Antipsychotics and stroke risk

“Antipsychotic drug ‘stroke risk’” is the headline on the BBC News website. A study has found that all forms of antipsychotics boost the risk of stroke in all patients. It reports that research in 2002 raised concerns about the harmful effects of atypical antipsychotics (a newer generation of antipsychotic drug) in people with dementia; since then drug watchdogs have recommended they not be used in this patient group. Over time, it has been suggested that the studies which demonstrated this link may have been affected by confounders (that is, that other differences between patients, rather than the drugs, explained the results). This study confirms that this is unlikely to be the case, and supports the recommendation that atypical antipsychotics should not be used by people with dementia.

Importantly, this study alone cannot conclude that antipsychotics increase the absolute risk of stroke compared with not taking antipsychotics (as some news reports imply), because it only looked at those people who ended up having a stroke. The study suggests that patients who do end up having a stroke are more likely to have it while taking antipsychotics than when they are not. Based on the fact that people with dementia are more likely to have a stroke than people without dementia, and in light of previous studies, the researchers conclude that antipsychotics, and in particular atypical antipsychotics, should if possible be avoided in patients with dementia.

Where did the story come from?

Drs Ian Douglas and Liam Smeeth from the London School of Hygiene and Tropical Medicine carried out this study. Dr Smeeth is supported by a research fellowship from the Wellcome Trust. It was published in the peer-reviewed British Medical Journal .

What kind of scientific study was this?

This study was described by the researchers as a ‘within person case series’. By this they mean that they compared the effects of the medication on stroke risk in individual patients during periods when they were using antipsychotics with the risk in periods when they weren’t using antipsychotics. The study was conducted to work out whether results from previous research may have been due to differences between the patients enrolled (i.e. unmeasured confounding such as a difference in risk of cardiovascular events at baseline) and to see whether there were differences in stroke risk between users of typical and atypical antipsychotics. The researchers were also interested in assessing whether the risk of stroke was different in people with dementia than in patients using antipsychotics for other diagnoses.

Anonymous patient data came from a large database of over 6 million British adults called the GP Research Database (GPRD). This records continuous information from adults registered with more than 400 GP practices in this country. Consultations, diagnoses, prescribed medication and demographic data are recorded in the database. Data from the GPRD has been used in many studies and it is described as being representative of the population of England and Wales and of the UK in terms of age and sex.

The patients of interest for this study were:

Enrolled in the database prior to 2003.
Had an incident (first diagnosis) stroke in the 12 months after they were first registered in the database and before December 2002.
Had been prescribed at least one antipsychotic medicine before December 2002.

The prescriptions for all antipsychotics were identified for all patients. Researchers used information on the size of the drug pack and on dosing frequency to determine the length of time that the patient was likely to have taken the antipsychotic after being prescribed it. They then divided the follow-up time of each individual patient into periods when they were ‘exposed’ (taking antipsychotics) and ‘unexposed’ (when they weren’t taking antipsychotics). As data was not available on when exactly the patients stopped taking antipsychotics, the ‘exposed’ category included a period of up to 175 days on top of the likely dosing schedule to account for the time taken to return to a fully unexposed state.

To determine the effect of exposure on stroke risk, researchers assessed the rate ratio (defined as the ratio of stroke events in exposed periods to stroke events in unexposed periods) overall, and also compared this between different types of antipsychotic, and between people with and without dementia.

What were the results of the study?

Overall, stroke among all patients was 1.7 times more common during an ‘exposed’ period compared with an ‘unexposed’ period. This result was statistically significant (95% CI 1.6 to 1.9).

For all patients, typical antipsychotics increased rates of stroke by 1.7 times while atypical antipsychotics increased rates by 2.3 times. In patients with dementia (1,423 in total), exposure to any antipsychotic increased rate of stroke by 3.5 times and by 1.4 times in people without dementia.

The atypical antipsychotics appeared to increase stroke risk more in people with dementia with an increase in rate of 5.9 compared with an increase of 3.3 with typical antipsychotics. All of these results were statistically significant and the researchers note that the difference in stroke rates between exposed and unexposed groups fell towards zero after treatment.

What interpretations did the researchers draw from these results?

The researchers say that results from previous studies which have linked antipsychotic use with increased risk of stroke are not due to differences in the baseline cardiovascular risk between patients. They conclude this because their study used a ‘within individual’ design which eliminates potential confounding due to differences between individuals. Atypical antipsychotics increased risk slightly more than typical ones and the risk is “more than twice as great among people with dementia compared with those without”.

What does the NHS Knowledge Service make of this study?

This retrospective study uses a self controlled case series design. As the researchers note, the benefit of this design is that cases act as their own controls and factors (which don’t vary over time) are accounted for. Differences between patients at baseline also become irrelevant. On this basis, the results support the conclusion of previous studies finding that an increased risk of stroke with use of antipsychotics was probably not confounded by differences in cardiovascular risk between patients at baseline.

Some other points to highlight:

In studies that rely on records, there is an obvious concern about quality of underlying data. The researchers note that the validity of the data in the GPRD has been shown to be consistently high and the use of detailed prescription data that is recorded by GP practices meant that recall bias (of relying on someone to remember their prescription) was not an issue.
An editorial which accompanies this publication suggests that the greatest weakness with these types of studies is if the likelihood of being exposed is affected by some event in the past. For example in this case, if having had a stroke meant that patients were less likely to continue taking antipsychotics or to be re-prescribed them. The researchers have attempted to minimise this potential bias by not including patients having a stroke after December 2002. Prescribing patterns may have changed after this point because around this time the first major concerns about the use of antipsychotics in patients with dementia emerged.
The researchers note another potential weakness: their inability to control for within patient confounders, that is, factors that change over time and may increase stroke risk alongside use of antipsychotics. As an example they say that initiation of antipsychotics may be linked with a change in an independent risk factor for stroke, such as smoking.
The study only looked at people who ended up having a stroke. By itself, therefore, it cannot draw any conclusions about the effects of antipsychotics in people who never end up with a stroke. In other words, it cannot make any conclusions about the effects of antipsychotics on the absolute risk of stroke in a patient.

Notwithstanding the limitations highlighted above, this study confirms that among people who end up having a stroke, it is more likely to happen while taking antipsychotics, particularly atypical antipsychotics. In light of the increased stroke risk in people with dementia, and in light of previous studies, the authors conclude that prescriptions should be avoided wherever possible in this patient group. Importantly, the researchers note that there is a “much more modest” link between use of antipsychotics and stroke in people who don’t have dementia and, in these patients, their use may be acceptable. Health professionals will always ensure that prescriptions take into account all potential risks and benefits.

Sir Muir Gray adds...

All medicine can do harm as well as good; the more powerful the potential for benefit, the more powerful the potential for harm, unfortunately, so careful research always has to focus on both.