Less sedentary time is associated with a more favourable glucose-insulin axis in obese pregnant women—a secondary analysis of the DALI study

Gestational diabetes mellitus (GDM) is a common pregnancy complication in Europe (≈6.1% overall, with wide variation) and is linked to adverse short-term outcomes (fetal overgrowth, higher cesarean rates) and long-term risks (offspring obesity, maternal type 2 diabetes). Rising maternal obesity contributes to increasing GDM prevalence; preconception obesity is the key modifiable risk factor, yet it cannot be addressed once pregnancy begins. Therefore, modifiable behaviors during pregnancy, particularly physical activity (PA) and sedentary time (ST), are of interest. PA in pregnancy improves glucose uptake and lowers circulating insulin, and meta-analyses indicate that PA before and during pregnancy reduces GDM risk. However, most prior studies assessed PA before or early in pregnancy, not accounting for typical declines in activity as gestation progresses, which may influence associations with glucose-insulin dynamics. Sedentary behavior independently predicts type 2 diabetes, cardiovascular disease, and mortality outside pregnancy, but evidence regarding sedentary behavior and the glucose-insulin axis in pregnancy is inconsistent, likely due to heterogeneous definitions and measurements. The DALI randomized trials promoted PA and reduced ST, showing reduced ST, increased MVPA, and lower neonatal adiposity, but no improvements in maternal glucose or insulin; intergroup differences in activity were small. Analyzing all randomized participants as one cohort with objective measures enables assessment across a broader activity range. The hypothesis was that reduced ST and increased MVPA would improve the glucose-insulin axis. The study aimed to investigate longitudinal associations between objectively measured PA/ST and the glucose-insulin axis in obese pregnant women.

Prior evidence shows PA improves glucose uptake and reduces circulating insulin; a meta-analysis reported lower GDM risk with PA before and during pregnancy. Activity typically decreases as pregnancy progresses, which may modify associations. Sedentary behavior is an independent risk factor for type 2 diabetes, cardiovascular disease, and premature mortality outside pregnancy, but limited pregnancy studies report inconsistent associations with glucose-insulin outcomes due to heterogeneous definitions and measurement methods. Objective, longitudinal assessments of PA and sedentary behavior during pregnancy are needed to clarify their roles in maternal glucose-insulin regulation.

Design and setting: Secondary analysis of the DALI (vitamin D And Lifestyle Intervention for gestational diabetes mellitus prevention) multi-centre randomized controlled trials conducted 2012–2015 across 11 sites in 9 European countries (Austria, Belgium, Denmark [Odense, Copenhagen], Ireland, Italy [Padua, Pisa], Netherlands, Poland, Spain, UK). Trials registered ISRCTN70595832; ethics approvals obtained; written informed consent provided. For this analysis all randomized participants were pooled and analyzed observationally. Participants: Pregnant women <20 weeks’ gestation, singleton pregnancy, age ≥18 years, pre-pregnancy BMI ≥29 kg/m², and without GDM at baseline OGTT (IADPSG criteria). Exclusions: pre-existing diabetes, chronic medical diseases, abnormal calcium metabolism or calcium measurements (for vitamin D trial). Inclusion in this analysis required at least two valid accelerometer assessments across the three pregnancy time points to enable longitudinal analysis. Randomization groups (pilot: HE, PA, HE&PA; lifestyle: HE, PA, HE&PA, control; vitamin D: vitamin D±HE&PA or placebo±HE&PA) were combined into one cohort for observational analyses. Data collection time points: <20 weeks (T1), 24–28 weeks (T2), 35–37 weeks (T3), plus delivery. Standardized 75 g OGTTs were performed at each time point in women without GDM, with blood sampled fasting, 60, and 120 min post-load. Laboratory measures: Glucose (hexokinase method; sensitivity 0.1 mmol/L). Insulin (ADVIA Centaur sandwich immunoassay; analytical sensitivity 0.5 mU/L; intra-assay CV 3.3–4.6%; inter-assay CV 2.6–5.9%). GDM at T2/T3 defined by IADPSG: fasting ≥5.1 mmol/L and/or 1 h ≥10 mmol/L and/or 2 h ≥8.5 mmol/L. Insulin resistance: HOMA-IR = (glucose × insulin) / 22.5 (validated in pregnancy). Beta-cell function proxies: Stumvoll first phase = 1194 + 4.724×insulin0 − 117×glucose60 + 1.414×insulin60; Stumvoll second phase = 295 + 0.349×insulin60 − 25.72×glucose60 + 1.107×insulin0 (not validated in pregnancy). Other measures: Questionnaires captured age, ethnicity, parity, marital status, employment, smoking, alcohol, pre-pregnancy weight. Height measured at first visit; weight measured at T1–T3. Pre-pregnancy BMI calculated kg/m². Offspring sex from medical records. Physical activity assessment: ActiGraph accelerometers (GT1M, GT3X+, or ActiTrainer) worn during waking hours for 3 days at each time point on an elastic belt over the right hip; removed for water-based activities (time/reason recorded). Non-wear: ≥90 min of zero counts. Valid day: wear time >480 min; required 3 valid days per time point. Minutes/day in sedentary (<100 cpm), light (100–1951 cpm), and MVPA (>1951 cpm) defined by Freedson cut-points; swimming time from diary added to MVPA minutes. Statistical analysis: Descriptive comparisons between included (≥2 accelerometer assessments) and excluded participants using t-tests/chi-square. Outcomes (fasting glucose, fasting insulin, HOMA-IR, Stumvoll first/second phase) and activity variables (ST, MVPA, wear time) summarized by time point; paired t-tests for temporal changes. Skewed variables (MVPA, insulin, HOMA-IR, Stumvoll indices) log-transformed (natural log) as needed. Primary longitudinal analyses used linear mixed models with observations nested within individuals (random intercept and slope). Time modeled using gestational age in weeks; quadratic term used where appropriate (Stumvoll indices). ST and MVPA were included together for independent estimates, scaled per 10 min/day, and adjusted for wear time. Covariates: maternal age, education, ln pre-pregnancy BMI, randomization group, and country. Potential effect modification by offspring sex and pre-pregnancy BMI tested via interaction terms with ST and MVPA (p<0.10 considered significant for interactions). Within-between (hybrid) random-effects models decomposed within-person changes from between-person differences. Sensitivity analyses: (1) Complete cases (3 accelerometer measurements); (2) excluding women who developed GDM; (3) modeling time categorically (T1, T2, T3); (4) multiple imputation by chained equations (20 datasets) respecting two-level structure and using transform-then-impute for derived variables. Two-tailed p<0.05 was significant. Analyses in R 3.6.1 with lme4; multiple imputation with mice; plotting with dotwhisker and ggplot2.

