Results From a European Multicenter Randomized Trial of Physical Activity and/or Healthy Eating to Reduce the Risk of Gestational Diabetes Mellitus: The DALI Lifestyle Pilot

Gestational diabetes mellitus (GDM) is associated with increased risk of adverse perinatal outcomes, with a continuous relationship between glycemia on oral glucose tolerance test (OGTT) and pregnancy outcome (1). Obesity in pregnancy and excess gestational weight gain (GWG) are associated with similar adverse pregnancy outcomes, which are additional to any harm from antenatal hyperglycemia (2).

Although obesity in pregnancy, excess GWG, and GDM are associated with adverse outcomes, the evidence that interventions to reduce these harms are beneficial is mixed (3). In a meta-analysis, combined lifestyle intervention (physical activity [PA] with healthy eating [HE]) was associated with reduced GWG (overall and in some studies), but not with improvement in perinatal outcomes (macrosomia or cesarean delivery) or significant reduction in GDM rates (3). More recent large randomized controlled trials (RCTs) have shown no reduction in the incidence of GDM with lifestyle intervention, with limited differences in GWG between intervention and control women (4–6), but potential reductions in birth weight (5). Although prevention of excessive GWG and macrosomia are key goals in such interventions, there has been concern that too little maternal weight gain could be associated with an increased incidence of small-for-gestational-age babies and reduced offspring fat-free mass (7). With these cautions, it is clear that further trials are needed to investigate efficacious approaches that will reduce GWG and progression to GDM, particularly in obese women.

We now report on the findings from the lifestyle pilot for DALI (Vitamin D And Lifestyle Intervention for GDM prevention) (8), undertaken to test the procedures, methodology, and lifestyle interventions for the forthcoming full RCT within a European multicenter RCT of three different lifestyle approaches for the prevention of GDM. The pilot reports on risk of GDM rather than GDM itself and does not include the vitamin D intervention.

The pilot study for DALI is a multicenter RCT of three different lifestyle approaches that could prevent GDM. The pilot was undertaken across 10 European centers in nine countries, each with its own nutrition and PA cultural norms. This diversity in background populations was included to ensure that the intervention was applicable in different settings and populations. The study was approved by the relevant ethics committees. The pilot was registered as an RCT (ISRCTN70595832). The primary outcomes were maternal weight gain (defined as the weight change from baseline measurement to the last measurement at 35–37 weeks of gestation), fasting glucose, and insulin sensitivity as derived from the HOMA at 35–37 weeks (9). The pilot study was devised to recruit four women for each of the three interventions in each center (total n = 120).

Pregnant women aged ≥18 years, before 20 weeks of gestation with a prepregnancy BMI ≥29 kg/m² (based on feasibility following a review of European obesity prevalence), were eligible for inclusion. Women were excluded from the study if they were diagnosed with GDM by OGTT before randomization, using World Health Organization (WHO) 2013 criteria (fasting venous plasma glucose ≥5.1 mmol/L and/or 1-h glucose ≥10 mmol/L and/or 2-h glucose ≥8.5 mmol/L) (10). Other exclusion criteria were pre-existing diabetes, inability to walk ≥100 m safely, multiple pregnancy, requiring a complex diet, a significant chronic medical condition or psychiatric disease, and inability to speak the major language of the country of recruitment fluently or to converse with the lifestyle coach in another language for which translated materials existed.

The 10 centers were Cambridge (coordinating center), Galway, Amsterdam, Leuven, Poznan, Pisa/Padova (as one center), Barcelona, Vienna, Odense, and Copenhagen. A common information sheet, in the local language, was given to eligible participants for consideration and the trial explained by the research midwife/nurse. Written consent was obtained for participation. The pilot study started in January 2012, and the last participant gave birth in August 2013.

After signed consent, assessments were made by the research midwife/nurse before 20 weeks (screening/baseline), 24–28 weeks, and 35–37 weeks of gestation and after delivery. The lifestyle intervention was delivered by a lifestyle coach. Where GDM developed after baseline, women were managed according to local practice. Supplementary Fig. 1 provides a diagrammatic overview of assessments and intervention.

Eligible women were randomly allocated to one of the three interventions (HE, PA, or both [HE+PA]) following an allocation schedule prestratified for intervention center, using a computerized random number generator. Sealed opaque envelopes (containing the intervention arm to which each participant was allocated with a randomization number) were sent to each site by the DALI trial coordinator (D.S.). After assurance of eligibility, the personal details of the participant were provided to the lifestyle coach who opened the next randomization envelope. Those involved with measurements were kept blinded to the intervention through noninvolvement and explicit training. Separation between coach and measurements team was assessed during site monitoring visits.

