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
Primary Subject Public health Heading:
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. HMG CoA reductase), therapeutic subgroup (e.g. 50 51 lipid modifying agents) and anatomical group (e.g. cardiovascular system). Compound medications were 52 divided into single chemical substances (e.g. CoAprovel into irbesartan and hydrochlorothiazide). In UK 53 Biobank, diagnosis was recorded according to hierarchical tree systems. For analysis, Level 2 diagnosis 54 was used for non cancer illness (UK Biobank data coding 6: e.g. Top level = cardiovascular, Level 1 = 55 cerebrovascular disease, Level 2 = stroke); top level diagnosis was used for cancer (UK Biobank data 56 coding 3: e.g. breast cancer, skin cancer). Further details of the UK Biobank procedures are available at 57 58 the official website (http://www.ukbiobank.ac.uk). Demographic variables beyond age and gender included 59 household income, education and ethnicity. UK Biobank assessment centre was also recorded. 60 4 For peer review only − http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 5 of 16 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-012177 on 30 November 2016. Downloaded from Statistical analyses 1 2 Analyses were conducted using linear modelling for each cognitive test and medication combination 3 separately. For each cognitive test medication combination, the study group identified for analysis was all 4 participants with complete cognitive and covariate data. The modelling strategy was to relate cognitive 5 6 performance to medication with adjustment for confounding by co morbidity (diagnosis), other medications, 7 relevant demographics and assessment centre. To simplify the problem, for each medication of interest, the 8 10 diagnoses and the 10 medications which were most strongly co occurring (largest odds ratios) with the 9 medication of interest, were identified and added as co variants in the linear model.t. 10 11 Cognitive outcomes and age were modelled as continuous variables. Sex, medication and 12 diagnosis were modelled as two level factors. Education was modelled as a 5 level factor (A levels and 13 above, O levels or equivalent, professional qualifications only, none of the above, prefer not to answer), 14 15 and annual householdFor income aspeer a 7 level factor review (£<18K, 18 30.9K, 31K 51.9K,only 52 100K, >100K, do not 16 know, prefer not to answer).Full adjustment was made in all models. Analyses were conducted at three 17 medication classification levels (chemical substance, chemical subgroup and therapeutic subgroup) 18 separately and are presented according to functional system (e.g. nervous, cardiovascular, gastrointestinal 19 and metabolism, immune, others). Results are presented as effect sizes with 95% confidence intervals. 20 21 To detect an effect size of 0.25 SD with 80% power at a significance level of 0.05, chemical 22 substance or its subgroups taken by 125 or more participants were required. This criterion was met for 368 23 24 chemical substances, 165 chemical subgroups and 68 therapeutic subgroups. The statistical power of a 25 medication used by ‘n’ UK Biobank participants was approximated as the power of a binomial test with two 26 samples of unequal size – one sample of size ‘n’ and the other of size equal to the total number of 27 participants minus ‘n’.[7] False Discovery Rate was used to correct for multiple comparisons. 28 29 30 31 Results 32 33 Table 1 presents demographic characteristics of UK BioBank participants at baseline assessment that were
34 included into the statistical analyses. Figure 1 show the effect size for the 71 medicines that show some http://bmjopen.bmj.com/ 35 evidence of independent association with cognitive performance. Results for all 368 medicines may be 36 37 found in supplementary materials Appendix 1. Most drugs were not associated with significant changes in 38 any of the cognitive tests (262 of 368) after adjusting for potential confounding factors. Among those that 39 were, the majority associated with poorer cognitive performance i.e. fewer correct answers on the 40 reasoning test, more errors on the memory test, and longer RT. 41 42 Medications prescribed for nervous system disorders showed the strongest associations with on September 26, 2021 by guest. Protected copyright. 