OUP user menu

Risk of congenital heart defects associated with assisted reproductive technologies: a population-based evaluation

Karim Tararbit, Lucile Houyel, Damien Bonnet, Catherine De Vigan, Nathalie Lelong, François Goffinet, Babak Khoshnood
DOI: http://dx.doi.org/10.1093/eurheartj/ehq440 500-508 First published online: 7 December 2010

Abstract

Aims To estimate the risk of congenital heart defects (CHD) associated with assisted reproductive technologies (ART).

Methods and results We used data from the Paris Registry of Congenital Malformations on 5493 cases of CHD and 3847 malformed controls for which no associations with ART were reported in the literature. Assisted reproductive technologies included inductors of ovulation only, in vitro fertilization, and intracytoplasmic sperm injection. Exposure to ART was higher for cases than controls (4.7 vs. 3.6%, P= 0.008) and was associated with a 40% increase in the maternal age, socioeconomic factors, and year of birth-adjusted odds of CHD without chromosomal abnormalities [adjusted odds ratio (OR) 1.4, 95% confidence interval (CI) 1.1–1.7]. Assisted reproductive technologies were specifically associated with significant increases in the odds of malformations of the outflow tracts and ventriculoarterial connections (adjusted OR 1.7, 95% CI 1.2–2.4) and of cardiac neural crest defects and double outlet right ventricle (adjusted OR 1.7, 95% CI 1.1–2.7). In general, we found specific associations between methods of ART and subcategories of CHD.

Conclusion Cases with CHD were more likely to have been conceived following ART when compared with malformed controls. This higher risk for CHD varied specifically according to the method of ART and the type of CHD and may be due to ART per se and/or the underlying infertility of couples.

  • Congenital heart defects
  • Assisted reproductive technologies
  • Risk factor
  • Epidemiology

Introduction

Congenital heart defects (CHD) are the most prevalent birth defects and comprise the most important cause of malformation-related infant mortality.1,2 Despite the progress in prenatal diagnosis, medical and surgical management of infants with CHD3,4 substantial risks of morbidity and mortality remain for severe cases of CHD. Risk factors of CHD include inherited5 and non-inherited causes,6 among which the role of assisted reproductive technologies (ART) remains uncertain.

Assisted reproductive technologies include various techniques used to achieve pregnancy in the case of male or female infertility and comprise inductors of ovulation, conventional in vitro fertilization (IVF), and IVF with intracytoplasmic sperm injection (ICSI). These techniques are increasingly used in many countries7,8 in part due to trends towards delayed childbearing.9 For example, in France, nearly 2.4% of all children born in 2006 were conceived following ART.10

Children conceived following ART are known to be at higher risk for adverse birth outcomes1115 including those related to multiple births, preterm delivery, and intrauterine growth retardation. There are more uncertainties regarding the risks for birth defects,13,1623 and in particular for specific defects such as CHD.12,2426 The meta-analysis by Hansen et al.18 showed a moderate increase in the overall risk for birth defects in children born after IVF or ICSI compared with children born spontaneously [odds ratio (OR) = 1.40, 95% confidence interval (CI) 1.28–1.53]. However, insufficient data exist regarding specific risks for CHD that may be associated with ART.

Previous studies included a relatively small number of cases of CHD and showed inconsistent results for risks of CHD in relation to ART.12,2426 Moreover, specific associations between subcategories of CHD and different methods of ART have not been adequately assessed in the literature. This is important as most known teratogens and risk factors for birth defects are associated with one or a few specific anomalies.

Using population-based data from the Paris Registry of Congenital Malformations including more than 5000 cases of CHD, we estimated the risk for CHD in relation to different methods of ART for: all CHD, CHD without chromosomal abnormalities, and subcategories of CHD defined based on anatomo-embryological criteria.

Methods

Data

We used data from the population-based Paris Registry of Congenital Malformations,4,27 which registers all cases of birth defects and chromosomal abnormalities among live-borns, still-borns (≥22 weeks of gestation), and pregnancy terminations. The Registry covers the population of women who live in Greater Paris area (Paris and its surrounding suburb) and deliver or have a pregnancy termination in a Parisian maternity unit. The annual number of deliveries in our population is about 38 000.

