OBJECTIVE

To evaluate changes in insulin physiology in euglycemic pregnancy and gestational diabetes mellitus (GDM).

RESEARCH DESIGN AND METHODS

Participants underwent oral glucose tolerance tests at ≤15 weeks’ gestation (early pregnancy), 24–32 weeks’ gestation (mid-late pregnancy), and 6–24 weeks postpartum. We evaluated longitudinal changes in insulin secretory response (log Stumvoll first-phase estimate) and insulin sensitivity (log Matsuda index) using linear mixed models. We then evaluated participants who met GDM criteria in early pregnancy (early GDM) and mid-late pregnancy (classic GDM) separately from those without GDM. We derived the pregnancy insulin physiology (PIP) index to quantify β-cell compensation for insulin resistance.

RESULTS

Among 166 participants, 21 had early GDM and 24 developed classic GDM. Insulin sensitivity was reduced slightly in early pregnancy (β = −0.20, P < 0.001) and substantially in mid-late pregnancy (β = −0.47, P < 0.001) compared with postpartum. Insulin secretory response (adjusted for insulin sensitivity) was augmented in early pregnancy (β = 0.16, P < 0.001) and mid-late pregnancy (β = 0.16, P = 0.001) compared with postpartum. Compared with postpartum, the PIP index was augmented in early pregnancy (β = 215, P = 0.04) but not mid-late pregnancy (β = 55, P = 0.64). Early GDM was distinguished by a substantial reduction in early pregnancy insulin sensitivity (β = −0.59, P < 0.001) compared with postpartum. Both early and classic GDM lacked evidence of early pregnancy augmentation of insulin secretory response (adjusted for insulin sensitivity) and the PIP index (P > 0.1 vs. postpartum). Early pregnancy PIP index predicted GDM independent of participant characteristics (area under the curve without PIP index 0.70 [95% CI 0.61–0.79], area under the curve with PIP index 0.87 [95% CI 0.80–0.93]).

CONCLUSIONS

β-Cell function is enhanced in early pregnancy. Deficient first-trimester β-cell function predicts GDM.

Gestational diabetes mellitus (GDM), hyperglycemia first recognized in pregnancy, is conventionally diagnosed at 24–28 weeks’ gestation. By this gestational age, dramatic shifts in insulin physiology can be observed, including markedly increased insulin resistance and a concomitant increase in insulin secretory response (1,2). A better understanding of the insulin physiology underlying GDM could facilitate new approaches to its prevention and treatment and might also provide insights that are relevant to other forms of diabetes.

GDM has been conceptualized as a manifestation of a chronic maternal disease, whereby there is subclinical pancreatic β-cell dysfunction that is present prior to pregnancy and is discovered in the context of insulin resistance of late pregnancy and universal screening for glucose intolerance (1,3). However, several pieces of evidence suggest that there is also a pregnancy-related component to the disorder. These include the observations that GDM is more common in twin pregnancies and has incomplete recurrence in subsequent pregnancies (4,5). Using data from a longitudinal study conducted >20 years ago that enrolled participants prior to pregnancy, we previously demonstrated that early pregnancy insulin secretory response to intravenous glucose is augmented prior to the onset of gestational insulin resistance (6). It follows that a defect in the β-cell’s ability to respond to the pregnancy hormonal milieu might contribute to GDM, but to our knowledge this has not been rigorously examined in a contemporary cohort.

We enrolled pregnant individuals with diabetes risk factors in a longitudinal study to test the hypothesis that there is an insulin resistance–independent augmentation of insulin secretory response to oral glucose in early pregnancy that is deficient in GDM. Because some participants already had hyperglycemia in the first trimester and others developed hyperglycemia later in gestation, we studied the longitudinal changes in insulin physiology in these phenotypes separately. Prior research on insulin physiology is predicated upon a rectangular hyperbolic relationship between insulin resistance and secretory response (79); that is, for individuals with the same level of β-cell function, the product of insulin sensitivity and insulin secretory response measures (deemed the disposition index [DI]) is constant. To rigorously examine gestational β-cell function, we evaluated the rectangular hyperbolic assumption and introduced a robust modeling approach that is correctly specified under a wider range of hyperbolic relationships. Using this approach, we present the pregnancy insulin physiology (PIP) index as a valid generalization of the DI concept and test its ability to predict GDM.

Participants

The Study of Pregnancy Regulation of Insulin and Glucose (SPRING) is a longitudinal cohort that enrolled pregnant women between 2016 and 2021. Inclusion criteria were pregnancy at ≤15 weeks’ gestation along with the presence of one or more GDM risk factors (either personal history of GDM alone, family history of diabetes or GDM alone, or BMI ≥25 kg/m2 plus one additional risk factor as described by the American Diabetes Association) (10). Exclusion criteria were known pregestational diabetes, use of medications that affect glucose levels, and a history of bariatric surgery.

Study Procedures

Participants attended study visits at ≤15 weeks’ gestation (early pregnancy), 24–32 weeks’ gestation (mid-late pregnancy), and 6–24 weeks postpartum. They completed questionnaires regarding age, personal and family health history, race and ethnicity, employment, education, and marital status. Height was measured at the first visit, and weight was measured at each visit. At each visit participants underwent a 75-g oral glucose tolerance test (OGTT); we drew blood samples before (fasting) and 30, 60, and 120 min after the glucose load. Glucose results were reported to obstetric providers. Initially, all participants whose glucose values met the International Association of the Diabetes and Pregnancy Study Groups (IADPSG) criteria (11) were diagnosed with GDM and referred for treatment regardless of gestational age; those who met these criteria at the early pregnancy visit were excluded from the mid-late pregnancy visit OGTT. After July 2018, participants with an isolated mild elevation in fasting glucose (92 and 99 mg/dL) at the first study visit were not diagnosed with GDM and could participate in the mid-late pregnancy OGTT.

Laboratory Analysis

Blood samples for glucose were drawn in sodium fluoride tubes and taken promptly to the hospital’s chemistry laboratory for analysis, mirroring the procedure used clinically at the study site. Glucose was measured using the Cobas 8000 analyzer (Roche Diagnostics; intra-assay coefficient of variation <1%). Serum insulin was measured in two laboratories (intra-assay coefficient of variation for both assays <5%). Initially, fresh serum was assayed using an electrochemiluminescent immunoassay on the Roche automated platform at Laboratory 1. After July 2018, frozen serum was assayed using Beckman Coulter Access ultrasensitive insulin assay at Laboratory 2. Results from samples assayed using both methods were highly correlated (r2 = 0.97) but required a linear transformation (Supplementary Methods).

Statistical Analysis

Insulin Secretory Response and Sensitivity Indices

We measured insulin secretory response with the Stumvoll first-phase estimate (12):

graphic
where the subscript corresponds to the minutes postload of the measurement (0 indicates fasting). Stumvoll is well correlated with the first-phase insulin secretory response measured by intravenous glucose tolerance test in pregnancy (13).

We measured insulin sensitivity (the opposite of insulin resistance) with the Matsuda index (14):

graphic

Matsuda has been validated against euglycemic clamp measures of insulin sensitivity in pregnancy (15).

In the formulae above, glucose is measured in mg/dL and insulin is measured in µIU/mL.

Longitudinal Analyses

At each visit, we calculated observed population means and SD of glucose and insulin measures as well as Stumvoll and Matsuda indices. In unadjusted analyses, we used paired t tests to compare within-participant differences in these measures at each pregnancy visit with postpartum measures.

