Results

Study selection and characteristics

The study identification and inclusion flow chart were prepared in compliance PRISMA guidelines.32Fig.1 showed the PRISMA flowchart. Our search yielded 1558 records of which 39 proved potentially eligible on the basis of titles and abstracts (see Supplementary Table S2 for the reasons and lists of references excluded in full-text screening) and 19 articles proved eligible.33-51 There were 11 cohort studies,33-435 case-control studies,44-48 2 case report studies49-50 and 1 case series study.51 Supplementary Table S2 showed the excluding studies.
Ten of the including studies originated from USA, three from Canada, two from Sweden, other apart from Turkey, Australia, Canada, Denmark and Israeli. The eligible studies also used different types of control groups which included chlorpromazine, unexposed, other anti-emetics, metoclopramide, NTE, meclizine and RxAE. Both approaches aimed to rule out confounding by indication. Colvin et al., Berard et al., Asker et al., Dormuth et al., Lemon et al., Huybrechts et al., Suarez et al., Pasternak et al., Werler et al., Zambelli-Weiner et al. had one control group, unexposed for comparison. ÖZDEMİRCİ et al had one control group, Chlorpromazine for comparison. Fejzo et al. and Einarson et al. had two different control groups, diseased-matched and unexposed, for comparison. Parker et al. had two different control groups, RxAE and unexposed, for comparison. Sakran et al. evaluated the rates of malformations in the metoclopramide-exposed pregnancies within the same cohort as the ondansetron exposed group, while Danielsson et al. also took a similar aprroach and asessed the outcomes of the meclizine-exposed pregnancies. There were also two case report studies and one case series study. Table 1 presents the characteristics of 19 studies.

Methodological quality of individual study

Most (94.74%,18/19) of the eligible studies were categorized as being high methodological quality by assessment of with the Newcastle-Ottawa scale, and the only one (5.26%) low quality study was received by Einarson et al..40 Three cohort studies33,35,40 and three case-control studies44,46,48 at high risk of bias for assessment of exposure because they did not report any information or citations on the validity of the ondansetron measure or measured only at baseline. Ten cohort studies at high risk of bias for comparability of cohorts on the basis of the design or analysis34-43, and three case-control studies44,47-48 at high risk of bias for study controls for any additional factor. Four (36.36%) cohort studies33-34,37,40 at high risk of bias for assessment of outcome. One cohort study42 at high risk of bias for adequacy of follow up of cohorts. Two case-report studies and one case-series report study did not assess the risk of bias. Supplementary Fig S1 showed the risk of bias of the including studies.

Overall congenital malformations

Nine studies assessing total number of 293,134 ondansetron-exposed and 8,443,119 control infants reported congenital malformations in this analysis.35-36,38-39,41,43,45,47-48 No significant increase in the rate of overall congenital malformation was detected following ondansetron use during pregnancy (OR=1.10,95% CI:0.94–1.29, Low certainty) in our primary analysis (Fig 2). We observed no suggestion of a subgroup effect (Supplementary Fig S2). No statistically significant publication bias was found for overall congenital malformations (p =0.60) (Supplementary Fig S3). Sensitivity analysis by excluding anyone study showed similar results.

Heart defects

Five studies assessing a total number of 104,763 ondansetron-exposed and 2,687,298 control infants reported heart defects in this analysis.39,41,45,46,48Significant increase in the rate of heart defects was detected following ondansetron use during pregnancy (OR=1.06,95% CI:1.02–1.11, Moderate certainty) in our primary analysis (Fig 3). To confirm the types of heart disease, we conducted subgroup analyses of the different types of congenital heart disease (Fig S4). Further research showed that significant increase only in the rate of other circulatory defects was detected following ondansetron use during pregnancy (OR=1.11,95% CI:1.02–1.20). For heart defects, the pooled effect of studies with ORs was not different from those with RRs (Supplementary Fig S5). Sensitivity analysis by excluding anyone study showed similar results.

Cleft palate

Three studies assessing a total number of 100,837 ondansetron-exposed and 2,656,466 control infants reported cleft palate in this analysis.41,45,48 No significant increase in the rate of cleft palate was detected following ondansetron use during pregnancy (OR=0.78,95% CI:0.37–1.64, Very Low certainty) in our primary analysis (Fig 4). For cleft palate, the pooled effect of studies was all with ORs. The sensitivity analysis lines were identical after excluding any of the studies.

Other organ malformations

Three studies assessing a total number of 78,307 ondansetron-exposed and 2,299,834 control infants reported other organ malformations in this analysis.45,47,48 Significant increase in the rate of other organ malformations was detected following ondansetron use during pregnancy (OR=1.09,95% CI:1.03–1.16, Moderate certainty) in our primary analysis (Fig 5). To confirm the types of organ malformations, we conducted subgroup analyses of the different types of organ malformations (Fig S6). Significant increase in the rate of craniosynostosis, diaphragmatic hernia, laryngeal cleft, relatively severe malformation and other defects were detected. The pooled effect of studies was all with ORs. The sensitivity analysis lines were identical after excluding any of the studies.

Stillbirth or Miscarriage

Six studies assessing a total number of 190,193 ondansetron-exposed and 4,034,373 control infants reported stillbirth or miscarriage in this analysis.33,35,38,42,43,46 Significant decrease in the rate of stillbirth or miscarriage was detected following ondansetron use during pregnancy (OR=0.60,95% CI:0.40–0.91, Low certainty) in our primary analysis (Fig 6). For stillbirth or miscarriage, the pooled effect of studies with ORs was not different from those with RRs/HRs (Supplementary Fig S6). No statistically significant publication bias was found for stillbirth or miscarriage (p=0.62) (Supplementary Fig S7). The sensitivity analysis lines were identical after excluding any of the studies.

Preterm birth

Three studies assessing a total number of 4,661 ondansetron-exposed and 9,102 control infants reported preterm birth in this analysis.33,42,43 No significant increase in the rate of cleft palate was detected following ondansetron use during pregnancy (OR=0.78,95% CI:0.37–1.64, Low certainty) in our primary analysis (Fig 7). For stillbirth or miscarriage, the pooled effect of studies with ORs was not different from those with HRs (Supplementary Fig S8). Sensitivity analysis by excluding anyone study showed similar results.

Other pregnancy outcomes

Three studies assessing a total number of 3,842 ondansetron-exposed and 104,778 control infants reported other pregnancy outcomes in this analysis.35,42,43 No significant increase in the rate of other pregnancy outcomes was detected following ondansetron use during pregnancy (OR=1.08,95% CI:0.92–1.27, Low certainty) in our primary analysis (Fig 8). To confirm the types of other pregnancy outcomes, we conducted subgroup analyses of the different types of other pregnancy outcomes (Fig S9). Significant increase in the rate of birth length were detected. For other pregnancy outcomes, the pooled effect of studies with ORs was not different from those with HRs (Supplementary Fig S10). Sensitivity analysis by excluding anyone study showed similar results. The included case-report studies reported the outcomes included ALT/AST, pharmacokinetic changes, intrauterine growth retardation [IUGR], transient tachypnea, a mild hydrocele and extrarenal pelvis.