Prenatal Exposure to Valproic Acid DuringPregnancy and Limb Deficiencies:A Case-Control Study

Elvira Rodríguez-Pinilla,1* Ignacio Arroyo,2 Juan Fondevilla,3 Maria Jose García,4 andMaria Luisa Martínez-Frías1,5

1Estudio Colaborativo Español de Malformaciones Congenitas (ECEMC), Facultad de Medicina, UniversidadComplutense, Madrid, Spain

2Servicio de Pediatría, Hospital General San Pedro de Alcántara, Cáceres, Spain3Servicio de Pediatría, Hospital Juan Ramon Jimenez, Huelva, Spain4Servicio de Pediatría, Hospital Virgen de la Luz, Cuenca, Spain5Departamento de Farmacología, Facultad de Medicina, Universidad Complutense, Madrid, Spain

We conducted a case-control study usingdata from the Spanish Collaborative Studyof Congenital Malformations (ECEMC) onthe relationship between prenatal exposureto valproic acid (VPA) and the presence oflimb deficiencies in newborn infants.Among a total of 22,294 consecutive mal-formed infants (once we excluded geneticsyndromes) and 21,937 control infants withspecified data on antiepileptic drugs duringgestation, 57 malformed infants and 10 con-trol infants were exposed to VPA during thefirst trimester of pregnancy. Of the total ofmalformed infants exposed to VPA, 36.8%(21/57) presented with congenital limb de-fects of different types (including overlap-ping digits, talipes, clubfoot, clinodactyly,arachnodactyly, hip dislocation, pre- andpostaxial polydactyly, etc.), three of themhaving limb deficiencies. The result of thecase-control analysis shows a risk for limbdeficiencies of odds ratio = 6.17 [confidenceinterval (CI) 1.28–29.66, P = 0.023], after con-trolling for potential confounder factors. Ifwe consider that in our population theprevalence at birth of this type of defect is6.88 per 10,000 livebirths (95% CI 6.43–7.36)we can estimate that the risk for womentreated with VPA of having a baby with limbdeficiencies would be around 0.42%. Thelimb deficiencies in the three patients ex-posed to VPA were the following: the first

case was a newborn infant with hypoplasiaof the left hand, the second patient was anewborn infant with unilateral forearm de-fect and hypoplastic first metacarpal bonein the left hand, and the third patient pre-sented with short hands with hypoplasticfirst metacarpal bone, absent and hypoplas-tic phalanges, retrognathia, facial asymme-try, hypospadias, teleangiectatic angioma inskull, and hypotonia. Am. J. Med. Genet. 90:376–381, 2000. © 2000 Wiley-Liss, Inc.

KEY WORDS: valproic acid; congenitalmalformations; limb defi-ciencies

INTRODUCTION

The risk for neural tube defects after prenatal expo-sure to valproic acid (VPA) is well documented [Robertand Guibaud, 1982; Bjerkedal et al., 1982; Mastroia-covo et al., 1983; Lindhout and Meinardi, 1984; Carterand Stewart, 1988; Martínez-Frías et al., 1989; Prietoand Martínez-Frías, 1999]. Moreover, a “fetal valproatesyndrome” has been described and delineated [DiLib-erti et al., 1984; Jager-Roman et al., 1986; Ardinger etal., 1988; Martínez-Frías, 1990]. Prominent metopicridge and outer orbital ridge deficiency or bifrontal nar-rowing seem to be especially characteristic of fetal VPAexposure, but other minor craniofacial anomalies (ab-normalities in position or configuration of the ears,long and flat philtrum, thin upper lip, micrognathia,midface hypoplasia, short nose with broad and flat na-sal bridge, epicanthal folds, etc.) are also seen in in-fants with VPA exposure. In relation to other majoranomalies, prenatal exposure to VPA has been associ-ated with respiratory tract anomalies, limb anomaliessuch as talipes equinovarus, genital anomalies, andcardiac defects [Ardinger et al., 1988]. More recently,

Contract grant sponsor: Instituto de Salud Carlos III, Ministe-rio de Sanidad y Consumo (Spain).

