in 40 children. The congenital anomalies from the proband case with VACTERL (child 19 in Table 1) have not been included in this table.
Regarding the specific congenital anomalies in the FDA database (excluding our proband case), there were 34 different types of birth defects (Table 2). Nineteen of 34 (56%) anomalies are those that are part of the VACTERL spectrum. In order to correct for multiple comparisons (n = 31), we considered a p value as significant only if it was < 0.001. Using this strict criterion, 4/19 (21%) were statistically significantly increased compared to historical controls, with a very low chance of Type 1 error (3.05%). Nine of 19 (47%) of these VACTERL-type defects occurred more than historical controls, with a p value of < 0.01. It is also important to note that there are several congenital heart defects in Table 2 that are not included in the heading “congenital heart disease” (e.g., VSD). If all of these heart defects are included in the heading “congenital heart disease,” then congenital heart defects also occurred significantly more compared to historical controls (p = 0.007).
Twenty-four of 41 (59%) children had 1 or more congenital anomalies that are part of VACTERL; 7 had 2 or more (Table 1). However, only 1 was diagnosed with VACTERL (Subject 19 in Table 1). Of all the reported congenital anomalies, 37/61 (61%) are consistent with the birth defects that encompass VACTERL. If the proband case is not included, 30/54 (56%) of the congenital anomalies are consistent with those seen in VACTERL. In spite of the fact that anal atresia and esophageal atresia are part of the VACTERL spectrum, there were none observed in this database (excluding the proband case).
In 24/41 cases (59%) the mother was taking no other concomitant medications (Table 1). For those mothers taking concomitant medications, the most common medications used included prednisone (8), azathioprine (5), methotrexate (5), ibuprofen (4), and hydrochlorothiazide (3). Of these, only methotrexate is clearly teratogenic. There was no trend in observed anomalies in the 5 women who were taking concomitant methotrexate.
Regarding perinatal demise, there were 3 cases of spontaneous abortions (Table 1; Subjects 4, 11, and 14) and 2 induced abortions (Table 1; Subjects 17 and 21). All were in etanercept-treated mothers.
Of note, 2 of these 41 children were twins (Table 1; Subjects 10 and 11). Both had the same congenital anomalies (congenital heart disease and pulmonary malformation), suggesting a common teratogenic exposure. As stated, 1 of these was spontaneously aborted. The mother was taking etanercept without concomitant medications.
DISCUSSION
There have been a significant number of congenital anomalies in children born to mothers taking TNF-α antagonists reported to the FDA. A seemingly high number of these birth defects (~60%) include anomalies that are part of the well documented non-random association known as VACTERL association. Further, these data could be considered a conservative estimate, since the specific congenital anomaly is not specified in 5 cases. As stated, we recently had another case of a child with tracheal stenosis, bronchomalacia, patent ductus arteriosis, and a skeletal disorder (probable VACTERL association) born to a mother taking adalimumab throughout her entire first trimester. This second case was not included in the FDA database or any of the statistical analysis. This database also includes a high percentage of children (32%) born with more than 1 congenital anomaly (7 of whom had 2 anomalies that are part of VACTERL). As a comparison, approximately 14% of children born with congenital anomalies have multiple congenital anomalies11.
Since congenital anomalies are present in 3%–5% of all live births12 and VACTERL occurs in 1.6/10,000 live births5, you would expect to see ~1.6 cases of VACTERL association in every 300–500 children born with congenital anomalies. We have now seen 2/42 (4.8%) cases of VACTERL [compared to historical controls of 0.005% (1.6/300); p < 0.001; 0.9984 cumulative Poisson probability and 23/40 (58%) others with similar defects.
Cytokines play a critical role in both normal and abnormal human fetal development. In particular, TNF-α has been identified in the ovary, uterus, placenta, and embryo13. Previously, it was felt that the only role of TNF-α in human gestation was that of triggering immunological pregnancy loss and as a mediator of embryopathic stresses13. However, it is now known that TNF-α has a dual, almost paradoxical, role in embryological development. It also protects embryos from developmental toxicants13–16. This multifunctional cytokine is a powerful activator of both apoptotic and antiapoptotic signaling cascades involved in embryological development13,15.While the precise role of TNF-α in human fetal development is not fully elucidated, it is clear that this cytokine will not only boost cell death to kill the embryo if harmful assaults produce structural anomalies, but it also stimulates protective mechanisms preventing birth defects as the fetus continues to develop13–16. Interestingly, TNF-α single nucleotide polymorphisms have been linked to human conotruncal heart defects17 (tetralogy of Fallot, VSD, etc.) and limb deficiencies18. Therefore, drugs that decrease embryologic TNF-α levels can interfere with these signaling cascades, its protective effects, and potentially increase the risk for anomalies.
Maternal IgG readily crosses the placenta19. Most IgG antibodies, in general, cross the placenta and IgG1 antibodies are preferentially transferred20. The Fc region of IgG is required for its transport across the placenta19. Therefore it would be logical that an Fc receptor-construct fusion protein (etanercept) and monoclonal IgG1 antibodies (infliximab and adalimumab) would as well. Indeed it has been demonstrated that these drugs cross the placenta to the fetus21,22. The timing of when they are first able to cross is not completely understood.
