Pregnancy after bariatric surgery and adverse perinatal outcomes: A systematic review and meta-analysis

1Obesity is a global public health challenge with over 650 million adults affected worldwide, and prevalence continues to rise, making obesity the most common medical condition in women of reproductive age. Maternal obesity, defined as prepregnancy body mass index (BMI) ≥ 30 kg/m2, has severe implications for both mother and baby. Maternal risks include higher likelihood of gestational diabetes, preeclampsia, and cesarean section. For the neonate, there is increased risk of pre- and postterm birth, small and large for gestational (SGA/LGA), congenital anomalies, and perinatal mortality. Interventions to reduce maternal obesity are important not only to improve pregnancy outcomes but also to reduce the long-term health burden on the mother and offspring, including cardiovascular disease and insulin resistance.

2Bariatric surgery is the most effective treatment for long-term weight loss, and over half of surgeries are performed on women of reproductive age. Women who undergo bariatric surgery prior to pregnancy are less likely to experience comorbidities associated with obesity, such as gestational diabetes and hypertension. However, micronutrient deficiencies are increased after bariatric surgery and may therefore have implications for fetal environment. Maternal deficiencies in folate, iron, and vitamin D, for example, are all linked with adverse perinatal outcomes including neural tube defects, preterm birth, and low birth weight. Malabsorptive procedures such as Roux-en-Y gastric bypass (RYGB) and biliopancreatic diversion (BPD) reduce the absorption of micronutrients because part of the small intestine is bypassed, whereas restrictive procedures such as laparoscopic adjustable gastric banding (LAGB) and sleeve gastrectomy (SG) reduce stomach capacity. There have been multiple case reports of congenital anomalies occurring after malabsorptive procedures because of maternal malnutrition; however, the evidence from observational studies is conflicting.

3Previous meta-analyses on pregnancy after bariatric surgery have focused on maternal outcomes, and there is limited evidence on perinatal outcomes other than size for gestational age and preterm birth. The aim of this systematic review and meta-analysis was to compare adverse perinatal outcomes among women who underwent bariatric surgery prior to pregnancy with those who had not. When possible, the difference in effect size between malabsorptive and restrictive procedures was explored.

4Searches were conducted in Medline, Embase, PsycINFO, CINAHL, Scopus, Google Scholar, and relevant e-journals from inception to June 3, 2019. We included observational studies published in the English language, involving women who had undergone bariatric surgery prior to pregnancy, and compared them to women without a history of bariatric surgery. We included studies that combined all types of bariatric surgery or provided data for RYGB, LAGB, SG, or BPD separately. The following perinatal outcomes were included: perinatal mortality (including stillbirth), congenital anomalies, preterm birth, postterm birth, SGA, LGA, neonatal intensive care unit (NICU) admission, birth weight, and gestational age.

5The search strategy (S1 Table) included a mixture of keywords and MeSH headings: (pregnan* or mother* or matern*) and (bariatric surgery or weight loss and surgery or gastric bypass or gastric band* or sleeve or biliopancreatic diversion or LAGB or RYGB) and (death or mortality or newborn* or fetal or congenital or stillbirth or defect* or perinatal or obstetric or neonat* or outcome* or birth). Reference lists and citations were searched for all included primary studies and for relevant reviews identified by the database searches. Authors were contacted if additional data were required for inclusion in meta-analysis. Screening, data extraction, and quality assessment were carried out in duplicate.

6This review was conducted in line with the PRISMA and MOOSE guidelines (S1 PRISMA Checklist). The protocol is published on PROSPERO (CRD42017051537).