*Correspondence to: Dr. Elvira Rodríguez-Pinilla, ECEMC,Facultad de Medicina, Universidad Complutense, Madrid 28040,Spain. E-mail: luisama@eucmos.sim.ucm.es

Received 18 May 1999; Accepted 26 October 1999

American Journal of Medical Genetics 90:376–381 (2000)

© 2000 Wiley-Liss, Inc.

limb defects (mainly radial ray defects) have been re-ported in a small number of VPA-exposed human foe-tuses [Brons et al., 1990; Verloes et al., 1990; Grahamet al., 1991; Sharony et al., 1993; Langer et al., 1994;Clayton-Smith and Donnai, 1995; Okada et al., 1995].

Here we present a case-control study conducted us-ing data from the Spanish Collaborative Study of Con-genital Malformations (ECEMC) on the relationshipbetween prenatal exposure to VPA and the presence oflimb anomalies in the newborn infants, particularlylimb deficiencies.

MATERIALS AND METHODS

Data are derived from the ECEMC, which is an on-going hospital-based case-control study and surveil-lance system. All children born in 77 participating hos-pitals from all over Spain are examined by physicianswho, being interested in the problem of congenital de-fects, collaborate with the ECEMC program and acceptits uniquely strict methodology. They examine the chil-dren within the first three days of life to identify majorand/or minor/mild defects. For each malformed baby,the next nonmalformed infant of the same sex born inthe same hospital is selected as a control subject. Oncethe case and control infants have been identified(blinded to any type of prenatal exposure), the samephysicians interview their mothers to gather informa-tion on obstetrical data, family history, and prenatalexposures. In many instances, photographs, karyo-types, imaging studies, pathology reports, and resultsof other studies are also available for review. The in-formation on drugs used during pregnancy comes fromspecific questions, one of them being whether anticon-vulsant drugs were used. For each drug, proprietaryname, month or week of intake, duration of treatment,and daily and total dosage are specified. Detailed de-scriptions of the ECEMC methodology have been pub-lished elsewhere [Martínez-Frías et al., 1997, 1998].

To codify the different drugs, we used a modified andextended EPhMRA classification system, which issimilar to the Anatomical Therapeutic Chemical (ATC)classification system [ATC Classification, 1991]. Tocodify the different types of congenital defects, wemodified the ICD-8 codes for congenital anomalies byadding two more digits in order to obtain greater speci-ficity.

Between April 1976 and June 1997, the ECEMC sur-veyed a total of 1,431,368 liveborn infants. Of those,25,967 (1.8%) were malformed infants and 25,326 wereselected as controls. For the case-control analysis weexcluded 2,984 cases with known genetic syndromes,as well as their respective controls (a total of 2,866controls). Thus, we analysed 22,983 malformed casesand 22,460 controls. Of these, 3.0% of the cases and2.3% of the controls had unspecified data about anti-epileptic drug use during pregnancy (689 cases and 523controls), which were also excluded.

We considered as exposed to VPA those infants(cases and controls) who where exposed during the firsttrimester of pregnancy. Infants who were not exposedto VPA during the first trimester of pregnancy wereconsidered as unexposed.

For the analysis we used two groups of controls: first,the specific controls of the cases (paired controls), andsecond, the controls born at the same hospital as eachcase during a period ± 45 days from the case’s date ofbirth, in order to increase the statistical power. Thesecontrols represent an unbiased sample and have thecharacteristics of appropriate controls.

The overall analysis was done using 2 × 2 tables ofcase-control infants, estimating the odds ratio (OR)with 95% confidence intervals (CIs)and the Fisher ex-act test. A logistic regression analysis was performed tocontrol for the following potential confounder factors:maternal and paternal age, maternal and paternaleducation level, consanguinity, ingestion of vitaminsand/or iron (as indicator of a good medical control ofpregnancy), and maternal treatment with antiepilepticdrugs other than VPA during the first trimester ofpregnancy.

