BMJ

Confidential: For Review Only

The association between anticholinergic medication and dementia incidence varies with pattern of exposure and medication class: A nested case-control study using the Clinical Practice Research Datalink

Journal: BMJ

Manuscript ID BMJ.2017.042315

Article Type: Research

BMJ Journal: BMJ

Date Submitted by the Author: 10-Nov-2017

Complete List of Authors: Richardson, Kathryn; University of East Anglia, ; Fox, Chris; Norwich Medical School, Department of Psychological Sciences Maidment, Ian; University of Aston, Steel, Nicholas; University of East Anglia, Norwich Medical School Loke, Yoon; University of East Anglia, School of Medicine Arthur, Antony; University of East Anglia, School of Health Sciences University of East Anglia, Norwich Research Park myint, phyo; University of Aberdeen, ; University of East Anglia, Norwich Medical School Grossi, Carlota; University of East Anglia, School of Health Sciences Mattishent, Katharina; University of East Anglia, Norwich Medical School Bennett, Kathleen; Royal College of Surgeons in Ireland (RCSI), Population Health Sciences Campbell, Noll; Purdue University College of Pharmacy, Department of Pharmacy Practice Boustani, Malaz; Indiana University, Center for Aging Research Robinson, Louise; Institute for Health and Society, Newcastle University Brayne, Carol; University of Cambridge, Institute of Public Health Matthews, Fiona; Newcastle University, Institute for Health and Society, Biomedical Research Building, Campus for Ageing and Vitality Savva, George; Quadram Institute Bioscience, Analytical Sciences Unit; University of East Anglia, School of Health Sciences

Keywords: dementia, pharmacoepidemiology, anticholinergic, Primary care

https://mc.manuscriptcentral.com/bmj Page 1 of 70 BMJ

1 2 3 The association between anticholinergic medication and dementia incidence varies with 4 pattern of exposure and medication class: A nested case-control study using the Clinical 5 Practice Research Datalink 6 7 Kathryn Richardson, Chris Fox, Ian Maidment, Nicholas Steel, Yoon K Loke, Antony Arthur, Phyo K 8 9 Myint, Carlota M Grossi, Katharina Mattishent, Kathleen Bennett, Noll Campbell, Malaz Boustani, 10 Louise Robinson, Carol Brayne, Fiona E Matthews, George M Savva 11 Confidential: For Review Only 12 Kathryn Richardson, School of Health Sciences, University of East Anglia, Norwich, NR4 7TJ, UK 13 Research Fellow in Statistics 14 Chris Fox, Norwich Medical School, University of East Anglia, Norwich, NR4 7TJ, 15 16 Clinical Professor 17 Ian Maidment, School of Life and Health Sciences, Aston University, Birmingham, B4 7ET, UK 18 Senior Lecturer in Clinical Pharmacy 19 20 Nicholas Steel, Norwich Medical School, University of East Anglia, Norwich, NR4 7TJ 21 Clinical Professor in Public Health 22 Yoon K Loke, Norwich Medical School, University of East Anglia, Norwich, NR4 7TJ 23 Professor of Medicine and Pharmacology 24 25 Antony Arthur, School of Health Sciences, University of East Anglia, Norwich, NR4 7TJ, UK 26 Professor of Nursing Science, 27 Phyo K Myint, School of Medicine and Dentistry, University of Aberdeen, Aberdeen, AB25 2ZD, UK 28 Clinical Chair in Medicine of Old Age, 29 30 Carlota M Grossi, School of Health Sciences, University of East Anglia, Norwich, NR4 7TJ, UK 31 Senior Research Associate, 32 Katharina Mattishent, Norwich Medical School, University of East Anglia, Norwich, NR4 7TJ 33 34 Clinical Research Fellow, 35 Kathleen Bennett, Division of Population Health Sciences, Royal College of Surgeons in Ireland, 36 Dublin, Ireland Associate Professor in Pharmacoepidemiology 37 38 Noll Campbell, Department of Pharmacy Practice, College of Pharmacy, Purdue University, US 39 Assistant Professor 40 Malaz Boustani, School of Medicine, University of Indiana, Indiana United States 41 Professor of Medicine 42 43 Louise Robinson, Newcastle University Institute for Ageing, Newcastle University, NE4 5PL, UK 44 Professor of Primary Care and Ageing 45 Carol Brayne, Professor of Public Health Medicine, Cambrdge Institute of Public Health, University of 46 Cambridge, CB2 0SR, UK 47 48 Fiona E Matthews, Institute for Health and Society, Newcastle University, NE4 5PL, UK 49 Professor of Epidemiology 50 51 George M Savva, School of Health Sciences, University of East Anglia, NR4 7TJ, UK 52 Senior Lecturer in Applied Statistics 53 54 Correspondence to: K Richardson [email protected] 55 Word count: 4999 56 57 58 59 60 https://mc.manuscriptcentral.com/bmj BMJ Page 2 of 70

1 2 3 4 5

6 7 8 Key words: dementia, Alzheimer’s disease, pharmacoepidemiology, case-control study, 9 anticholinergic, medication use, antimuscarinic 10