7The Cochrane Cohort Study data extraction tool was adapted to meet the requirements of this review. Study characteristics extracted included study design, study location, type of bariatric surgery, and control group. Frequencies, effect sizes, and confidence intervals (CIs) of adverse perinatal outcomes were also extracted. For continuous outcomes, means and standard deviations were extracted. When multiple studies reported data from the same cohort with the same participant inclusion criteria, the decision was made to include the study with the larger sample size for the exposed group. Studies with duplicate data were only included if they reported different perinatal outcomes and were therefore included in separate meta-analysis. The Newcastle-Ottawa quality assessment scale was used to appraise the quality of the included studies out of a maximum of eight points (S1 Fig). The studies were assessed for representativeness of the exposed cohort, selection of the nonexposed cohort, ascertainment of exposure and outcome, study design and analysis, and adequacy of follow-up.

8A meta-analysis was used to calculate a pooled odds ratio (OR) and 95% CI when there were at least three studies reporting the same outcome. For continuous perinatal outcomes, a weighted mean difference (WMD) and 95% CI were calculated. DerSimonian and Laird random-effects model was used to take clinical heterogeneity into account such as unreported differences between surgical procedures (e.g., technique and limb length) and different levels of patient postsurgery and preconception care. When a study reported data on multiple control groups, a hierarchy was developed to firstly include the most comparable BMI group to the postbariatric patient, which was prepregnancy BMI matched, then obesity. When there was evidence of moderate heterogeneity (I2 > 40%), subgroup analysis by type of surgery or comparison group, as defined a priori, was carried out if three or more studies existed for each group. Any remaining heterogeneity was explored through meta-regression for factors including location, sample size, publication date, and quality. Publication bias was investigated using Egger’s test and funnel plots. For studies reporting adjusted results, their crude and adjusted ORs were compared to determine whether adjustments affected the effect size. Sensitivity analysis was performed for each meta-analysis by excluding one study at a time to identify the effect of any individual study on the pooled effect size and between-study heterogeneity. All analyses were conducted in Stata/SE 15.0.

9Database searches identified 3,470 results for title and abstract screening, of which 141 studies underwent full-text assessment (Fig 1). The kappa statistic for inter-rater agreement of study inclusion between authors was 0.84 (scores > 0.81 are considered excellent). Thirty-seven studies met the inclusion criteria, but four were excluded because they reported the same cohort, participant inclusion criteria, and outcomes as another study. This resulted in 33 studies that reported original data on perinatal outcomes (14,880 pregnancies after bariatric surgery and 3,979,978 pregnancies without bariatric surgery, Table 1). Fifteen of the included studies were conducted in Europe, 10 were conducted in the United States, three in Israel, two in each Australia and Brazil, and one in Canada. Studies were published between 1998 and 2018. All studies scored over five out of eight for quality, with 20 studies scoring at least seven (S2 Table). Many studies conducted more than one analysis with multiple surgery types or control groups. Sixteen analyses combined all bariatric surgery patients, whereas 14 studies were restricted to RYGB, six analyses included only LAGB, one included only SG, and one included BPD. Nine analyses compared women’s postsurgical pregnancies to pre/early-pregnancy BMI–matched controls, and 14 used obesity controls (which were ≥30 kg/m2, 35 kg/m2, or 40 kg/m2) in line with their relevant bariatric surgery guidelines, or matched for presurgical BMI. Eleven analyses compared pregnancies before and after bariatric surgery, nine compared outcomes to the general population, and five used healthy BMI as the control group.

10Perinatal mortality or stillbirth was reported in 10 studies. The pooled odds were significantly increased post-bariatric surgery compared to women without prior bariatric surgery (OR 1.38, 95% CI 1.03–1.85, p = 0.031) (Fig 2A). Ten studies reported on congenital anomalies, which were also found to have significantly increased odds post-bariatric surgery (OR 1.29, 95% CI 1.04–1.59, p = 0.019) (Fig 2B). There was no significant heterogeneity for either outcome (I2 = 12.1%, 95% CI 0.0–53.1, p = 0.331 and I2 = 28%, 95% CI 0.0–65.5, p = 0.186, respectively).

