Probiotics for the Prevention of Gestational Diabetes Mellitus in Overweight and Obese Women: Findings From the SPRING Double-Blind Randomized Controlled Trial

The incidence of gestational diabetes mellitus (GDM) is increasing in parallel with the rise in overweight and obesity in the obstetric population (1). GDM is associated with adverse maternal and neonatal outcomes. Women with GDM have an increased risk of preeclampsia and delivery by cesarean section (2,3). Maternal GDM increases the risk of excessive fetal growth and adiposity, shoulder dystocia, admission to the neonatal intensive care unit, and neonatal hypoglycemia (4,5). Furthermore, GDM is associated with increased later risk of obesity, type 2 diabetes, and cardiovascular disease in both mother and baby (6–8).

Women who are overweight and obese are at higher risk of GDM (9). GDM therapy is currently focused on normalizing glycemia. While this prevents or reduces the impact of short-term complications, this may not be true for the longer-term complications (10,11). Ideally, GDM would be prevented. A number of strategies to prevent GDM during pregnancy have been explored, including the use of metformin, lifestyle interventions, probiotics, myo-inositol, and vitamin D (12). Preconception GDM prevention strategies such as weight loss and lifestyle interventions seem logical but have been hampered by substantial barriers to engagement and behavioral change (13–15).

In recent years, the role of the gut microbiota in regulating metabolism has become a topic of investigation. The composition of the gut microbiota is determined by a large number of factors including disease states (16) and dietary intake (17). The gut microbiota are altered in obesity, although the extent and direction of the changes are variable (18). Pregnancy is associated with large changes to the gut microbiota with diversity decreasing as pregnancy progresses (19). In addition, probiotics—live bacteria with a known beneficial effect on the host—may be able to improve gut microbiota function and thereby the health status of the host (20). Interventions such as probiotics are attractive, as acceptability and compliance appear to be much better than with lifestyle interventions (21). There have been promising data to suggest that probiotics may positively alter measures of glucose metabolism or prevent GDM (22–24). Luoto et al. (24) reported 64% relative reduction in GDM frequency (13% in diet and probiotics, 36% in diet and placebo, and 34% in control groups) in a cohort of 256 pregnant women with a mean BMI of 23.6 kg/m². Given the higher risk of GDM in overweight and obese women and the effectiveness of probiotics in normal-weight women (15), we designed and conducted a randomized controlled trial (RCT) of probiotics in overweight and obese women.

SPRING (Study of PRobiotics IN Gestation) was a prospective double-blind RCT of probiotics versus placebo to examine whether probiotics prevented GDM in overweight and obese women following Consolidated Standards of Reporting Trials (CONSORT) guidelines (25) (as per Fig. 1). Detailed methodology including information regarding inclusion and exclusion criteria, sample size, and power calculations has been published (26). Inclusion criteria included a singleton pregnancy at <20 weeks’ gestation, BMI of >25 kg/m², >18 years of age, able to read and understand English, and able to provide informed consent. Gestational age was determined clinically by the woman’s pregnancy care provider, based on the first date of the last menstrual period or, where that was not known, estimated gestational age based on the earliest pregnancy ultrasound scan. As outlined in the published protocol (26), all women underwent a random venous plasma glucose (RVPG) prior to enrollment. Those with RVPG ≥8.0 mmol/L proceeded to a 75-g oral glucose tolerance test (OGTT) and were excluded if any values met or exceeded criteria for GDM. Exclusion criteria included gestational age >20 weeks at recruitment, multiple pregnancy, known preexisting diabetes, impaired fasting glucose or impaired glucose tolerance, taking medications that may influence glucose metabolism (metformin, glucocorticoids, immunosuppressants, antipsychotics), medical conditions known to alter glucose metabolism, known major fetal abnormality on 11- to 13-week ultrasound scan, and known ingestion of probiotics via capsules or sachets. Overweight and obese women (BMI ≥25 kg/m², based on height and weight measured at their first study visit) were randomly assigned to probiotics or placebo prior to 20 weeks’ gestation. The composition of the probiotics was identical that used by Luoto et al. (24) and consisted of a mixture of Lactobacillus rhamnosus (LGG) and Bifidobacterium animalis subspecies lactis (BB-12) (Chr. Hansen A/S, Hørsholm, Denmark) at a dose of >1 × 10⁹ colony-forming units each per day or matched placebo (microcrystalline cellulose and dextrose anhydrate capsules; Chr. Hansen) (27). Capsules were to be taken once daily from enrollment until birth.

