Materials and Methods
Study design and setting
This was a single-center, double-blind, placebo-controlled, randomized trial with maternal written consent, conducted in accordance with the standards of the National Maternity Hospital (NMH) Ethics Committee, which granted full ethical approval for the trial in November 2011. The trial is registered on Current Controlled Trials (ISRCTN97241163 Part B).
Patient selection and recruitment
The Probiotics in Pregnancy Study population was a presenting sample of pregnant women attending the NMH who were newly diagnosed with either IGT (1 raised value) or GDM (≥2 raised values) following a 3-hour 100-g oral glucose tolerance test in the current pregnancy. Routine clinical care for diagnosed women attending the hospital includes advice on a low glycemic index diet, self-monitoring blood glucose levels using glucometers, and fortnightly attendance at the diabetes clinic for monitoring of glucose control. At follow-up clinic visits, if a patient has a fasting plasma glucose >5 mmol/L (90 mg/dL) and/or 1-hour postprandial plasma glucose >7 mmol/L (126 mg/dL) following a standard breakfast, and the patient is considered to be compliant to the low glycemic index diet at home, dietary control of glucose is deemed inadequate and treatment with metformin or insulin is commenced.
Women were approached by the research dietitian during the dietary and lifestyle education class, which is run weekly in the NMH for women with IGT or GDM. Information sheets were provided and written consent was obtained from eligible and willing participants. Inclusion criteria were a new diagnosis of IGT or GDM in the current pregnancy, age >18 years, <34 weeks’ gestation, singleton pregnancy and adequate English to enable full understanding of the study. Women were excluded if they had pregestational diabetes, were aged <18 years, were ≥34 weeks’ gestation, had a multiple pregnancy or fetal anomaly, were commenced on insulin or metformin therapy immediately after diagnosis, or had a poor understanding of the English language.
Blinding, masking, and randomization
The probiotic and placebo capsules were produced and supplied by Alimentary Health Ltd, Cork, Ireland, and anonymously labeled as “A” or “B.” Each active probiotic capsule contained 100 mg of Lactobacillus salivarius UCC118 at a target dose of 10 9 colony-forming units. Further details of the probiotic and placebo capsule contents and packaging have been previously described.
Allocation to either one of the capsules was conducted by an independent researcher using a computer-generated simple randomization process in a ratio of 1:1. No stratification factors were applied. The allocation sequence was concealed from the research dietitian enrolling and assessing the participants in sequentially numbered, sealed, opaque envelopes. After written informed consent was obtained and baseline assessments were completed, the envelope corresponding to each participant study identification number was opened to reveal the allocation to capsule A or B. Although, the research dietitian was then aware that all participants allocated to one of the capsules were all in the same treatment arm, the identity of the treatment arm remained unknown. To minimize risk of bias, all clinical and laboratory staff who were involved with care of study participants or analysis of samples remained blinded to the allocation sequence.
Data collection, intervention, and trial management
On recruitment, the research dietitian provided each participant with an information sheet outlining fermented and probiotic-containing foods and supplements to avoid throughout the remainder of their pregnancy to minimize the risk of confounding from the ingestion of other probiotics. Suitable nonprobiotic yogurts were recommended that were also low in added sugar to ensure compliance to the low glycemic index diet.
Following an overnight fast, baseline blood samples were collected approximately 1 week after recruitment when participants routinely attended the diabetes clinic. Women were provided with a bottle of their allocated capsules (A or B), each bottle containing a 2-week supply (15 capsules), and instructed to take 1 capsule daily after a meal of their choice and to keep the bottle refrigerated. Subsequent bottles were provided at follow-up visits to the diabetes clinic and daily capsules were taken until delivery. Compliance to the capsule intervention was monitored by pill counts.
Four to 6 weeks after commencement on the capsules, a second fasting blood sample was obtained at a routine clinic visit to measure postintervention metabolic parameters. Details of any pharmacological therapy (insulin or metformin) among study participants for management of blood glucose levels were noted. Participant self-reported use of antibiotic therapies during the intervention period was also recorded.
At delivery, cord blood samples were collected where possible. Details of delivery method, gestational age, infant sex, birthweight, and length and head circumference were recorded. Birthweight centiles were calculated using the Bulk Centile Calculator Version 6.6 (January 2013; Gestation Network, grow@gestation.net ). The trial steering committee met bimonthly and an independent assessor reviewed the safety of data.
