Using longitudinal real-world data, investigators identified three distinct metformin responder phenotypes based on long-term HbA1c trajectories: good responders, poor responders, and non-responders. Good responders comprised 63.6% of the cohort, poor responders 8.9%, and non-responders 27.5%. Good responders maintained glycaemic control on metformin monotherapy. Poor responders initially achieved targets but subsequently lost control within approximately 2.5 years. Non-responders exhibited minimal HbA1c reduction after metformin initiation and reached glycaemic failure within about 1 year.
The study analyzed 59,881 longitudinal HbA1c measurements from 7,105 patients with type 2 diabetes receiving metformin monotherapy. Data were drawn from Kaiser Permanente Northern California electronic health records with follow-up up to six years. The analysis integrated demographic, clinical, genetic, and pharmacological factors, although the abstract emphasizes demographic and clinical predictors; detailed genetic findings were not reported in the abstract.
Trajectories were derived from long-term HbA1c measurements to define phenotypes by temporal patterns of glycaemic control on metformin monotherapy. Definitions reflected both magnitude and timing of HbA1c change: non-responders showed little HbA1c reduction and early failure (median ~1 year), poor responders achieved early improvement but lost control within ~2.5 years, and good responders sustained control throughout follow-up. These trajectory-based phenotypes were used as the primary outcome to characterize response durability to metformin.
Five baseline clinical variables—baseline HbA1c, age at diagnosis, body mass index (BMI), sex, and estimated glomerular filtration rate (eGFR)—were sufficient to classify trajectory phenotypes with good discrimination. Model performance, as reported, achieved an area under the receiver operating characteristic curve (AUC) of 0.84, indicating strong separation among phenotypes using routinely available baseline data.
Incorporating on-treatment HbA1c measurements further improved early identification of non-responders beyond baseline predictors alone. This suggests that observing early response while on therapy can add prognostic value for identifying patients unlikely to benefit from continued metformin monotherapy.
Among patients classified as suboptimal responders (poor and non-responders), analyses quantified the effects of modifiable factors. Improved medication adherence and weight control were associated with a delay in time to glycaemic failure by approximately seven months. Despite this delay, many suboptimal responders ultimately experienced glycaemic failure on metformin monotherapy, indicating that these interventions may prolong but not necessarily prevent progression to inadequate control for a sizable subset.
Key clinical implications from the analysis include:
Baseline clinical characteristics can be used to stratify risk of suboptimal response to metformin, enabling earlier consideration of intensified or alternative therapies for patients predicted to be poor or non-responders.
Early on-treatment HbA1c monitoring enhances detection of non-response and may inform timely treatment adjustments.
Behavioral and care-delivery interventions that improve adherence and support weight control can meaningfully delay glycaemic failure, but clinicians should anticipate that many suboptimal responders may still require additional pharmacologic strategies.
The majority of patients in this real-world cohort did achieve durable control on metformin, reinforcing its role as first-line therapy while highlighting a substantial minority who may need personalized treatment plans.
This work is a preprint and has not undergone peer review; it should not be used alone to guide clinical practice. The analysis used de-identified electronic health record data provided under an institutional data use agreement. The authors declared no competing interests. The abstract reports integration of genetic and pharmacological data but does not provide detailed genetic associations or specific pharmacologic modifiers in the abstract; those details were not reported in the abstract. The reported funding source included support from the National Institute of General Medical Sciences (R01GM117163).
Overall, the study characterizes three clinically relevant HbA1c trajectory phenotypes among patients initiating metformin monotherapy and demonstrates that a small set of baseline clinical features, augmented by early on-treatment HbA1c, can identify patients unlikely to achieve durable glycaemic control with metformin alone. Modifiable factors such as adherence and weight control can delay, but may not entirely prevent, progression to glycaemic failure in suboptimal responders.