Systematic Review With Meta-analysis

The Risks of Proton Pump Inhibitors During Pregnancy

Cheng Mei Li; Alexandra Zhernakova; Lars Engstrand; Cisca Wijmenga; Nele Brusselaers


Aliment Pharmacol Ther. 2020;51(4):410-420. 

In This Article


This comprehensive meta-analyses based on 26 studies assessing the risk of different pregnancy outcomes among women exposed to PPI or H2RA during pregnancy is the first meta-analysis providing evidence that PPI use during pregnancy may be linked to congenital malformations in humans, with significant pooled results in case-control studies and borderline significant in cohort studies. These results should be without doubt interpreted with caution, since the data are based on published results (no individual patient data were available), originating from a relatively heterogeneous group of observational studies, yet statistical heterogeneity seemed limited. Our main analyses combined all studies reporting on at least one congenital malformation, yet different subgroups were assessed, by study design, type of malformation, type of drug and timing of exposure. Although power was low in these analyses, the results seemed consistent in the direction of effect.

For any teratogenic exposure, it is unexpected that the risk of all congenital malformations would be increased (complicating detection[61]), yet the low risk of each individual malformation combined with the low prevalence of exposure (<1% in many of the included studies) makes it even more challenging to detect any potential association. Therefore, potential associations with congenital malformations based on case-control studies are considered closer to reality than cohort studies.[61]

Although quality was assessed to be good for most studies, most studies do have some methodologic limitations, in particular relating to selection bias, since only 10 studies were population-based covering a well-described geographical region.[39,41,45,46,50–54,56] Therefore, some selection process occurred when selecting exposed and non-exposed women in the cohort studies, or cases but especially controls in case-control studies. An important limitation for assessing the risk of congenital malformations, is that timing of exposure was not always clearly mentioned, although generally known to be only of importance if exposure occurred during the critical organ-forming period (1st trimester). Therefore, even if exposure was not clearly defined, it could be assumed that the majority of the exposure occurred during the first trimester since most drug-safety studies during pregnancy focus on congenital malformations. For assessing the other outcomes, the critical exposure window has not yet been established. Recall bias seemed to be of importance in the case-control studies, were information on drug exposure was collected in five out of six studies after the outcome had occurred (potentially contributing to higher pooled risk estimates for case-control studies compared to cohort studies);[61] and over-the-counter use may have been missed when based on prescribed records (13 out of 25 studies were based on any use), while compliance to prescriptions (actual use) is not reported. Insufficient information was provided to allow assessment of (accumulated) dosage or indication of use. The effect of pre-gestational exposure could not be assessed in the current study because only reported in sufficient detail in one study,[54] although five studies considered the 30 days before pregnancy as gestational exposure.[41,48,51,53,56] Congenital malformations were without doubt the most commonly reported adverse event. Yet, taking into account the rarity of congenital malformations (approximately 2.5%[62]) combined with the low prevalence of PPI use during pregnancy (<1% in several of the included studies), many of the studies were clearly underpowered to detect differences. Although we restricted our inclusion criteria to English articles only, no eligible articles were identified in other languages, so the effect of language bias should be negligible. One additional yet important limitation of this meta-analysis is that a relatively large proportion of the included studies (7/25) reported some conflict of interest through pharmaceutical sponsoring or connections; which may contribute to the fact that none of the single studies showed strong evidence for an association with these outcomes before. We cannot rule out a negative publication bias for sponsored studies that did find an association, also since they would go against the common belief that these drugs are safe. Therefore, we searched any unpublished trials through, NIH Clinical Research Studies and the EU Clinical trial registry, but did not identify any relevant studies. Nevertheless, a reporting bias may also occur, where positive associations were more likely to have reported.

To assess the clinical importance, it remains important to note that the absolute risk of congenital malformations is low. Globally, only one in 200–300 boys develops hypospadias, and even with a 46% risk as estimated in this paper (although not significant), this would correspond to only 0.5% of all cases of hypospadias being related to PPI exposure (population attributable fraction), and 31% of all cases of hypospadias occurring to mothers exposed to PPIs (attributable fraction), assuming causality and a 1% exposure to PPIs in the pregnant population. Since this risk would still "only" be one in approximately 200, this is clearly less than high-risk teratogens affecting at least 25% of all exposed foetuses, and moderate teratogens with a relative risk between 2 and 10.[61]

Previous meta-analyses showed results in line with the presented findings, but failed to reach statistical significance.[11,12,16] Yet, we identified 13 studies which have been published since, increasing power and enabling subgroup analyses (although power was still limited). Yet, power was still limited for the different subgroup analyses, which may also explain why no significant increased risk of congenital malformations was found when comparing PPIs with H2RA directly. Power calculations (assuming alpha = 0.05 and power = 0.80) showed that, assuming a 1% use of PPIs among pregnant women, 0.85 million pregnancies are needed to detect a 45% increased risk of hypospadias (assuming ~1/200) or 0.48 million for detecting a 30% increase for all congenital malformations assuming ~2/100). With only two studies over 800 000 pregnancies,[41,54] and three more over 200 000,[41,45,46] it is not surprising that most individual studies did not reach statistical significance, and that case-control studies are usually preferred for investigating congenital malformations. Although the described potential associations are unlikely due to direct effects of PPIs, indirect effects may occur through changes in the maternal microbiome, in the gut, vagina or other niches.[8–10] The increasing amount of studies linking all sorts of outcomes to long-term PPI use in adults (incl. cancer, kidney disease, mortality, cardiac events, …),[63] combined with potential associations with other outcomes described for in-utero exposure (including asthma[14]), we believe a risk-benefit assessment for use during pregnancy is warranted, and more active discouragement of unsupervised over-the-counter use.

Future research is needed to establish or narrow down the harmful window of exposure, in particular for non-teratogenic outcomes such as preterm birth; and the effect of pre-gestational exposure. This would ideally be based on sufficiently powered nationwide complete data on PPI use (both prescribed and over-the-counter), with exact information on compliance, dosage and timing of exposure, enabling assessment of competing effects of the different outcomes such as spontaneous abortions and termination of pregnancy (for medical reasons).

To conclude, this meta-analysis is the first to show an increased risk of congenital malformations associated with PPI use during pregnancy, in particular when based on case-control studies, although power was low in all subgroup analyses. These findings suggest that PPIs should ideally be restricted to well-described indications if no other available treatment options are available and if the benefits outweigh the risks. No conclusions could be drawn for the other pregnancy outcomes.