The Paris Registry is a member of the European Network of Registries of Congenital Malformations (EUROCAT) and of the International Clearinghouse for Birth Defects Surveillance and Research. The Registry follows the EUROCAT methodology and quality of data is routinely monitored by both EUROCAT1 and the National Committee of Registries in France. The review of procedures regarding confidentiality of data is overseen by both the National Committee of Registries and the National Committee of Informatics and Freedom. Data are based on medical records and are collected from several sources including maternity units, neonatology wards, and cytogenetic and pathology services.

Data for the study population corresponded to the period 1987–2006 as the first case of a malformation with exposure to IVF occurred in 1987 and 2006 was the last year for which data were available.

Methods

A case–control study with malformed controls was conducted to estimate the risk of CHD in relation to ART. Cases were children/foetuses with a diagnosis of CHD. For the malformed controls, following Hook's recommendations,28 we included a wide spectrum of malformations for which no association with ART was reported in the literature. The malformed controls were foetuses/children with isolated club-foot, angioma, skin abnormality, polydactyly, syndactyly, or congenital hip dislocation.

We estimated the risk (odds) of CHD when compared with malformed controls in relation to different methods of ART. This risk was estimated for the following outcomes: (i) all CHD combined, (ii) CHD without chromosomal abnormalities, (iii) CHD without chromosomal abnormalities and excluding isolated ventricular septal defects (VSD), and (iv) 10 mutually exclusive subcategories of CHD (Table 1, subcategories 1–10) defined based on anatomo-embryological criteria and classified by consensus by two paediatric cardiologists (L.H. and D.B.). Three additional subcategories (Table 1, subcategories 11–13) were defined by re-grouping, based on their common developmental and genetic origins,29 certain types of CHD included in the 10 preceding subcategories.

View this table:
Table 1

Subcategories of congenital heart defects defined according to anatomo-embryological criteriaa

Subcategories of CHDMalformations includedICD 10
Anomalies of heart positionHeterotaxy syndromeb/isomerism of atrial appendagesQ206
Mirror-image arrangementQ893
Malformations of the outflow tracts and ventriculoarterial connectionsTransposition of great arteries (complete)Q203
Vascular malpositions
 Double outlet right ventricleQ201
 Double outlet left ventricleQ202
Cardiac neural crest defects
 Common arterial trunkQ200
 Tetralogy of FallotQ213
 Aortopulmonary septal defectQ214
 Pulmonary valve atresia + VSDQ220 + Q210
 Atresia of pulmonary arteryQ255
 Interrupted aortic arch/atresia of aortaQ252
 Overriding aortaQ2542
Defects of aortic valves/left outflow tract
 Stenosis of aortic valveQ230
 Insufficiency of aortic valveQ231
Defects of pulmonary valves/right outflow tract
 Pulmonary valve atresiaQ220
 Pulmonary valve stenosisQ221
 Pulmonary valve insufficiencyQ222
 Other defects of pulmonary valveQ223
 Pulmonary infundibular stenosisQ243
Other defects of great vesselsQ258/Q259
Malformations of the atrioventricular valves and atrioventricular connectionsDefects of tricuspid valve
 Tricuspid valve stenosis/atresiaQ224
 Ebstein's anomalyQ225
 Other defects of tricuspid valveQ228
 Unspecified tricuspid valve defectQ229
Defects of mitral valve
 Mitral stenosisQ232
 Mitral insufficiencyQ233
 Unspecified mitral defectsQ239
Common atrium/cor triloculare biventriculareQ2115
Atrioventricular septal defectsQ212
Functionally univentricular CHDDouble inlet ventricleQ204
Hypoplastic left heart syndromeQ234
Hypoplastic right ventricle
 Hypoplastic right heart syndromeQ226
 Cor biloculareQ208
Anomalies of the great arteriesAnomalies of ascending aorta
 Hypoplasia of ascending aortaQ2540
 Supravalvular aortic stenosisQ253
 Aneurysm or dilatation of aortaQ2545
 Aneurysm of sinus of ValsalvaQ2543
Coarctation of aortaQ251
Anomalies of aortic archQ2541/Q2544
Anomalies of pulmonary artery
 Stenosis of pulmonary arteryQ256
 Other defects of pulmonary arteryQ257
Anomalies of coronary vesselsDefect of coronary vesselsQ245
Ventricular septal defectsVentricular septal defectsQ210
Anomalies of the atria and interatrial communicationsCor triatriumQ242
Interatrial communicationsQ2110/Q2112/Q2113/Q2114/Q2118
Anomalies of venous connectionsAnomalous systemic venous connection
 Persistent left superior vena cavaQ261
 Other defects of great veinsQ260/Q265/Q266/Q268/Q269
Anomalous pulmonary venous connection
 Total anomalous pulmonary venous connectionQ262
 Partial anomalous pulmonary venous connectionQ263
 Unspecified anomalous pulmonary venous connectionQ264
Discordant atrioventricular connectionsDiscordant atrioventricular connectionQ205
Unspecified defects of cardiac chamber and connectionsQ209
TGA, heterotaxy syndromeb and discordant atrioventricular connectionsQ206/Q203/Q205/Q209
Cardiac neural crest defects and double outlet right ventricle without ventricular hypoplasiaQ200/Q213/Q214/Q220 + Q210/Q252/Q255/Q2542/Q201 without Q226 and Q234
Isolated atrioventricular septal defectsQ212
  • aSee the Methods section for details.