In statistical analyses, we used IADPSG criteria to define GDM at both pregnancy time points (11). Participants were stratified into subgroups based on if and when they met GDM criteria (11): GDM at the early pregnancy visit (“early GDM”), GDM at the mid-late pregnancy visit but not the early pregnancy visit (“classic GDM”), or neither (“no GDM”). Participants with early GDM were excluded from the analyses of glucose and insulin physiology in mid-late pregnancy. Participants who were not diagnosed with GDM at the early pregnancy visit and missed the mid-late pregnancy visit were excluded from stratified analyses. We compared unadjusted glucose, insulin, Stumvoll, and Matsuda values between these subgroups at each visit using two-sample t tests.

To assess whether insulin secretory response and insulin sensitivity differed between visits, we fit confounder-adjusted linear mixed-effects models with a categorical variable for visit (postpartum reference) and adjusted for participant characteristics, including age, family history of diabetes, Hispanic ethnicity, marital status, completion of college, and BMI at each visit (Supplementary Methods). We modeled Stumvoll and Matsuda separately, which were both log transformed to account for skewness. To assess whether differences in insulin secretory response between visits were independent of changes in insulin sensitivity, we modeled log Stumvoll as the outcome and adjusted for time-varying log Matsuda. Analyses were repeated after stratification by GDM status; we excluded the mid-late pregnancy visit from models in the early GDM group due to limited sample size from per-protocol exclusion from this visit.

Between-Group Comparisons

To compare outcomes measured at each visit across subgroups, we fit separate confounder-adjusted linear regression models for each visit with group membership as a categorical variable with a no GDM group reference. We also investigated whether observed changes in insulin sensitivity between early pregnancy and postpartum differed between GDM subgroups by fitting a linear regression model where the outcome was log Matsuda at early pregnancy minus the postpartum value, with adjustment for subgroup membership and the same participant characteristics as the longitudinal models (including BMI in early pregnancy).

Missing Data

Multiple imputation was used to account for missing data at attended visits (Supplementary Methods). Stumvoll and Matsuda indices were calculated using the imputed values.

DI and PIP Index

We examined a DI-like measure (defined as the product of Stumvoll and Matsuda) as a strategy for modeling insulin secretion and sensitivity simultaneously. We first evaluated whether a rectangular hyperbolic relationship between Matsuda and Stumvoll was present by plotting the log-transformed values against one another, with evaluation of whether the 95% CI for the slope includes −1 (16). This analysis was restricted to the early pregnancy visit in the no GDM group. To account for measurement error in Matsuda and Stumvoll, we utilized Deming regression, assuming a variance in the error of 15.8% for Matsuda and 18.1% for Stumvoll (17,18).

In the Supplementary Methods, we demonstrate that a model with DI as the outcome is misspecified if the relationship between Matsuda and Stumvoll is not rectangular hyperbolic. The longitudinal model that we fit using log Stumvoll as the outcome and adjusting for log Matsuda is correctly specified across a wide range of hyperbolic relationships. As an alternative to the DI or DI-like measures, we propose the PIP index as a measure for characterizing β-cell function that better accounts for the hyperbolic relationship between insulin resistance and secretory response. The process we used to derive the PIP index is described in the Supplementary Methods. The derived formula is described in the Results section.

We modeled the PIP index as the outcome in confounder-adjusted linear mixed-effects models. We also calculated percent change in PIP index for each participant, comparing the postpartum visit to each pregnancy visit. We calculated the proportion of participants overall and within each subgroup whose PIP index was more than 10% lower, more than 10% higher, or less than 10% different at each visit compared with postpartum; we used the χ2 test to compare the distribution of these categories in the early and classic GDM subgroups to the no GDM subgroup. Percent change calculations were restricted to participants for whom PIP index could be directly calculated using observed data.

We also fit a series of logistic regression models for the outcome of GDM (either early or classic), which included the following sets of predictors: 1) participant characteristics (age, family history of diabetes, race and ethnicity, marital status, educational attainment, and BMI) alone, 2) PIP index alone, and 3) participant characteristics plus PIP index. We also examined models that included the early pregnancy and postpartum PIP index separately and together. Fasting glucose at the early pregnancy visit was also used as a predictor for comparison. Area under the receiver operating characteristic curve (AUC) and 95% CI were estimated for each model. We repeated this analysis after excluding participants with early GDM.

Sensitivity Analyses

We conducted a series of sensitivity analyses. First, we restricted analyses to participants who attended all visits. Two additional versions of the confounder-adjusted models were fit, one without BMI adjustment and one that adjusted for exclusive breastfeeding at the postpartum visit.

Statistical Software

Analyses were conducted in R version 4.1.1 (19). Linear mixed-effects models were fit using the lme function in the nlme package (20). Multiple imputation was conducted using the MICE package (21). Likelihood ratio tests of nested linear mixed-effects models were conducted using the lmtest package (22). Deming regression was conducted using the mcr package (23). Animations were generated using the gganimate package (24).

Participant Characteristics

The cohort consisted of 166 participants, 127 of whom completed the mid-late pregnancy visit and 114 of whom completed the postpartum visit (Supplementary Fig. 1).

Table 1 presents the characteristics of the participants, who were all cis-gendered women. Twenty-one participants (12.6%) met criteria for GDM in early pregnancy. Participants with early GDM were older than participants who never met criteria for GDM. They had average BMI similar to those of individuals who never developed GDM. A larger proportion of them had a personal history of GDM compared with those who did not develop GDM.