43 cognitive performance relative to the other systems. For example, users of users of levetiracetam, 44 toparimates and dopamine derivates had one more error on the memory test, and users of risperidone, 45 46 benzamides and primidone had a 50ms slower mean RT when compared to non users. Conversely 47 dopamine agonists, often prescribed for early Parkinson’s disease, were associated with faster RT (effect 48 size ± 95% confidence interval: 0.02±0.28/ 0.30±0.29/ 35±0, for reasoning, memory and RT, 49 respectively).Among antipsychotics, risperidone were associated with slower RT (effect size ± 95% 50 confidence interval: 0.55±0.40/0.36±0.41/37±1, for reasoning, memory and RT, respectively). Among 51 anti epileptics, topiramate was associated with poor cognitive scores ( 0.75±0.41/1.47±0.38/7± 2) as was 52 53 levetiracetam ( 0.09±0.35/0.81±0.38/34±1). 54 55 Among cardiovascular medications, first line antihypertensives such as ACE inhibitors were 56 associated with more correct responses in the reasoning test (e.g. perindopril, 0.17±0.08). However, beta 57 blocking agents, calcium blocking channels (e.g. amlodipine, 0.13±0.05/0.10±0.05/5±1) and some diuretics 58 (e.g. furosemide, 0.13±0.11/ 0.07±0.10/10±1) were associated with poorer cognitive function. 59 60 5 For peer review only − http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 6 of 16 BMJ Open: first published as 10.1136/bmjopen-2016-012177 on 30 November 2016. Downloaded from In the gastroinstestinal (GI) tract and metabolism system, drugs for diabetes were adversely 1 associated with cognitive performance. Insulin was related to slower RT ( 0.06±0.11/0.11±0.11/15±1). 2 Proton pump inhibitors (PPI) were adversely related to reasoning, memory and RT (e.g. omeprazole, 3 4 0.15±0.04/0.10±0.04/7±1). But these effects were not found in H2 receptor antagonists. Laxatives 5 (including contact and osmotic but not bulk form laxatives) were associated with slower RT. 6 7 Immunomodulating medications were among the very few drugs that showed beneficial cognitive 8 associations. For example, ibuprofen was related to significantly better reasoning test and RT (0.07±0.03/ 9 0.05±0.03/ 5±0), as were some cancer medications when tested at the subgroup level (e.g. aromatase 10 inhibitors in reasoning, 0.30±0.15). Similar but non significant beneficial trends were found in anti 11 inflammatory/antineoplastic medication such as meloxicam, methotrexate and TNF alpha inhibitors. 12 13 However these effects were not found in other anti inflammatory medications e.g. diclofenac and 14 corticosteroids, and were reversed in some non inflammatory analgesics (e.g. paracetamol, 15 0.20±0.03/0.10±0.03/6±1).For peer review only 16 17 Some health supplements (e.g. glucosamine, 0.12±0.05/ 0.08±0.05/ 6±0) showed beneficial 18 cognitive association. But supplements used for anaemia (e.g. iron 0.24±0.16/0.12±0.16/0.93±5.27) or 19 osteoporosis (e.g. calcium in reasoning and RT, 0.05±0.12/ 0.10±0.11/2± 2) were related to poor cognitive 20 function. 21 22 23 24 Discussion 25 26 This analysis of a very large volunteer cohort finds considerable evidence for an association between 27 28 commonly prescribed medications and cognition within the normal range. Of interest is whether these 29 associations reflect an effect of the medication on cognition or reflect a wider causal context, such as an 30 effect of the underlying disorder, comorbidity or socioeconomic context. 31 32 Examples of associations where there is an a priori reason to suspect a primary relationship with an 33 underlying disease includes antipychotics, drugs used to treat Parkinson’s disease and antiepileptics. In
34 each case these classs of compound are most commonly used for diseases known to have an element of http://bmjopen.bmj.com/ 35 cognitive impairment as part of the core syndrome [8–10] although, in addition, antipsychotics and 36 37 antiepileptics are known to have independent effects on cognition.[11–14] 38 39 In other cases we report an association with medications where there is a plausible association with 40 a cognitive component of the underlying disease but where cognition is not a necessary or well established 41 part of the underlying syndrome. Examples include anti diabetes medications (e.g. insulin and 42 sulfonylurea), antihypertensives (e.g. calcium channel blockers and diuretics) and anaemia medications. on September 26, 2021 by guest. Protected copyright. 43 Both diabetes and hypertension are associated with cognitive impairment.