  • bDefinition of heterotaxy syndrome was based on the article by Jacobs et al.44

The main predictor variable was exposure to ART, which included the following: inductors of ovulation only (IO), IVF, and ICSI. Exposure to ART was assessed: (i) as a binary (ART yes/no) variable and (ii) as a variable in four categories (no ART, IO, IVF, and ICSI), and for IVF and ICSI combined (IVF + ICSI).

Variables considered as potentially confounding factors included year of birth, maternal age, occupation, and geographic origin. These factors are known to be related to both exposure to ART and prevalence of birth defects22,30 even if their specific associations with CHD are not well documented. Maternal occupation was coded according to the French National Institute of Statistics and Economic studies (INSEE) classification: professional, intermediate, administrative/public service, other, and none. These categories represent a gradient from the highest to the lowest occupational group in France. The ‘other' group comprised: artisan/small business owner, shop keeper/shop assistant, service worker, skilled worker, and unskilled worker, for which each category represented a small number of deliveries in our population. Maternal geographic origin was coded as: French, North African, other African, and other countries.

Analyses for the largest three groups of cases (all CHD, CHD without chromosomal abnormalities, and CHD without chromosomal abnormalities and excluding isolated VSD) were also done separately for singletons.

Power

Assuming a type-I error of 0.05, we had a power of 80% to detect an OR of 1.5 for the overall risk of CHD in relation to IVF (exposure ≈2%). Assuming a case–control ratio of 1:4, we had a power of 80% to detect an OR of 2.0 associated with IVF exposure for subcategories of 1000 cases. For subcategories of 100 cases, we had an 80% power to detect an OR of 3.0 or more. For exposure to ICSI (≈0.6%), we had a power of 80% to detect an OR of 2.0 for the overall risk of CHD and 3.0 for subcategories of 1000 cases.

Statistical analyses

The odds of CHD in relation to ART was estimated using logistic regression models for each outcome, after taking into account year of birth, maternal age, occupation, and geographic origin. The adjustment for maternal age was made using fractional polynomials.31

Separate logistic regression models were estimated for each of the following outcomes: (i) all CHD, (ii) CHD without chromosomal abnormalities, (iii) CHD without chromosomal abnormalities and excluding VSD, and (iv) each of the 13 CHD subcategories that included at least 100 cases. The following CHD subcategories were not analysed separately as they included less than 100 cases: anomalies of heart position, anomalies of coronary arteries, anomalies of venous connections, discordant atrioventricular connections, and isolated atrioventricular septal defects. However, these cases were included in the analyses for all CHD/CHD without chromosomal abnormalities/excluding VSD.

