Statistical Analysis
Analysis was based on the “intention to treat”. We report baseline characteristics and outcomes as mean (SD), median (IQR), or number and percentage. We report the effects of the intervention as relative risks (RRs), with their 95% confidence intervals (CI). The statistical significance of the differences was tested using the Fisher exact test. Differences in means of continuous variables were tested with the Student T-test. We used SPSS (versions 18 and 20, IBM, Chicago, USA) and Stata (version 15, StataCorp, College Station, USA) for data management and statistical analysis.
Based on data from our clinic, we expected an incidence of 40% of insulin prescription in women with gestational diabetes treated initially with diet only (control group). For a clinically relevant reduction in insulin prescription with exercise to 20% (number-needed to treat of 5), the sample size (α=0.05 and power of 80%) was 91 patients in each group. We had to stop the trial before reaching the complete sample size because of low recruitment, low compliance to the exercise sessions and lack of funding for the physical therapist salary.