11Preterm birth was reported in 20 studies, with 19 eligible for meta-analysis. The overall odds of preterm birth were significantly increased post-bariatric surgery compared to women without prior bariatric surgery (OR 1.35, 95% CI 1.14–1.60, p = 0.001) (S2 Fig). There was significant heterogeneity (I2 = 50.1%, 95% CI 15.3–70.6, p = 0.007), which remained significant after subgroup analyses by control group but was reduced after subgrouping by type of surgery (Fig 3A). There were significantly increased odds of preterm birth after bariatric surgery in the ‘all bariatric surgery’ group (OR 1.57, 95% CI 1.38–1.79, p < 0.001). The association was not significant for subgroups ‘RYGB’ (OR 1.14, 95% CI 0.89–1.46, p = 0.289) or ‘LAGB or SG’ (OR 0.88, 95% CI 0.58–1.34, p = 0.565). The study excluded from the meta-analysis because of lack of crude data reported an adjusted OR for preterm birth of 1.43 (95% CI 1.01–2.03) post-bariatric surgery (n = 293) compared to general population controls (n = 656,353). Postterm birth was reported in five studies, and the odds more than halved after bariatric surgery (OR 0.46, 95% CI 0.35–0.60, p < 0.001) (Fig 3B). There was no significant heterogeneity (I2 = 7.2%, 95% CI 0.0–80.7, p = 0.366).

12Despite the results of increased preterm birth and decreased postterm birth, the WMD of 13 studies reporting continuous gestational age did not reach statistical significance (WMD −0.16 weeks, 95% CI −0.38 to 0.06, p = 0.156) (S3 Fig). Heterogeneity between studies was substantial and did not reduce with subgroup analyses for type of bariatric surgery. Meta-regression revealed that the following factors did not contribute to heterogeneity: type of surgery, control group, publication year, continent, sample size, or quality score (S3A Table).

13SGA, intrauterine growth restriction, and low birth weight were investigated in 22 studies, and 21 of these were eligible for meta-analysis. The odds of an SGA baby post-bariatric surgery were more than doubled (OR 2.13, 95% CI 1.80–2.52, p < 0.001) (S4 Fig). There was significant evidence of heterogeneity (I2 = 47.0%, 95% CI 11.8–68.2, p = 0.009), which was reduced by subgroup analyses by surgery type (Fig 4A). Odds of SGA were significantly increased for the ‘all bariatric surgery’ group (OR 1.87, 95% CI 1.61–2.17, p < 0.001) and were further increased for ‘RYGB or BPD’ (OR 2.72, 95% CI 2.32–3.20, p < 0.001). There was no association between SGA and ‘LAGB or SG’ (OR 1.25, 95% CI 0.62–2.51, p = 0.533). The study excluded from the meta-analysis reported an adjusted OR of 2.69 (95% CI 1.96–3.69) post-bariatric surgery (n = 293) compared to general population controls (n = 656,353).

14LGA and macrosomia were investigated in 22 studies, and 21 were eligible for meta-analysis. The ORs of an LGA baby post-bariatric surgery were more than halved (0.42, 95% CI 0.34–0.54, p < 0.001) (S5 Fig). There was substantial evidence of heterogeneity (I2 = 69.5%, 95% CI 52.4–80.5, p < 0.001). Subgroup analyses by type of surgery identified that the ‘RYGB or BPD’ group was associated with the biggest decrease in odds of LGA (OR 0.24, 95% CI 0.14–0.41, p < 0.001), in comparison with ‘all bariatric surgery’ (OR 0.51, 95% CI 0.39–0.67, p < 0.001), and ‘LAGB or SG’, which was not significant (OR 0.59, 95% CI 0.30–1.14, p = 0.116) (Fig 4B). Heterogeneity did not decrease in these subgroup analyses. Meta-regression revealed that sample size was significantly contributing to heterogeneity (residual I2 = 61.21, coefficient = 0.249, p = 0.031) (S3B Table). The study excluded from the meta-analysis reported an adjusted OR of 0.03 (95% CI 0.01–0.21) for LGA post-bariatric surgery (n = 293) compared to general population controls (n = 656,353).