We tested whether the effects associated with ART (specifically IO and IVF for which data were available for the entire study period) changed over time; i.e. we tested for interaction effects between IO/IVF and time using nested models with likelihood ratio tests. None of the interaction effects between ART and time were statistically significant (detailed results not shown; available from authors). We also tested for any interaction effects between ART and singleton/multiple births.

The statistical significance level was set at α = 0.05 and all tests were two-sided.

As recommended by Rothman32 and Savitz and Olshan33 in the case of observational studies, such as ours, aimed at detecting patterns of (specific) associations, no adjustment was made for multiple comparisons in analysing the associations between the a priori chosen subcategories of CHD and ART.

Analyses were done with Stata 9 software (Statacorp, TX, USA).

Results

Study population

After excluding data with missing information on ART (2% of cases), the study population included 5493 cases of CHD, 4459 cases of CHD without chromosomal abnormalities, and 3104 without chromosomal abnormalities and excluding VSD (Table 2).

View this table:
Table 2

Number of cases and controls and proportions of children/foetuses conceived after assisted reproductive technologies

CategoryaSubjects with complete data on ART (n)Exposure to ART (%)P-valueb
Controls38473.6
All CHD54934.70.008
CHD without chromosomal abnormalities44594.90.003
CHD without chromosomal abnormalities and excluding VSD31045.00.005
Malformations of the outflow tracts and ventriculoarterial connections10885.60.003
Malformations of the atrioventricular valves and atrioventricular connections6082.60.231
Functionally univentricular CHD4022.50.253
Anomalies of the great arteries3715.60.331
Ventricular septal defects22485.00.006
Anomalies of the atria and interatrial communications1244.80.463
TGA, heterotaxy syndrome, and discordant atrioventricular connections4751.30.362
Cardiac neural crest defects and double outlet right ventricle without ventricular hypoplasia5371.80.014
  • aSubcategories of CHD with less than 100 cases are not shown.

  • bComparison of the proportion of children/foetuses conceived after ART between the subcategory of CHD and the controls.

The subcategories of CHD comprised from 124 to 2248 cases (Table 2; subcategories of CHD with less than 100 cases are not shown—detailed data available from authors). The largest subcategory was the VSD (n= 2248) representing nearly 40% of all cases. The subcategory malformations of the outflow tracts and ventriculoarterial connections was the second largest subcategory and comprised 1088 cases.

After excluding controls with missing data for ART (3.5% of controls), the study population included a total of 3847 malformed controls, comprising isolated congenital hip dislocation (n= 1299), polydactyly (n= 769), club foot (n= 733), angioma (n= 515), skin abnormality (n= 367), or syndactyly (n= 164).

Maternal age was missing for 35 (0.6%) cases and 23 (0.6%) controls. Data on maternal geographic origin were missing for 1.7% of the cases and 2.1% of the controls, and information on maternal occupation was missing for 7.9% of the cases and 3.1% of the controls.

Most socio-demographic characteristics were different between cases and controls. Mothers of cases were older, less often of French origin, and were more likely to be in the occupational category ‘none' than mothers of controls. On the other hand, mothers who had conceived following ART were more likely to be of French origin and to be in the occupational category ‘professional' (highest category) than mothers who had conceived spontaneously (detailed results not shown; available from authors).

Risk of congenital heart defects associated with assisted reproductive technologies

Table 2 shows the proportions of cases and controls that were exposed to ART. Overall, CHD cases were more likely to have been conceived following ART when compared with controls (4.7 vs. 3.6%, respectively, P= 0.008). Exposure to the different methods of ART was also significantly different between cases and controls (P= 0.013); in particular, 1.9% of the cases were born following IVF vs. 1.3% of the controls and 0.6% of the cases were born following ICSI vs. 0.3% of the controls.

