Comparative effectiveness of alternative second-line oral glucose-lowering therapies for type 2 diabetes: a precision medicine approach applied to routine data.

Aims/hypothesis: National clinical guidelines recommend that second-line treatment for type 2 diabetes mellitus is chosen according to individuals' characteristics but there is limited evidence available to inform this choice. This paper's aim is to compare the effects on HbA_1c of sulfonylureas (SU), dipeptidyl peptidase-4 inhibitors (DPP4i) or sodium-glucose cotransporter-2 inhibitors (SGLT2i) added to metformin as second-line oral glucose-lowering treatments according to an individual's age, baseline HbA_1c and presence of multiple long-term conditions (MLTCs).

Methods: We accessed primary care-hospital linked data for 41,790 individuals from the Clinical Practice Research Datalink (CPRD) in England who initiated second-line treatment after metformin between 2015 and 2021. We combined target trial emulation with instrumental variable analysis to reduce the risk of confounding. The outcome was change in HbA_1c between baseline and 1 year follow-up. We reported results stratified by age (18-49 years, 50-69 years and ≥70 years), baseline HbA_1c (<67 mmol/mol [<8.3%], 67-77 mmol/mol [8.3-9.2%] and >77 mmol/mol [>9.2%]) and presence of MLTCs.

Results: The mean (95% CI) difference in HbA_1c change for SGLT2i vs SU was larger for people aged 18-49 years (-5.74 mmol/mol [-7.47, -4.01]) (-0.5% [-0.7, -0.4]) than for those aged 50-69 years (-4.03 mmol/mol [-5.61, -2.44]) (-0.4% [-0.5, -0.2]) and for those aged 70 years or over (-2.68 mmol/mol [-4.50, -0.86]) (-0.3% [-0.4, -0.07]). The mean (95% CI) difference in HbA_1c change for SGLT2i vs DPP4i was -5.80 mmol/mol (-7.60, -4.00) (-0.5% [-0.7, -0.4]) for those aged 18-49 years, -4.13 mmol/mol (-5.82, -2.45) (-0.4% [-0.5, -0.2]) for those aged 50-69 years and -3.13 mmol/mol (-5.01, -1.24) (-0.3% [-0.4, -0.1]) for those aged ≥70 years. The mean difference (improvement) in HbA_1c was similar across subgroups defined by baseline HbA_1c or presence of MLTCs. For SGLT2i vs SU, the mean (95% CI) difference was -5.37 mmol/mol (-7.13, -3.62) (-0.5% [-0.6, -0.3]) for people without MLTC and -3.72 mmol/mol (-5.34, -2.10]) (-0.3% [-0.5, -0.2]) for people with MLTC. For SGLT2i vs DPP4i the corresponding estimated differences (95% CI) were -5.44 mmol/mol (-7.27, -3.61) (-0.5% [-0.7, -0.3]) for those without MLTC and -3.93 mmol/mol (-5.64, -2.21) (-0.3% [-0.5, -0.2]) for those with MLTC.

Conclusions/interpretation: Second-line treatment with SGLT2i is more effective than SU or DPP4i in reducing HbA_1c across subgroups of people defined by age, baseline HbA_1c and presence of MLTCs. Our evidence complements RCTs in using routinely available information on demographic characteristics, biomarkers and comorbidities to inform an individualised approach.