Table 1

Participant characteristics, overall and stratified by GDM subgroup

OverallNo GDMEarly GDMClassic GDMP value
Participants (n166 99 21 24  
Attended mid-late pregnancy visit (n127 99 24  
Attended postpartum visit (n114 73 18 21  
Age at early pregnancy visit (years) 32.9 (4.68) 32.8 (4.31) 35.3 (3.88) 33.9 (4.78) 0.030 
Gestational age at early pregnancy visit (weeks) 12.6 (1.60) 12.7 (1.46) 12.0 (1.79) 12.8 (1.60) 0.210 
Gestational age at mid-late pregnancy visit (weeks) 26.6 (1.88) 26.5 (1.88) 25.6 (1.61) 27.1 (1.90) 0.286 
Weeks postpartum at postpartum visit 10.8 (4.91) 10.6 (4.67) 11.6 (5.40) 10.7 (5.60) 0.802 
BMI (kg/m2     
 Early pregnancy visit 29.6 (6.78) 29.9 (5.91) 30.1 (7.14) 26.7 (6.21) 0.086 
 Mid-late pregnancy visit 31.7 (6.12) 32.1 (5.60) 36.6 (12.28) 29.2 (6.39) 0.185 
 Postpartum visit 29.8 (5.66) 30.8 (5.18) 29.1 (6.71) 26.7 (5.55) 0.015 
Nulliparous 84 (50.6%) 50 (50.5%) 11 (52.4%) 10 (41.7%) 0.706 
Family history of diabetes 65 (39.2%) 34 (34.3%) 9 (42.9%) 12 (50.0%) 0.328 
Personal history of GDM among parous participants 20 (24.4%) 5 (10.2%) 7 (70.0%) 6 (42.9%) <0.001 
Family history of GDM 29 (17.5%) 20 (20.2%) 5 (23.8%) 3 (12.5%) 0.597 
Race/ethnicity     0.006 
 Hispanic/Latina 34 (20.5%) 19 (19.2%) 2 (9.5%) 4 (16.7%)  
 Non-Hispanic/Latina      
  White 89 (53.6%) 57 (57.6%) 10 (47.6%) 15 (62.5%)  
  Black 16 (9.6%) 12 (12.1%) 2 (9.5%) 0 (0.0%)  
  Asian 17 (10.2%) 4 (4.0%) 7 (33.3%) 4 (16.7%)  
  None of the above 10 (6.0%) 7 (7.1%) 0 (0.0%) 1 (4.2%)  
Employed full-time 117 (70.5%) 69 (69.7%) 15 (71.4%) 19 (79.2%) 0.654 
Married 123 (74.1%) 77 (77.8%) 15 (71.4%) 19 (79.2%) 0.792 
Completed college 145 (87.3%) 87 (87.9%) 19 (90.5%) 23 (95.8%) 0.514 
Breastfeeding, postpartum visit     0.746 
 Exclusively breastfeeding 56 (49.6%) 36 (50.0%) 9 (50.0%) 10 (47.6%)  
 Some breastfeeding and some formula 39 (34.5%) 24 (33.3%) 8 (44.4%) 7 (33.3%)  
 Exclusively formula 18 (15.9%) 12 (16.7%) 1 (5.6%) 4 (19.0%)  
OverallNo GDMEarly GDMClassic GDMP value
Participants (n166 99 21 24  
Attended mid-late pregnancy visit (n127 99 24  
Attended postpartum visit (n114 73 18 21  
Age at early pregnancy visit (years) 32.9 (4.68) 32.8 (4.31) 35.3 (3.88) 33.9 (4.78) 0.030 
Gestational age at early pregnancy visit (weeks) 12.6 (1.60) 12.7 (1.46) 12.0 (1.79) 12.8 (1.60) 0.210 
Gestational age at mid-late pregnancy visit (weeks) 26.6 (1.88) 26.5 (1.88) 25.6 (1.61) 27.1 (1.90) 0.286 
Weeks postpartum at postpartum visit 10.8 (4.91) 10.6 (4.67) 11.6 (5.40) 10.7 (5.60) 0.802 
BMI (kg/m2     
 Early pregnancy visit 29.6 (6.78) 29.9 (5.91) 30.1 (7.14) 26.7 (6.21) 0.086 
 Mid-late pregnancy visit 31.7 (6.12) 32.1 (5.60) 36.6 (12.28) 29.2 (6.39) 0.185 
 Postpartum visit 29.8 (5.66) 30.8 (5.18) 29.1 (6.71) 26.7 (5.55) 0.015 
Nulliparous 84 (50.6%) 50 (50.5%) 11 (52.4%) 10 (41.7%) 0.706 
Family history of diabetes 65 (39.2%) 34 (34.3%) 9 (42.9%) 12 (50.0%) 0.328 
Personal history of GDM among parous participants 20 (24.4%) 5 (10.2%) 7 (70.0%) 6 (42.9%) <0.001 
Family history of GDM 29 (17.5%) 20 (20.2%) 5 (23.8%) 3 (12.5%) 0.597 
Race/ethnicity     0.006 
 Hispanic/Latina 34 (20.5%) 19 (19.2%) 2 (9.5%) 4 (16.7%)  
 Non-Hispanic/Latina      
  White 89 (53.6%) 57 (57.6%) 10 (47.6%) 15 (62.5%)  
  Black 16 (9.6%) 12 (12.1%) 2 (9.5%) 0 (0.0%)  
  Asian 17 (10.2%) 4 (4.0%) 7 (33.3%) 4 (16.7%)  
  None of the above 10 (6.0%) 7 (7.1%) 0 (0.0%) 1 (4.2%)  
Employed full-time 117 (70.5%) 69 (69.7%) 15 (71.4%) 19 (79.2%) 0.654 
Married 123 (74.1%) 77 (77.8%) 15 (71.4%) 19 (79.2%) 0.792 
Completed college 145 (87.3%) 87 (87.9%) 19 (90.5%) 23 (95.8%) 0.514 
Breastfeeding, postpartum visit     0.746 
 Exclusively breastfeeding 56 (49.6%) 36 (50.0%) 9 (50.0%) 10 (47.6%)  
 Some breastfeeding and some formula 39 (34.5%) 24 (33.3%) 8 (44.4%) 7 (33.3%)  
 Exclusively formula 18 (15.9%) 12 (16.7%) 1 (5.6%) 4 (19.0%)  

Continuous variables are presented as mean (SD). Binary and categorical variables are presented as count (percent). For binary variables (nulliparous, family history of diabetes, etc.), the counts presented are for participants who had this characteristic. Counts and percentages for personal history of GDM are restricted to parous participants. Percentages are calculated relative to the number of participants for whom the given characteristic was not missing. P values are calculated using ANOVA, not assuming equal variances for continuous variables, and χ2 tests for binary and categorical variables. Breastfeeding practice was obtained only at the postpartum visit. Participants in the no GDM group are those who attended the early and mid-late pregnancy visits and did not meet GDM criteria at either visit. This excludes the 22 participants who did not meet GDM criteria at the early-pregnancy visit and missed the mid-late pregnancy visit.

Of participants who did not meet criteria for GDM in early pregnancy, 123 attended the mid-late pregnancy study visit, 24 (19.5%) of whom newly met criteria for GDM. These participants with classic GDM were slightly older than participants who never developed GDM. Their average BMI was lower than the BMI of those who never developed GDM, and they were less likely to be nulliparous; a larger proportion of them had a personal history of GDM compared with those who did not develop GDM.

Glucose and Insulin Levels

Table 2 shows unadjusted fasting and postload glucose and insulin values at each study visit. Glucose and hemoglobin A1c values were previously reported in most of the cohort (25,26). No participants met criteria for overt diabetes by hemoglobin A1c in early pregnancy. No participants in the early GDM group met criteria for overt diabetes by OGTT postpartum.

Table 2

Observed levels of glucose, insulin, Stumvoll, and Matsuda for each visit and measurement time point (if applicable)

