The study’s design deserves careful attention because it shapes how much confidence the findings can carry. The researchers drew on the records of 131 hospitalised adults with schizophrenia, each diagnosis confirmed by an independent re-review of the case record against the DSM-5 criteria rather than taken at face value from the chart. Sixty-seven patients received risperidone plus olanzapine, while 64 received risperidone alone, and 124 of the 131 had complete week-eight data available for analysis. Because treatment assignment was not randomised, the two groups could differ in ways that both influenced the choice of combination therapy and shaped the outcomes, a limitation the authors themselves flag prominently.

One of the most consequential imbalances was not in baseline characteristics, which were well matched, including treatment history, but in the total amount of antipsychotic drug each group actually received. When total exposure was converted into olanzapine-equivalent daily doses, the combination group received 77 percent more antipsychotic overall, 30.71 milligrams per day versus 17.35 milligrams per day in the monotherapy arm. Using a common pharmacological model that assumes a maximal effect at full receptor blockade, this translates into hypothetical striatal dopamine D2 receptor occupancy of 87.3 percent for the combination group versus 83.0 percent for monotherapy. This matters because D2 occupancy is thought to be a central driver of antipsychotic efficacy, and it means the apparent symptomatic advantage of the combination cannot be cleanly separated from the simple fact that patients in that arm were receiving more total drug.

The symptomatic results were measured with the Scale for the Assessment of Positive Symptoms and the Scale for the Assessment of Negative Symptoms, two instruments that have long served as standard tools for quantifying the distinct dimensions of schizophrenia psychopathology. Positive symptoms encompass hallucinations, delusions, and disorganised thinking, while negative symptoms include blunted affect, social withdrawal, apathy, and poverty of speech, which are often harder to treat and contribute heavily to functional disability. Over the eight weeks, the combination group showed a mean reduction of 33.35 points on the SAPS compared with 24.16 points for monotherapy, a difference of 9.19 points with a 95 percent confidence interval of 13.18 to 5.20. On the SANS, the corresponding reductions were 27.11 versus 18.33 points, a difference of 8.78 points with a confidence interval of 12.48 to 5.08. Both differences were statistically significant after adjustment for multiple comparisons using the Hochberg procedure.

Response rates told a similar story in more clinically interpretable terms. For positive symptoms, 57.1 percent of the combination group met the response threshold compared with 32.8 percent of the monotherapy group. For negative symptoms the gap was wider still, with 46.0 percent versus 13.1 percent achieving response. The negative-symptom finding is particularly striking because negative symptoms have historically been resistant to most pharmacological interventions, and any signal in this domain attracts attention. However, the same confounding by total antipsychotic exposure applies here as well, and it remains possible that simply pushing total D2 blockade higher, whether with one drug or two, would produce a comparable effect.

The metabolic toll was the study’s other co-primary concern, and here the picture was unambiguous. All eight metabolic parameters examined, spanning anthropometric, glycaemic, and lipid measures, deteriorated more in the combination group, with every comparison reaching statistical significance at the adjusted threshold of 0.004 or lower. Weight gain was the most visible manifestation, averaging 3.58 kilograms in the combination arm versus 1.42 kilograms with monotherapy, a difference of 2.16 kilograms over just eight weeks. Clinically meaningful weight gain, defined as an increase of at least 7 percent of body weight, occurred in 39.7 percent of the combination group versus 9.8 percent of the monotherapy group, yielding a number needed to harm of just 3. In practical terms, for every three patients treated with the combination rather than monotherapy, one additional patient would be expected to gain a clinically significant amount of weight within two months.

Olanzapine is well known among the second-generation antipsychotics for its particularly strong association with weight gain and metabolic disturbance, an effect attributed to a combination of histamine H1 receptor antagonism, serotonin 5-HT2C antagonism, and direct effects on appetite regulation and glucose metabolism. Risperidone carries a comparatively lower metabolic burden, which makes the choice to add olanzapine to it a decision that inherently stacks a high-risk agent on top of a lower-risk one. The eight-week window of this study, while sufficient to capture the steepest phase of antipsychotic-associated weight gain, likely understates the longer-term metabolic trajectory, since weight gain and dyslipidaemia tend to continue accumulating well beyond the first two months of exposure.

An important secondary observation was that weight gain was unrelated to symptom improvement within either treatment group. This dissociation is noteworthy because it argues against the possibility that patients who improved more simply ate more or moved less as a byproduct of feeling better, and it also suggests that the metabolic and symptomatic effects of the combination operate through partly independent mechanisms. The exploratory cognitive measure, the Montreal Cognitive Assessment, showed no difference between groups, though the authors caution that this brief screening instrument, readministered in identical form at each visit, has limited sensitivity to change and was not designed as a primary endpoint.