RESULTS

Among the total of 22,294 consecutive malformed in-fants (once we excluded genetic syndromes) and 21,937controls, with specified data on antiepileptic drugs dur-ing gestation, 57 malformed infants and 10 control chil-dren were exposed to VPA during the first trimester ofpregnancy. Thus, the crude OR for prenatal exposureto VPA during the first trimester is 5.62 (CI 95% 2.78–11.71, P <0.0000001) for any type of congenital defect.

Among the malformed infants exposed to VPA,36.8% (21/57) presented congenital limb defects of dif-ferent types, including overlapping digits, talipes,clinodactyly, arachnodactyly, hip dislocation, pre- andpostaxial polydactyly, and limb deficiencies. Neverthe-less, it is important to note that in 5 of the 21 caseswith limb anomalies, the infants also had spina bifida.In these cases, the limb anomalies (all of them derivedfrom immobility caused by a neural tube defect, such astalipes equinovarus) were secondary to the primary de-fect (spina bifida). If we exclude these cases and theirrespective controls, we obtain a crude risk for any typeof limb anomaly after prenatal exposure to VPA duringthe first trimester of an OR 4 3.95 (CI 95% 1.24–13.94,P 4 0.01) (Table I).

Table II depicts the results for limb deficiencies, forwhen we used the paired controls and the controls whowere born in the same hospital as cases but in a periodof ± 45 days from the case’s date of birth. As shown inTable II, the two values of the ORs are not statisticallydifferent, although that obtained with the paired con-trols does not reach the level of statistical significance,

TABLE I. Odds Ratios (95% Exact CI) for Any Type of LimbAnomaly After Exposure to VPA During the First Trimester

of Pregnancy (Excluding Those Derived From ImmobilityCaused by a Neural Tube Defect)

Exposed Nonexposed OR 95% CI P-value

Casesa 16 9,1063.95 1.24–13.94 0.015

Pairedcontrols 4 8,983

aMalformed infants with limb anomalies.

Valproic Acid and Limb Deficiencies 377

possibly due to the sample size. If we make a pair-matched analysis, the OR is in the same range (OR 43.0). These results suggest that mothers exposed toVPA have a risk ranging from 3 to 5 for having a childwith limb deficiencies.

In order to control for some confounder factors, weperformed an analysis using different multiple logisticregression models: first, including all cases with limbcongenital defects (once we excluded those secondary toneural tube defects), and second, considering only caseswith limb deficiencies. The logistic regression analyses,controlling maternal and paternal age, maternal andpaternal educational level, consanguinity, ingestion ofvitamins and/or iron, and maternal treatment with an-tiepileptic drugs other than VPA during the first tri-mester of pregnancy, show similar results in both situ-ations. In the first one, including cases with all type oflimb anomalies, prenatal exposure to VPA implies arisk of OR 4 3.59 (CI 1.10–11.78, P 4 0.035). Theresult for limb deficiencies shows a risk of OR 4 6.17(CI 1.28–29.66, P 4 0.023), also after controlling forthe same potential confounder factors.

The limb deficiencies in the three patients exposed toVPA were the following: The first case was a newbornfemale with radius hypoplasia, radioulnar subluxation,and hypoplasia of the left hand (exposed to 1,000 mg/day of VPA in monotherapy from the second month tothe end of pregnancy) (Fig. 1), the second was a new-born male with right forearm defect, rudimentary ra-dius and ulna, and hypoplastic first metacarpal bone inthe left hand (exposed to 1,500 mg/day of VPA in mono-therapy throughout the pregnancy) (Figs. 2 and 3), andthe third had short hands with hypoplastic first meta-carpal bone, absent or hypoplastic phalanges, retro-gnathia, facial asymmetry, hypospadias, teleangiec-tatic angioma of the skull, and hypotonia (exposed to1,000 mg/day of VPA, 800 mg/day of carbamazepine,and 6 mg/day of clonazepam throughout the preg-nancy) (Fig. 4).

