In this study, we found that the overall incidence of congenital anomalies in pregnancies where prescriptions for statins had been filled in the first trimester of pregnancy (3/64, 4.69%) was not statistically greater than the incidence in those pregnancies where prescriptions for fibrates only had been filled in the first trimester (3/14, 21.43%), or where the statins had been stopped at least 1 month before conception (7/67, 10.45%). There was no evident pattern in the types of congenital anomalies among the live births. However, the sample size of our study was small and lacked sufficient power to detect small increases in overall risk among those taking statins during pregnancy. Furthermore, the number of cases ascertained was also very small. We acknowledge that this is a limitation in drawing inferences from the statistical tests of significance. Nevertheless, this is an issue common to many such studies published so far.
The congenital anomalies detected in the three cases where statin prescriptions had been filled in the first trimester involved lovastatin, simvastatin and atorvastatin. No malformations were detected among the 11 infants exposed to pravastatin. One prevailing theory is that statins with high affinity for lipid environments (including, simvastatin > lovastatin > atorvastain > cerivastatin > fluvastatin) more readily enter extrahepatic tissues, including the embryo during pregnancy, and thus have a greater potential to downregulate cholesterol biosynthesis [11
]. In case reports of statin-exposed pregnancies and outcomes [11
], all cases of adverse outcomes at birth were associated with lipophilic statins; no malformations were reported among infants exposed to pravastatin.
Studies on humans to date have not shown statins to be potent teratogens or to cause particular patterns of congenital anomalies. In our study, the 3/64 cases of live babies with congenital anomalies associated with filling prescriptions for statins in the first trimester of pregnancy consisted of an unspecified anomaly of the heart, a ventricular septal defect and an atrial septal defect. Similarly, another study [11
] involving the evaluation of reports on first-trimester statin exposures from the FDA's Medical Products Reporting Program reported 18 live births with structural anomalies. Detected anomalies included: limb deficiencies (n
= 2), cleft palate (n
= 1), cleft lip (n
= 2), oesophageal atresia (n
= 1), spina bifida (n
= 1), duodenal atresia (n
= 1), polydactyly (n
= 1), hypospadias (n
= 1), constriction of pyelourethral junction (n
= 1), clubbed foot (1), vertebral, anal, tracheo-oesophageal, renal (VATER) (n
= 1), aqueductal stenosis (n
= 1), ‘severe deformity’ (n
= 1), microtia (n
= 1), ‘cardiovascular defect’ (n
= 1) and an atrial/ventricular septal defect/aortic hypoplasia (as subsequently corrected [12
]). The statins implicated were atorvastatin, simvastatin and lovastatin.
Most studies of associations of human birth defects, such as ours, are undertaken in live births. However, live births represent only part of the population in which such adverse outcomes may be detected [28
]. Although the proportion of those congenital anomalies that can be attributed to medication exposure is low compared with other environmental causes [29
], it is still feasible that some anomalies not captured by our analysis, which considered only live births, could also have been associated with medication use. For example, if the number of cases in the pregnancies lost to ascertainment from our study group A exceeded the number of observed cases, the opposite results to what we reported could have been produced. This is an important limitation, but a problem common to many studies of the associations of birth defects in humans. Under these circumstances the possibility exists for the Yule–Simpson effect (Simpson's paradox) [30
], a statistical paradox in which an effect in two groups measured separately (such as congenital anomalies among nonlive births and among live births) is reversed when the groups are combined. This effect has been examined quantitatively [28
], illustrated by considering maternal female sex hormone exposure and congenital cardiovascular defects. Clearly, congenital anomalies in nonlive births could have an important impact on OR estimates.
To our knowledge, there is no extensive evidence suggesting that the indication for antilipaemic treatment could itself be a cause of adverse fetal outcome other than a case report of possible intrauterine growth restriction [8
]. However, common co-morbidities associated with the use of antilipaemic medications such as hypertension, hypothyroidism and diabetes could increase the risk of adverse fetal outcome. Diabetes, for example, is known to be strongly associated with major congenital anomalies and perinatal mortality [31
] and is an aetiological factor for abortion [29
]. In this study, care was taken to select groups of women that were comparable to account for any possible effects of indication; two comparator groups were selected, with both groups consisting of women who had previously had or currently had an indication for antilipaemic medication treatment.
In this database study we were able to assemble a cohort of all women known to have been prescribed statins or fibrates within a population, and ascertain pregnancy status. This method allowed us to ascertain a denominator (a defined population of exposed individuals) from which the cases (the numerator) are drawn. Controlled epidemiological studies are considered to be more appropriate than uncontrolled case reports in establishing a relationship between gestational drug exposure and pregnancy outcomes in humans [11
]. Nevertheless, there is a need for caution in trying to identify patterns, particularly those concerning anomalies that are relatively common, when study populations are small.
Pharmacoepidemiolgical studies based on administrative databases are important and may yield valuable information despite limitations of the incompleteness of information on potential confounders or on certain clinical variables of interest. In this study, we did not have access to information such as the reasons for the termination of a pregnancy. Information on other potential confounding variables such as smoking and folic acid intake were not available. We also acknowledge that an assumption of our study was that those who filled a prescription were also sufficiently exposed to the medicines, but absolute noncompliance with medications obtained by pregnant women is known to be low, at about 8% [34
An advantage of database studies that include the routine collection of information on dispensed drugs, including name, dose and amount dispensed, is the avoidance of limitations associated with the need for long-term recall by study subjects (that may result in recall bias). They allow the rapid assembly of cohorts that would otherwise be costly and time consuming if done prospectively. Such database research provides valuable information in the investigation of associations that might have an important public health impact.
The prevention of severe hypercholesterolaemia during pregnancy remains desirable. However, for the majority of women discontinuation of antilipaemic drugs such as statins should have little impact on them or their babies. Given the continuing uncertainty over the safety of statins for the fetus, the avoidance of their use during pregnancy remains the best advice; consequently, women on statins should plan their pregnancies to guarantee the best outcome.
In conclusion, we did not detect a pattern in congenital anomalies or find evidence of an increased risk of fetal congenital anomalies in the live-born infants of women filling prescriptions for statins in the first trimester of pregnancy, compared with women on fibrates or those who stopped statins before pregnancy. However, studies of larger populations that include more information on nonlive births are necessary before confirming or refuting an association between statins and fetal congenital anomalies.