This study was conducted at the Royal Brisbane and Women’s Hospital (RBWH), Redcliffe Hospital, and the Mater Mothers’ Hospital in Brisbane, Australia. The trial commenced recruitment in November 2012.

Participants were randomized using computer-generated random number codes sealed in opaque envelopes. Participants were stratified by center and by BMI category (BMI >25–30, >30–40, and >40 kg/m²).

Matching placebo and probiotic capsules were identically packaged in the RBWH pharmacy by independent pharmacists. All study staff and participants were blinded to the randomized allocation.

Compliance to study medication adherence was monitored through interviews with all participants. For the subset of participants who supplied a fecal sample at 28 weeks’ gestation, presence of BB-12 DNA was tested by end-stage PCR. In brief, DNA was isolated with the repeated bead beating and column (RBB+C) method and a Qiagen AllPrep DNA extraction kit as previously described (28). End-stage PCR was performed using the Bflact2 and Bflact5 primer pair (29). Presence of a band at 680 base pairs was considered confirmation of the presence of BB-12 in the gut microbiota.

We had initially planned to recruit women prior to 16 weeks’ gestation. However, owing to changes in hospital policies regarding the timing of first antenatal visit, recruitment was extended up to 20 weeks’ gestation in May 2015 after 310 women had been recruited. Sixty-three participants (14.4%) were recruited between 16 and 19 weeks’ gestation. A fecal sample was requested at enrollment and at 28 weeks. The requirement for fecal samples was noted to be a barrier to recruitment, and this was removed in May 2015. Initial power calculations based on a 50% reduction in GDM frequency with the use of probiotics (from 18 to 9%), and allowing for 20% attrition, called for the recruitment of 540 women (452 completers). In the end, 436 women were randomized and 411 participants completed the study, reducing the power from 80 to 76% to find the originally predicted 50% reduction in GDM in the probiotic arm. Limited funding prevented prolongation of the study.

The primary outcome of the study was the frequency of GDM at 28 weeks’ gestation by a 75-g OGTT using the World Health Organization–endorsed criteria of the International Association of the Diabetes and Pregnancy Study Groups (30).

Prespecified maternal secondary outcomes included gestational weight gain, preeclampsia, hypertensive disorders of pregnancy, cesarean delivery, and gestational age at delivery. Weight gain based on self-report at the beginning of pregnancy and measured weight at 36 weeks’ gestation was categorized as excessive or inadequate based on the Institute of Medicine guidelines (31). Planned neonatal outcomes included prematurity, neonatal special care admission, jaundice, hypoglycemia, birth weight, small for gestational age (SGA), large for gestational age, stillbirth, birth injury, congenital anomaly, fat-free mass, and percentage fat (measured in the PEA POD Infant Body Composition System [COSMED, Rome, Italy], using air displacement plethysmography). Australian national birth weight percentiles by sex and gestational age were used. Standard definitions were used for other outcomes (32,33).

Baseline characteristics were summarized for each intervention group as n (%) for categorical variables, mean (SD) for normally distributed continuous variables, and median (interquartile range [IQR]) for nonnormally distributed continuous variables. Intervention group comparisons were analyzed using binary logistic regression or general linear models as appropriate with adjustment for center and BMI category. Adjusted odds ratios and difference of means with 95% CIs were reported. Where a model did not converge due to small cell counts, the effect estimate and P value were not reported. Statistical significance was specified at α < 0.05. Data were analyzed using IBM SPSS Statistics 23. All analyses were conducted by an independent statistician at the QIMR Berghofer Medical Research Institute.

Participants provided written informed consent. The study was approved by the human research ethics committees of the RBWH, Mater Health Services, and The University of Queensland. The clinical trial management committee conducted interim analysis to ensure there were no adverse events and, along with all relevant ethics committees, approved all changes to the clinical trial protocol to assist with recruitment.

There were 204 women in the placebo group and 207 women in the probiotics group. Compliance with study medication, assessed by participant interview and monthly telephone follow-up, was reported as >90%.

The presence of the probiotic organism BB-12 in the feces of a subset of study participants randomized to probiotic therapy was confirmed by targeted PCR. Of the women with stool samples obtained at 28 weeks’ gestation (N = 215 [105 probiotics and 110 placebo]), detection of fecal BB-12 in the probiotics group was similar between women who developed GDM (12 of 16 [75%]) and those who did not (71 of 89 [80%], P = 0.74).