After randomization, five face-to-face and four optional telephone coaching sessions were scheduled for each participant, as shown in Supplementary Fig. 1. As previously described (8), coaching involved discussion of seven HE and/or five PA “messages” based upon previous work (11), as shown in Table 1. Messages were supported by a “toolkit” for the participant, including participant handbook, educational materials (e.g., adapted F.I.T.T. model [frequency, intensity, time, type]) based on American College of Obstetricians and Gynecologists (ACOG) guidelines (12), pedometers (Yamax Digiwalker SW-200, Tokyo, Japan), and flexible elastic dynabands (Thera-Band, Akron, OH). The message delivery was built upon behavioral theory principles (13), including patient empowerment and cognitive behavioral techniques inspired by motivational interviewing (14). At least four face-to-face coaching sessions were expected to take place before the second measurement session, and the intervention was completed by 35 weeks of gestation. Standardization of the intervention was achieved through a training program concluding with an observed session with an actor, provision of a desk file with all materials and methods, and use of a personal digital assistant (PDA; HTC HD7 Windows phone), with bespoke software to provide a framework for the session. A “paper” PDA including all fields was available for when there were any problems with the PDA.

A key component was for women to strive to achieve a maximum GWG of 5 kg, as this is the lower limit of the weight gain recommended by the Institute of Medicine (IOM) for those with a BMI ≥30 (15) and following the observation of better outcomes among Danish obese women with this degree of weight gain (16). If GWG was already beyond this before the start of the intervention or GWG exceeded 5 kg, women were supported to minimize their GWG thereafter. The coaches had scales available to assist women with their weight management when scales were not available in the home.

At the three OGTT sessions (Supplementary Fig. 1), after a 10-h fast, women completed questionnaires (including, e.g., demographics, prepregnancy weight, lifestyle, past/current medical and obstetric history, and medication use) and anthropometric measurements between blood tests. Local laboratories were used to rapidly obtain results of the OGTT to assess eligibility for the study. Blood samples were centrifuged, and separated serum and plasma aliquots (1,000 or 250 µL) were placed in microrack tubes and stored at −20°C or −80°C until further analysis in the central trial laboratory in Graz, Austria, certified according to ISO 9001 standards.

Glucose was measured using the hexokinase method (DiaSys Diagnostic Systems, Holzheim, Germany) with a lower limit of sensitivity of 0.1 mmol/L. Insulin was quantified by a sandwich immunoassay (ADVIA Centaur; Siemens Healthcare Diagnostics Inc., Vienna, Austria) with an analytical sensitivity of 0.5 mU/L, intra-assay coefficients of variation (CVs) of 3.3–4.6%, and interassay CVs of 2.6–5.9%. Leptin was measured using a sandwich ELISA (DRG Instruments, Marburg, Germany). The assay had an analytical sensitivity of 1.0 ng/mL, intra-assay CVs of 6.0–7.0%, and interassay CVs of 8.7–11.6%. All assays were performed according to the instructions of the manufacturer. HOMA was calculated as [glucose · insulin]/22.5 (9).

Height was measured at baseline with a stadiometer (SECA 206; SECA, Birmingham, U.K.; Leicester Height Measure), and the average value of two measurements was used. Women were weighed on calibrated electronic scales (SECA 888 and SECA 877) wearing no shoes and light clothes, to the nearest 0.1 kg; the average value of two measurements was used. Weight gain was defined as the change in objectively measured weight and was calculated for three periods: baseline to 24–28 weeks, baseline to 35–37 weeks, and 24–28 to 35–37 weeks. Data from the medical records were obtained regarding comorbidities, obstetric and perinatal outcomes, and birth weight.

As primary outcomes, GWG, fasting glucose, and insulin sensitivity (HOMA) at 35–37 weeks were used as they will be in the main trial. Secondary outcomes were the number of women who developed GDM, birth weight, insulin concentration, and leptin concentration.

Trial data were entered into a bespoke web-based electronic database using the Microsoft.NET development environment. For the assessment at 35–37 weeks, fasting glucose and insulin and insulin sensitivity were carried forward from 24–28 weeks if GDM was diagnosed. Women diagnosed with GDM at 24 weeks (n = 44) were excluded from the 35–37-week GWG analyses. Insulin and HOMA data were log transformed because of skewness. As a pilot, no power calculations were performed to decide on sample size, and it was arbitrarily agreed that four women would be recruited per intervention per site to test local procedures while minimizing pilot size.