Assessment of dietary intakes
Women were asked to record their dietary intakes using a 3-day food diary within the first 2 weeks of commencing on the capsules. They were instructed to record in detail all food and drink consumed on 3 consecutive days, to include 1 weekend day. Dietary data were entered into the nutritional analysis software WISP version 3.0 (Tinuviel Software, Llanfechell, Angelsey, UK) for energy and nutrient intake analysis.
Blood analyses
All maternal and cord samples for plasma glucose were analyzed following centrifugation by hospital laboratory staff at the shortest possible interval following sample collection using the AU680 Chemistry analyzer (Beckman Coulter Inc., High Wycomb, UK) and the hexokinase method. Additional sera collected at baseline and postintervention were centrifuged at 4°C within an hour of collection for 5 minutes at 1409 g. The separated serum was immediately frozen at –20°C, with subsequent transfer to a –80°C freezer. Serum samples from cord blood were treated similarly although the interval from time of collection until centrifugation was, in some instances, several hours later. Maternal and cord samples were later analyzed for c-peptide, triglycerides, total cholesterol, and high-density lipoprotein (HDL) and low-density lipoprotein (LDL) cholesterol. Insulin was also analyzed in maternal samples. Insulin and c-peptide were quantified by automated immunoassay (Roche Analytics E170; Roche Diagnostics, Basel, Switzerland) with typical CVs <5%. Total cholesterol, HDL cholesterol, and triglycerides were analyzed on a Roche Cobas 702 analyzer (Roche Diagnostics). LDL cholesterol levels were estimated using the equation of Friedewald et al. The homeostasis model assessment-insulin resistance index was calculated using the formula: fasting glucose (mmol/L) × fasting insulin (μU/mL)/22.5.
Outcome measures
The primary outcome was difference in postintervention maternal fasting glucose between the probiotic and placebo groups, adjusting for baseline. Secondary outcomes were requirement for pharmacological therapy and neonatal birthweight. These outcomes were assessed by intention-to-treat analysis, which included all subjects enrolled in the study.
Other maternal outcomes assessed included metabolic parameters (insulin, homeostasis model assessment-insulin resistance, c-peptide, and lipids), gestational weight gain, hypertensive disorders, and delivery complications. Neonatal outcomes assessed were: cord blood metabolic parameters (glucose, c-peptide, and lipids), 5-minute Apgar score, and admission to neonatal intensive care unit. As postintervention blood samples were not collected on all enrolled participants due to study withdrawal and loss to follow-up, metabolic parameters were analyzed among a “per-protocol” cohort, which excluded women with incomplete data collection and those commenced on pharmacological therapy for blood glucose management. The primary outcome analysis was also repeated among this cohort as pharmacological therapy directly influences fasting glucose levels.
Sample size
Previously conducted research with the UCC118 probiotic among healthy nonpregnant women resulted in a 0.4-mmol/L (7.2 mg/dL) reduction in fasting glucose with SD of 0.7, compared to placebo (unpublished data). It is hypothesized that a similar reduction in fasting glucose would be observed with probiotic treatment among the target pregnant population in this study, which is considered clinically meaningful based on the results of the Hyperglycemia and Adverse Outcomes Study. Thus, power analysis indicated that a total sample of 100 subjects, 50 in each group, was required to detect a 0.4-mmol/L reduction in fasting glucose at the .05 level of significance with at least 80% power. Researchers aimed to recruit up to 150 participants to account for attrition and commencement of pharmacological therapy, which would directly influence the primary outcome measure.
Statistical analyses
All statistical analyses were performed using IBM SPSS software for Windows version 20.0 (SPSS Inc, Chicago, IL). For the intention-to-treat analysis of the primary outcome, missing data on maternal fasting glucose were handled through multiple imputation. Variables were graphically assessed for normality using histograms. Baseline demographics, dietary intakes, metabolic parameters, and pregnancy outcomes within each group were reported as mean (SD) and n (%). For dietary intakes, pregnancy outcome, and fetal metabolic parameters, differences between the groups were assessed using the independent samples t test and χ 2 test for continuous and categorical variables, respectively. For the primary outcome, analysis of covariance was used to assess differences between the groups in postintervention fasting glucose, adjusting for baseline. Statistical significance was set at P < .05. This method was also employed to assess differences in other maternal metabolic parameters in the per-protocol cohort, while within group differences for each metabolic parameter were assessed using the paired samples t test.