[15,16]The fact that we find an 44 association between drugs used to treat anaemia iron and thiamine and poor cognition does lend weight 45 46 to those studies that do suggest that poor cognition is associated with anaemia.[17] It is noteworthy 47 however, that most of these studies reported such an association in the elderly and the majority of 48 participants analysed in the current study are between 40 and 70 years old. These findings suggest that 49 disorders known to have a relationship with pathological cognitive impairment and dementia in the elderly 50 might also have a relationship with poor cognitive performance within the normal range and in middle aged 51 people. 52 53 However some of the associations between poor cognition and medication we find do not have a 54 55 clear link with underlying diagnosis and have not received much attention until recently. Examples here 56 include PPIs, calcium/vitamin D3 and laxatives. Recent studies have found that PPIs were associated with 57 an increased risk of dementia.[18,19] Low blood vitamin D level and osteoporosis, the most common 58 indication for calcium and vitamin D3, were also correlated with an increased risk of dementia.[20–23] In 59 60 6 For peer review only − http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 7 of 16 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-012177 on 30 November 2016. Downloaded from the present study, participants in their middle age taking PPIs, calcium or vitamin D3 showed poorer 1 cognitive performance compared to non takers. Additionally many laxatives showed a strong association 2 with increased RT. It is unclear whether these associations were due to the side effects of medications or 3 4 underlying syndromes for which they were prescribed. However clinicians may need to consider having a 5 careful discussion with their patients about the potential ‘cognitive’ side effects of these medications. 6 7 Perhaps not surprisingly, we find very few associations between drugs and better cognitive 8 performance. However, it is interesting that there is a small but significant effect of Ibuprofen and 9 comparable but non significant trends with other anti inflammatories including meloxicam and 10 11 methotrexate. There is an important role of inflammation in Alzheimer’s disease and other dementias [24] 12 and these data might be in line with the known effect of anti inflammatory medication including nonsteroidal 13 anti inflammatory medication in reducing incidence of dementia [25,26]. Nonetheless, if this is an 14 explanation of the association we observe, then it is remarkable that the effect is detectable within the 15 normal range of cognitionFor and inpeer people in mid life review and before the onset only of dementia. Alternatively, it might 16 be that these anti inflammatory medications have a dementia independent beneficial effect on cognition 17 18 and that the previous epidemiological evidence suggesting a protective effect in relation to dementia was a 19 reflection of such an effect. This is clearly a critical important distinction as is the observation that the effect 20 we observe is not present in all anti inflammatory medications including, for example acetic acid derivatives 21 and fenamates. 22 23 In terms of the broader causal environment, attempts were made to control for the effects of 24 25 underlying disorder, comorbidity and socioeconomic context. That consistent patterns of association were 26 not always found within chemical or therapeutic subgroups argues against all the associations detected 27 being confounded. Nevertheless, residual confounding cannot be discounted and more detailed analyses 28 using diverse datasets are required for specific causal pathways to be identified. Of particular interest 29 would be longitudinal data demonstrating an impact of medication on cognitive change. Nevertheless, 30 these data illustrate the opportunity and the challenges provided by real world data to investigate the 31 32 broader impact of medicines. 33 In summary, using a very large volunteer cohort, we find multiple associations between a range of 34 http://bmjopen.bmj.com/ 35 medications and poorer cognition, albeit within the normal range. Many of these associations are likely to 36 reflect the underlying disease or syndrome for which the medication is prescribed. Nonetheless it is 37 remarkable that this is apparent in relatively young and relatively healthy volunteers. In other cases the link 38 to underlying disease is less obvious and even where there is a putative link to the underlying disease we 39 40 do not find association with cognition with all drugs used to treat that disease. These findings raise the 41 possibility that some drugs may have hitherto unsuspected negative effects on cognition and a very few 42 compounds, most especially anti inflammatory agents, might even have a positive effect. Further replication on September 26, 2021 by guest. Protected copyright. 