Journals Library

Do antipsychotic medications increase the risk of cardiovascular disease? Cohort study in the THIN database comparing different methods for exploring long term prescriptions in primary care

Project title

Do antipsychotic medications increase the risk of cardiovascular disease? Cohort study in the THIN database comparing different methods for exploring long term prescriptions in primary care

Project reference


Final report date

01 May 2015

Project start date

01 March 2013

Project end date

01 October 2014

Project duration

19 months

Project keywords

Antipsychotic Medication Cardiovascular Disease; Electronic Patient Records; Schizophrenia; Bipolar Disorder; Dementia; Depression; Personality Disorder; Bias; Propensity Score

Lead investigator(s)
  • Professor David Osborn, Faculty of Brain Sciences, UCL
NIHR School Collaborators
  • Professor Irwin Nazareth, Faculty of Population Health Sciences, UCL
  • Dr Kate Walters, Faculty of Population Health Sciences, UCL
  • Dr Irene Petersen, Faculty of Population Health Sciences, UCL
  • Dr Louise Marston, Faculty of Population Health Sciences, UCL (Statistician)

Project objectives

We aimed to answer the following research questions using the THIN UK primary care database:

  1. Is antipsychotic dose and class associated with incident cardiovascular disease?
  2. Does any association with CVD vary according to different methods of categorising the degree of exposure to antipsychotics? 

Changes to project objectives

There was a delay between submitting the original proposal and signing contracts / employing statistical research staff for the project.

During this time there was some criticism of methods used to assess the long term impact of antipsychotics. In particular, papers using Scandinavian databases were criticised for failing to deal with antipsychotics in a sophisticated manner when assessing their impact on mortality. However they were widely cited.

This led us to change and improve our study design so as to minimise the likelihood of bias and the possible misinterpretation of results regarding the benefits and harms of antipsychotics.

Therefore, we commenced with a careful descriptive study of antipsychotic prescribing across national primary care. This work has been published. We then explored some of the problems of trying to model associations between antipsychotic medications and cardiovascular events. Antipsychotics are often prescribed for long periods of time, from early adulthood, with periods when different agents are prescribed at different doses and also with periods when drugs maybe discontinued by the patient with or without their GP. However cardiovascular events occur later in life. There may be multiple changes in antipsychotic over this time and so it is difficult to model any association.

We therefore used our research time to demonstrate these difficulties and to assess whether routine primary care data could be used to make head to head comparisons of antipsychotics, if analyses were restricted to those receiving monotherapy during follow-up.

Our findings are currently being written up for publication and are described below. 

Brief summary

Work completed and published

We have published descriptive work from this project in BMJ Open. The abstract is below (for access to the full article see the dissemination section).


To examine the recorded indication for antipsychotic prescriptions in UK primary care.


Cohort study.


Primary care.


Individuals prescribed antipsychotics between 2007 and 2011.


The proportion of individuals prescribed antipsychotics with a diagnosis of (1) psychosis and bipolar disorder, (2) other diagnoses including depression, anxiety and dementia and (3) none of these diagnoses.


We identified 47 724 individuals prescribed antipsychotic agents. 13 941 received first-generation agents and 27 966 received second-generation agents. The rates of prescribing were higher in females (incidence rate ratio (IRR) 1.092 (95% CI 1.088 to 1.095), older people (80+ vs 40–49; IRR 2.234 (2.222 to 2.246)) and in those from the most deprived areas (most deprived vs least deprived IRR 3.487 (3.567 to 3.606). Of those receiving first-generation antipsychotics, less than 50% had a diagnosis of psychosis/bipolar disorder. For the second-generation agents, the numbers ranged from 4824 (36%) for quetiapine to 7094 (62%) for olanzapine. In patients without psychosis/bipolar disorder, common diagnoses included anxiety, depression, dementia, sleep and personality disorders. For example, in risperidone users, 14% had an anxiety code, 22% depression, 12% dementia, 11% sleep disorder and 4% personality disorder. The median daily doses and duration of treatment were greater in those with schizophrenia (eg, risperidone median daily dose 4 mg; IQR 2–6: median duration 1.2 years) than in those with non-psychotic/bipolar disorders such as depression or anxiety (eg, risperidone 1 mg; IQR 1–2: 0.6 years). A relatively large proportion (between 6% and 17%) of people receiving individual antipsychotics had none of the diagnoses stated above.


In UK primary care, a large proportion of people prescribed antipsychotics have no record of psychotic or bipolar disorder. They are often older people with conditions including dementia, non-psychotic depression, anxiety and sleep disorders.

Ongoing work

We are currently preparing our work regarding the viability of exploring cardiovascular events in people prescribed antipsychotics. A preliminary abstract is included below.