Results
Recruitment of study participants commenced in March 2012 and ended in May 2014 and the final delivery occurred in July 2014. During this period, 371 women entering the care of the antenatal diabetes team were assessed for eligibility and of these, 149 women provided informed consent, were randomized to the probiotic or placebo group, and included in the intention-to-treat analysis. The participant flow from initial screening to final analysis is displayed in the Figure . A total of 115 study participants completed the study (57 probiotic group, 58 placebo group) with full blood sample collection data. Among these, 15 women commenced pharmacological treatment and were thus excluded from subsequent per-protocol analysis.
Baseline demographics of the intention-to-treat cohort are presented in Table 1 . There was a slightly lower rate of Caucasian ethnicity and obesity and a higher rate of primiparity in the probiotic compared to placebo group, although these differences were not significant. Among women who completed the study (N = 115), the mean duration of capsule consumption was 8 weeks (SD 2.5), which did not differ between groups. Dietary intake data were available for 103 participants and no differences were detected between the intervention and control groups ( Appendix ; Supplementary Table 1 ).
| Characteristic | Total population (n = 149) | Probiotic group (n = 74) | Placebo group (n = 75) | P value |
|---|---|---|---|---|
| Mother’s age, y | 33.0 (4.8) | 33.5 (5.0) | 32.6 (4.5) | .237 |
| Booking weight, kg | 78.23 (18.28) | 78.51 (18.99) | 77.95 (17.66) | .851 |
| Height, m | 1.64 (0.06) | 1.64 (0.06) | 1.64 (0.07) | .919 |
| BMI, kg/m 2 | 29.00 (6.23) | 29.06 (6.70) | 28.94 (5.79) | .913 |
| Weight at randomization, kg | 84.72 (18.09) | 85.29 (19.07) | 84.14 (17.17) | .708 |
| Gestational age at diagnosis, wk | 29.6 (2.5) | 29.8 (2.5) | 29.5 (2.4) | .553 |
| Gestational age at randomization, wk | 31.5 (2.2) | 31.6 (2.4) | 31.5 (2.1) | .837 |
| Ethnicity, n (%) | .824 | |||
| Caucasian Irish | 101 (67.8) | 48 (64.9) | 53 (70.7) | |
| Non-Irish Caucasian | 24 (16.1) | 11 (14.9) | 13 (17.3) | |
| Asian | 15 (10.1) | 10 (13.6) | 5 (6.7) | |
| African | 6 (4.0) | 4 (5.4) | 2 (2.6) | |
| Latino | 3 (2.0) | 1 (1.4) | 2 (2.7) | |
| Caucasian ethnicity, n (%) | 125 (83.9) | 59 (79.7) | 66 (88.0) | .170 |
| BMI category, n (%) | .329 | |||
| Normal weight (18.5–24.9) | 45 (30.2) | 22 (30.1) | 23 (30.7) | |
| Overweight (25.0–29.9) | 41 (27.5) | 24 (32.9) | 17 (22.7) | |
| Obese (>30.0) | 62 (41.6) | 27 (37.0) | 35 (46.7) | |
| Complete third-level education (n = 102), n (%) | 60 (58.8) | 32 (64.0) | 28 (53.8) | .298 |
| Smoking in pregnancy (n = 115), n (%) | 5 (3.4) | 2 (3.5) | 3 (5.2) | .662 |
| Smoking prepregnancy (n = 94), n (%) | 10 (6.7) | 5 (10.9) | 5 (10.4) | .943 |
| Primiparous, n (%) | 76 (51.0) | 43 (58.9) | 33 (44.0) | .070 |
| Diagnosis on glucose tolerance test, n (%) | .368 | |||
| GDM | 73 (49.0) | 39 (52.7) | 34 (45.3) | |
| IGT | 76 (51.0) | 35 (47.3) | 41 (54.7) | |
| Commenced pharmacological therapy, n (%) | 22 (14.8) | 12 (16.9) | 10 (14.1) | .643 |
| Supplement user, n (%) | 139 (93.3) | 70 (97.2) | 69 (95.8) | .649 |
| Family history of diabetes, n (%) | 68 (45.6) | 32 (44.4) | 36 (48.6) | .611 |
| Antibiotic user, n (%) | 5 (4.3) | 2 (3.4) | 3 (5.1) | .662 |
Primary outcome
The results of the intention-to-treat analysis are presented in Table 2 . Mean ± SD maternal fasting glucose decreased from preintervention to postintervention within both the probiotic (4.90 ± 0.62 to 4.65 ± 0.49 mmol/L) and placebo (5.01 ± 0.77 to 4.65 ± 0.53 mmol/L) groups but no difference in baseline-adjusted postintervention levels between groups was detected. Analysis of fasting glucose among the per-protocol cohort excluding women with incomplete data collection and those receiving pharmacological therapy ( Table 3 ) also revealed a significant decrease from preintervention to postintervention within both the probiotic and placebo groups, but no significant between group differences.