Baseline characteristics for each intervention group are outlined in Table 1. Women were recruited into the study at a mean of 15.9 weeks (SD 1.5) and a median of 15.9 weeks (IQR 14.9–16.7). Prerandomization median RVPG was 80 mg/dL (IQR 79–94) in the placebo group and 86 mg/dL (IQR 81–95) in the probiotics group. Fifty-five women in the placebo group and 56 in the probiotics group underwent an early OGTT, and mean fasting glucose values in both groups were identical at 77 mg/dL (SD 7).

The primary outcome and underpinning glucose measures relating to the primary outcome for this study are presented in Table 2. Rates of GDM were 25 of 204 (12.3%) in the placebo arm and 38 of 207 (18.4%) in the probiotics arm of the study (P = 0.10). Mean fasting glucose was higher in women randomized to probiotics (79.3 mg/dL) compared with placebo (77.5 mg/dL) (P = 0.049) (difference of means 1.53 [95% CI 0.0095–3.06]). One- and two-hour glucose levels were similar between the groups. There was no indication that rates of GDM were lower in the probiotics arm, and if anything, the results suggested a trend in the opposite direction.

Maternal and neonatal secondary outcomes are presented in Table 3. In women taking probiotics, 9.2% developed preeclampsia compared with 4.9% in the placebo group (0 = 0.09). Excessive weight gain occurred in 32.5% of women in the probiotics group compared with 46% of women in the placebo group (P = 0.01). However, there was no difference in mean weight gain in the two groups. There were no other statistically significant differences in any maternal secondary outcome measures between probiotics and placebo. The only statistically significant difference in neonatal outcomes related to SGA (<10th percentile). SGA occurred in 13 of 199 infants in the placebo arm (6.5%) and 5 of 205 infants in the probiotics arm (2.4%) (P = 0.042).

We conclude that probiotics, administered throughout pregnancy from the first half of the second trimester, do not reduce the frequency of GDM at 28 weeks’ gestation.

We also conclude that this study provides evidence that daily ingestion of LGG and BB-12 at a dose of >1 × 10⁹ colony-forming units each per day does not alter any of the primary or secondary maternal outcomes examined in overweight and obese women. There was a statistically significant reduction in SGA (<10th percentile) in infants born to women taking probiotics.

While there was no statistical difference in GDM between the arms of this study, we noted a higher fasting glucose in women who were randomized to the probiotics arm. Although statistically significant, this small increase in fasting glucose may not be clinically significant. However, if carried through to the population level, even a small increase in fasting glucose could result in higher GDM prevalence. The data from this and other studies will need to be further explored when the Cochrane Review of this topic is updated (34). Although the total number of participants was slightly lower than initially planned, the results clearly do not support efficacy of probiotics.

It is hard to reconcile the lower rates of excessive weight gain in women on probiotics with the other findings from this study. Further, there was no difference in overall weight gain between groups or weight gain per week. However, the role of probiotics in prevention of excessive weight gain will need further exploration.

The lower rates of SGA in infants in the probiotic arm may have related to the fact that 18.4% of these pregnancies were exposed to GDM compared with 12.3% of infants born to women in the placebo group. The higher fasting glucose level in the probiotics arm may have also influenced the lower rate of SGA in these infants. The higher rate of inadequate weight gain and lower rate of excessive weight gain with probiotics may also have contributed. However, the overall rate in both arms was much less than the expected 10%, so the SGA rate is overall low rather than high. No other measure of body composition was consistent with this lower rate of SGA. The alternative explanation is that probiotics have a role in the prevention of SGA. This will need to be explored in future meta-analyses.

For many years, there has been widespread speculative commentary regarding the potential for probiotics to be an elegant and easy solution to GDM prevention (35). They appeared safe and easy to take without any obvious maternal or neonatal side effects (36). Despite reassuring safety data, and the logical reasons that probiotics might improve glucose metabolism, our study clearly argues against efficacy of probiotics in GDM prevention.

