For Peer Review Only

BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-012177 on 30 November 2016. Downloaded from Commonly prescribed drugs associated with cognition: a cross-sectional study in UK Biobank ForJournal: peerBMJ Open review only Manuscript ID bmjopen-2016-012177 Article Type: Research Date Submitted by the Author: 06-Apr-2016 Complete List of Authors: Nevado-Holgado, Alejo; University of Oxford, Psychiatry Kim, Chi-Hun; University of Oxford, Psychiatry Winchester, Laura; University of Oxford, Psychiatry Gallacher, John; University of Oxford, Psychiatry Lovestone, Simon; University of Oxford, Psychiatry <b>Primary Subject Public health Heading</b>: Secondary Subject Heading: Mental health, Pharmacology and therapeutics, Neurology Keywords: PUBLIC HEALTH, MENTAL HEALTH, Cognition, UK biobank http://bmjopen.bmj.com/ on September 26, 2021 by guest. Protected copyright. For peer review only − http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 1 of 16 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-012177 on 30 November 2016. Downloaded from 1 2 3 Commonly prescribed drugs associate with cognition: a cross-sectional study 4 in UK Biobank 5 6 7 8 Authors 9 10 Alejo J Nevado-Holgado*, Chi-Hun Kim*, Laura Winchester, John Gallacher, Simon Lovestone 11 12 13 14 15 Address For peer review only 16 17 Department of Psychiatry, University of Oxford, Warneford Hospital, Oxford OX3 7JX, UK 18 19 20 21 22 23 Authors’ names and positions 24 25 Alejo J Nevado-Holgado*: Postdoctoral researcher 26 27 Chi-Hun Kim*: Postdoctoral researcher 28 29 Laura Winchester: Postdoctoral researcher 30 31 John Gallacher: Professor 32 33 Simon Lovestone: Professor 34 http://bmjopen.bmj.com/ 35 *These authors contributed equally to this work. 36 37 38 39 40 Correspondence to: S Lovestone [email protected] 41 42 on September 26, 2021 by guest. Protected copyright. 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 1 For peer review only − http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 2 of 16 BMJ Open: first published as 10.1136/bmjopen-2016-012177 on 30 November 2016. Downloaded from Abstract 1 2 Objective To investigate medications associated with cognitive function 3 4 Design Population-based cross-sectional cohort study 5 6 Setting UK Biobank 7 8 Participants UK Biobank participants aged 37-73 years who completed cognitive tests at the baseline visit 9 in 2006-2010 10 11 Main outcome measures Cognitive test outcomes on verbal-numerical reasoning test (n=165493), 12 memory test (n=482766) and reaction time test (n=496813). 13 14 Results Most drugs were not associated with significant changes in any of the cognitive tests (262 of 368) 15 For peer review only 16 after adjusting for age, gender, education, household income, ethnicity, assessment centre and concurrent 17 diagnoses and other medications. Among those that were, the majority were associated with poorer 18 cognitive performance. Drugs used for nervous system disorders showed the largest effect size with a 19 direction of effect associated with poorer cognition (effect size ± 95% confidence interval) for reasoning (as 20 assessed by number of corrects), memory (assessed by number of errors) and reaction time (ms) 21 respectively: antipsychotics e.g. risperidone -0.55±0.40/0.36±0.41/37±1, antiepileptics e.g. topiramate - 22 23 0.75±0.41/1.47±0.38/7±-2. Other drugs used for non-nervous system conditions also showed significant 24 negative association with cognitive score, including those where such an effect might have been predicted 25 (antidiabetics e.g. insulin -0.05±0.11/0.11±0.11/14±1, antihypertensive e.g. amlodipine - 26 0.13±0.05/0.10±0.05/5±1) and others where such an association was a surprising observation (proton 27 pump inhibitors e.g. omeprazole -0.15±0.04/0.10±0.04/7±1, calcium - -0.12±0.08/-0.01±0.08/6±1). Finally, 28 only a few medications and health supplements showed association towards a positive effect on cognition 29 30 (anti-inflammatory agents e.g. ibuprofen 0.07±0.03/-0.05±0.03/-5±0, glucosamine 0.12±0.05/-0.08±0.05/- 31 6±0). 32 33 Conclusions In this large volunteer study, some commonly prescribed medications in the UK were 34 associated with poor cognitive performance. Some associations may reflect underlying diseases for which http://bmjopen.bmj.com/ 35 the medications were prescribed, although the analysis controlled for the possible effect of diagnosis. 36 Other drugs, whose association cannot be linked to the effect of any disease, may need vigilance for their 37 implications in clinical practice. 38 39 40 41 42 on September 26, 2021 by guest. Protected copyright. 