Data were analyzed according to the intention-to-treat principle. Analyses were performed blind to the intervention group allocation. Analyses for the RCT were performed according to an a priori statistical analysis plan, and hence no adjustment was made to significance levels for multiple comparisons. Differences between total study sample and number of subjects dropping out of the study were assessed using Student t test (normally distributed continuous variables), Mann-Whitney U test (skewed continuous variables), or χ² test (categorical variables). To assess differences between intervention groups, multilevel analyses were performed, with two levels (coach and individual). This takes into account possible clustering effects, since the success of treatment can be dependent on the skills of the provider as well as specific characteristics of a treatment center. Analyses were adjusted for baseline values of the outcome or for BMI at baseline for the analyses with GWG outcomes. In the regression analyses, all three intervention groups were entered simultaneously to calculate differences between groups. In the multilevel models, log-transformed data for insulin and HOMA were used. For dichotomous variables, logistic regression models were used to assess differences between intervention groups. A bilateral P < 0.05 was taken as significant.

Figure 1 shows the flow of the participants throughout the study. A large proportion (22%) had GDM on the initial OGTT and hence were not included in the RCT. The numbers in each group and their baseline characteristics were similar (Table 2). Women dropping out (35; 23%) before the final OGTT had higher post-OGTT load glucose and insulin concentrations and BMI at baseline (Table 2). Numbers per site ranged from 3 in Odense to 26 in Belgium. Seven of the 150 women showed a weight gain of 5 kg or more between reported prepregnancy weight and baseline measurements, but many women were either unaware of or likely approximated their prepregnancy weight.

On average, women gained 8.6 kg (SD 4.6) from baseline to 35–37 weeks of gestation, and only 20% of women achieved the <5 kg GWG target from baseline across the three intervention groups. HE alone was associated with less weight gain by 24–28 weeks and 35–37 weeks than PA (–2.6 kg [95% CI −4.9, −0.2] and −1.6 kg [−3.0, −0.2], respectively) (Table 3). No differences were observed between HE+PA and the other two groups.

There were no significant differences in glucose, insulin, or leptin measurements across groups at baseline or by 24–28 weeks' gestation (Table 3). However, by 35–37 weeks' gestation, HE had significantly lower fasting glucose and 2-h insulin concentrations with nonsignificantly lower GDM incidence than PA. No significant differences between HE+PA and the other two groups were observed. There were no significant differences in HOMA or leptin (Table 3). At 24–28 weeks, 24 of 130 women (19%) had developed GDM, and by 35–37 weeks, this increased to 36 of 112 women (32%), of whom 3 received insulin treatment.

Procedures generally worked well (8), although recruitment overshot (to 150 women) due to the lag between baseline data collection, receiving OGTT results/confirming eligibility, and informing all sites that recruitment was completed. There was no expectation that the study would be large enough to show a difference in the primary outcomes comparing different interventions. The significant differences in GWG, fasting glucose, and 2-h insulin levels (albeit trivial) between HE and PA were therefore surprising, although even with the DALI interventions, 32% of women developed GDM. The data suggest that the HE intervention was more efficacious than the PA intervention in reducing GDM risk, although the higher GWG in PA may reflect an increase in muscle mass, not necessarily in adiposity.

Previous RCTs initiated before 20 weeks' gestation have largely shown no impact on GDM prevalence, fasting glucose (3–6,17), or insulin concentrations. Exceptions are the BAMBINO PA intervention pilot, which saw a reduction in fasting glucose by 28 weeks' gestation, but with GWG unreported (18), and the Lifestyle in Pregnancy (LiP) study, which showed lower insulin concentrations at 28–30 weeks with the intervention, but no significant reduction in fasting glucose (19).

The LiP study showed a GWG difference of only 1.4 kg between intervention and control women despite a very ambitious intervention program, including repeat visits to a dietitian, a weekly fitness class, and fitness center membership. Thus, a time-consuming program might be less effective than a simple one. One of the key issues with lifestyle interventions in general has been their limited impact on GWG when compared with control subjects. Several of the RCTs did have lower GWG in the intervention group (4–6), but control subjects had a lower GWG than anticipated, resulting in a lesser difference between the groups. A meta-analysis showed that differences between intervention and control groups across 10 RCTs achieved a mean difference in GWG of −2.21 kg (3). In spite of this, few reported a significant reduction in glycemia or GDM. The study by Quinlivan et al. (20) (n = 124) was the only trial in the meta-analysis that achieved a significant reduction in GDM (odds ratio [OR] 0.18) and achieved a 6.80-kg GWG difference. The study by Thornton et al. (21) (n = 232) also achieved a substantially lower GWG (−9.07 kg) but with a nonsignificant reduction in GDM (OR 0.53).