- Sample: 232 women (from 740 randomized) with ≥2 valid accelerometer assessments; included vs excluded were generally similar, though included had fewer Caucasian women and more were employed. - Temporal patterns: Wear time and MVPA decreased across pregnancy; sedentary minutes/day remained ~constant but sedentary percentage of wear time rose (from 69.7% at <20 weeks to 73.0% at 35–37 weeks). MVPA median decreased from 38.1 to 29.6 min/day. Fasting glucose slightly decreased to mid-pregnancy; 1 h and 2 h glucose increased over pregnancy. Fasting and postprandial insulin, HOMA-IR, and Stumvoll first/second phase increased throughout. - Between-person associations (averaged across time): Each 10 min/day more sedentary time was associated with higher fasting glucose (Estimate 0.008; 95% CI 0.002, 0.014 mmol/L scale), higher ln fasting insulin (0.011; 0.002, 0.019), higher ln HOMA-IR (0.012; 0.004, 0.021), and higher ln Stumvoll first and second phase (0.008; 0.001, 0.014 and 0.007; 0.001, 0.014). More MVPA was associated with lower ln Stumvoll first and second phase (−0.137; −0.210, −0.064 and −0.133; −0.202, −0.063). - Within-person changes over gestation: Increases in sedentary time were associated with increases in ln Stumvoll first and second phase (both 0.006; 0.000, 0.011); no significant within-person associations for other outcomes or for MVPA changes were reported. - Overall, sedentary time showed stronger and more consistent associations with the glucose–insulin axis than MVPA in this obese pregnant cohort.

The study addressed whether objectively measured sedentary time and MVPA are longitudinally related to markers of glucose-insulin regulation in obese pregnant women. Findings indicate that women who are more sedentary have higher fasting glucose, insulin, insulin resistance (HOMA-IR), and higher estimated first and second phase insulin secretion, and that increases in sedentary time during pregnancy track with higher estimated insulin secretion. Higher MVPA was associated with lower estimated insulin secretion between individuals, but sedentary time showed broader associations across glycemic and insulinemic measures. These results suggest that sedentary behavior may exert a stronger influence on the glucose-insulin axis than MVPA in this high-risk population, potentially reflecting increased insulin demand and resistance with greater inactivity. The observed decline in MVPA and rise in sedentary proportion over gestation highlight the importance of early behavior change. Although prior DALI trial group comparisons showed modest changes without glycemic improvements, analyzing the cohort across a wider activity range revealed clinically relevant associations, supporting recommendations to reduce sedentary time alongside promoting MVPA during pregnancy.

In obese pregnant women, sedentary time is more strongly and consistently associated with adverse glucose–insulin measures than MVPA. Reducing sedentary time, in addition to increasing MVPA, may benefit maternal glucose-insulin regulation. Given that activity declines during pregnancy, interventions targeting reductions in sedentary behavior and increases in MVPA should begin in early pregnancy or preconception. Future research should test causal effects of sedentary time reduction on maternal glycemic outcomes and GDM incidence, refine objective assessment protocols across pregnancy, and determine optimal timing and components of behavior-change interventions.

- The Stumvoll first and second phase insulin release indices used as proxies for beta-cell function are not validated in pregnancy, which may affect interpretation. - Only 232 of 740 randomized women had ≥2 valid accelerometer assessments, introducing potential selection bias; included participants differed somewhat (fewer Caucasian, more employed). - Participants who developed GDM did not undergo subsequent OGTTs, creating informative missingness; addressed via sensitivity analyses and multiple imputation under a missing-at-random assumption, which may not fully hold. - Physical activity was measured over three valid days per time point and excluded most water-based activities (except swimming added from diaries), possibly limiting capture of habitual activity patterns. - Light physical activity was not modeled due to multicollinearity with sedentary time and MVPA, potentially omitting relevant information. - Despite adjustment for key covariates and country, residual confounding cannot be excluded in this observational analysis pooling randomized groups.