All participantsNo GDMEarly GDMClassic GDM
Mean (SD)P (PP)Mean (SD)P (PP)Mean (SD)P (PP)P (NG)Mean (SD)P (PP)P (NG)
Glucose (mg/dL)           
 Early pregnancy           
  Fasting 81.50 (7.56) 0.014 79.02 (5.74) 0.002 93.81 (7.99) 0.220 <0.001 81.96 (4.80) 0.163 0.014 
  30 min 127.92 (26.55) 0.264 120.01 (23.69) 0.127 157.48 (22.58) 0.564 <0.001 138.09 (19.29) 0.402 <0.001 
  60 min 124.85 (36.51) 0.378 113.24 (30.56) 0.513 174.52 (35.14) 0.042 <0.001 139.71 (19.54) 0.712 <0.001 
  120 min 105.77 (29.72) 0.076 96.28 (24.68) 0.711 144.95 (33.68) 0.021 <0.001 112.75 (21.01) 0.795 0.002 
 Mid-late pregnancy           
  Fasting 80.05 (6.75) <0.001 78.55 (5.28) <0.001 91.75 (7.54)   84.29 (8.40) 0.656 0.003 
  30 min 133.35 (23.08) 0.917 129.20 (21.52) 0.738 134.75 (37.66)   149.04 (20.32) 0.223 <0.001 
  60 min 138.64 (33.16) <0.001 129.34 (25.57) 0.007 129.25 (36.47)   177.04 (32.80) <0.001 <0.001 
  120 min 114.33 (28.93) <0.001 106.47 (21.01) 0.003 124.00 (43.69)   143.79 (34.80) <0.001 <0.001 
 Postpartum           
  Fasting 84.30 (8.35)  82.62 (6.96)  90.33 (11.14)  0.011 84.33 (8.01)  0.381 
  30 min 132.69 (25.30)  128.64 (26.16)  146.29 (21.82)  0.014 139.67 (20.40)  0.063 
  60 min 126.31 (35.81)  117.82 (31.79)  149.71 (34.40)  0.002 141.30 (38.32)  0.018 
  120 min 102.98 (29.46)  96.54 (23.75)  116.61 (33.21)  0.025 116.05 (37.60)  0.038 
Insulin (μIU/mL)           
 Early pregnancy           
  Fasting 6.84 (6.13) 0.534 6.15 (5.17) 0.445 9.51 (5.34) 0.012 0.007 7.50 (10.62) 0.632 0.729 
  30 min 61.30 (39.43) 0.008 62.73 (42.77) 0.038 65.06 (36.12) 0.003 0.794 45.53 (27.59) 0.938 0.078 
  60 min 63.99 (48.67) <0.001 61.05 (49.81) <0.001 93.43 (56.26) <0.001 0.016 59.15 (35.51) 0.013 0.757 
  120 min 50.86 (43.89) <0.001 45.68 (41.86) <0.001 84.41 (57.83) 0.009 0.006 46.23 (39.40) 0.109 0.734 
 Mid-late pregnancy           
  Fasting 8.81 (5.11) <0.001 8.67 (5.03) <0.001 13.25 (6.92)   8.67 (5.03) <0.001 0.994 
  30 min 69.12 (40.56) <0.001 71.89 (43.18) <0.001 45.20 (11.56)   62.62 (30.91) 0.015 0.354 
  60 min 81.08 (60.34) <0.001 78.92 (56.53) <0.001 49.72 (33.56)   94.83 (75.62) <0.001 0.273 
  120 min 66.49 (55.91) <0.001 65.54 (60.06) <0.001 50.23 (31.82)   72.96 (40.17) <0.001 0.217 
 Postpartum           
  Fasting 5.84 (3.40)  5.97 (3.56)  5.95 (3.58)  0.871 5.04 (2.74)  0.392 
  30 min 46.44 (29.30)  49.59 (30.41)  36.54 (23.04)  0.182 41.98 (28.27)  0.265 
  60 min 46.33 (33.89)  46.57 (35.64)  48.13 (33.49)  0.621 44.29 (30.58)  0.846 
  120 min 29.82 (22.95)  26.56 (21.42)  32.93 (22.86)  0.488 34.13 (21.93)  0.174 
Stumvoll           
 Early pregnancy 1,126 (514) <0.001 1,187 (536) 0.012 1,000 (485) 0.001 0.127 889 (454) 0.233 0.010 
 Mid-late pregnancy 1,216 (482) <0.001 1,275 (507) <0.001 1,042 (261)   1,019 (346) 0.002 0.005 
 Postpartum 903 (365)  972 (355)  698 (366)  0.019 774 (316)  0.028 
Matsuda           
 Early pregnancy 10.22 (8.06) 0.131 11.82 (9.11) 0.778 5.09 (3.75) <0.001 <0.001 8.90 (5.06) 0.154 0.041 
 Mid-late pregnancy 6.67 (5.06) <0.001 7.03 (5.34) <0.001 5.23 (3.17)   5.50 (3.96) <0.001 0.124 
 Postpartum 10.93 (6.85)  11.62 (7.56)  8.71 (4.33)  0.045 10.68 (5.63)  0.546 
All participantsNo GDMEarly GDMClassic GDM
Mean (SD)P (PP)Mean (SD)P (PP)Mean (SD)P (PP)P (NG)Mean (SD)P (PP)P (NG)
Glucose (mg/dL)           
 Early pregnancy           
  Fasting 81.50 (7.56) 0.014 79.02 (5.74) 0.002 93.81 (7.99) 0.220 <0.001 81.96 (4.80) 0.163 0.014 
  30 min 127.92 (26.55) 0.264 120.01 (23.69) 0.127 157.48 (22.58) 0.564 <0.001 138.09 (19.29) 0.402 <0.001 
  60 min 124.85 (36.51) 0.378 113.24 (30.56) 0.513 174.52 (35.14) 0.042 <0.001 139.71 (19.54) 0.712 <0.001 
  120 min 105.77 (29.72) 0.076 96.28 (24.68) 0.711 144.95 (33.68) 0.021 <0.001 112.75 (21.01) 0.795 0.002 
 Mid-late pregnancy           
  Fasting 80.05 (6.75) <0.001 78.55 (5.28) <0.001 91.75 (7.54)   84.29 (8.40) 0.656 0.003 
  30 min 133.35 (23.08) 0.917 129.20 (21.52) 0.738 134.75 (37.66)   149.04 (20.32) 0.223 <0.001 
  60 min 138.64 (33.16) <0.001 129.34 (25.57) 0.007 129.25 (36.47)   177.04 (32.80) <0.001 <0.001 
  120 min 114.33 (28.93) <0.001 106.47 (21.01) 0.003 124.00 (43.69)   143.79 (34.80) <0.001 <0.001 
 Postpartum           
  Fasting 84.30 (8.35)  82.62 (6.96)  90.33 (11.14)  0.011 84.33 (8.01)  0.381 
  30 min 132.69 (25.30)  128.64 (26.16)  146.29 (21.82)  0.014 139.67 (20.40)  0.063 
  60 min 126.31 (35.81)  117.82 (31.79)  149.71 (34.40)  0.002 141.30 (38.32)  0.018 
  120 min 102.98 (29.46)  96.54 (23.75)  116.61 (33.21)  0.025 116.05 (37.60)  0.038 
Insulin (μIU/mL)           
 Early pregnancy           
  Fasting 6.84 (6.13) 0.534 6.15 (5.17) 0.445 9.51 (5.34) 0.012 0.007 7.50 (10.62) 0.632 0.729 
  30 min 61.30 (39.43) 0.008 62.73 (42.77) 0.038 65.06 (36.12) 0.003 0.794 45.53 (27.59) 0.938 0.078 
  60 min 63.99 (48.67) <0.001 61.05 (49.81) <0.001 93.43 (56.26) <0.001 0.016 59.15 (35.51) 0.013 0.757 
  120 min 50.86 (43.89) <0.001 45.68 (41.86) <0.001 84.41 (57.83) 0.009 0.006 46.23 (39.40) 0.109 0.734 
 Mid-late pregnancy           
  Fasting 8.81 (5.11) <0.001 8.67 (5.03) <0.001 13.25 (6.92)   8.67 (5.03) <0.001 0.994 
  30 min 69.12 (40.56) <0.001 71.89 (43.18) <0.001 45.20 (11.56)   62.62 (30.91) 0.015 0.354 
  60 min 81.08 (60.34) <0.001 78.92 (56.53) <0.001 49.72 (33.56)   94.83 (75.62) <0.001 0.273 
  120 min 66.49 (55.91) <0.001 65.54 (60.06) <0.001 50.23 (31.82)   72.96 (40.17) <0.001 0.217 
 Postpartum           
  Fasting 5.84 (3.40)  5.97 (3.56)  5.95 (3.58)  0.871 5.04 (2.74)  0.392 
  30 min 46.44 (29.30)  49.59 (30.41)  36.54 (23.04)  0.182 41.98 (28.27)  0.265 
  60 min 46.33 (33.89)  46.57 (35.64)  48.13 (33.49)  0.621 44.29 (30.58)  0.846 
  120 min 29.82 (22.95)  26.56 (21.42)  32.93 (22.86)  0.488 34.13 (21.93)  0.174 
Stumvoll           
 Early pregnancy 1,126 (514) <0.001 1,187 (536) 0.012 1,000 (485) 0.001 0.127 889 (454) 0.233 0.010 
 Mid-late pregnancy 1,216 (482) <0.001 1,275 (507) <0.001 1,042 (261)   1,019 (346) 0.002 0.005 
 Postpartum 903 (365)  972 (355)  698 (366)  0.019 774 (316)  0.028 
Matsuda           
 Early pregnancy 10.22 (8.06) 0.131 11.82 (9.11) 0.778 5.09 (3.75) <0.001 <0.001 8.90 (5.06) 0.154 0.041 
 Mid-late pregnancy 6.67 (5.06) <0.001 7.03 (5.34) <0.001 5.23 (3.17)   5.50 (3.96) <0.001 0.124 
 Postpartum 10.93 (6.85)  11.62 (7.56)  8.71 (4.33)  0.045 10.68 (5.63)  0.546 

