Clone-censor-weight to evaluate the effect of timing of treatment initiation in individuals with new-onset diabetes eligible for treatment

Background: Individuals with new-onset diabetes are eligible for glucose-lowering treatment as soon as their glycated haemoglobin (HbA1c) exceeds the diagnostic threshold for diabetes, 48 mmol/mol. The effect of therapeutic inertia, i.e., deferred versus early initiation of glucose-lowering treatment, on clinical outcomes is unknown.

Objectives: To emulate a target trial and use the clone-censor-weight method to evaluate the effect of timing of treatment initiation among individuals eligible for first glucose-lowering treatment.

Methods: In this nationwide population-based study, we used HbA1c from routine laboratory databases linked to Danish healthcare registries during 2013-2021. Danish adults were included at the time of their first-ever HbA1c >=48 mmol/mol and restricted to those with modestly elevated HbA1c at diagnosis (48-52 mmol/mol). Individuals with prior glucose-lowering treatment were excluded. We defined 5 treatment initiation strategies: immediate initiation within 14 days; initiation within 3 months; 6 months; 1 year; or no treatment initiation during the 1-year treatment assessment period. All individuals were cloned and assigned to all treatment initiation strategies. Clones were censored if they deviated from a certain treatment strategy. The effect of timing of treatment initiation will be estimated by comparing risk estimates for each treatment strategy estimated from the complement of a weighted Kaplan-Meier estimator, where weights represent the inverse probability of censoring.

Results: A total of 83,214 individuals were identified with a first-ever HbA1c above the diagnostic threshold for diabetes (48 mmol/mol). Of these, 51,122 (61%) had modestly elevated HbA1c (48-52 mmol/mol) and comprised the study cohort. A total of 3,997 (8%) initiated glucose-lowering treatment within 14 days, 6,084 (12%) within 3 months, 2,902 (6%) within 6 months, and 3,684 (7%) within 1 year. The median time to first glucose-lowering treatment initiation was 50 days (IQR 16-160). A high proportion, 34,455 (67%), remained untreated during the first year. In preliminary analyses, during a median of 4.3 years (IQR 2.9-5.0) of follow-up, 35,358 (69%) experienced a community- or hospital-treated infection and 7,807 (15%) died. Further analyses are to be performed on these outcomes.

Conclusions: Our unique population-based laboratory data provide a setup where the clone-censor-weight method allows for evaluation of the effect of deferred versus early treatment initiation strategies on clinical outcomes among individuals who become eligible for diabetes treatment. The results can therefore provide valuable information for clinicians on the effects of timing initial glucose-lowering treatment.