Exenatide, Dapagliflozin, or Phentermine/Topiramate Differentially Affect Metabolic Profiles in Polycystic Ovary Syndrome

Karen E. Elkind-Hirsch; N. Chappell; Ericka Seidemann; John Storment; Drake Bellanger


J Clin Endocrinol Metab. 2021;106(10):3019-3033. 

In This Article

Collateral Research

To evaluate abdominal adiposity, a WC greater than 35 inches (89 cm), WHR greater than 0.8, and WHtR greater than 0.5 were considered to be elevated, indicating increased cardiometabolic risk.[16,27] Fasting blood glucose (FBG) and mean blood glucose (MBG) concentrations were calculated in milligrams per deciliter (mg/dL) from glucose levels obtained during the OGTT. MBG concentrations were calculated by summing glucose values obtained at 0, 30, 60, and 120 minutes during the OGTT and dividing by 4. Hyperinsulinemia was considered when fasting levels were greater than 10 mU/L and greater than 40 mU/L, 2 hours post load. The FAI was calculated from the TT concentration (nanomole per liter [nmol/L])/concentration of SHBG (nanomolar per liter [nM/L]) × 100.[28] Dyslipidemia was defined as the presence of at least one lipid parameter abnormality on the described standard lipid panel. A TRG/HDL-C ratio greater than 3.0 was used as an indirect measure of insulin resistance.[29]

Safety and tolerability was assessed by collating data on treatment-emergent adverse events (AE), laboratory tests, physical examinations, and vital signs. Prevention of pregnancy was monitored monthly both by laboratory and home pregnancy testing. All participants were educated about not becoming pregnant and performed monthly urine home pregnancy tests.

Sample Size Justification/Statistical Analyses

The measurement of changes in insulin action as a primary outcome in randomized, controlled trials of prediabetic women with PCOS after treatment using DAPA, EQW, or combination treatments has not been previously reported. Sample size was calculated using a mean difference formula based on the assumption that the smallest mean difference between treatment groups is 20%, with an average SE of 2.2%. The sample size had a statistical power of 0.80 and 2-sided P value of less than .05 and was calculated to be 22 for each group. The final sample size required was 130; there were 26 participants in each group, allowing for a 20% dropout rate.

The general features of the participants (quantitative variables) are presented as number of cases and mean and SEM unless otherwise mentioned. Categorical variables are presented as frequencies or percentages. Continuous variables were tested for normality of distribution using the Kolmogorov-Smirnov test. When necessary, nonnormally distributed data were subjected to logarithmic transformation to obtain a normal distribution where necessary for subsequent analyses. We determined P less than .05 (2-sided) to define statistical significance. The primary outcome was comparison of the oral disposition index (IS-SI index) determined from a 75-g OGTT before and after treatment. For continuous variables, data were analyzed using a repeated-measures general linear model (Subjects/drug treatments × repeated-measures analyses of variance) including the arm of drug treatment as the between-subjects effect, and the visit (baseline and 24 weeks) as the within-subjects effect. To evaluate the differences in the response to each of the 5 treatments over visits, the interaction effect was calculated. Only when a statistically significant interaction effect was found (P < .05) was the contrast test applied to locate the differences between the 5 medication groups. Baseline comparisons between groups (intent-to-treat and completers) and change from baseline with treatment (absolute and percentage) to compare means between groups for continuous variables were made using one-way analyses of variance; if the differences between groups were significant (P < .05 for 2-sided), post hoc comparisons were performed with the Bonferroni test to analyze the variation among the 5 groups. Simple linear associations between quantitative anthropometric and DXA variables were evaluated using Pearson correlation coefficients with high collinearity defined as r ≥ 0.4. All analyses were conducted using IBM SPSS Statistics for Windows, version 25.0. The completer population was defined as all randomly assigned individuals who completed treatment through weeks 20 to 24.