Results are presented as mean (SD). Only observed values (i.e., preimputation) are included. Values obtained during visits in which participants were excluded per protocol are omitted. P (PP) refers to the P value obtained from a paired t test that compares the value of a given measure (e.g., fasting glucose) at a given visit with that at the postpartum visit within the same group of participants. P (NG) refers to the P value obtained from a two-sample t test that compares the value of a given measure at a particular visit for a given subgroup with that for the no GDM subgroup. There is no P value for the rows and columns corresponding to the mid-late pregnancy visit for early GDM, as very few of these individuals attended those visits, per protocol. Insulin values were log transformed prior to performing t tests and obtaining P values. Mean and SD insulin values are presented prior to transformation.

Insulin Physiology Indices in the Full Cohort

Unadjusted insulin secretory response (Stumvoll first-phase estimate) and insulin sensitivity (Matsuda index) are shown in Table 2. Insulin secretory response was higher in both early pregnancy and mid-late pregnancy than it was postpartum. Insulin sensitivity was similar in early pregnancy and lower in mid-late pregnancy compared with postpartum.

Longitudinal model results, after adjusting for participant characteristics and multiple imputation, are shown in Fig. 1 and Supplementary Table 1. In full cohort analyses before adjustment for insulin sensitivity, the insulin secretory response (log-transformed Stumvoll first-phase estimate) was increased in both early pregnancy (β = 0.21, P < 0.001) and in mid-late pregnancy (β = 0.27, P < 0.001) compared with postpartum. Insulin sensitivity (log-transformed Matsuda index) was reduced slightly in early pregnancy (β = −0.20, P < 0.001) and substantially in mid-late pregnancy (β = −0.47, P < 0.001) compared with postpartum. Changes in insulin sensitivity were independent of BMI (Supplementary Table 7). After adjustment for insulin sensitivity, insulin secretory response was significantly greater in both early pregnancy (β = 0.16, P < 0.001) and mid-late pregnancy (β = 0.16, P = 0.001) than it was postpartum; the average magnitude of the insulin secretory response increase over the postpartum level was similar in early and mid-late pregnancy.

Figure 1

Estimated mean values for insulin secretory response, insulin sensitivity, and PIP index at each study visit. Each bar represents the mean values at each study visit (E, early pregnancy; ML, mid-late pregnancy; PP, postpartum) obtained from covariate-adjusted linear mixed-effects models for insulin secretory response, modeled as log Stumvoll (A); insulin sensitivity, modeled as log Matsuda (B); insulin secretory response additionally adjusted for insulin sensitivity, modeled as log Stumvoll with additional adjustment for log Matsuda (C); and PIP index (D). The estimated mean value at each visit corresponds to a participant who is the mean age of the participant population, has the mean BMI of all participants at the early pregnancy visit, has no family history of diabetes, is not Hispanic, is married, and has completed college. In C, the reference individual also has the mean observed value of log Matsuda at the visit being modeled. Results are presented for all participants and then stratified by subgroup: no GDM, early GDM, and classic GDM. Separate models for each outcome were fit within each subgroup. As such, estimated means in different subgroups at the same visit are not directly comparable. An asterisk indicates that the P value corresponding to the statistical test comparing the value of a measure at a given visit with the postpartum visit is less than 0.05.

Figure 1

Estimated mean values for insulin secretory response, insulin sensitivity, and PIP index at each study visit. Each bar represents the mean values at each study visit (E, early pregnancy; ML, mid-late pregnancy; PP, postpartum) obtained from covariate-adjusted linear mixed-effects models for insulin secretory response, modeled as log Stumvoll (A); insulin sensitivity, modeled as log Matsuda (B); insulin secretory response additionally adjusted for insulin sensitivity, modeled as log Stumvoll with additional adjustment for log Matsuda (C); and PIP index (D). The estimated mean value at each visit corresponds to a participant who is the mean age of the participant population, has the mean BMI of all participants at the early pregnancy visit, has no family history of diabetes, is not Hispanic, is married, and has completed college. In C, the reference individual also has the mean observed value of log Matsuda at the visit being modeled. Results are presented for all participants and then stratified by subgroup: no GDM, early GDM, and classic GDM. Separate models for each outcome were fit within each subgroup. As such, estimated means in different subgroups at the same visit are not directly comparable. An asterisk indicates that the P value corresponding to the statistical test comparing the value of a measure at a given visit with the postpartum visit is less than 0.05.

Close modal

Following measurement error correction, the estimated slope from the linear regression of log-transformed Stumvoll on log-transformed Matsuda was −0.40 (95% CI −0.49 to −0.30). Because this interval excluded −1, the relationship did not meet criteria for a rectangular hyperbola, indicating that their product (the DI) was not a valid index of their relationship. From this regression, we define the PIP index as Stumvoll × [MatsudaΔ], setting Δ = 0.4 in our data set. Observed participant-level Stumvoll and Matsuda values at each visit are plotted in Fig. 2, over which a range of PIP index curves are superimposed. In full cohort analyses, the PIP index was higher in the first trimester than it was postpartum (β = 215, P = 0.04) but was not different from postpartum in mid-late pregnancy (β = 55, P = 0.62) (Fig. 1D). Animated versions of Fig. 2, in which individual-level changes in Stumvoll, Matsuda, and PIP index can be visualized, can be found in the Supplementary Videos.

Figure 2

Individual Stumvoll and Matsuda indices at each visit, with PIP index curves. A–C: Observed participant-level values of Matsuda and Stumvoll at each visit (early pregnancy [A], mid-late pregnancy [B], and postpartum [C]). Individual data points are categorized by color and shape based on which GDM subgroup they ultimately belong to. PIP index curves (dashed lines) are superimposed over these points, where points along the same curve have the same PIP index value. PIP index curves closer to the origin correspond to lower values (light blue), and PIP index curves farther from the origin correspond to higher values (dark blue). Animated versions of these figures demonstrating individual-level changes can be found in Supplementary Video 1.

Figure 2

Individual Stumvoll and Matsuda indices at each visit, with PIP index curves. A–C: Observed participant-level values of Matsuda and Stumvoll at each visit (early pregnancy [A], mid-late pregnancy [B], and postpartum [C]). Individual data points are categorized by color and shape based on which GDM subgroup they ultimately belong to. PIP index curves (dashed lines) are superimposed over these points, where points along the same curve have the same PIP index value. PIP index curves closer to the origin correspond to lower values (light blue), and PIP index curves farther from the origin correspond to higher values (dark blue). Animated versions of these figures demonstrating individual-level changes can be found in Supplementary Video 1.