15Birth weight mean and standard deviation for babies born after maternal bariatric surgery and controls were reported in 17 studies. WMD was significantly lower post-bariatric surgery (WMD −242.42 g, 95% CI −307.43 g to −177.40 g, p < 0.001) (S6 Fig). Heterogeneity was substantial (I2 = 75.7%, 95% CI 61.1–84.8, p < 0.001) but reduced after subgroup analyses by surgery type. RYGB resulted in the largest reduction in birth weight (WMD −226.10 g, 95% CI −273.43 g to −178.78 g, p < 0.001), compared with ‘all bariatric surgery’ (WMD −223.71 g, 95% CI −273.68 g to −173.74 g, p < 0.001), and ‘LAGB’, for which the reduction was not significant (WMD −135.14 g, 95% CI −289.17 g to 18.90 g, p = 0.086). One study investigated only BPD, for which the mean difference was −500 g (95% CI −570.85 g to −429.15 g, p < 0.001).

16NICU admission was reported in nine studies with babies born post-bariatric surgery being significantly more likely to be admitted to NICU (OR 1.41, 95% CI 1.25–1.59, p < 0.001) (Fig 5). There was no evidence of heterogeneity (I2 = 0.0%, 95% CI 0.0–64.8, p = 0.808).

17There was no evidence of small study effects for any outcome except LGA (p = 0.021), which may signal publication bias (S7 Fig, S4 Table). A subset of studies reported both crude and adjusted data for the adverse perinatal outcomes, but when compared, there was little difference in size or direction of associations (S8 Fig). Sensitivity analyses revealed that the results were robust, with only small changes in pooled effect sizes when meta-analysis were repeated with one study excluded (S5 Table).

18This systematic review and meta-analysis has demonstrated that perinatal mortality, congenital anomalies, preterm birth, SGA, and NICU admission are associated with increased odds in women who have had bariatric surgery prior to pregnancy compared to women without prior bariatric surgery. Postterm birth and LGA, however, are associated with decreased odds after bariatric surgery. Malabsorptive procedures were associated with a significant increase in SGA and decrease in LGA, whereas restrictive procedures were not. Subgrouping by type of surgery significantly reduced heterogeneity for the outcomes with a high I2 value, whereas subgrouping by control group did not. There was no evidence of publication bias for any outcome except LGA.

19The increase in adverse perinatal outcomes could be related to malnutrition. Unlike restrictive procedures, which reduce stomach size and appetite, malabsorptive procedures bypass a portion of the small intestine where many vitamins and minerals are absorbed, making these patients particularly susceptible to nutrient deficiencies that may negatively affect a subsequent pregnancy. The association between folic acid intake and neural tube defects is well established, and there are links between iron deficiency and preterm birth and between calcium and birth weight. Impaired nutrient transport across the placenta is also associated with perinatal morbidity; however, there is limited evidence regarding placental function after bariatric surgery. The studies reporting data on congenital anomalies in pregnancy with and without prior bariatric surgery did not subgroup by type of anomaly—this would be valuable for future research to pinpoint the mechanism behind the anomalies. Another factor that may explain the increase in SGA infants is the increased glycaemic variability and postprandial hypoglycaemia observed after RYGB, as fetal growth has been found to be associated with maternal glucose nadir levels during oral glucose tolerance testing in pregnancy.

20The strengths of this systematic review and meta-analysis include the thorough search strategy of multiple databases and supplementing this with hand searches of reference lists, citations, and relevant journals. All screening, data extraction, and quality assessment was carried out in duplicate to minimise human error. There are no randomised controlled trials, because of the nature of this research question, but all included studies were medium- to high-quality observational studies. This is the first meta-analysis, to our knowledge, to report significantly increased odds of perinatal mortality and congenital anomalies after bariatric surgery. This is also the first meta-analysis, to our knowledge, to investigate postterm birth after bariatric surgery, for which a significant decrease was found. The results for SGA, LGA, preterm birth, and NICU admission confirm the findings of past meta-analyses but with stronger associations than previously reported and the inclusion of 12 additional studies.