43 and extension studies are warranted and will become increasingly possible as large cohort data and 44 routine, real world data from clinical records become more widely available. 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 7 For peer review only − http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 8 of 16 BMJ Open: first published as 10.1136/bmjopen-2016-012177 on 30 November 2016. Downloaded from This research has been conducted using the UK Biobank resource. 1 2 3 4 Contributors: ANH, CHK, LW, JG and SL contributed to study design and wrote the manuscript; ANH, 5 CHK and LW contributed to data acquisition; ANH and CHK conducted data processing and statistical 6 analyses; SL is the guarantor. 7 8 9 10 Funding: We received funding for this analysis through the Medical Research Council Dementias Platform 11 UK (MR/L015382/1) and the Wellcome Trust Consortium for Neuroimmunology for Mood Disorders and 12 13 Alzheimer’s Disease (104025/Z/14/Z). The funders had no influence on study design, data collection and 14 analysis, decision to publish, or preparation of the manuscript. 15 For peer review only 16 17 18 Competing interests: All authors have completed the ICMJE uniform disclosure form at 19 www.icmje.org/coi_disclosure.pdf and declare: no support from any organisation for the submitted work ; 20 no financial relationships with any organisations that might have an interest in the submitted work in the 21 previous three years; no other relationships or activities that could appear to have influenced the submitted 22 work. 23 24
25 26 Ethical approval: This study has been conducted using UK Biobank data. UK Biobank has received ethics 27 28 approval from the UK Biobank Research Ethics Committee (ref 11/NW/0382). 29 30 31 32 Transparency statement: The lead author (the manuscript’s guarantor) affirms that this manuscript is an 33 honest, accurate, and transparent account of the study being reported; that no important aspects of the
34 study have been omitted; and that any discrepancies from the study as planned (and, if relevant, http://bmjopen.bmj.com/ 35 registered) have been explained. 36 37 38 39 Data sharing: No additional data are available.. 40 41 42 on September 26, 2021 by guest. Protected copyright. 43 Copyright The Corresponding Author has the right to grant on behalf of all authors and does grant on 44 behalf of all authors, a worldwide licence to the Publishers and its licensees in perpetuity, in all forms, 45 formats and media (whether known now or created in the future), to i) publish, reproduce, distribute, display 46 and store the Contribution, ii) translate the Contribution into other languages, create adaptations, reprints, 47 48 include within collections and create summaries, extracts and/or, abstracts of the Contribution, iii) create 49 any other derivative work(s) based on the Contribution, iv) to exploit all subsidiary rights in the Contribution, 50 v) the inclusion of electronic links from the Contribution to third party material where ever it may be located; 51 and, vi) licence any third party to do any or all of the above. All research articles will be made available on 52 an Open Access basis (with authors being asked to pay an open access fee. The terms of such Open 53 Access shall be governed by a Creative Commons licence—details as to which Creative Commons licence 54 55 will apply to the research article are set out in our worldwide licence referred to above. 56 57 58 59 60 8 For peer review only − http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 9 of 16 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-012177 on 30 November 2016. Downloaded from References 1 2 1 Chung F H, Chiang Y R, Tseng A L, et al. Functional Module Connectivity Map (FMCM): a framework 3 for searching repurposed drug compounds for systems treatment of cancer and an application to 4 colorectal adenocarcinoma. PloS One 2014;9:e86299. doi:10.1371/journal.pone.0086299 5 6 2 Corbett A, Pickett J, Burns A, et al. Drug repositioning for Alzheimer’s disease. Nat Rev Drug Discov 7 2012;11:833–46. doi:10.1038/nrd3869 8 9 3 Corbett A, Williams G, Ballard C. Drug repositioning in Alzheimer’s disease. Front Biosci Sch Ed 10 2015;7:184–8. 