Table 3 shows the crude and adjusted associations between the overall risk of CHD and ART (all methods combined). Exposure to ART was associated with a 1.3-fold increase in the maternal age, socioeconomic factors, and year of birth-adjusted odds of all CHD (adjusted OR 1.3, 95% CI 1.0–1.6). For CHD without chromosomal abnormalities and CHD without chromosomal abnormalities and excluding isolated VSD, estimates were comparable to (slightly higher than) those observed for all CHD (adjusted OR 1.4, 95% CI 1.1–1.7 and adjusted OR 1.5, 95% CI 1.1–1.9, respectively).

View this table:
Table 3

Logistic regression analyses of the associations between assisted reproductive technologies (all methods combined) and congenital heart defects

CasesCrude ORa95% CIAdjustedb ORa95% CI
AllAll CHD1.0Ref.1.0Ref.
1.31.1–1.61.31.0–1.6
CHD without chromosomal abnormalities1.0Ref.1.0Ref.
1.41.1–1.71.41.1–1.7
CHD without chromosomal abnormalities and excluding VSD1.0Ref.1.0Ref.
1.41.1–1.81.51.1–1.9
Singletons onlyAll CHD1.0Ref.1.0Ref.
1.10.8–1.51.10.8–1.5
CHD without chromosomal abnormalities1.0Ref.1.0Ref.
1.10.8–1.51.20.9–1.6
CHD without chromosomal abnormalities and excluding VSD1.0Ref.1.0Ref.
1.10.8–1.51.20.8–1.6
  • aOdds ratios represent the odds of a birth (including live births, stillbirths, and pregnancy terminations) with congenital heart disease (cases) relative to the odds of a birth with one of the malformed controls (see the Methods section for details).

  • bAdjusted for maternal age, geographic origin, occupation, and year of birth.

Table 4 shows the crude and adjusted ORs between the risk of CHD and the different methods of ART. The adjusted ORs for IVF and ICSI were similar and the combined IVF + ICSI exposure was associated with a 1.4-fold increase in the odds of CHD (adjusted OR 1.4, 95% CI 1.0–2.9). The combined IVF + ICSI exposure was also associated with a 1.5-fold increase in the odds of CHD without chromosomal abnormalities (adjusted OR 1.5 95% CI 1.1–2.1) and a 1.7-fold increase in the odds of CHD without chromosomal abnormalities and excluding VSD (adjusted OR 1.7 95% CI 1.2–2.4). In contrast, we did not find any statistically significant association between IO and CHD with or without chromosomal abnormalities.

View this table:
Table 4

Logistic regression analyses of the associations between different methods of assisted reproductive technologies and congenital heart defects

Casesmethod of ARTcrude ORa95% CIadjustedb ORa95% CI
All
 All CHDNone1.0Ref.1.0Ref.
Inductors of ovulation only1.10.8–1.51.20.9–1.6
IVF1.41.0–2.01.41.0–2.0
ICSI2.31.1–4.41.40.7–2.7
IVF + ICSI1.61.2–2.11.41.0–1.9
 CHD without chromosomal abnormalitiesNone1.0ref.1.0ref.
Inductors of ovulation only1.20.9–1.61.20.9–1.7
IVF1.51.1–2.11.61.1–2.3
ICSI2.41.2–4.81.40.7–2.9
IVF + ICSI1.71.2–2.31.51.1–2.1
 CHD without chromosomal abnormalities and excluding VSDNone1.0ref.1.0ref.
Inductors of ovulation only1.10.8–1.61.30.9–1.8
IVF1.61.1–2.31.71.2–2.6
ICSI2.51.2–5.21.50.7–3.2
IVF + ICSI1.71.3–2.41.71.2–2.4
Singletons only
 All CHDNone1.0Ref.1.0Ref.
Inductors of ovulation only0.90.6–1.41.00.7–1.5
IVF1.40.8–2.41.20.7–2.1
ICSI1.40.6–3.21.20.5–2.7
IVF+ICSI1.40.9–2.21.20.8–1.9
 CHD without chromosomal abnormalitiesNone1.0ref.1.0ref.
Inductors of ovulation only1.00.7–1.41.10.7–1.6
IVF1.40.8–2.41.30.7–2.3
ICSI1.50.6–3.41.30.5–3.0
IVF + ICSI1.40.9–2.21.30.8–2.1
 CHD without chromosomal abnormalities and excluding VSDNone1.0ref.1.0ref.
Inductors of ovulation only0.80.6–1.31.00.7–1.6
IVF1.40.8–2.51.50.8–2.7
ICSI1.40.6–3.41.20.5–3.1
IVF + ICSI1.40.9–2.31.40.8–2.3
  • aOdds ratios represent the odds of a birth (including live births, stillbirths, and pregnancy terminations) with congenital heart disease (cases) relative to the odds of a birth with one of the malformed controls (see the Methods section for details).