21The results from our study are limited by the small sample sizes of some of the included studies. Multiple studies reported few, or even zero, cases of perinatal mortality or congenital anomalies and have therefore resulted in large CIs. Larger epidemiological studies or individual patient data (IPD) meta-analyses need to be carried out for this rare exposure and rare outcome combination. Additionally, there are no large studies exploring congenital anomalies and perinatal mortality specifically after restrictive surgery such as LAGB or SG, which may not have a detrimental effect. A number of studies have reported several adverse perinatal outcomes, many of which are linked, which may result in a loss of statistical and clinical independence. We were unable to include non–English language studies, and one non–English language study meeting our inclusion criteria was excluded. This study from France identified a significant decrease in macrosomia, as our meta-analysis did; however, it also found a decrease in SGA in contrast to the significant increase we found.

22Women that become pregnant post-bariatric surgery tend to be older than the general population of pregnant women. Many women also still have a BMI > 30 kg/m2 despite the weight loss from surgery. There is also evidence that alcohol use and smoking are increased after bariatric surgery. The combination of increased maternal age, high BMI, and unhealthy behaviours in women after bariatric surgery plays a role in the development of adverse perinatal outcomes, in addition to the malnutrition. These are important confounders to consider when investigating perinatal outcomes in this group. When comparing ORs with adjustments made for these factors to unadjusted ORs, we did not see a change in the results. However, in a clinical setting, these factors and behaviours are important for the healthcare provider to take into account because of the evidence of the link with adverse perinatal outcomes. As with all meta-analyses of observational data, unmeasured confounding in the included studies may have implications on the results. Gestational weight gain (GWG) is another factor associated with perinatal outcomes such as birth weight; however, further research is required to determine how the relationship between GWG and pregnancy outcomes differs for women after bariatric surgery and whether current GWG guidelines can apply to this population.

23The LAGB subgroup analyses tended to have larger CIs than any other subgroup. This may be due to smaller sample sizes or differences in LAGB band management. Some clinics actively manage gastric bands during pregnancy by deflating in cases of nausea or vomiting and inflating in cases of excess GWG. Future studies should explore how band management could be used to achieve optimal pregnancy outcomes. The studies that combined all types of bariatric surgery drastically differed in surgery type composition, with studies reporting from 13.3% RYGB to 98% RYGB in their cohorts. It would be useful for future studies to separate outcomes by type of surgery or to conduct IPD meta-analyses on the existing data, which would enable standardisation of categories across studies.

24Future studies should explore the effect of time to conception after different types of bariatric surgery, especially considering gestational weight loss and advanced maternal age. Many women that are previously considered to be infertile experience increased fertility after bariatric surgery, which may result in unexpected pregnancies immediately after surgery in the rapid weight loss phase. Many clinics recommend waiting 12–18 months to conceive post-surgery, but the evidence base is limited for this.

25Bariatric surgery prior to pregnancy is promising for reducing obesity-related comorbidities for the mother, and benefits include reduced risks of gestational diabetes and preeclampsia, which are both serious complications associated with adverse maternal and fetal outcomes. Our meta-analysis has shown that the risks of postterm birth and LGA babies are reduced after bariatric surgery; however, we have also identified adverse outcomes for the baby and efforts now need to be focused on how to reduce these. Internationally, guidelines exist for a variety of high-risk pregnancy groups such as those with diabetes, hypertension, and obesity. This study confirms that bariatric surgery patients that become pregnant are also a high-risk group, and guidelines for health professionals need to be developed as obesity and bariatric surgery increases. The current evidence base could be used to inform risk communication about potential future pregnancies with women of reproductive age prior to surgery. For women with a history of bariatric surgery, preconception nutritional support should be offered, and increased fetal, nutrition, and glucose monitoring is required throughout pregnancy. Further studies are required to determine whether restrictive surgery results in better perinatal outcomes than malabsorptive surgery without compromising maternal outcomes, and if so, these may be the preferred surgery for women of reproductive age.