Consistent with our findings, Lindsay et al. (37) examined the use of Lactobacillus salivarius UCC118 for 4 weeks from 24 weeks’ gestation in a population of obese women in Ireland and found no difference in any measure of glucose metabolism. Asemi et al. (23) compared probiotic yogurt (Lactobacillus acidophilus LA5 and BB-12) with conventional yogurt over a 9-week period during pregnancy. Consistent with our study, there was no improvement in fasting glucose or systolic or diastolic blood pressure in the probiotic yogurt arm.

In the Finnish study, which inspired the current study, probiotic supplementation with LGG and BB-12 reduced the risk of elevated maternal glucose concentrations (38) and the incidence of GDM in women with a mean BMI of 23.6 kg/m² (SD 3.8) (24). Our active treatment arm included the same probiotic preparation. There were several key differences between our study and the Finnish study. Firstly, women were selected into the Finnish study based on a personal history of atopic disease or at least one relative with atopic disease; 79% of women in this study had a history of atopic disease. In contrast, participants in our study were selected on the basis of BMI. Women in the Finnish study were relatively lean in comparison with the overweight and obese women in our study. Interestingly, the rate of GDM in our population was much lower than that in the Finnish study (34% in the control arm). GDM rates in the Finnish study were likely to be higher, due to the use of different diagnostic criteria, and this may well have been driven by the lower fasting glucose cutoff (39). In the Finnish study, probiotics were administered in addition to a dietary intervention, whereas in our study, only limited dietary advice was provided: to ensure probiotics were taken with milk or water and not with hot or acidic drinks. It is likely that the women in the Finnish study differed substantially from the women in our population in terms of genetics, obesity, immune system characteristics, dietary patterns, and levels of exercise. These factors could individually and collectively affect both the baseline microbiome as well as changes during pregnancy. It has been shown that there is large variability in which bacterial species become more or less abundant in pregnancy (19). Therefore, multiple factors may have influenced the conflicting outcomes and will need to be explored through future research.

Consistent with the Finnish study, a recent New Zealand study of women or partners at risk for atopic disease who were administered Lactobacillus rhamnosus HN001 in pregnancy found that the rate of GDM was 13.8% in the placebo arm and 8.2% in the probiotic arm, with a relative risk of 0.59 (95% CI 0.32–1.08) (22). This study used the same criteria for GDM as was used in our study. The trend toward a lower rate of GDM in this New Zealand study may suggest that probiotic species other than the one used in our study may have different effects. The results of these two studies could suggest that probiotics may be more effective in those with a background of atopic disease compared with the overweight and obese women recruited to our study.

In a recently published meta-analysis (40), unpublished data were obtained from authors of two studies (41,42) and combined with the work of Lindsay et al. (37) to address the issue of probiotics in the prevention of GDM. The combined outcomes of GDM in 365 women across these three different studies were consistent with our large study and demonstrated no effect of probiotics for the prevention of GDM. Surprisingly, two of the studies that assessed GDM outcomes in this meta-analysis commenced probiotics at 32 weeks’ (41) and 36 weeks’ (42) gestation, respectively. It is unclear how an intervention commenced after 32 weeks could be expected to prevent an outcome that is usually diagnosed by an OGTT at 26–28 weeks’ gestation.

In contrast, another recent meta-analysis found an improvement in metabolic outcomes in pregnant women (43). In this meta-analysis, only the Lindsay study, already discussed above, related to GDM prevention in overweight and obese women. This meta-analysis included five studies of probiotics in healthy pregnant women and four studies on the effect of probiotics in women who already had GDM. Probiotics were associated with lower fasting serum insulin, no change in fasting plasma glucose, and conflicting results with the two approaches to measuring insulin resistance.

Our study was designed to be pragmatic with regard to probiotic species, dose, and gestational age at commencement of the intervention. These decisions were guided by feasibility and practicality. Therefore, while our study did not show benefit, other probiotic species might be efficacious in the prevention of GDM. The question of dose of probiotics for pregnant overweight and obese women was not examined in this study, and it is possible that outcomes may have differed (in either direction) with an increased dose of probiotics. In our analysis of the presence of the bacteria in the feces of the women, adherence was 75% and 80% in the women on probiotics who developed GDM and remained normoglycemic, respectively. This suggests that adherence was similar and therefore that the probiotic intervention likely does not affect the development of GDM. The difference in adherence between the fecal analysis (79%) and the self-reporting of intake by the women (90%) could reflect inaccurate self-reporting. However, we have not examined how long the bacteria are detectable in the feces after cessation of the intervention nor do we know the exact sensitivity threshold for the appearance of a band on the agarose gel. Therefore, it is possible that the difference in adherence based on fecal analysis versus self-report is not as large as it appears.