The authors are careful about what their data can and cannot support. Because the comparison was between a two-drug regimen and a one-drug regimen without randomisation, and because the combination group received substantially more total antipsychotic, they conclude that the study does not establish an advantage of combining two agents over adequate dosing of a single one. This distinction is more than semantic. If the benefit of the combination is attributable simply to higher total D2 occupancy, then the same symptomatic gains might be achievable with careful dose optimisation of risperidone alone, potentially sparing patients the metabolic consequences of adding olanzapine. Conversely, if olanzapine’s distinct receptor profile contributes something that risperidone dose escalation cannot, that would be a genuine pharmacological argument for the combination, but the current design cannot distinguish between these possibilities.

The study was conducted at a single centre, the Third People’s Hospital of Fuyang District in Hangzhou, Zhejiang Province, and the retrospective design means that the decision to add olanzapine was made by treating clinicians based on their judgement of insufficient response, introducing the possibility of confounding by indication. Patients selected for combination therapy may have been, on average, more ill or more treatment-resistant at the point of escalation, though the balance in baseline characteristics and treatment history suggests this concern was at least partially addressed. The ethics committee waived written informed consent because the study involved only retrospective analysis of de-identified, routinely collected clinical data with no patient contact, and all data were anonymised before analysis.

For clinicians, the practical message is one of caution rather than prohibition. The combination of risperidone and olanzapine did produce larger symptomatic improvements across both positive and negative symptom domains in this cohort, and the effect sizes were not trivial. But the metabolic penalty was equally consistent, affecting every parameter measured, and the magnitude of weight gain observed in nearly 40 percent of combination-treated patients within eight weeks is a signal that demands active monitoring of weight, glucose, and lipids whenever this strategy is employed. The finding that total antipsychotic exposure differed so substantially between groups also serves as a reminder that in real-world prescribing, the question is rarely simply whether to combine drugs but how much total pharmacological burden a patient is being asked to carry, and whether that burden could be achieved more safely with a single well-titrated agent.

Subject of Research: Symptomatic benefit and metabolic cost of combining olanzapine with risperidone in schizophrenia

Article Title: Symptomatic benefit and metabolic cost of adding olanzapine to risperidone in schizophrenia: an 8-week retrospective cohort study

Article References: Shen, L., Han, X., Shen, J., Shen, F., Li, Y., & Li, H. (2026). Symptomatic benefit and metabolic cost of adding olanzapine to risperidone in schizophrenia: an 8-week retrospective cohort study. BMC Psychiatry. https://doi.org/10.1186/s12888-026-08766-8

Image Credits: AI Generated

DOI: 10.1186/s12888-026-08766-8

Keywords: schizophrenia, antipsychotic polypharmacy, olanzapine, risperidone, metabolic side effects, weight gain, negative symptoms, D2 receptor occupancy, psychopharmacology, retrospective cohort study, Symptomatic, benefit

The research team, led by investigators at Dongzhimen Hospital and Beijing University of Chinese Medicine together with colleagues at Ningbo Municipal Hospital of Traditional Chinese Medicine and Tsinghua University Yuquan Hospital, conducted a systematic review and network meta-analysis, a statistical technique that allows researchers to compare multiple treatments simultaneously, even when some treatments have never been tested head-to-head in the same trial. The team searched four major biomedical databases, PubMed, Embase, the Cochrane Library, and Web of Science, for randomized controlled trials published through December 5, 2025. To be included, studies had to evaluate metformin or a GLP-1 receptor agonist for at least twelve weeks in patients continuing antipsychotic treatment, ensuring that any measured metabolic changes reflected an intervention layered on top of ongoing psychiatric care rather than a change in the psychiatric medication itself.

The search ultimately captured twenty-nine randomized controlled trials encompassing 1,721 patients. The typical participant was young, with a median study-level age of 36.2 years, and slightly less than half of the pooled population, 47.7 percent, was male. The median treatment duration across the trials was sixteen weeks, long enough to capture meaningful changes in body weight and glucose metabolism but short enough to leave questions about durability unanswered. The researchers assessed the risk of bias in each trial using the Cochrane Risk of Bias 2.0 tool, the current gold standard for judging whether randomized studies were designed and conducted in ways that protect against inflated or misleading results.

At the heart of the analysis was a random-effects frequentist network meta-analysis performed in Stata 17.0 MP, a framework that pools direct comparisons between treatments with indirect ones to estimate how any two interventions would compare even without direct trial evidence. To rank the competing treatments, the team calculated SUCRA values, the surface under the cumulative ranking curve, a metric that expresses the probability that a given treatment ranks best across outcomes. They also deployed univariate network meta-regression to probe whether study-level characteristics, such as baseline weight or antipsychotic type, modified treatment effects, and they graded confidence in the findings using the CINeMA framework, short for Confidence in Network Meta-Analysis, which evaluates how much trust each estimate deserves based on the structure and consistency of the evidence network.

The headline result concerns semaglutide, the GLP-1 receptor agonist that has transformed obesity medicine in recent years. Compared with control groups, semaglutide was associated with a reduction in body mass index of 3.55 kilograms per square meter, with a 95 percent confidence interval spanning from 4.27 down to 2.84, a magnitude that in clinical terms represents a substantial improvement for a patient population in which weight gain of twenty or thirty kilograms is not uncommon. The drug also produced meaningful reductions in waist circumference, averaging 6.34 centimeters, in glycated hemoglobin A1c, the three-month average of blood sugar, which fell by 0.44 percentage points, and in fasting blood glucose, which dropped by 0.53 millimoles per liter. Together, these figures paint semaglutide as a broad-acting agent against the weight and glycemic dimensions of antipsychotic metabolic harm.