Close modal

Forty-six percent of participants had a PIP index that was >10% higher in early pregnancy than postpartum, while 30% of participants had a PIP index in early pregnancy that was >10% lower than postpartum and 24% had a PIP index that was similar in early pregnancy and postpartum (Fig. 2 and Supplementary Table 2). In mid-late pregnancy, 45% of participants had PIP index more than 10% greater than that in postpartum, while 38% of participants had a PIP index more than 10% lower than that in postpartum.

Changes in insulin physiology across gestation were similar in those without GDM (N = 99) to those seen in the full cohort (Fig. 1 and Supplementary Table 1).

Insulin Physiology Indices in GDM

Among participants with early GDM, before insulin sensitivity adjustment, the insulin secretory response was greater in early pregnancy than it was postpartum (β = 0.43, P = 0.002) (Fig. 1A). However, insulin sensitivity in participants with early GDM was markedly lower in early pregnancy (β = −0.59, P < 0.001) (Fig. 1B and Supplementary Table 1); the decrement in insulin sensitivity in early pregnancy compared with that postpartum was significantly greater in early GDM than in the no GDM group (P = 0.002). The difference in insulin sensitivity between early pregnancy and postpartum in the group with early GDM remained significant after exclusion of participants who enrolled at 14–15 weeks’ gestation (mean [SD] early pregnancy Matsuda 4.81 [3.66], mean [SD] postpartum Matsuda 8.50 [4.54]; P = 0.005) After adjustment for insulin sensitivity, insulin secretory response in participants with early GDM was no longer significantly greater in early pregnancy than it was postpartum (β = 0.25, P = 0.14) (Fig. 1C); consistent with this, the PIP index was not significantly higher in early pregnancy than it was postpartum in early GDM (β = 131, P = 0.55) (Fig. 1D).

Compared with those without recent GDM, postpartum participants with early GDM had a lower insulin secretory response after adjustment for insulin sensitivity (β = −0.42, P = 0.006) (Supplementary Table 1) as well as a lower PIP index (β = −716, P = 0.009). Insulin sensitivity was not significantly lower postpartum in those with recent early GDM than in those without recent GDM (β = −0.23, P = 0.12) (Supplementary Table 1).

Among participants with classic GDM, before adjustment for insulin sensitivity, insulin secretory response was not different between early pregnancy and postpartum (β = 0.14, P = 0.18) (Fig. 1A and Supplementary Table 1). In contrast, there was greater insulin secretory response in mid-late pregnancy than postpartum before insulin sensitivity adjustment (β = 0.24, P = 0.03). Changes in insulin sensitivity in those with classic GDM followed a pattern similar to that in the full cohort (Fig. 1B). After adjustment for insulin sensitivity, the pregnancy-associated increase in insulin secretory response over postpartum that we saw in those without GDM was completely absent in participants with classic GDM (early pregnancy: β = 0.04, P = 0.70; mid-late pregnancy: β = −0.02, P = 0.85) (Fig. 1C). In classic GDM, the PIP index was also no higher (Fig. 1D) in early pregnancy (β = −11, P = 0.97) or mid-late pregnancy (β = −158, P = 0.55) compared with that postpartum. The proportion of participants with PIP index values that were stable or more than 10% higher in early pregnancy compared with values postpartum was lower in classic GDM than with the no GDM group (47% vs. 75%, P = 0.04) (Fig. 2 and Supplementary Table 2).

Postpartum, there was no difference in insulin sensitivity in those with recent classic GDM and those without GDM before adjustment for BMI (Supplementary Table 8) (P = 0.77), but after adjustment for the lower BMI in the classic GDM group, there was a trend toward lower insulin sensitivity (β = −0.26, P = 0.05) (Supplementary Table 1). Postpartum, participants with classic GDM had an insulin secretory response (adjusted for insulin sensitivity) that was not significantly lower than the insulin secretory response in those without GDM (β = −0.19, P = 0.16); the PIP index was also not significantly different postpartum in individuals with recent classic GDM compared with those without GDM (β = −464, P = 0.13).

Prediction of GDM Using the PIP Index

A lower early-pregnancy PIP index predicted GDM status in a univariate model (P < 0.001, AUC 0.83, 95% CI 0.76–0.89) (Supplementary Table 3). In a model that also included participant characteristics, the early pregnancy PIP index remained a significant predictor; addition of PIP increased the predictive ability of a model that contained participant characteristics alone (AUC without PIP 0.70 [95% CI 0.61–0.79], AUC with PIP 0.87 [0.80–0.93]). In univariate GDM prediction models, the AUC was nominally higher for the early pregnancy PIP index (0.83 [0.76–0.89]) than for the postpartum PIP index (0.72 [0.60–0.81]). In models that included both the early pregnancy and postpartum PIP indices, the postpartum PIP index was not a significant GDM predictor (Supplementary Table 3). These findings did not change when we excluded participants with early GDM (Supplementary Table 3). The AUCs for models incorporating the early pregnancy PIP as a GDM predictor were nominally higher than those of models that instead included early-pregnancy fasting glucose (Supplementary Table 3).

Sensitivity Analyses

Results in individuals who attended all study visits did not substantively differ from those reported above (Supplementary Tables 4–6). Adjustment for exclusive breastfeeding also did not alter findings (Supplementary Table 8).

Here, in a cohort of pregnant women with diabetes risk factors, we provide novel observations about longitudinal changes in insulin physiology in pregnancy and GDM. We first demonstrate that there is an insulin resistance–independent enhancement of insulin secretory response to oral glucose in early gestation, consistent with both our prespecified hypothesis and earlier work in a smaller cohort that used intravenous glucose loading (6). We then show that this pregnancy-mediated increase in β-cell function is deficient in GDM by demonstrating that individuals with GDM do not have a significantly higher Matsuda-adjusted Stumvoll estimate in early pregnancy compared with postpartum, unlike their counterparts without the disorder. To capture β-cell compensation for insulin resistance in pregnant or postpartum individuals in a single measure, we derived the PIP index, which is analogous to the DI (9). The PIP index is augmented in the first trimester, and a lower first-trimester PIP index predicts the development of GDM, independent of clinical risk factors and of the postpartum PIP index. Thus, while there is evidence of chronic (postpartum) β-cell dysfunction in those with GDM (3,2729), knowledge of whether the β-cell responds to the early pregnancy hormonal milieu (independent of insulin resistance) provides valuable prognostic information. Our findings illuminate insulin physiology adaptations in euglycemic pregnancy and highlight the absence of apparent augmentation of β-cell function in early pregnancy as a physiologic contributor to GDM.

By following insulin physiology trajectories longitudinally and establishing a valid measure (the PIP index) that describes the relationship between insulin secretory response and insulin sensitivity in pregnancy, we document that this relationship evolves in many pregnant individuals across gestation. A minority have a stable PIP index across pregnancy and postpartum; the plurality has a higher PIP index in early pregnancy than postpartum, indicative of enhanced β-cell function. In a prior smaller longitudinal study of 34 participants that used intravenous glucose tolerance tests and euglycemic clamps to assess insulin secretory response and insulin sensitivity across gestation, we demonstrated, as independently replicated here, that there is an augmentation of insulin secretory response in early pregnancy that occurs prior to and in excess of the response to gestational insulin resistance (6). Unlike the current study, this prior study did not find deficiencies in the augmentation of insulin secretory response in pregnant women with GDM (n = 15) compared with those without (n = 19). However, in that study, by design, pregnant individuals with GDM had multiple diabetes risk factors, whereas those without GDM were free of risk factors, leaving open the possibility of unmeasured confounders that differed between the two groups. In the current study, the diabetes risk factors were present both in participants who did and did not develop GDM. This, along with the differing insulin physiology assessment methods and comparison with a prepregnancy (rather than postpartum) physiology in the prior study, might account for the differences between the two reports. Most other longitudinal studies of insulin physiology in individuals with GDM have focused on late, rather than early, pregnancy (3,8,2729). Such studies may have missed the early-pregnancy insulin resistance–independent augmentation of the insulin secretory response and therefore could not have identified its absence as a contributor to GDM.