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 2 For peer review only − http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 3 of 16 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-012177 on 30 November 2016. Downloaded from Strengths and limitations of this study 1 2 • Using a very large UK population cohort, this study was sufficiently powered to perform a systematic 3 investigation for a range of medications and its association with cognition. 4 5 • We report multiple associations between commonly prescribed medications and poorer or better 6 cognitive performance, which may have implications for clinicians and patients. 7 • Due to the cross-sectional design, it is difficult to make causal inferences between medication and 8 cognition. 9 • Although trained nurses interviewed participants to obtain medications and diagnoses, this self- 10 11 reported nature may limit the accuracy of information. 12 13 14 15 For peer review only 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 http://bmjopen.bmj.com/ 35 36 37 38 39 40 41 42 on September 26, 2021 by guest. Protected copyright. 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 3 For peer review only − http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 4 of 16 BMJ Open: first published as 10.1136/bmjopen-2016-012177 on 30 November 2016. Downloaded from Introduction 1 2 A significant number of drugs are used for therapeutic indications other than those they were either 3 designed, or first used for. Well known examples include Viagra, developed for cardiac indications but used 4 for erectile failure, and aspirin, used in an increasingly wide range of indications. This phenomenon has 5 6 prompted systematic efforts to repurpose compounds already in clinical practice.[1,2] In the field of 7 cognitive disorders, repurposing approaches have included consensus building using understanding from 8 epidemiology to molecular sciences,[3] and informatics driven approaches using platforms such as the 9 connectivity map,[1] which derives gene expression signatures from in vitro cells exposed to drugs and 10 matches these to disease associated signatures. As part of a process seeking drugs to promote healthy 11 brain ageing, and potentially counter disorders of cognitive decline, as well as to identify compounds that 12 13 might have a detrimental effect on cognition, we have used the UK Biobank [4] dataset to identify drugs 14 associated with cognitive performance. 15 For peer review only 16 17 18 Methods 19 20 UK Biobank is a prospective study of a half million participants across the UK with extensive data from 21 questionnaires, interviews, physical measures, biological samples and imaging.[4] This study used baseline 22 data from 502647 subjects aged between 37 and 73 years collected at 21 assessment centres in 2006- 23 2010. 24 25 Outcome and other measures 26 27 Three cognitive tests, which have been described previously [5], were used as outcome variables. In brief, 28 29 cognitive tests were administered using a touch screen device with trained researchers present for 30 assistance. Memory was assessed using the pairs matching test in which subjects had to remember 6 pairs 31 of shapes and their locations displayed for 5 seconds on the screen. Performance was assessed as the 32 total number of errors made in matching all 6 pairs. For verbal-numerical reasoning, subjects were asked to 33 solve as many multiple choice questions as possible (maximum 13) within two minutes. Performance was 34 http://bmjopen.bmj.com/ assessed as the total number of correct responses. Reaction time (RT) was measured by pressing a button 35 36 as quickly as possible when two identical shapes were presented on the screen. Performance was 37 assessed as the mean RT for correctly identified matching pairs. Participants completed each cognitive test 38 were included for the analysis of that cognitive test: verbal-numerical reasoning test (n=165493), memory 39 test (n=482766) and RT test (n=496813). The verbal-numeric reasoning test was introduced later to the 40 study and the numbers available for analysis are lower. 41 42 Trained nurses interviewed subjects to obtain diagnoses given by doctors and medications. Only on September 26, 2021 by guest. Protected copyright. 43 regular medications and health supplements taken weekly, monthly or three monthly were recorded. 44 45 Medications were recorded using 6745 categories adapted from Read code version 3 (CTV3) used in the 46 general practice in the UK. Of these, 1192 medications were taken by 30 or more participants, and were 47 classified using the Anatomical Therapeutic Chemical (ATC) classification [6] as a backbone. For example, 48 brand names with different doses were allocated into their chemical substance (e.g. Lipitor and atorvastatin 49 were treated as atorvastatin), chemical subgroup (e.g.

Load more