It is also possible that probiotics started prior to pregnancy, or in very early pregnancy, have an effect different from that observed here. Women in the study by Wickens et al. (22) were randomized to probiotics at 14–16 weeks’ gestation. Women in the study by Luoto et al. (24) had their first study visit at a median of 14 weeks’ gestation (range 7–18) (44). Our median gestation at commencement of intervention was 15.9 weeks’ gestation. It is possible that a longer period of intervention may influence outcomes. While earlier commencement of probiotics during pregnancy would be important to explore, this poses practical difficulties in routine clinical practice. Even if very early institution of probiotics was found to be useful, there would be significant logistic barriers to the translation of positive findings. This is a question that will need to be explored in other studies.

The strengths of this study are that it is the largest double-blind RCT conducted to date examining the role of probiotics for the prevention of GDM in a high-risk group. It used a probiotic that had previously been shown to prevent GDM (24). The results are quite clearly negative.

There are some limitations to our study. Our protocol did not require a formal OGTT for every participant prior to enrollment, so although available measures of baseline glycemia were similar between the groups, we are not able to exclude more subtle differences. It is possible, given the low rates of smoking and high rates of tertiary education, that the overweight and obese women in this study were healthier than average. Therefore, the trial population may not be representative of all overweight and obese women. Our original recruitment targets were not achieved, but it is very unlikely that recruitment of additional participants would have altered the conclusions of this study. Owing to slow recruitment and altered hospital routines, we changed the gestational age at enrollment to include participants up to 20 weeks’ gestation, which was approved and supported by all of the relevant regulatory authorities and the trial management committee. This lowered the total exposure time to probiotics for some women (14.4% of the total study population) by 1–4 weeks. However, given the outcomes of this study, it is hard to envisage that a longer period of probiotic exposure in the intervention arm would have altered our findings. Nonetheless, the impact of probiotics from very early in pregnancy remains an unanswered question.

Based on the findings of this double-blind randomized placebo controlled trial, we conclude that probiotics administered from the second trimester of pregnancy do not prevent GDM, or improve secondary outcomes, in overweight and obese pregnant women.

Acknowledgments. The authors thank Chr. Hansen A/S, which donated the probiotic and matching placebo capsules, with a legal agreement between The University of Queensland and Chr. Hansen A/S that guaranteed investigator independence. The authors thank all staff and patients who participated in the trial and also thank all staff within the Redcliffe Hospital, Mater Health, and RBWH who assisted in the conduct of this study. The authors are grateful for the work of Brooke Berry (Queensland Health Pathology Service), Meredith Shallcross (Redcliffe Hospital SPRING trial coordinator), and Sharney Grant (Redcliffe Hospital SPRING trial research midwife) in contributing to this research.

Funding. This study was funded by National Health and Medical Research Council grant APP1028575 and the Royal Brisbane and Women’s Hospital Foundation.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. L.K.C., H.D.M., H.L.B., K.F., A.T., B.E.L., J.M.T., S.W., A.K., M.M., P.O., A.P., and M.D.N. interpreted data. L.K.C., H.D.M., H.L.B., K.F., A.T., B.E.L., J.M.T., S.W., A.K., M.M., P.O., A.P., and M.D.N. wrote the manuscript. L.K.C., H.D.M., H.L.B., K.F., A.T., B.E.L., J.M.T., S.W., A.K., M.M., P.O., A.P., and M.D.N. reviewed and approved the final manuscript. L.K.C., H.D.M., H.L.B., K.F., A.T., B.E.L., S.W., A.K., M.M., and M.D.N. performed data collection. L.K.C., H.D.M., H.L.B., K.F., A.T., M.M., P.O., A.P., and M.D.N. performed data analysis. L.K.C., H.D.M., H.L.B., B.E.L., J.M.T., S.W., M.M., P.O., and M.D.N. designed the study. L.K.C. affirmed that this manuscript is an honest, accurate, and transparent account of the study reported; that no important aspects of the study have been omitted; and that all discrepancies from the study as planned and registered have been reported on and explained in the manuscript. L.K.C. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented in abstract form at the 77th Scientific Sessions of the American Diabetes Association, San Diego, CA, 9–13 June 2017.