Antipsychotics may confer long term benefits and risks for mortality. Several studies using routine clinical data have explored these associations but confounding by indication impedes interpretation of results. We used large UK general practice database to determine whether relative cardiovascular risks/benefits could be established for the three commonest second generation antipsychotics. We aimed to determine risk of a CVD event in similar groups of people only prescribed a single antipsychotic.


We included all people over 18 prescribed olanzapine, risperidone or quetiapine between 1995 and 2011 in THIN. We excluded people with pre-existing CVD and those prescribed multiple antipsychotics. We created  propensity scores for being prescribed each antipsychotic using socio-demographic, diagnostic, biometric and clinical variables. We used propensity score matching to select similar people, and then calculated the statistical power to detect a difference in risk of new CVD events in head to head comparisons.


We identified 31,751 people ever prescribed these 3 agents of whom 18,319 ( 58%) received a single antipsychotic during follow-up (n=5090 risperidone, 7797 olanzapine and 4613 quetiapine). Propensity score matching identified the following numbers of people suitable for comparing the proportions developing CVD events: olanzapine (n=4557) versus risperidone (n=4753), with a CVD event rate of 2.19% and 2.78% respectively, giving 44% power. For olanzapine (n=3789) versus quetiapine (n=4133), the CVD event rate was 2.14% and 1.98% respectively (table), giving only 8% power.


Even when large clinical databases are used for research into drugs such as antipsychotics, it is difficult to conduct methodologically sounds studies regarding relatively rare events. The statistical power we determined was insufficient to make robust conclusions regarding antipsychotics and CVD events. We will now finalise out results and publish a paper demonstrating the difficulties of establishing long terms possible harm associated with drugs such as antipsychotics which are prescribed for long periods of time. 

Plain English summary

There is a well-documented 10-20 year age difference in mortality between people with severe mental illnesses such as schizophrenia.

While there are many possible reasons for this, such as high rates of smoking, poor diet and sedentary lifestyles, there is also a concern regarding the short term and possible longer term effects of antipsychotic medications.

While these medications are effective for the treatment of psychosis and severe affective disorders, the newer agents also have adverse effects of weight gain and  abnormal metabolism including sugars and fats.


We used a national anonymous primary care database with routine clinical data (THIN) to explore whether we could answer questions regarding possible hard associated with antipsychotics- especially cardiovascular disease.

Firstly we described who received antipsychotics in UK primary care between 1995 and 2011. We then went on to assess whether it was possible to use these data to make ’head to head’ comparisons of different antipsychotics to compare the proportions who developed cardiovascular disease while on the drugs.


We identified large numbers of people who were prescribed antipsychotics in primary care. We were surprised to discover that around half of those receiving these prescriptions did not have a condition for which they are usually recommended. Many people had diagnoses of depression, anxiety, sleep disorders, personality disorders or none of thee in their GP record.

This work has been published in an open access journal (BMJ Open) and attracted attention in the British media as well as social media before Christmas 2014.

Our work regarding CVD events in people receiving different antipsychotics is being written up.

Our preliminary results suggest that even large database such as THIN may not allow us to answer these questions reliably.

When we compared individuals receiving one drug only, eg olanzapine, against people on another drug (eg risperidone) the number of people developing heart attacks or stokes was still too small to allow us to make valid statistical comparisons.


In UK primary care, antipsychotics are prescribed to a range of people who do not have records for a severe mental illness such as psychosis or affective disorders.

Although the databases contain thousands of people receiving these drugs, the numbers who develop conditions like CVD are still small. This makes it difficult to make accurate comparisons of the cardiovascular risk of these drugs using these data. We will publish this work to caution other researchers from making biased or invalid comparisons of these drugs using routine clinical data. 


Published articles

  1. Marston, L., Nazareth, I., Petersen, I., Walters, K., & Osborn, D. P. (2014). Prescribing of antipsychotics in UK primary care: a cohort study. BMJ Open, 4 (12), e006135-?

Planned articles

We also plan to publish a study summarising the difficulties of assessing rarer, distal events such as cardiovascular disease in people prescribed antipsychotics, despite the large numbers of people who receive these medications within database such as The Health Improvement Network. 

Public involvement

We received helpful input regarding our research questions from the UCL/ Camden and Islington NHS Foundation Trust Service user research forum.

We are now beginning to disseminate the results and will re-engage with our PPI advisors for advice.

In other projects we collaborate closely with the McPiN foundation and we would recommend them for involvement in future research regarding mental health- in particular for their commitment to realistic service user involvement, advice and help with dissemination. 


The initial paper from this research has only been published for four months but has high altmetrics (see BMJ Open paper at above doi). 

This project was funded by the National Institute for Health Research School for Primary Care Research (project number 96)

Department of Health Disclaimer

The views and opinions expressed therein are those of the authors and do not necessarily reflect those of the NIHR School for Primary Care Research, NIHR, NHS or the Department of Health.