Rapid metabolic change followed antipsychotic initiation in severe mental illness
Starting a second-generation antipsychotic was followed by a sharp early metabolic shift in people with severe mental illness in England, with HbA1c and weight rising fastest in the first 90 days after treatment began. The pattern matters because the early post-initiation period is often when patients are most likely to be focused on psychiatric stabilization rather than metabolic surveillance, even though the biological changes are already underway.
The analysis drew on electronic health records from the Clinical Practice Research Datalink Aurum database and included adults aged 18 to 59 years with severe mental illness who had a recorded oral antipsychotic initiation and at least one further prescription within 90 days. People with dementia, type 2 diabetes, or antipsychotic exposure before age 18 were excluded. In total, 96,849 individuals met eligibility criteria. Slightly more than half were male, the mean age was 37.2 years, and the cohort was ethnically diverse, although ethnicity was missing for a substantial minority. Weight was available for most participants, while HbA1c data were recorded for roughly one in four, reflecting the unevenness of cardiometabolic monitoring in routine care.
The first 90 days carried the steepest glycaemic and weight rise
Before treatment began, HbA1c rose only marginally, at 0.01 mmol/mol per week, and weight increased by 0.02 kg per week. After antipsychotic initiation, both trajectories steepened quickly. During days 0 to 90, HbA1c increased at 0.12 mmol/mol per week and weight at 0.21 kg per week. From day 91 to 2 years, the HbA1c slope returned to the pre-treatment rate, while weight continued to climb more slowly at 0.03 kg per week.
The shape of those curves is clinically important. The early surge suggests that metabolic effects begin soon after exposure, before long-term weight gain alone can fully explain the glycaemic deterioration. Antipsychotics can alter appetite, insulin sensitivity, energy balance, and possibly peripheral glucose handling through receptor-mediated effects, which helps explain why HbA1c can rise in parallel with, but not entirely because of, weight gain. A delayed response limited only to later follow-up would miss the window when the trajectory is changing most rapidly.
Olanzapine produced the largest early metabolic signal
The drug-specific analysis showed the strongest early increases with olanzapine, which was associated with HbA1c rising at 0.18 mmol/mol per week and weight increasing at 0.28 kg per week. In contrast, quetiapine was linked to weight gain of 0.19 kg per week but minimal HbA1c change, suggesting that weight gain and glycaemic worsening do not always move in lockstep across the class.
This dissociation is a reminder that weight alone is an incomplete metabolic surrogate. Some agents may push appetite and adiposity more strongly, while others may exert a more direct effect on glycaemic regulation. For clinicians, the practical consequence is straightforward: a patient whose weight is still stable can already be accumulating glycaemic risk, and a patient whose weight has increased modestly may not show the full extent of metabolic injury if HbA1c is not checked.
Women experienced larger early increases than men, with HbA1c rising at 0.14 mmol/mol per week and weight at 0.22 kg per week. The study did not attempt to reduce that sex difference to a single mechanism, but the finding argues for closer early monitoring in women receiving antipsychotic therapy, especially when other risk factors for dysglycaemia are present.
Baseline prediabetes sharply increased two-year diabetes incidence
Among 918 individuals with baseline prediabetes, 126 developed type 2 diabetes, with an estimated 2-year cumulative incidence of 19.0%. In comparison, 152 of 9,171 individuals with normal baseline HbA1c developed type 2 diabetes, corresponding to a 2-year cumulative incidence of 3.0%.
The contrast is stark. Pre-existing dysglycaemia marked a subgroup whose risk after antipsychotic initiation was far higher than that of people starting from a normal glycaemic baseline. In routine practice, that means the first prescription is not just a psychiatric decision; it is also a metabolic risk decision. When baseline HbA1c already sits in the prediabetic range, the threshold for choosing a lower-risk antipsychotic, intensifying lifestyle counseling, or arranging earlier laboratory follow-up should be lower.
Routine monitoring needs to begin early and continue beyond the initial spike
The investigators concluded that the combination of rapid early HbA1c and weight increases, followed by slower but persistent change, supports baseline and 12-week monitoring of both outcomes, then risk-stratified follow-up thereafter. The time course fits a model in which the earliest months after initiation represent the period of greatest metabolic inflection, while the later phase reflects ongoing cumulative exposure.
In practice, this means metabolic assessment should be treated as part of antipsychotic prescribing rather than a separate later task. Baseline weight and HbA1c establish the starting point. A 12-week review captures the period when the slope can change most abruptly. After that, monitoring should be individualized according to the antipsychotic used, sex, and baseline glycaemic status. The study’s uneven HbA1c capture across the cohort also reflects a broader problem: the very people most likely to benefit from early surveillance are often the ones least consistently followed.
People with lived experience contributed to the research question, the study approach, interpretation, and dissemination. That involvement is clinically relevant because antipsychotic treatment sits at the intersection of symptom control, adherence, and long-term physical health. A medication that stabilizes psychosis but drives avoidable metabolic injury can narrow future treatment options and compound cardiovascular risk over time.
The main limitation of the analysis lies in real-world data completeness, especially for HbA1c, which was recorded in only a minority of participants. Even so, the large cohort and individual-level time-series design give the findings substantial weight. The unanswered clinical issue is how best to operationalize risk-stratified monitoring across different prescribing settings so that the earliest metabolic changes are not missed.