DISCUSSION

The “fetal valproate syndrome” includes epicanthalfolds, flat nasal bridge, small nose with anteverted nos-trils, long upper lip with relatively shallow philtrum,relatively small mouth with downturned angles, andthin upper vermilion border [DiLiberti et al., 1984;Lammer et al., 1987; Winter et al., 1987; Ardinger etal., 1988; Martínez-Frías, 1990; Clayton-Smith andDonnai, 1995]. This craniofacial appearance has beenassociated with other abnormalities including low birthweight and development delay in some cases, and aspectrum of minor and major abnormalities such as

neural tube defects [Robert and Guibaud, 1982; Mas-troiacovo et al., 1983; Lindhout and Meinardi, 1984;Carter and Stewart, 1988; Martínez-Frías et al., 1989;Prieto and Martínez-Frías, 1999], congenital heart de-

TABLE II. Odds Ratios (95% Exact CI) for Limb Deficiencies After Exposure to VPA During the FirstTrimester of Pregnancy

Type of controlsCasesa

exposed/nonexposedControls

exposed/nonexposed OR 95% CI P-value

Paired controls 3/735 1/724 3.0 0.28–73.85 0.6Controls ±45 daysb 3/735 6/7,677 5.2 1.04–23.41 0.04

aMalformed infants with limb deficiencies.bControls born at the same hospital as each case during a period ±45 days from the case’s date of birth.

Fig. 1. Radiograph of Patient 1 showing radius hypoplasia and theradioulnar subluxation.

378 Rodríguez-Pinilla et al.

fects [Robert and Rosa, 1983], and defects of hands,feet, and genitalia [Bailey et al., 1983; DiLiberti et al.,1984; Winter et al., 1987].

The congenital limb defects more frequently de-scribed in malformed babies exposed to VPA have beenpreaxial and postaxial polydactyly, clinodactyly, tali-pes, broad big toes and fingerlike thumbs, contracturesof the fingers, arachnodactyly, overlapping digits, andsyndactyly [Lammer et al., 1987; Winter et al., 1987;DiLiberti et al., 1984]. The present study supports thepreviously suspected association between prenatal ex-posure to VPA and any type of limb anomaly, quanti-fying the magnitude of the risk at around 4.

In relation to the limb deficiencies, these have beenless frequently described. In the experimental study ofHendrickx et al. [1988] on the developmental toxicity ofprenatal VPA exposure in rhesus monkeys, craniofa-cial and skeletal defects were the most significant find-ings. Aplastic and hypoplastic phalanges are describedin the exposed monkeys as one of the most commonappendicular skeletal defects. Paulson et al. [1985] alsoobserved skeletal defects, namely syndactyly and par-tial hemimelia, in mouse foetuses exposed to high dosesof VPA.

Data on limb deficiencies in infants prenatally ex-posed to VPA are scanty. As far as we know, there arethe following reported cases: seven foetuses with radialray deficiencies published by Brons et al. in 1990, onechild with bilateral radial ray aplasia and unilateralproximal phocomelia of the upper limb [Verloes et al.,1990], one case with marked symmetric preaxial de-fects with absent radii and tibiae [Graham et al., 1991],two additional cases of severe limb defects published bySharony et al. [1993], one case with an isolated bilat-

eral radial ray reduction [Langer et al., 1994], a pair ofsibs described by Clayton-Smith and Donnai [1995]with typical fetal valproate facial changes and radialray defects, and one infant with a multiple congenitalanomaly pattern, with deficiency of the upper limb, de-scribed by Okada et al. [1995]. Most limb deficiencies inthese cases are of the preaxial ray, as are the three newcases that we describe in this article. In all of our pa-tients a careful and detailed clinical analysis was doneto rule out syndromes such as Fanconi anaemia or Rob-erts syndrome. Chromosome analysis was performed intwo of the three patients, with normal (46,XY) results.The case without karyotype (the first case described inthis article), was a boy born at term (40 week of gesta-tion) with a birthweight of 3,360 g (50th centile), andwithout other clinical signs to point to a chromosomalanomaly. In the same way, there were no other expo-sures to known teratogenic agents, early chorionic vil-lus sampling, or drugs with vasoconstriction effects.Two of the three mothers of the cases with limb defi-ciencies were also exposed to a compound with iron.

