Antipsychotics in dementia carry a wider range of harms than warnings say

Antipsychotic drugs given to people with dementia are linked to a broader set of serious harms than regulators have flagged, and the risks run highest in the days just after treatment begins, according to a matched cohort of 173,910 people with dementia in England [s1]. Compared with non-use, current antipsychotic use was associated with raised risks of pneumonia (hazard ratio 2.19), acute kidney injury (1.72), venous thromboembolism (1.62), stroke (1.61), fracture (1.43), myocardial infarction (1.28) and heart failure (1.27) [s1].

That list is longer than the one behind the warnings clinicians have worked to for two decades, which centred on stroke and death. The study does not show that the drugs cause these events, but its size, its design and a built-in check against hidden bias make it hard to dismiss.

What the study measured

The analysis used linked primary care, hospital and mortality records from the Clinical Practice Research Datalink in England, covering adults aged 50 and over with a dementia diagnosis recorded between 1 January 1998 and 31 May 2018 [s1]. Of the 173,910 people with dementia, 63.0% were women [s1]. Each of the 35,339 new antipsychotic users was matched with up to 15 non-users of the same underlying risk using incidence density sampling [s1].

The researchers tracked eight outcomes: stroke, venous thromboembolism, myocardial infarction, heart failure, ventricular arrhythmia, fracture, pneumonia and acute kidney injury [s1]. Compared with non-use, any antipsychotic use was associated with increased risks of all of them except ventricular arrhythmia, and the hazard ratios above are for the period of current use, defined as the 90 days after a prescription [s1]. The 95% confidence intervals were tight: pneumonia 2.10 to 2.28, acute kidney injury 1.61 to 1.84, venous thromboembolism 1.46 to 1.80, stroke 1.52 to 1.71, fracture 1.35 to 1.52, myocardial infarction 1.15 to 1.42 and heart failure 1.18 to 1.37 [s1].

Turning ratios into people

A relative risk means little without a baseline rate, and here the baseline is not small. In the 90 days after starting an antipsychotic, the cumulative incidence of pneumonia among users was 4.48% (95% CI 4.26% to 4.71%), against 1.49% (1.45% to 1.53%) in the matched non-users — an absolute difference of 2.99 percentage points (2.77 to 3.22) [s1]. That is roughly one extra case of pneumonia for every 33 people treated over three months, concentrated in the frailest patients and in the first weeks of use [s1].

The authors state plainly that the range of adverse outcomes was wider than previously highlighted in regulatory alerts, and that the highest risks came soon after initiation [s1].

Why an observational finding here carries weight

The obvious objection to a study like this is confounding by indication: people prescribed antipsychotics are agitated, sicker, closer to the end of life, and might suffer more of these events anyway. The design tries to blunt that by matching on background risk, but it cannot remove it. What it can do is test for it. The investigators pre-specified an unrelated negative-control outcome — appendicitis and cholecystitis combined — that antipsychotics have no plausible reason to cause. No increased risk was seen for that outcome, which argues against a large pool of unmeasured confounding inflating the other associations [s1].

Set against what the drugs achieve

The harms have to be read against the benefit, and the benefit is modest. The CATIE-AD trial randomised 421 outpatients with Alzheimer's disease and psychosis, aggression or agitation to olanzapine, quetiapine, risperidone or placebo, and found no significant difference between the drugs and placebo on the main effectiveness measure: improvement on a global clinical scale at 12 weeks was seen in 32% of the olanzapine group, 26% of the quetiapine group, 29% of the risperidone group and 21% of the placebo group (P = 0.22) [s3]. More patients stopped the drugs than placebo because of side effects — 24%, 16% and 18% respectively, versus 5% on placebo (P = 0.009) — leading the authors to conclude that adverse effects offset the drugs' efficacy advantages [s3].

The mortality signal that first prompted regulators to act is older still. A 2005 meta-analysis of 15 placebo-controlled trials found death in 3.5% of patients randomised to atypical antipsychotics versus 2.3% on placebo, an odds ratio of 1.54 (95% CI 1.06 to 2.23; P = 0.02) [s2]. That finding underpins the boxed warnings these drugs have carried since.

What it changes

Guidelines already position antipsychotics as a last resort for the behavioural and psychological symptoms of dementia, to be used at the lowest dose for the shortest time and reviewed often. The new cohort strengthens the case for that caution and extends the list of harms clinicians and families weigh — pneumonia and acute kidney injury now sit alongside stroke and death [s1]. It also sharpens the timing: if risk is front-loaded into the first weeks, the argument for starting only when non-drug measures have genuinely failed, and for reviewing early, is stronger, not weaker.

None of the sources here supports a decision for or against treatment in any individual, and stopping an antipsychotic abruptly carries its own risks. This article describes what the evidence shows and is not medical advice.

Sources

• Multiple adverse outcomes associated with antipsychotic use in people with dementia: population based matched cohort study — BMJ, 2024-04-17
• Risk of Death With Atypical Antipsychotic Drug Treatment for Dementia: Meta-analysis of Randomized Placebo-Controlled Trials — JAMA, 2005-10-19
• Effectiveness of Atypical Antipsychotic Drugs in Patients with Alzheimer's Disease (CATIE-AD) — New England Journal of Medicine, 2006-10-12