11 12 4 Sudlow C, Gallacher J, Allen N, et al. UK Biobank: An Open Access Resource for Identifying the 13 Causes of a Wide Range of Complex Diseases of Middle and Old Age. PLoS Med 2015;12:e1001779. 14 doi:10.1371/journal.pmed.1001779 15 For peer review only 16 5 Hagenaars SP, Harris SE, Davies G, et al. Shared genetic aetiology between cognitive functions and 17 physical and mental health in UK Biobank (N=112 151) and 24 GWAS consortia. Mol Psychiatry 18 Published Online First: 26 January 2016. doi:10.1038/mp.2015.225 19 20 6 WHO Collaborating Centre for Drug Statistics Methodology, Guidelines for ATC classification and DDD 21 assignment 2016. Oslo: 2016. http://www.whocc.no/atc_ddd_publications/guidelines/ 22 23 7 Cohen J. Statistical Power Analysis for the Behavioral Sciences. L. Erlbaum Associates 1988. 24 25 8 Barch DM, Ceaser A. Cognition in schizophrenia: core psychological and neural mechanisms. Trends 26 Cogn Sci 2012;16:27–34. doi:10.1016/j.tics.2011.11.015 27 28 9 Svenningsson P, Westman E, Ballard C, et al. Cognitive impairment in patients with Parkinson’s 29 disease: diagnosis, biomarkers, and treatment. Lancet Neurol 2012;11:697–707. doi:10.1016/S1474 30 4422(12)70152 7 31 32 10 Berg AT. Epilepsy, cognition, and behavior: The clinical picture. Epilepsia 2011;52 Suppl 1:7–12. 33 doi:10.1111/j.1528 1167.2010.02905.x
34 http://bmjopen.bmj.com/ 35 11 Bédard MA, Schérer H, Stip E, et al. Procedural learning in schizophrenia: further consideration on the 36 deleterious effect of neuroleptics. Brain Cogn 2000;43:31–9. 37 38 12 Reilly JL, Harris MSH, Keshavan MS, et al. Adverse effects of risperidone on spatial working memory in 39 first episode schizophrenia. Arch Gen Psychiatry 2006;63:1189–97. doi:10.1001/archpsyc.63.11.1189 40 41 13 Mula M, Trimble MR. Antiepileptic drug induced cognitive adverse effects: potential mechanisms and 42 contributing factors. CNS Drugs 2009;23:121–37. on September 26, 2021 by guest. Protected copyright. 43 44 14 Park S P, Kwon S H. Cognitive Effects of Antiepileptic Drugs. J Clin Neurol Seoul Korea 2008;4:99– 45 106. doi:10.3988/jcn.2008.4.3.99 46 47 15 Sadanand S, Balachandar R, Bharath S. Memory and executive functions in persons with type 2 48 diabetes: a meta analysis. Diabetes Metab Res Rev 2016;32:132–42. doi:10.1002/dmrr.2664 49 50 16 Elias MF, Goodell AL, Dore GA. Hypertension and Cognitive Functioning A Perspective in Historical 51 Context. Hypertension 2012;60:260–8. doi:10.1161/HYPERTENSIONAHA.111.186429 52 53 17 Andro M, Le Squere P, Estivin S, et al. Anaemia and cognitive performances in the elderly: a 54 systematic review. Eur J Neurol 2013;20:1234–40. doi:10.1111/ene.12175 55 56 18 Gomm W, von Holt K, Thomé F, et al. Association of proton pump inhibitors with risk of dementia: A 57 pharmacoepidemiological claims data analysis. JAMA Neurol Published Online First: 15 February 2016. 58 doi:10.1001/jamaneurol.2015.4791 59 60 9 For peer review only − http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 10 of 16 BMJ Open: first published as 10.1136/bmjopen-2016-012177 on 30 November 2016. Downloaded from 19 Haenisch B, Holt K von, Wiese B, et al. Risk of dementia in elderly patients with the use of proton pump 1 inhibitors. Eur Arch Psychiatry Clin Neurosci 2014;265:419–28. doi:10.1007/s00406 014 0554 0 2 3 20 Chang K H, Chung C J, Lin C L, et al. Increased risk of dementia in patients with osteoporosis: a 4 population based retrospective cohort analysis. AGE 2013;36:967–75. doi:10.1007/s11357 013 9608 x 5 6 21 Littlejohns TJ, Henley WE, Lang IA, et al. Vitamin D and the risk of dementia and Alzheimer disease. 7 Neurology 2014;:10.1212/WNL.0000000000000755. doi:10.1212/WNL.0000000000000755 8 9 22 Miller JW, Harvey DJ, Beckett LA, et al. VItamin d status and rates of cognitive decline in a multiethnic 10 cohort of older adults. JAMA Neurol 2015;72:1295–303. doi:10.1001/jamaneurol.2015.2115 11 12 23 Zhou R, Zhou H, Rui L, et al. Bone loss and osteoporosis are associated with conversion from mild 13 cognitive impairment to Alzheimer’s disease. Curr Alzheimer Res 2014;11:706–13. 