Study Design and Participants
This prospective, observational cohort study was registered at Clinical Trials.gov (NCT04411277). A local institutional review board (IRB) at the Universidade Positivo approved the protocol and informed consent. Participants were identified by medical records from the Hospital da Cruz Vermelha Brasileira-Filial do Paraná and at the neuro geriatrician and endocrinology private clinic. Individuals were invited to participate by phone, and interested participants were referred to the Cline Research Center, where the study procedures were carried out. In this way, the sample included patients from the public and private health systems.
Data collection process
At the first clinic visit to the research center, after signing the informed consent form, subjects were evaluated for their educational level, and presence of comorbidities, including cardiovascular disease, hypertension, and dyslipidemia. Microvascular complications were assessed by a previous eye examination and the presence of retinopathy diagnosis, foot examination for neuropathy, and calculated estimated glomerular filtration rate (eGFR) using the CKD-EPI formula.
Screening for the presence of sarcopenia was performed with the application of the SARC-F questionnaire; measuring muscle strength with grip strength by Saehan dynamometer; the functional test of getting up from the chair without supporting the arms and walking for 3 meters, turning 180º and returning to sitting position. In addition, body composition analysis using the InBody 270 multifrequency bioimpedance was used to calculate the appendicular skeletal muscle mass.
Inclusion criteria were age > 65 years, a history of type 2 diabetes, BMI between 18·5 and 35kg/m², HbA1c between 7% and 9%, stable diabetes medication regimen, and body weight over the prior three months and treatment with insulin with or without oral agent treatment.
We excluded subjects diagnosed with type 1 diabetes taking glucagon-like peptide 1 receptor agonists (GLP1-RA), eGFR <30 ml.min.1.73 m², anemia (hemoglobin <11 grams), chronic liver disease (ALT > 3x upper limits of normal), use of glucocorticoids within three months prior to the study period, history of neoplasia treatment, inability by the patient or caretaker to commit to the CGM instructions.
Eligible patients had the CGM (FreeStyle Libre Flash CGM system, Abbott Diabetes Care) sensor inserted for the first time in the posterior upper arm and were recommended to scan the CGM as many times as possible during the day, avoiding scan intervals longer than 8 hours. Participants were instructed to return to the research center every 2 weeks, where the CGM data was downloaded, and glucose reports were generated for review by the primary provider. The research team and healthcare professionals were instructed not to change the treatment regimen except for safety reasons. In the last case, participants were excluded from the data analysis.
Outcomes Measures
The primary outcome was time in range (TIR) between 70-180 mg/dl, and secondary outcomes were time below range (TBR) (<70mg/dL), Time above range (TAR) (> 180mg/dL, and > 250mg/dL), Glucose variability (GV) (calculated by % coefficient of variation, %CV= (standard deviation (SD)/mean glucose x 100%)), and adherence to CGM use (the percentage of captured sensor data). The TIR, TBR, and TAR were expressed as % of the time by day.
Statistical Analysis
We used measures of central tendency (mean and median), variability (standard deviation (SD), interquartile ranges (IQR)), absolute (n), and relative (%) frequency to describe the participant characteristics and glucose profile. A set of six linear regressions with random intercept (participants) and slopes (visits) verified the changes in TBR, TIR, TAR (>180 - 250 mg/dl and >250 mg/dl), GV, and GMI across the three visits (baseline, visit 1 and 2) controlled by gender, age, educational level, and health system (private or public). We further explored the covariance between intercept and slope to identify the influence of the baseline values on the changes over time. The results were expressed in regression coefficients (β) and 95% confidence intervals (95%CI). We assumed a significance level of 95% (p<0·05). The sample size showed a power of 0·98 to identify regression coefficients equal to or higher than ±0·99 with an alpha of 95%. The software Stata MP 14·1 ran all the statistical analyses.