  • bAdjusted for maternal age, geographic origin, occupation, and year of birth.

When analyses were restricted to singletons, ORs for IVF and ICSI remained >1, but the magnitude of associations decreased and CIs included the null value (Tables 3 and 4). However, test of the interaction effect between ART and singleton/multiple births was not statistically significant (P = 0.67).

Analyses for subcategories of congenital heart defects defined according to anatomo-embryological criteria

Table 5 shows the results of the analyses for the subcategories of CHD. Assisted reproductive technologies were associated with significant increases in the odds of malformations of the outflow tracts and ventriculoarterial connections (adjusted OR 1.7, 95% CI 1.2–2.4) and of cardiac neural crest defects and double outlet right ventricle (adjusted OR 1.7, 95% CI 1.1–2.7).

View this table:
Table 5

Logistic regression analyses of the associations between assisted reproductive technologies (all methods combined) and subcategories of congenital heart defects

SubcategoriesCrude ORa95% CIAdjustedb ORa95% CI
Malformations of the outflow tracts and ventriculoarterial connections1.0Ref.1.0Ref.
1.61.2–2.21.71.2–2.4
Malformations of the atrioventricular valves and atrioventricular connections1.0Ref.1.0Ref.
0.70.4–1.20.60.4–1.2
Functionally univentricular CHD1.0Ref.1.0Ref.
0.70.4–1.30.60.3–1.3
Anomalies of the great arteries1.0Ref.1.0Ref.
1.30.8–2.21.30.8–2.3
Ventricular septal defects1.0Ref.1.0Ref.
1.41.1–1.81.31.0–1.6
Anomalies of the atria and interatrial communications1.0Ref.1.0Ref.
1.40.6–3.22.00.8–5.0
TGA, heterotaxy syndrome, and discordant atrioventricular connections1.0Ref.1.0Ref.
1.20.8–2.01.30.8–2.3
Cardiac neural crest defects and double outlet right ventricle without ventricular hypoplasia1.0Ref.1.0Ref.
1.61.1–2.51.71.1–2.7
  • aOdds ratios represent the odds of a birth (including live births, stillbirths, and pregnancy terminations) with congenital heart disease (cases) relative to the odds of a birth with one of the malformed controls (see the Methods section for details).

  • bAdjusted for maternal age, geographic origin, occupation, and year of birth.

The adjusted ORs for ICSI were in general similar to those for IVF except for the subcategories malformations of atrioventricular valves and atrioventricular connections, functionally univentricular CHD and transposition of the great arteries (TGA), heterotaxy syndrome, and discordant atrioventricular connection. The combined IVF + ICSI category was associated with a 1.8-fold increase in the odds of malformations of the outflow tracts and ventriculoarterial connections (adjusted OR 1.7 95% CI 1.1–2.8) and a 1.8-fold increase in the odds of cardiac neural crest defects and double outlet right ventricle without ventricular hypoplasia (adjusted OR 1.8 95% CI 1.0–3.3). IO were associated with a 2.5-fold higher odds of anomalies of atria and interatrial communications (adjusted OR 2.5, 95% CI 0.7–8.7). In general, the estimates of the associations between the different methods of ART and CHD subcategories varied across the different subcategories. However, CIs for smaller subcategories were wide, reflecting the relative imprecision of the estimates (detailed results of the associations between different methods of ART and subcategories of CHD are not shown in the printed version; these results are available from authors/or as Supplementary material online, Table S3).