Our findings of more/better effects of diet alone are in line with a previous review, finding larger effects of dietary intervention alone on GWG, compared with PA or a combined approach (22). We speculate that a combined intervention might dilute the message or require too much change and then be demotivating. Another interpretation would be that a combined intervention does not mean “complete” but “partial” HE and PA; PA intervention being less effective, HE+PA ranks intermediate between HE and PA. The lack of effect of the PA interventions does not imply no effect of PA on pregnancy and offspring outcomes per se, as this pilot had no control subjects. Interventions to increase PA in pregnancy are well known to be difficult (23), especially in the third trimester. Although we are being as careful as possible to avoid the kinds of GWG convergence seen in some other major RCTs, it might be difficult to prevent a low GWG in the control group among women motivated to come into a lifestyle trial. Including multiple European cultures might generate a range of lifestyle responses in control subjects as well as intervention participants. Such differences between sites, with their very different underlying cultural landscapes (and climates) and associated differences in nutrition and PA habits, will be of interest in the main trial. At this point, it would appear that our generic approach to the lifestyle change messages (Table 1) has been successful. These were originally based on an intervention among New Zealand Maori (11) and hence expected to be cross cultural to some extent. We found the motivational interviewing approach to be well suited to the intervention. Even so, only 29% of participants achieved the <5 kg GWG goal even in the HE group of the pilot, a proportion comparable to those in other RCTs (4).

One other difference in the DALI approach from other RCTs is the exclusion of women with GDM by WHO 2013 criteria (22%) prior to recruitment. It may be that in the other trials, it was never possible to reverse pre-existing GDM in these women. However, there is debate over the validity of the new criteria in early pregnancy (10), which were built around data at 24–28 weeks' gestation (1). Data from China suggest that a significant proportion of women with GDM on the fasting glucose in early pregnancy have no GDM using the new criteria at 24–28 weeks (24).

The study has a number of weaknesses. It was established as a pilot to a full RCT of 440 women, and hence there were no power calculations performed to decide on the number of participants required and there was no untreated control group. As a pilot, the expertise of the intervention coaches would have improved over time, in preparation for the main trial. Numbers were still too small to look for adherence to lifestyle change or adverse pregnancy outcomes. The main trial will report on accelerometer and food diary data. We have not adjusted the significance level for multiple comparisons; however, the comparisons followed our a priori statistical analysis plan, and the difference in GWG is in line with other RCTs, which speaks against the role of chance.

In conclusion, an HE intervention among obese European women led to lower GWG, fasting glucose, and 2-h insulin concentrations by 35–37 weeks' gestation than a PA intervention or a combination of the two. Although a larger trial is still clearly needed, these pilot findings are promising and support the use of early HE interventions in obese pregnant women.

Acknowledgments. The authors thank the coaches, research midwives and nurses, women, and health professionals collaborating in the recruitment for this study.

Funding. This project has received funding from the European Community's 7th Framework Programme (FP7/2007-2013) under grant agreement 242187. In the Netherlands, additional funding was provided by the Netherlands Organisation for Health Research and Development (ZonMw) (grant 200310013). In the U.K., the DALI team acknowledges the support received from the NIHR Clinical Research Network: Eastern, especially the local diabetes clinical and research teams based in Cambridge. In Spain, additional funding was provided by CAIBER 1527-B-226.

The funders had no role in any aspect of the study beyond funding.

Duality of Interest. D.M.J. is part of an advisory board for research meetings arranged by Novo Nordisk. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. D.S. contributed to the conception and design of the trial, wrote the initial draft of the paper, read and corrected draft versions of the manuscript, and approved the final manuscript. J.G.M.J. performed the analyses, read and corrected draft versions of the manuscript, and approved the final manuscript. S.G., G.J., and J.H. read and corrected draft versions of the manuscript and approved the final manuscript. R.D., A.v.A., R.C., J.M.A., F.D., G.D., A.K.-W., P.D., E.R.M., D.M.J., L.L.A., A.L., M.D., A.B., E.W.-O., A.Z., D.H., P.R., and F.J.S. contributed to the conception and design of the trial, read and corrected draft versions of the manuscript, and approved the final manuscript. M.N.M.v.P. contributed to the conception and design of the trial, performed the analyses, read and corrected draft versions of the manuscript, and approved the final manuscript. D.S. 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 75th Scientific Sessions of the American Diabetes Association, Boston, MA, 5–9 June 2015.