Metformin, by contrast, showed a different signature. The decades-old biguanide, which works largely by suppressing glucose production in the liver and improving insulin sensitivity, was associated with reductions in lipid metabolism markers, specifically total cholesterol and triglycerides, the blood fats most closely tied to cardiovascular risk. Intriguingly, the analysis also flagged an exploratory potential benefit for psychiatric symptom scores, with a standardized mean difference of 0.31 in favor of metformin, a small-to-moderate effect that, if confirmed, would suggest the drug might ease the burden of psychotic symptoms themselves rather than merely offsetting the metabolic cost of their treatment. The authors were careful to frame this finding as preliminary, but it adds to a growing body of speculation about links between insulin signaling, inflammation, and brain function in psychiatric illness.

The divergence in outcome profiles between the two drug classes is arguably the study’s most clinically useful message. GLP-1 receptor agonists mimic the gut hormone glucagon-like peptide-1, slowing gastric emptying, enhancing satiety, and prompting insulin release in a glucose-dependent manner, which explains their strength on weight and glycemic endpoints. Metformin’s mechanism, centered on AMPK activation and hepatic glucose suppression, aligns more naturally with lipid effects. For clinicians, this suggests the choice of adjunctive therapy could be tailored to a patient’s dominant metabolic problem: a patient ballooning in weight on olanzapine or clozapine might be a stronger candidate for a GLP-1 agonist, while one with worsening triglycerides might do better on metformin, at least until direct comparative trials settle the question.

That caveat matters, and the authors emphasize it forcefully. Because most trials in the network compared each drug against placebo rather than against each other, the estimates linking semaglutide and metformin are indirect, and the transitivity assumption, the premise that the trials being connected are similar enough in their populations and designs to be meaningfully compared, remains uncertain. The team explicitly describes their conclusions as exploratory evidence intended to inform clinical discussion and future research, not as definitive treatment guidance. The CINeMA-based confidence ratings reflect this caution, and the fact that the article was shared early as a peer-reviewed, accepted manuscript subject to further editorial processing adds another layer of provisional status to the findings.

Even with those limitations, the stakes of the research are difficult to overstate. People with schizophrenia die on average fifteen to twenty years earlier than the general population, and cardiovascular disease driven by antipsychotic-associated metabolic dysfunction is a leading contributor. Adherence is another casualty: patients who gain alarming amounts of weight frequently stop taking the very medications that keep their psychosis at bay, triggering relapse, hospitalization, and worsening long-term outcomes. An effective, well-tolerated metabolic countermeasure could therefore ripple across the entire trajectory of severe mental illness, improving not just waistlines and lab values but treatment persistence, relapse rates, and quality of life.

The study also arrives at a moment of intense public fascination with GLP-1 drugs, which have been hailed as near-miraculous for obesity and are being tested for everything from addiction to neurodegeneration. Applying them to antipsychotic-induced metabolic disturbances is a natural extension, but questions remain about cost, access, tolerability in psychiatric populations, and the theoretical concern that GLP-1 agents’ appetite suppression could interact with the weight dynamics of patients on antipsychotics. The authors’ funding, drawn from Chinese national and institutional sources including traditional Chinese medicine endocrinology programs, underscores the global nature of the effort to solve this problem. What the analysis delivers is a rigorous map of the current randomized evidence and a clear agenda: head-to-head trials directly comparing metformin and GLP-1 receptor agonists, longer follow-up to test durability, and studies powered to confirm whether metformin’s apparent psychiatric benefit is real. For the millions of patients whose lifesaving medications come with a metabolic tax, that research cannot come soon enough.

Subject of Research: Comparing metformin and GLP-1 receptor agonists for treating antipsychotic-induced metabolic disturbances

Article Title: Metformin and GLP-1 receptor agonists for antipsychotic-induced metabolic disturbances: a systematic review and network meta-analysis

Article References: Chen, Y.-X., Yang, Q.-W., Lin, M.-X., Sun, M.-H., Dong, Y.-Y., Yu, R.-D., Mao, D.-D., Zhao, Y., Zhang, L., Zhao, J.-X., Zhang, Y.-F., & Xu, J. (2026). Metformin and GLP-1 receptor agonists for antipsychotic-induced metabolic disturbances: a systematic review and network meta-analysis. BMC Medicine. https://doi.org/10.1186/s12916-026-05280-2

Image Credits: AI Generated

DOI: 10.1186/s12916-026-05280-2

Keywords: metformin, GLP-1 receptor agonists, semaglutide, antipsychotics, metabolic side effects, network meta-analysis, schizophrenia, obesity, BMI, HbA1c, psychiatry, BMC Medicine