Although animal reproduction studies were performed prior to each of these drugs being approved, animal studies are not always predictive of human response. Etanercept was studied at 60–100 times the standard human dose; however, the area under the curve systemic exposure levels were estimated to be 4 times the normal human dose23. Further, these tests were performed on only 8 rats and 6 rabbits24. Since infliximab does not cross-react with TNF-α in species other than humans and chimpanzees, animal reproduction studies have not been conducted with this drug25. In spite of these obvious limitations, the 3 TNF-α antagonists are rated category B by the FDA, meaning they are presumed to be safe based on animal studies.
Other drugs that inhibit TNF-α have been linked to congenital anomalies. The 2 most notable are thalidomide and valproic acid. Importantly, the anomalies that these drugs cause are the same as many of those that encompass VACTERL association. Of note, early animal reproduction studies revealed no signs of teratogenicity with thalidomide26. It is now well known that pre-axial limb defects are most characteristic of thalidomide embryopathy6. However, other anomalies caused by thalidomide include spinal defects, cardiac malformations (usually conotruncal defects), anal atresia, and renal anomalies27. Thalidomide has also been associated with birth defects not traditionally associated with VACTERL, but birth defects do appear in our data, namely strabismus, Duane’s syndrome (a congenital disorder of eye movement), and congenital hearing loss6. It is also important to note that congenital anomalies do not occur in all cases exposed to thalidomide. There is approximately a 20% risk of thalidomide embryopathy if a fetus is exposed during the critical period28. If the mother stops the drug before that crucial period, the risk is less. Valproic acid has a definite association with neural tube defects8,9 and has been suggested to cause limb malformations29, cardiac anomalies29, tracheomalacia30, and hypospadias31. This same drug also has anti-TNF-α properties10. Valproic acid doubles the risk of neural tube defects to the fetus, but the likelihood is still rare (2% vs 1%)29.
There are existing data regarding TNF antagonists and pregnancy. OTIS is a collaborative research group prospectively following women exposed to TNF-α antagonists during pregnancy. To date, they have published data on 44 women exposed to etanercept or adalimumab during their first trimester and there have been 6 congenital anomalies in 5 children2,3. In the adalimumab registry, specifically, 21.6% of the women experienced spontaneous abortions versus 6.4% in the disease control group3. Another study tracked 22 pregnancy outcomes in women after exposure to TNF-α antagonists, without any reported congenital anomalies, but only 2 patients were exposed to the drug until pregnancy confirmation32. A database review of 96 women exposed to infliximab during pregnancy has been published33. Fifty-eight of 96 women (60%) received infliximab during their first trimester and none of the women received the drug throughout the duration of their pregnancy. Five infants were born with complications including tetralogy of Fallot and intestinal malrotation. A British register has collected data on 23 women exposed to anti-TNF-α therapy at the time of conception34. All but 2 of these patients discontinued their therapy during the first trimester. There was a high spontaneous miscarriage rate of 6/23 (26%), 3 elective terminations, and 14 live births, the latter without any “major fetal abnormalities.” Another retrospective series of 10 women with Crohn’s disease exposed to infliximab throughout pregnancy yielded 3 premature births with complications but no congenital malformations35. Finally, another group reported the findings of 3 women who unexpectedly became pregnant while receiving anti-TNF therapy. One patient elected for therapeutic termination; of the remaining 2, 1 had a child born with adrenal congenital hyperplasia36. As stated, the majority of pregnant mothers exposed to thalidomide, and the vast majority of those exposed to valproic acid, have normal healthy children in spite of the fact that these are known teratogenic drugs. Therefore case series or small registries of pregnant mothers who took TNF-α antagonists that show a limited number of birth defects, or even none at all, need to be interpreted in the proper context.
Our study has several limitations. The most obvious is that the number of pregnant women who have been treated with TNF-α antagonists is unknown. Further, there is no active control group to determine an odds ratio, so historical controls were utilized. As with any voluntary postmarketing adverse event reporting database, there is reporter bias. Adverse events are likely to be reported, whereas pregnant women exposed to TNF-α antagonists without adverse events will not report this. Also, the database itself is quite limited in its content. VACTERL is thought to occur in the children of diabetic mothers at increased frequency37; there is no maternal medical history given in this database, but the medications are listed and only one mother (Subject 9, Table 1) was taking any type of glucose-lowering therapy. Also, it is difficult to know if this same study method were applied to the FDA database for other medications known to have no association with VACTERL, what proportion of children would have anomalies that are part of VACTERL. Finally, several of the birth defects in this database can occur without any association with VACTERL (e.g., VSD, ASD), but are considered part of this spectrum for the analysis here. In spite of these limitations, these data do raise concerns.
Taken together, these data tell us that a significant number of congenital anomalies in children of pregnant mothers taking a TNF-α antagonist have been reported to the FDA. The majority of these anomalies are consistent with those seen in VACTERL association. Historical controls suggest that these anomalies are occurring more frequently than expected. Other drugs that are known to inhibit TNF-α are known teratogens and cause similar anomalies. It is most prudent to err on the side of caution when dealing with pregnancy and potentially teratogenic drugs. These data suggest a possible causative effect of the TNF-α antagonists. Although questions remain, we suggest that clinicians should not prescribe TNF antagonists to women during pregnancy.
Footnotes
- Accepted for publication September 17, 2008.
