Background: To evaluate the protective effect of oral antidiabetic drugs in a large cohort of elderly patients with type 2 diabetes differing for age, clinical status, and life expectancy, including patients with multiple comorbidities and short survival.

Methods: A nested case-control study was carried out by including the cohort of 188,983 patients from Lombardy (Italy), aged ≥ 65 years, who received ≥ 3 consecutive prescriptions of antidiabetic agents (mostly metformin and other older conventional agents) during 2012. Cases were the 49,201 patients who died for any cause during follow-up (up to 2018). A control was randomly selected for each case. Adherence to drug therapy was measured by considering the proportion of days of the follow-up covered by the drug prescriptions. Conditional logistic regression was used to model the risk of outcome associated with adherence to antidiabetic drugs. The analysis was stratified according to four categories of the clinical status (good, intermediate, poor, and very poor) differing for life expectancy.

Results: There was a steep increase in comorbidities and a marked reduction of the 6-year survival from the very good to the very poor (or frail) clinical category. Progressive increase in adherence to treatment was associated with a progressive decrease in the risk of all-cause mortality in all clinical categories and at all ages (65-74, 75-84 and ≥ 85 years) except for the frail patient subgroup aged ≥ 85 years. The mortality reduction from lowest to highest adherence level showed a tendency to be lower in frail patients compared to the other categories. Similar although less consistent results were obtained for cardiovascular mortality.

Conclusions: In elderly diabetic patients, increased adherence to antidiabetic drugs is associated with a reduction in the risk of mortality regardless of the patients' clinical status and age, with the exception of very old patients (age ≥ 85 years) in the very poor or frail clinical category. However, in the frail patient category the benefit of treatment appears to be less than in patients in good clinical conditions.