Discussion

On the basis of data from the Paris Registry of Congenital Malformations including more than 5000 cases of CHD, we found a 40% increase in the overall risk of CHD without chromosomal abnormalities in children conceived following ART after taking into account maternal age, socioeconomic factors, and year of birth. Our results also suggest that specific associations exist between ART and subcategories of CHD. Moreover, although IVF and ICSI were associated with significant increases in the risk of CHD, we did not find a significant association between IO and the overall risk of CHD.

When analyses were restricted to singletons only, the ORs decreased and the CIs included the null value. This suggests that any effect of ART on CHD may be in part due to multiple births. On the other hand, test of the interaction effect between ART and singletons/multiple was not statistically significant, although this may have been due to limited power of our study for detecting any interaction effects that may have existed. In any case, it is certainly possible that multiple births may be on the causal pathway between ART and CHD. This hypothesis is consistent with Reefhuis's26 finding that ‘multiple births were more likely to have birth defects, regardless of conception mode'. It should be noted however that the public health impact of ART on the risk for birth defects includes all (singleton and multiple) births.

On the basis of our findings, we calculated attributable risk fractions, which would represent the proportion of cases that may be caused by ART, or equivalently, the proportion of cases that would be avoided were the exposure to ART removed ceteris paribus, ‘if' the associations we found between the risk of CHD and ART can be assumed to represent causal relations (this may of course not be the case for reasons that are discussed further below). The attributable risk fraction estimates suggest in particular that around 2% of the CHD without chromosomal abnormalities may be caused by ART and 1.2% by IVF + ICSI. These proportions are similar to those calculated for certain other major risk factors of CHD34 and may increase as exposure to ART is likely to increase over time.

To our knowledge, there are no embryological or physiopathological hypotheses to explain the specific associations between ART and the subcategories of CHD that we found in our study. In particular, genomic imprinting disorders found in children born after ART35 seem unlikely to explain all of the observed associations. Moreover, our results for CHD without chromosomal abnormalities suggest that associations between CHD and ART are not due to the association of the latter with chromosomal abnormalities. Our results may be helpful for generating hypotheses regarding underlying mechanisms for the association between the risk of CHD and different methods of ART.

Our study has certain limitations. Although our study included a large number of cases of CHD, exposure to ART, and particularly to ICSI, remains infrequent in the general population and in our Registry data. Consequently, the CIs for the estimates of the associations between subcategories of CHD and IVF and more so ICSI were wide, indicating the limited precision of some of our estimates.

Moreover, we made no adjustment for multiple comparisons in analysing the associations between the a priori chosen subcategories of CHD and ART. Although this is consistent with the recommendations by Rothman32 and Savitz and Olshan33 in the case of observational studies, such as ours, aimed at detecting patterns of specific associations, controversies remain among experts as to the right statistical approach to multiple comparisons. In any case, the specific associations we found can only be considered exploratory and need to be further investigated.

The choice of malformed controls may have been a source of selection bias36,37 due to associations that may exist between exposure to ART and malformations included as controls. In order to minimize such bias, following Hook's recommendations,28 we selected a wide spectrum of malformations for which no associations with ART were described in the literature. This methodology is often used in the field of birth defects38 and in particular in assessing the teratogenic effects of medications.39,40

Although this approach of choosing a heterogeneous group of malformations as controls can lower the risk of selection bias by diluting any effects due to unknown (i.e. hitherto unreported in the literature) associations that may exist between ART and one or more of the malformations selected as controls, the possibility of residual bias cannot be excluded. This bias may result in an underestimation of the true association (i.e. what one would observe with an ideal set of non-malformed controls) between ART and CHD if ART is associated with an increase in the risk of the malformations included as controls. Conversely, the overestimation of the associations between ART and CHD can occur if ART is associated with a decrease in the risk of malformations included as controls, although such an association seems unlikely. Moreover, a differential misclassification bias for exposure assessment cannot be excluded if exposure to ART is ascertained in a different way in the case of CHD compared with the malformations included as controls. However, we are not aware of any a priori reason or empirical evidence to suggest that such a bias may exist for ART.

