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


Study Selection

The systematic literature selection is presented in Supplement 2, and resulted in 26 eligible studies reporting on PPI and/or H2RA use during pregnancy and at least a single type of neonatal adverse events.[34–60] The reasons for exclusion of articles excluded during full text evaluation are reported in Supplement 1. Two of the included studies were only identified through backward and forward citation tracking,[48,52] and one study[59] was only referred to as personal communication in two previously published meta-analyses of which one included the same author.[11,12] Two included studies were only included as conference abstracts; we therefore contacted the authors (unsuccessfully) to complement the available data.[39,55]

Study Characteristics

Supplement 3 includes all extracted study and population characteristics. All 26 included studies (20 cohort studies, 6 case-control studies) were published between 1985 and 2019, of which more than half were published during the last decade. Of these 26 studies, 14 described the risk for neonatal adverse events among PPI users only,[39–41,46,47,51–55,57–60] five on H2RA users only,[38,42,44,49,50] and seven reported on both.[34,36,37,43,45,48,56] Fourteen different countries contributed with data, with five studies from the United States, four from Denmark,[41,52–54] three from Sweden[45,46,60] and Israel,[40,50,51] and four international collaborations.[40,44,47,56] Two Swedish registry-based studies had an overlap considering the covered time period, so the oldest study was only used to assess H2RA which was not covered in the more recent study.[45,46] Two Hungarian studies were also based on the same cohort, yet one assessed the risk in individuals with peptic ulcers,[43] and the other one with chronic dyspepsia,[34] so since the participants did not overlap, both were included in the analyses. Two Danish studies[33,46] and two American studies[36,48] reported on hypospadias based on the same nationwide data collection and study period, so the oldest one was excluded from the specific analyses on hypospadias.[41,54] The total number of pregnancies included in each study ranged between 42[38] and 852 483;[52] and the number of congenital malformations between 2[38] and 22 843.[43]

Risk of Bias of Included Studies. There were 20 cohort studies of which 16 (80.0%) had at least seven stars according to the NOS quality scale [37–42,44–47,49–54,56,58–60] and six case-control studies of which two (33.3%) had at least seven stars (Supplement 4).[34,36,43,48,55,57] The risk of bias has been assessed in more detail in Supplement 3: five studies recruited women through a teratogen information service, that is, women who were exposed to potentials teratogens called in for advice, with controls also being selected among callers exposed to non-teratogens.[40,44,47,49,59] Two studies were prescription-event monitoring studies,[38,58] one recruited symptomatic individuals (with hyperemesis gravidarum),[42] and 18 studies were based on national or regional data collection systems.[34,36,37,39,41,43,45,46,48,50–57,60]

Exposure information (drug use) was collected during pregnancy in all but six studies (including five out of six case-control studies),[34,36,43,48,57,58] and 14 studies define exposure as any use, while eight restricted to prescribed use only (unclear in four studies). Three studies looked exclusively at exposure during the first trimester,[36,45,52] while four did not report the exposure window of interest.[37,38,42,59] Indications of use were only reported in eight studies,[40,47,49] of which five were restricted to one diagnosis (nausea and vomiting,[36] hyperemesis gravidarum,[42] peptic ulcers,[37,43] chronic dyspepsia).[34]

Seven studies reported potential conflicts of interest or received funding from pharmaceutical companies.[34,38,47,50,56,57,59]

Congenital Malformations

In total, 23 of the included studies reported at least one type of congenital malformations, 18 assessing PPI use,[34,36,39–41,43,46–48,51–59] and 11 H2RA use.[34,36,38,42–45,47–50,56] Of these, one study reported no events in the exposed group,[58] and one in the non-exposed group.[38] Four studies only reported on specific congenital malformations, including hypospadias,[41,48] club foot[57] and hydrocephalus,[52] and one additional study only reported on a selected number of congenital malformations, but did not report the overall numbers or risks.[36]

The overall risk of congenital malformations was 30% higher among PPI users than in the control group (OR = 1.28, 95% CI 1.09–1.52, N = 18), yet no increased risk was found for H2RA users (OR = 1.05, 95% CI 0.83–1.34, N = 11) (Table 1). The risk was highest in the case-control studies (OR = 2.04, 95% CI 1.46–2.86, N = 6) and the confidence intervals were close to reaching statistical significance for the cohort studies (OR = 1.12 95% CI 0.99–1.27, N = 12) (Figure 1). Pooled effect sizes could be calculated for hypospadias,[41,48] cleft palate without cleft lip,[36,54] cardiovascular malformations[46,51,54,55] and clubfoot,[54,57] all with increased yet not significant effect sizes (Table 1). The subgroup analyses on timing of exposure or type of PPI did not reach significance since most studies did not report the necessary stratified information or did not specify the exposure window clearly. Yet, none of these subgroup analyses resulted in decreased effect sizes and statistical heterogeneity was low. The six studies comparing the risk of any congenital malformation for PPIs directly with H2RA, provided a pooled odds ratio of 1.22 (95% CI 0.67–2.23).[34,36,43,47,48,56]

Figure 1.

Forest plot showing the risk of congenital malformations in women exposed to proton pump inhibitors during pregnancy, by study design

The cumulative meta-analysis including all studies reporting at least one congenital malformation indicates that if this meta-analyses would have been done any time after the large Danish study in 2010,[54] a significantly increased risk of at least 20% would have been found (Figure 2). The influence analysis (Supplement 5) on the same group of studies shows that no single study tips the results into significance, since every single study can be removed from the analyses without losing significance. The funnel plot did not provide evidence for the presence of small study effects (Supplement 6).

Figure 2.

Cumulative meta-analysis showing the cumulative evidence for the association between proton pump inhibitor use during pregnancy and the risk of congenital malformations

Spontaneous Abortion and Stimulated Termination of Pregnancy

The association between PPI[47,58] or H2RA[44,47,49] use and spontaneous abortion or termination of pregnancy were only reported in three studies each (Table 2). For the PPI studies, heterogeneity was low, with none of the associations reaching statistical significance. For H2RA, lower risk for spontaneous abortion was found (OR 0.66, 95% CI 0.43–1.00).

Still Birth and Perinatal Mortality

The risk of stillbirth was assessed for PPI users in five studies,[40,53,56,58,60] and two studies for H2RA.[44,56] Although the risk seems increased for both groups, the confidence intervals were broad and the results were not significant (Table 2). Two studies reported on PPI use and perinatal mortality,[46,51] showing no evidence for an association.

Low Birth Weight and Small for Gestational Age

Five studies reported on low-birth weight or SGA for women exposed to PPIs.[46,51,53,56,60] No analyses reached statistical significance (Table 2).[46,53] Insufficient information was available on extreme low birth weight (only reported in one study[50]), or the average gestational weight for PPI/H2RA users and controls (mean and standard deviation only reported in two studies for PPIs[47,60] and one for H2RA[49]).

Preterm Birth

Seven studies reported on the association between PPI use and preterm birth [40,47,51,53,56,58,60] and five on H2RA use.[44,47,49,50,56] None of the analyses reached statistical significance, except for a potential increased risk for H2RA (1.25, 95% CI 1.02–1.56) (Table 2). Mean gestational length could not be assessed, since only two studies reported means and standard deviations for PPIs,[47,60] and one for H2RAs.[49] The funnel test did not provide evidence for the presence of small study effects (Egger test: P = 0.248).