| Pregnancy outcome | n | Probiotic group | n | Placebo group | P value |
|---|---|---|---|---|---|
| Postintervention maternal fasting glucose, mmol/L a | 74 | 4.65 (0.49) | 75 | 4.65 (0.53) | .373 |
| Total GWG, kg | 59 | 8.53 (5.67) | 63 | 7.96 (5.83) | .587 |
| Gestational age at delivery, d | 73 | 278.34 (12.25) | 74 | 277.23 (11.20) | .566 |
| Birthweight, kg | 73 | 3.57 (0.64) | 74 | 3.60 (0.57) | .845 |
| Birthweight centile | 56 | 47.49 (29.92) | 58 | 52.70 (29.18) | .349 |
| Neonatal length, cm | 64 | 51.09 (3.96) | 64 | 51.27 (3.01) | .771 |
| Neonatal head circumference, cm | 64 | 35.32 (2.88) | 65 | 35.09 (1.36) | .561 |
| 5-min Apgar score | 63 | 9.03 (0.25) | 63 | 8.68 (1.64) | .101 |
| Blood loss at delivery, mL | 63 | 416.38 (147.37) | 63 | 484.13 (320.68) | .131 |
| Postpartum hemorrhage, n (%) | 63 | 10 (15.9) | 63 | 13 (20.6) | .489 |
| Induction of labor, n (%) | 62 | 19 (30.6) | 65 | 15 (23.1) | .336 |
| Cesarean delivery, n (%) | 73 | 24 (32.9) | 74 | 21 (28.4) | .554 |
| Macrosomia (birthweight >4 kg), n (%) | 73 | 19 (26.0) | 74 | 17 (23.0) | .667 |
| LGA baby (>90th centile), n (%) | 56 | 5 (8.9) | 58 | 5 (8.6) | .954 |
| SGA baby (<10th centile), n (%) | 56 | 7 (12.5) | 58 | 7 (12.1) | .944 |
| Admission to NICU, n (%) | 71 | 14 (19.7) | 71 | 14 (19.7) | 1.000 |
| Male baby, n (%) | 73 | 33 (45.2) | 73 | 36 (49.3) | .619 |
| Preeclampsia/pregnancy induced HTN, n (%) | 68 | 6 (8.8) | 68 | 2 (2.9) | .145 |
| GWG below IOM guidelines, n (%) | 58 | 32 (55.2) | 63 | 32 (50.8) | .630 |
| GWG within IOM guidelines, n (%) | 58 | 11 (19.0) | 63 | 17 (27.0) | .296 |
| GWG above IOM guidelines, n (%) | 58 | 15 (25.9) | 63 | 14 (22.2) | .639 |
a Difference in postintervention fasting glucose calculated by analysis of covariance, adjusting for baseline glucose levels. For glucose unit conversion: 1 mmol/L = 18 mg/dL.