Our results support, from an epidemiological stand-point, the relationship between prenatal exposure toVPA and limb deficiencies in the newborn infants, oncewe controlled for potential confounder factors andquantify the risk. This association was only suspectedpreviously on the basis of clinical cases. However, wefailed to find any previous epidemiological study quan-tifying the risk of VPA producing limb deficiencies. It iswell accepted that epidemiology provides the only wayto quantify the magnitude of risk from specific agents[Khoury, 1995]. However, this is difficult when the se-

Fig. 3. Radiograph of Patient 2 showing the left hand with hypoplasticfirst metacarpal bone.

Fig. 2. Radiograph of Patient 2 showing the forearm amputation of theright upper limb, with rudimentary radius and ulna.

Valproic Acid and Limb Deficiencies 379

lected agent has a very low use in the population andthe prevalence of the related defect is also very low,large case-control studies being necessary to obtain sig-nificant results.

The fact that in the present analysis we have con-trolled confounder factors including maternal treat-ment with other antiepileptic drugs, together with thepreviously observed clinical cases with limb deficien-cies, as well as the limb deficiencies described in ani-mal models exposed to VPA [Paulson et al., 1985; Hen-drickx et al., 1988], makes the relationship betweenVPA and this type of defect likely.

If we consider that in our population the prevalenceat birth of limb deficiencies is 6.88 per 10,000 livebirths (95% CI 6.43–7.36) [Bermejo-Sánchez and Mar-tínez-Frías, 1997], and that the risk for this type ofdefect after prenatal exposure to VPA is 6.17, we canconclude that the risk for women treated with VPA ofhaving a baby with limb deficiencies would be around0.42%.

In relation to the dosage, all of our cases were ex-posed to therapeutic doses (between 1,000 and 1,500mg per day), as were the infants described by others[Clayton-Smith and Donnai, 1995; Graham et al.,1991]. Thus, we have identified a relationship betweenprenatal exposure to VPA and limb deficiencies due tothe therapeutic use of VPA in the treatment of epi-lepsy.

ACKNOWLEDGMENTS

We thank all the collaborators of the ECEMC, whocollected the information, and Dr. Díaz-Faes for hisassistance in the interpretation of the patients X-rays.This work was supported by a grant from Instituto deSalud Carlos III, Ministerio de Sanidad y Consumo(Spain).

REFERENCESArdinger HH, Atkin JF, Blackston RD, Elsas LJ, Clarren SK, Livingstone

S, Flannery DV, Pellock JM, Harrod MJ, Lammer EJ, Majewski F,Schinzel A, Toriello HV, Hanson JW. 1988. Verification of the fetalvalproate syndrome phenotype. Am J Med Genet 29:171–185.

ATC Classification. 1990. Guidelines for ATC classification (1990) WHOCollaborating Centre for Drug Statistic Methodology, Norway and Nor-dic Council on Medicines, Sweden.

Bailey CJ, Pool RW, Poskitt EME, Harris F. 1983. Valproic acid and fetalabnormality. Lancet 1:190.

Bermejo-Sánchez E, Martínez-Frías ML. 1997. Vigilancia epidemiologicade las malformaciones congenitas. Boletín del ECEMC: Revista de Dis-morfología y Epidemiología 4:43–114.

Bjerkedal TA, Czeizel A, Goujard J, Kallen B, Mastroiacovo P, Nevin N,Oakley G Jr, Robert E. 1982. Valproic acid and spina bifida. Lancet2:1096.

Brons JT, van der Harten HJ, van Geijn HP, Wladimiroff JW, NiermeijerMF, Lindhout D, Stuart PA, Meijer CJ, Arts NF. 1990. Prenatal ultra-sonographic diagnosis of radial-ray reduction malformations. PrenatDiagn 10:279–288.

Carter BS, Stewart JM. 1988. Valproic acid prenatal exposure. Associationwith lipomyelomeningocele. Clin Pediatr 28:81–85.

Fig. 4. A,B: Radiograph of the hands of Patient 3 showing hypoplastic first metacarpal bone and absent and hypoplastic phalanges.

