Summary: In a substudy of the large GRADE trial, adults with type 2 diabetes adding a second drug to metformin maintained high medication adherence over 3 years with only small differences between drugs, while lower adherence robustly predicted worse glycaemic control across all groups.
PICO Summary
| Element | Detail |
|---|---|
| Population | 1739 adults with type 2 diabetes (<10 years’ duration, HbA1c 6.8–8.5%, on metformin); 3-year adherence substudy of the GRADE comparative-effectiveness trial, USA. |
| Intervention | Addition of one of four second-line agents: insulin glargine, glimepiride, liraglutide, or sitagliptin, with adherence measured semiannually on a validated 0–100 scale. |
| Comparison | Differences in adherence, and in the adherence-glycaemia association, between the four arms. |
| Outcome | Overall adherence was high (mean 88.7/100) and fell only slightly by 3 years (-2.0; p<0.0001). No group differences appeared until year 3, when adherence was 5% and 3% higher for glimepiride and sitagliptin respectively than liraglutide (both p<0.05), but differences were small and of unclear clinical significance. A 10-point fall in adherence was associated with 15% and 19% higher risk of HbA1c ≥7.0% and >7.5% (both p<0.0001). |
Expert Commentary
This is a high-quality adherence analysis nested in one of the largest head-to-head trials of second-line diabetes drugs, and its strength is prospective, standardised measurement across four classes rather than the heterogeneous pharmacy-claims data that dominate this literature. Two messages stand out. First, adherence was uniformly high and the between-drug differences, including for injectable liraglutide versus oral agents, were small and of unclear clinical significance, which argues that drug choice should turn on efficacy, cardiovascular and weight effects, and hypoglycaemia risk rather than presumed adherence gaps. Second, the adherence-to-outcome link was robust, with each 10-point drop meaningfully raising the risk of missing glycaemic targets, which quantifies why ongoing adherence assessment matters. The limitations are important for generalisability: GRADE participants received intensive support and education that likely inflated adherence above routine practice, adherence was self-reported and may overstate real behaviour, baseline control was relatively good with short diabetes duration, and cost and insurance barriers were buffered by the trial. Can I use this with my patients? Yes. I would reassure patients worried about injection burden that adherence was comparable, choose agents on their clinical merits, and still actively monitor and support adherence, recognising real-world rates may be lower than this trial’s.
References
Gonzalez JS, Wen H, Butera NM, et al. Medication adherence in the Glycemia Reduction Approaches in Diabetes: a Comparative Effectiveness Study (GRADE). Diabetes Care. 2026;49(2):335–343. doi:10.2337/dc25-2008