| Metabolic parameter | Probiotic group (n = 48) | Placebo group (n = 52) | P value c | ||||
|---|---|---|---|---|---|---|---|
| Pre | Post | P value b | Pre | Post | P value b | ||
| Fasting glucose, mmol/L | 4.76 (0.45) | 4.57 (0.42) | .001 | 4.85 (0.58) | 4.58 (0.45) | < .001 | .588 |
| Insulin, mU/L | 13.88 (6.40) | 13.04 (5.08) | .271 | 14.61 (9.34) | 13.58 (7.73) | .234 | .927 |
| HOMA index | 2.95 (1.42) | 2.65 (1.06) | .097 | 3.27 (2.40) | 2.85 (1.78) | .105 | .875 |
| C-peptide, ng/mL | 2.95 (1.12) | 3.00 (0.94) | .675 | 2.98 (1.22) | 3.05 (1.06) | .609 | .843 |
| Total cholesterol, mmol/L | 6.26 (1.10) | 6.53 (0.96) | .001 | 6.24 (1.10) | 6.74 (1.12) | < .001 | .031 |
| HDL cholesterol, mmol/L | 1.73 (0.36) | 1.68 (0.38) | .167 | 1.70 (0.33) | 1.69 (0.35) | .782 | .341 |
| LDL cholesterol, mmol/L | 3.47 (1.10) | 3.55 (0.88) | .348 | 3.45 (1.08) | 3.76 (0.98) | < .001 | .011 |
| Triglycerides, mmol/L | 2.46 (0.84) | 2.85 (0.95) | < .001 | 2.40 (0.71) | 2.83 (0.86) | < .001 | .687 |
| Total:HDL cholesterol ratio | 3.75 (0.95) | 4.05 (1.02) | .001 | 3.80 (0.94) | 4.14 (1.05) | < .001 | .704 |
| LDL:HDL cholesterol ratio | 2.08 (0.73) | 2.20 (0.67) | .063 | 2.11 (0.81) | 2.32 (0.80) | < .001 | .244 |
a Per-protocol cohort excludes women who did not complete study and those commenced on pharmacological therapy
b Within group differences from preintervention to postintervention, calculated by paired sample t test
c Between group differences in postintervention parameters, calculated by analysis of covariance adjusting for baseline values.
Secondary outcomes
On intention-to-treat analysis, 12 women from the probiotic group (17%) and 10 from the placebo group (14%) required pharmacological therapy, but this difference was not significant ( P = .643) ( Table 1 ). Among these women, only 1 in the probiotic group and 2 in the placebo group commenced metformin while the remainder were treated with insulin. The mean gestational age at commencement of pharmacological therapy also did not differ between treated and placebo groups (33.2 ± 2.3 vs 31.9 ± 1.6 weeks; P = .155). Neonatal anthropometry, assessed by absolute birthweight, birthweight centile, small or large for gestational age (birthweight <10th and >90th centiles, respectively), macrosomia (birthweight >4 kg), head circumference, and length, did not differ between the probiotic and placebo groups in the analysis for the intention-to-treat cohort ( Table 2 ) or for the per-protocol cohort excluding women who did not complete the study and those treated with insulin or metformin ( Supplementary Table 2 ).
Maternal metabolic and pregnancy outcomes
Results of the maternal metabolic parameters measured preintervention and postintervention are presented in Table 3 for the per-protocol cohort. Within group increases in total cholesterol, triglycerides, total cholesterol to HDL cholesterol ratio, and LDL to HDL cholesterol ratio were observed for both groups. On comparison of parameters between probiotic and placebo groups, lower postintervention levels of total cholesterol (6.53 ± 0.96 vs 6.74 ± 1.12 mmol/L; P = .031) and LDL cholesterol (3.55 ± 0.88 vs 3.76 ± 0.98 mmol/L; P = .011) were observed in the probiotic vs placebo groups, after adjusting for baseline levels. However, these results would not have survived Bonferroni correction for analysis of multiple outcomes.
Other maternal outcomes pertaining to gestational weight gain and delivery did not differ between the groups in the intention-to-treat analysis ( Table 2 ) or in the per-protocol analysis ( Supplementary Table 2 ). There were no preterm deliveries.
Fetal outcomes
No differences were noted in admission rates to the neonatal intensive care unit or Apgar score ( Table 2 ). There were no instances of miscarriage or fetal anomalies but 2 stillbirths occurred in the placebo group. A postmortem was conducted in both instances and a placental cause was attributed to one, while the cause of the second stillbirth remained unexplained. Regarding cord blood metabolic parameters, there were no differences detected in glucose, c-peptide, or lipids between the groups ( Table 4 ).