Another major finding of our study is that there are distinct longitudinal trajectories of insulin resistance in early-onset versus classic-onset GDM. As described above, early- and classic-onset GDM seem to share deficient augmentation of β-cell function in response to pregnancy, although this deficit appeared more pronounced in classic-onset GDM. Both early- and classic-onset GDM also likely share reduced β-cell function in the postpartum state, although this was more apparent in early-onset GDM. These GDM phenotypes are distinguished by the timing of onset of gestational insulin resistance, with first-trimester insulin resistance apparent in early GDM. To our knowledge, this has not been previously described, but our longitudinal data are consistent with another study that identified substantial insulin resistance in GDM with onset prior to 20 weeks’ gestation (30,31). Without measurements outside of pregnancy, this prior study could not determine if the insulin resistance in this group was chronic or if it represented a difference in gestational physiology, the latter of which is suggested by our findings. Our study identifies early-onset gestational insulin resistance as a feature of first-trimester hyperglycemia but not of GDM with later onset. Whether this phenomenon is due to early placental production of insulin resistance-inducing factors or early maternal susceptibility to such factors is unknown.

Studies in rodent models provide evidence in support of the existence of pregnancy-related adaptations of β-cells that are distinct from those that respond to high-fat-diet–induced insulin resistance (32,33). These observations are consistent with our finding that individuals with both early- and classic-onset GDM were able to increase insulin secretion in response to insulin resistance despite a deficit in the insulin resistance–independent component of the β-cell adaptation to pregnancy. Additional research is needed to determine whether the apparent deficit in the pregnancy-induced augmentation of insulin secretory response is a feature of most types of chronic β-cell dysfunction or whether there are individuals who have a specific inability to augment insulin secretory response in pregnancy.

Strengths of our study include its longitudinal design, with the inclusion of an early pregnancy and postpartum time point, the similar risk factor profile in the participants with and without GDM (facilitating their comparison), and the stratification of the participants with GDM into early- and classic-onset groups (enabling more refined phenotyping). Limitations of our investigation include the exclusion of participants without diabetes risk factors, which was an intentional aspect of the study design. Thus, we cannot be certain that our findings, including the derivation of the PIP index, apply to pregnant individuals without diabetes risk factors. We derived the PIP index in our study cohort of 166 individuals by extending the methodology underpinning the DI. This was warranted because the 95% CI for the slope obtained after regressing the log of our insulin secretory response measure on the log of our insulin sensitivity measure did not include –1, which violates the key assumption underlying the DI. Our article introduces the methodology for defining similar PIP indices in other populations of pregnant individuals. We encourage other researchers to calculate analogous measures in other data sets. Additional evaluation of the reliability and repeatability of the PIP index as a predictor of GDM in different pregnancy populations is needed in future studies. With the need to measure both glucose and insulin on timed blood draws in the first trimester, the PIP index is not practical for use in predicting GDM clinically but is helpful for understanding inadequate augmentation of β-cell function in the first trimester as a risk factor for GDM. It would have been ideal to obtain prepregnancy data from our participants, but the feasibility challenges of recruiting individuals prior to pregnancy prevented this. Lactation may contribute to differences between prepregnancy and postpartum physiology, but our findings did not change when we adjusted for exclusive breastfeeding at the postpartum study visit. We had some disruption of data collection due to the COVID-19 pandemic and because of participant dropout. To robustly account for this, we used multiple imputation; a complete case analysis confirmed our key findings. Because we had a prespecified hypothesis with related secondary analyses, we did not perform adjustments for multiple hypothesis testing and our findings should be confirmed in independent cohorts (34). Finally, we did not use methods such as intravenous glucose tolerance tests or euglycemic clamps to assess insulin physiology, and we did not measure C-peptide levels, which could have provided insight into insulin secretion and clearance. However, our methodology allowed us to conduct physiologic profiling in a larger cohort and provides generalizable information about the response to an oral glucose load that may be similar to physiologic challenges encountered in daily life.

In conclusion, we observed an insulin resistance–independent increase in insulin secretory response associated with pregnancy that manifested early in gestation and was captured with the newly described PIP index. Individuals with GDM lack a pregnancy-mediated increase of the PIP index in the first trimester; a higher PIP index in early pregnancy protects against the development of GDM. Early- and classic-onset GDM are distinguished by the early onset of gestational insulin resistance in the former, suggesting a difference in the pathophysiology that contributes to these two phenotypes. Further investigations are needed to uncover hormonal mediators underlying the longitudinal changes in insulin physiology in euglycemic pregnancy and in both early- and classic-onset GDM.

See accompanying articles, pp. 2120 and 2147.

This article contains supplementary material online at https://doi.org/10.2337/figshare.22184218.

Acknowledgments. The authors thank the participants in the SPRING cohort and the research team involved in collecting the data, including Juliana Arenas, Michael J. Callahan, Nefeli Neamonitaki, Robin Azevedo, and Arantxa Medina Baez, all employees of the Diabetes Research Center, Massachusetts General Hospital.

Funding. This study was supported by the National Institute of Diabetes and Digestive and Kidney Diseases (K23DK113218), the Robert Wood Johnson Foundation’s Harold Amos Medical Faculty Development Program, and the Massachusetts General Hospital Claflin Distinguished Scholar Award. Data collection was also supported by Eunice Kennedy Shriver National Institute of Child Health and Human Development grant R01HD094150, as well as UL1TR001102 and UL1TR000170 to the Harvard Clinical and Translational Science Center from the National Center for Advancing Translational Science.

Duality of Interest. J.C.F. has received consulting honoraria from Novo Nordisk and AstraZeneca and speaking honoraria from Merck, Novo Nordisk, and AstraZeneca for research lectures, over which he had full control of content. C.E.P. has received fees and royalties from Mediflix and UpToDate (Wolters Kluwer), respectively, for presentations and articles related to diabetes, over which she had full control of content. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. T.T. planned and led the analysis, interpreted the findings, and drafted the manuscript. Z.S. performed aspects of the analysis and edited the manuscript. K.J. managed the database, performed aspects of the analysis, and edited the manuscript. J.C.F. participated in conceptualization of the study, interpreted the findings, and edited the manuscript. C.E.P. conceptualized the study, obtained funding, collected the data, planned the analysis, interpreted the findings, participated in drafting the manuscript, and supervised the work. C.E.P. 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.