380 Rodríguez-Pinilla et al.

Clayton-Smith J, Donnai D. 1995. Fetal valproate syndrome. J Med Genet32:724–727.

DiLiberti JH, Farndon PA, Dennis NR, Curry CJR. 1984. The fetal valpro-ate syndrome. Am J Med Genet 19:473–481.

Graham JM Jr, Sharony R, Garber A, Schreck R, Platt LD, Buehler BA.1991. Preaxial ray reduction defects valproate exposure during preg-nancy. Teratology 43:441.

Hendrickx AG, Nau H, Binkerd P, Rowland JM, Rowland JR, CukierskiMJ, Cukierski MA. 1988. Valproic acid developmental toxicity andpharmacokinetics in rhesus monkey: an interspecies comparison. Tera-tology 38:329–345.

Jager-Roman E, Deichl A, Jakob S, Hartmann AM, Koch S, Rating D,Steldinger R, Nau H, Helge H. 1986. Fetal growth, major malforma-tions, and minor anomalies in infants born to woman receiving valproicacid. J Pediatr 108:997–1004.

Khoury MJ. 1995. Commentary: Contributions of epidemiology to thestudy of birth defects in humans. Teratology 52:186–189.

Lammer EJ, Sever LE, Oakley GP Jr. 1987. Teratogen update: valproicacid. Teratology 35:465–473.

Langer B, Haddad J, Gasser B, Maubert M, Schlaeder G. 1994. Isolatedfetal bilateral ray reduction associated with valproic acid usage. FetalDiagn Ther 9:155–158.

Lindhout D, Meinardi H. 1984. Spina bifida and in-utero exposure to val-proate. Lancet 2:396.

Martínez-Frías ML, Rodríguez-Pinilla E, Salvador J. 1989. Valproate andspina bifida. Lancet 1:611.

Martínez-Frías ML. 1990. Clinical manifestation of prenatal exposure tovalproic acid using case reports and epidemiologic information. Am JMed Genet 37:277–282.

Martínez-Frías ML, Rodríguez-Pinilla E, Prieto L. 1997. Prenatal exposureto salicylates and gastroschisis: a case-control study. Teratology 56:241–243.

Martínez-Frias, ML, Rodríguez-Pinilla E, Bermejo E, Prieto L. 1998. Pre-natal exposure to sex hormones: a case-control study. Teratology 57:8–12.

Mastroiacovo P, Bertollini R, Morandini S, Segni G. 1983. Maternal epi-lepsy, valproate exposure, and birth defects. Lancet 2:1499.

Okada T, Tomoda T, Hisakawa H, Kurashige T. 1995. Fetal valproatesyndrome with reduction deformity of limb. Acta Paediatr Jpn 37:58–60.

Paulson RB, Sucheston ME, Hayes TG, Paulson GW. 1985. Teratogeniceffects of valproate in the CD-1 mouse fetus. Arch Neurol 42:980–983.

Prieto L, Martínez-Frías ML. 1999. Case-control studies using only mal-formed infants: are we interpreting the results correctly? Teratology60:1–2.

Robert E, Guibaud P. 1982. Maternal valproic acid and congenital neuraltube defects. Lancet 2:937.

Robert E, Rosa F. 1983. Valproate and birth defects. Lancet 2:1142.

Sharony R, Garber A, Viskochil D, Schreck R, Platt LD, Buehler BA, Gra-ham JM Jr. 1993. Preaxial ray reduction defects as part of valproic acidembryofetopathy. Prenat Diagn 13:909–918.

Verloes A, Frikiche A, Gremillet C, Paquay T, Decortis T, Rigo J, SenterreJ. 1990. Proximal phocomelia and radial ray aplasia in fetal valproicsyndrome. Eur J Pediatr 149:266–267.

Winter RM, Donnai D, Burn J, Tucker SM. 1987. Fetal valproate syn-drome: Is there a recognisable phenotype? J Med Genet 24:692–695.

Valproic Acid and Limb Deficiencies 381