When Should You Act on Antipsychotic Weight Gain — and Why Is One Month the Number That Matters?

A patient starts olanzapine, responds beautifully, and comes back at three months eleven kilos heavier. The reflex answer — note the weight, mention diet and exercise, review again next quarter — feels reasonable and is the reason the window closed. By the time weight gain is obvious enough to discuss, the trajectory that produced it was set weeks earlier. There is a usable number for when to act, and it arrives much sooner than most of us check.

🎯️ So when exactly do I intervene?

At the one-month weigh-in — if the patient has crossed 5% of their starting body weight, that is the moment to act rather than observe.

Antipsychotic weight gain is not a slow drift. With the higher-liability agents it is fastest in the first weeks to months and then continues more gradually, which means the early period carries most of the prognostic information. Gaining more than 5% in that first month predicts ≥15% gain at three months and ≥20% at twelve.

Acting means something concrete: reconsider the agent, add a preventive intervention, intensify structured lifestyle support. It does not mean writing "weight up, discussed" and booking a review. Weight prevented is far easier than weight reversed.

📊 How good is that rule, really?

Highly specific, only moderately sensitive — so treat it as a rule-in, not a rule-out.

Specificity runs at 88% for the three-month prediction and 89% for the twelve-month one. Sensitivity is lower, at 67% and 47% respectively. Read plainly: crossing 5% at one month rarely happens in someone who was going to stay stable, so when it happens it means something. But a patient who stays under 5% has not been cleared — roughly a third to a half of eventual heavy gainers will not have crossed the line yet.

That asymmetry is exactly why the rule is useful as a trigger and useless as a discharge criterion. It tells you when to escalate. It does not tell you when to stop weighing.

⚖️ Doesn't choosing a weight-neutral agent solve this?

No — and the ranking most of us carry comes from trials in chronic, previously treated patients, which is not who we are usually starting.

In antipsychotic-naive and first-episode patients the gaps between agents narrow sharply. Every agent produces substantial gain in this group, and the ones we describe as metabolically favourable still push a meaningful proportion of patients past clinically relevant thresholds. The practical consequence is that a first-episode patient started on aripiprazole needs watching as closely as one started on olanzapine — the ranking that justifies the choice does not license relaxing the monitoring.

Formulation may matter too, though here the evidence pulls in two directions. In one early-psychosis cohort, oral aripiprazole produced roughly 11 kg of gain against about 3.7 kg for the long-acting injectable. Against that, an 18-month randomised analysis found no metabolic advantage for long-acting injectables at all, and in that trial they were associated with greater weight gain. Treat the formulation question as open rather than settled.

Figures from Burschinski et al., World Psychiatry 2023 — a network meta-analysis of 31 antipsychotics over treatment longer than 13 weeks. Credible intervals shown where the ranking is close.

🩸 If the weight is stable, is the patient metabolically safe?

No. Glycemic harm shows up across the whole liability spectrum, including on the agents that barely move the scale.

A 2025 meta-analysis found antipsychotics raise HbA1c relative to placebo across low-, moderate- and high-weight-gain agents alike. That is the clearest available evidence that the scale is not a screening test, and it is why the monitoring schedule tracks glucose and lipids on their own timetable rather than inferring them from body weight.

It is also worth carrying the framing correction: the drug amplifies a pre-existing vulnerability rather than creating it. About 22% of patients already meet metabolic syndrome criteria at treatment initiation, rising to 32% by one year,. That argues for a baseline before the first dose — not for excusing the drug's contribution.

📅 So what should actually be on the schedule?

Weight at every early visit, and the bloodwork on its own clock — not triggered by what the scale is doing.

This is the consensus schedule. It is short enough to memorise, which is the point: build it into the template and it stops being a decision you make at each appointment.

Weigh more often than this in the first three months if the agent is high-liability or the patient is antipsychotic-naive. The four-week reading is the one that changes decisions, and it only works if somebody takes it.

⚠️ Is there a catch?

Three, and they are worth knowing before you lean on any of this too hard.

First, the threshold's evidence base is narrower than its confidence suggests. The figures come from a single prospective cohort of 351 psychiatric patients followed for a year at one Swiss centre, and the drugs studied were weight-gain-inducing psychotropics broadly rather than antipsychotics alone. Add the moderate sensitivity — a large minority of eventual heavy gainers pass through the first month unflagged — and it earns its place as a trigger, not as a prognostic verdict.

Second, the real-world constraint is execution rather than knowledge. Guideline monitoring schedules are clear and unambiguous, and screening rates in practice fall well short of them. Nothing in this post helps a patient whose weight is not actually measured at the one-month visit — which makes the habit, not the rule, the thing worth building.

Third, there is an uncomfortable correlation underneath all of it. Network meta-analysis finds that improvement in psychotic symptoms tracks with metabolic worsening: the patients doing best psychiatrically are often the ones paying the highest metabolic price. That is not an argument for under-treating the illness. It is the argument for running prevention alongside effective treatment from day one, rather than waiting for a problem to declare itself.