14 15 24 Heneka MT, CarsonFor MJ, Khoury peer JE, et al. Neuroinflammationreview in Alzheimer’sonly disease. Lancet Neurol 16 2015;14:388–405. doi:10.1016/S1474 4422(15)70016 5 17 18 25 Zandi PP, Anthony JC, Hayden KM, et al. Reduced incidence of AD with NSAID but not H2 receptor 19 antagonists: the Cache County Study. Neurology 2002;59:880–6. 20 21 26 Szekely CA, Breitner JCS, Fitzpatrick AL, et al. NSAID use and dementia risk in the Cardiovascular 22 Health Study: role of APOE and NSAID type. Neurology 2008;70:17–24. 23 doi:10.1212/01.wnl.0000284596.95156.48 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 10 For peer review only − http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 11 of 16 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-012177 on 30 November 2016. Downloaded from
Participants completed cognitive test 1 All
2 participants Verbal-numerical reasoning Memory Reaction time 3 (n = 502647) (n = 165493) (n = 482766) (n = 496813) 4 5.98 ± 2.16 4.25 ± 3.36 559 ± 118 Cognitive test score 5 (max. correct of 13) (no. of errors) (ms) 6 Age (years) 57.03 ± 8.09 57.20 ± 8.14 56.96 ± 8.08 57.02 ± 8.09 7 Gender (women) 55.0% 54.5% 54.5% 54.4% 8 Education 9 A levels and above 43.6% 45.7% 44.0% 43.4% 10 Other educational qualifications 33.2% 33.9% 33.4% 33.0% 11 Professional qualifications only 5.2% 5.2% 5.2% 5.1% 12 None of the above 17.2% 14.3% 16.4% 16.7% 13 Prefer not to answer 0.9% 0.9% 0.9% 1.7% Household income (GBP/year) 14 < 18,000 19.6% 18.8% 19.2% 19.2% 15 18,000 30,999 For peer21.8% review22.0% only21.9% 22.2% 16 31,000 51,999 22.3% 22.6% 22.6% 22.2% 17 52,000 100,000 17.4% 17.9% 17.7% 17.4% 18 > 100,000 4.6% 5.1% 4.7% 4.6% 19 Do not know 4.3% 3.9% 4.0% 4.1% 20 Prefer not to answer 10.0% 9.4% 9.7% 9.9% 21 22 Table 1. Baseline characteristics of participants The table shows summary statistics of all UK Biobank 23 participants (first column) and the participants who completed each cognitive test during their baseline visit 24 25 (second to fourth columns). Values indicate mean ± standard deviation for cognitive test score and age, or 26 percentages for education and household income. 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 11 For peer review only − http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 12 of 16 BMJ Open: first published as 10.1136/bmjopen-2016-012177 on 30 November 2016. Downloaded from
1 2 3 4 5 6 7 8 9 10 11 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 Figure 1. Medication use and cognitive performance according to anatomical system A summary of 49 selected medications are presented here (for all medications see Appendix 1). As analyses were conducted 50 at three medication classification levels (see Methods), numbers in parenthesis next to a medication name 51 52 on the left indicate the classification level of the medication: (1) chemical substance, (2) chemical subgroup 53 and (3) therapeutic subgroup. Relationship between medication and each cognitive test are represented by 54 effect size and 95% confidence intervals across three columns: verbal reasoning test (number of correct 55 responses), memory test (number of errors) and reaction time test (ms). The corresponding p values after 56 False Discovery Rate correction for multiple comparisons were indicated by filled circles with red line (p < 57 .005), empty circle (p < .05) and ‘x’ with black line (p > .05) *Abbreviations: ACE = angiotensin converting 58 59 enzyme, TNF = tumour necrosis factor, GI = gastroinstestinal. 60 12 For peer review only − http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 13 of 16 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-012177 on 30 November 2016. Downloaded from
1 2 Legend for Appendix 1 3 4 All medications are presented according to the three classification levels used for analysis (i.e. chemical 5 substance, chemical subgroup and therapeutic subgroup) and aggregated by their anatomical groups (see 6 Methods). Numbers in parenthesis next to a medication name on the left indicate the classification level of 7 the medication: (1) chemical substance, (2) chemical subgroup, (3) therapeutic subgroup and (4) 8 9 anatomical group. Some extra formatting is included in the Y axis to ease visual inspection of the drug 10 classification tree. First, all labels are sorted such that, for any given category, the descendants in the 11 classification tree appear below. Secondly, an arrow is included besides each label such that its length 12 represents how deep the drug or drug group is in the classification tree. Namely, no arrow for the 13 anatomical group, “<