1.
Catalano
PM
.
Trying to understand gestational diabetes
.
Diabet Med
2014
;
31
:
273
281
2.
Ryan
EA
,
O’Sullivan
MJ
,
Skyler
JS
.
Insulin action during pregnancy. Studies with the euglycemic clamp technique
.
Diabetes
1985
;
34
:
380
389
3.
Buchanan
TA
,
Xiang
A
,
Kjos
SL
,
Watanabe
R
.
What is gestational diabetes?
Diabetes Care
2007
;
30
(
Suppl. 2
):
S105
S111
4.
Retnakaran
R
,
Shah
BR
.
Impact of twin gestation and fetal sex on maternal risk of diabetes during and after pregnancy
.
Diabetes Care
2016
;
39
:
e110
e111
5.
Schwartz
N
,
Nachum
Z
,
Green
MS
.
The prevalence of gestational diabetes mellitus recurrence—effect of ethnicity and parity: a metaanalysis
.
Am J Obstet Gynecol
2015
;
213
:
310
317
6.
Powe
CE
,
Huston Presley
LP
,
Locascio
JJ
,
Catalano
PM
.
Augmented insulin secretory response in early pregnancy
.
Diabetologia
2019
;
62
:
1445
1452
7.
Bergman
RN
,
Phillips
LS
,
Cobelli
C
.
Physiologic evaluation of factors controlling glucose tolerance in man: measurement of insulin sensitivity and beta-cell glucose sensitivity from the response to intravenous glucose
.
J Clin Invest
1981
;
68
:
1456
1467
8.
Buchanan
TA
.
Pancreatic B-cell defects in gestational diabetes: implications for the pathogenesis and prevention of type 2 diabetes
.
J Clin Endocrinol Metab
2001
;
86
:
989
993
9.
Kahn
SE
,
Prigeon
RL
,
McCulloch
DK
, et al
.
Quantification of the relationship between insulin sensitivity and beta-cell function in human subjects. Evidence for a hyperbolic function
.
Diabetes
1993
;
42
:
1663
1672
10.
American Diabetes Association
.
2. Classification and diagnosis of diabetes: Standards of Medical Care in Diabetes—2022
.
Diabetes Care
2022
;
45
(
Suppl. 1
):
S17
S38
11.
Metzger
BE
,
Gabbe
SG
,
Persson
B
, et al.;
International Association of Diabetes and Pregnancy Study Groups Consensus Panel
.
International association of diabetes and pregnancy study groups recommendations on the diagnosis and classification of hyperglycemia in pregnancy
.
Diabetes Care
2010
;
33
:
676
682
12.
Stumvoll
M
,
Mitrakou
A
,
Pimenta
W
, et al
.
Use of the oral glucose tolerance test to assess insulin release and insulin sensitivity
.
Diabetes Care
2000
;
23
:
295
301
13.
Powe
CE
,
Locascio
JJ
,
Gordesky
LH
,
Florez
JC
,
Catalano
PM
.
Oral glucose tolerance test-based measures of insulin secretory response in pregnancy
.
J Clin Endocrinol Metab
2022
;
107
:
e1871
e1878
14.
Matsuda
M
,
DeFronzo
RA
.
Insulin sensitivity indices obtained from oral glucose tolerance testing: comparison with the euglycemic insulin clamp
.
Diabetes Care
1999
;
22
:
1462
1470
15.
Kirwan
JP
,
Huston-Presley
L
,
Kalhan
SC
,
Catalano
PM
.
Clinically useful estimates of insulin sensitivity during pregnancy: validation studies in women with normal glucose tolerance and gestational diabetes mellitus
.
Diabetes Care
2001
;
24
:
1602
1607
16.
Utzschneider
KM
,
Prigeon
RL
,
Faulenbach
MV
, et al
.
Oral disposition index predicts the development of future diabetes above and beyond fasting and 2-h glucose levels
.
Diabetes Care
2009
;
32
:
335
341
17.
Riggs
DS
,
Guarnieri
JA
,
Addelman
S
.
Fitting straight lines when both variables are subject to error
.
Life Sci
1978
;
22
:
1305
1360
18.
Utzschneider
KM
,
Prigeon
RL
,
Tong
J
, et al
.
Within-subject variability of measures of beta cell function derived from a 2 h OGTT: implications for research studies
.
Diabetologia
2007
;
50
:
2516
2525
19.
R Core Team
.
R: A Language and Environment for Statistical Computing
.
Vienna, Austria
,
R Foundation for Statistical Computing
,
2021
20.
Pinheiro
J
,
Bates
D
,
DebRoy
S
, et al
.
Package nlme. Linear and nonlinear mixed effects models
.
Vienna, Austria
,
The Comprehensive R Archive Network
,
2022
21.
Van Buuren
S
,
Groothuis-Oudshoorn
K
.
MICE: multivariate imputation by chained equations in R
.
J Stat Softw
2011
;
45
:
1
67
22.
Zeileis
A
,
Hothorn
T
.
Diagnostic checking in regression relationships
.
R News
2002
;
2/3
:
7
10
23.
Manuilova
E
,
Model
MF
.
Package mcr
.
Vienna, Austria
,
The Comprehensive R Archive Network
,
2021
.
24.
Pedersen
TL
,
Robinson
D
.
gganimate: a grammar of animated graphics 2022
.
Accessed 20 December 2022. Available from https://gganimate.com
25.
Bochkur Dratver
MA
,
Arenas
J
,
Thaweethai
T
, et al
.
Longitudinal changes in glucose during pregnancy in women with gestational diabetes risk factors
.
Diabetologia
2022
;
65
:
541
551
26.
Edelson
PK
,
James
KE
,
Leong
A
, et al
.
Longitudinal changes in the relationship between hemoglobin a1c and glucose tolerance across pregnancy and postpartum
.
J Clin Endocrinol Metab
2020
;
105
:
e1999
e2007
27.
Homko
C
,
Sivan
E
,
Chen
X
,
Reece
EA
,
Boden
G
.
Insulin secretion during and after pregnancy in patients with gestational diabetes mellitus
.
J Clin Endocrinol Metab
2001
;
86
:
568
573
28.
Kautzky-Willer
A
,
Prager
R
,
Waldhausl
W
, et al
.
Pronounced insulin resistance and inadequate beta-cell secretion characterize lean gestational diabetes during and after pregnancy
.
Diabetes Care
1997
;
20
:
1717
1723
29.
Ryan
EA
,
Imes
S
,
Liu
D
, et al
.
Defects in insulin secretion and action in women with a history of gestational diabetes
.
Diabetes
1995
;
44
:
506
512
30.
Egan
AM
,
Vellinga
A
,
Harreiter
J
, et al.;
DALI Core Investigator group
.
Epidemiology of gestational diabetes mellitus according to IADPSG/WHO 2013 criteria among obese pregnant women in Europe
.
Diabetologia
2017
;
60
:
1913
1921
31.
Harreiter
J
,
Simmons
D
,
Desoye
G
, et al.;
DALI Core Investigator group
.
IADPSG and WHO 2013 gestational diabetes mellitus criteria identify obese women with marked insulin resistance in early pregnancy
.
Diabetes Care
2016
;
39
:
e90
e92
32.
Pepin
ME
,
Bickerton
HH
,
Bethea
M
,
Hunter
CS
,
Wende
AR
,
Banerjee
RR
.
Prolactin receptor signaling regulates a pregnancy-specific transcriptional program in mouse islets
.
Endocrinology
2019
;
160
:
1150
1163
33.
Banerjee
RR
,
Cyphert
HA
,
Walker
EM
, et al
.
Gestational diabetes mellitus from inactivation of prolactin receptor and MafB in islet β-cells
.
Diabetes
2016
;
65
:
2331
2341
34.
Bender
R
,
Lange
S
.
Adjusting for multiple testing—when and how?
J Clin Epidemiol
2001
;
54
:
343
349
Readers may use this article as long as the work is properly cited, the use is educational and not for profit, and the work is not altered. More information is available at https://www.diabetesjournals.org/journals/pages/license.