The frequency of missing data on exposure was different between cases and controls. This may have resulted in biased (over or under) estimates of the associations between CHD and ART if the distribution of exposure was different between subjects with complete data and those with missing data. Nevertheless, the overall frequency of missing data in this study was low and we have no reason to believe that the distribution of exposure was indeed different for cases and controls with missing data.

We explored specific associations that may exist between different types of CHD and ART by conducting separate analyses for subcategories of CHD defined a priori based on anatomic and/or embryological criteria. An important caveat that needs to be considered is that our criteria for defining these subcategories can be arguable; alternative and equally, if not more, valid subcategories may be envisaged. Notwithstanding these considerations, our results suggest that indeed specific associations may exist between ART and risk of specific CHD subcategories, without necessarily implying that the subcategories investigated in our study are the most appropriate ones to use in this setting.

Another potential limit of our study is related to the effects of possible confounding factors that could not be taken into account.41 In particular, we did not have adequate data for folic acid and/or multivitamins intake for women in our population. Lack of adjustment for multivitamins use could have resulted in an underestimation of the risks associated with ART as: (i) multivitamins/folic acid intake has been shown to be associated with a lower risk of CHD42 and (ii) a higher proportion of women who conceived after ART may have had an adequate multivitamins/folic acid intake.26

Another potentially important confounding variable that we could not adjust for was paternal age. As paternal age is correlated with maternal age, we partially adjusted for paternal age by taking maternal age into account. Nevertheless, lack of full adjustment for paternal age may have resulted in an overestimation of risks, particularly for ICSI that is used more often, although not exclusively, in the case of male infertility.8

More broadly, in the association between ART and CHD, the question of the role of underlying infertility vs. that of any treatment effects of ART per se remains an open one.14,16,17,43 Some studies have attempted to separate the effects of underlying infertility from any treatment effects due to ART by adjusting for the duration of infertility/involuntary childlessness.17 This strategy has been criticized by Hansen et al.19 as this variable may be ‘synonymous with exposure'. This is consistent with the data from our Registry in which 99% of the cases/controls with exposure to ART had a duration of infertility of 2 years or more.

Although for most subcategories of CHD the associations between IVF and ICSI were similar, our findings suggest that ICSI may be associated with a higher risk for some subcategories of CHD when compared with conventional IVF. However, CIs for the estimates of the risks associated with ICSI for subcategories of CHD were wide. Hence, while suggestive, our results cannot lead to a definitive conclusion regarding higher risks due to ICSI vs. IVF for certain subcategories of CHD.

In conclusion, we found that cases with CHD were more likely to have been conceived following ART when compared with malformed controls. In particular, IVF and ICSI were associated with a 1.5-fold increase in the odds of CHD without chromosomal abnormalities after adjustment for year of birth, maternal age, occupation, and geographic origin. In contrast, we did not find a significant association between IO (alone) and the overall risk of CHD. In general, estimates suggested that there may be specific associations between different methods of ART and subcategories of CHD classified based on anatomo-embryological criteria. Such specific associations may reflect causal effects due to ART and/or the underlying infertility of couples who conceive following ART.

Funding

This work was supported in part by grants from the Assistance Publique—Hôpitaux de Paris Fonds d'Etude et de Recherche du Corps Médical (Paris, France) (to K.T.) and the Agence de Biomédecine (Saint-Denis La Plaine, France) (to B.K.). The Paris Registry of Congenital Malformations received financial support from INSERM (Paris, France) and the Institut de Veille Sanitaire (Saint-Maurice, France).

Conflict of interest: none declared.

References

View Abstract