11 Confidential: For Review Only 12 13 14 "What this paper adds" 15 16 Section 1: What is already known on this subject 17 18 • Use of medications with strong anticholinergic activity is associated with impaired 19 cognition in the short term. 20 21 • Published observational studies have found associations between anticholinergic use 22 and future cognitive decline and dementia incidence, but whether this is causal and 23 24 is directly attributable to anticholinergic activity is still unknown. 25 26 Section 2: What this study adds 27 28 • Our study confirms the previously observed association, but only for specific classes 29 of definite anticholinergics. Anticholinergic antidepressants, antiparkinsonians and 30 urologicals are linked to future dementia incidence, but antihistamines and 31 antispasmodics are not. 32 33 • Use of anticholinergic antidepressants and urologicals are positively associated with 34 dementia diagnosis 15-20 years after the exposure. 35 • Clinical practice and future research to reduce potentially harmful medication should 36 focus on risks associated with specific classes of anticholinergic medication. 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 https://mc.manuscriptcentral.com/bmj Page 3 of 70 BMJ

1 2 3 ABSTRACT 4 5 6 OBJECTIVES: To estimate how duration and level of exposure to different classes of 7 8 anticholinergic medication is associated with subsequent incident dementia. 9 DESIGN: Case-control study 10 11 SETTING:Confidential: General practices in the United For Kingdom Review contributing to the ClinicalOnly Practice 12 13 Research Datalink. 14 PARTICIPANTS: 40,770 patients aged 65-99 years and diagnosed with dementia between 15 16 April 2006 and July 2015, and 283,933 controls without dementia matched 7:1 on date, sex, 17 age, deprivation, and years of data history. 18 19 INTERVENTIONS: Daily defined doses (DDD) of anticholinergic medications coded according 20 to the Anticholinergic Cognitive Burden (ACB) scale; in total and grouped by subclass; 21 22 prescribed between 4-20 years prior to dementia diagnosis. 23 24 MAIN OUTCOME MEASURES: Odds ratios for incident dementia, adjusted for a wide range 25 of demographics and health-related covariates. 26 27 RESULTS: 14,453 (35%) cases and 86,403 (30%) controls were prescribed at least one potent 28 (ACB score 3, ACB3) anticholinergic during the exposure period. The adjusted odds ratio 29 30 (aOR) for ‘any ACB3’ was 1.11 (95% confidence interval 1.08 to 1.14), and an increasing odds 31 of dementia was significantly associated with an increasing average ACB score. When 32 33 considered by class, ACB3 antihistamines or antispasmodics were not significantly linked to 34 35 dementia. Within ACB3 antidepressants, urologicals, and antiparkinsonians, risks increased 36 with exposure with this increase observed for exposure 15-20 years before diagnosis. 37 38 CONCLUSIONS: In the largest study to date, a robust association between some classes of 39 anticholinergic medication use and future dementia incidence was observed. This could be 40 41 caused by a class-specific effect, or by medications being used for very early symptoms of 42 dementia. Future pharmacological and clinical research must examine subgroups of 43 44 anticholinergics as opposed to anticholinergic effects per se or summing scales for 45 46 anticholinergic exposure. 47 48 TRIAL REGISTRATION: EUPAS8705 (ENCePP e-register of studies) 49 50 51 52 53 54 55 56 57 58 59 60 https://mc.manuscriptcentral.com/bmj BMJ Page 4 of 70

1 2 3 Introduction 4 5 Dementia is a leading cause of disability and death worldwide (1), and its prevention is a 6 7 global public health priority. Dementia is caused by a number of different 8 9 neurodegenerative processes that contribute to irreversible cognitive decline and 10 11 Confidential: For Review Only 12 associated symptoms, progressive loss of independence and daily functioning. Mixed 13 14 dementias are more prevalent than is often recognised, with symptoms often more closely 15 16 linked to overall pathological burden as opposed to any specific disease process (2,3). No 17 18 19 disease modifying treatments for dementia exist, however, age-specific dementia incidence 20 21 across populations is declining, suggesting that changing lifestyles or environment may lead 22 23 to a meaningful change in dementia prevalence (4). Hence identifying and reducing 24 25 26 exposure to risk factors that can affect any aspect of long term brain health is important for 27 28 dementia prevention and cognitive health in the population (5). 29 30 31 Multiple medication use is increasing in middle aged and older populations (6,7), but the 32 33 potential harms of long term medication use are not well understood. Medications with 34 35 36 anticholinergic activity (henceforth anticholinergics) block the neurotransmitter 37 38 acetylcholine in the central or peripheral nervous system, and have diverse actions 39 40 depending on site. Anticholinergics are successfully used in the treatment of many 41 42 conditions including urinary incontinence, Parkinson’s disease, depression, epilepsy, 43 44 45 gastrointestinal disorders and to manage allergies. It is well known that anticholinergics 46 47 affect cognition (8), and guidelines suggest they are to be avoided among frail older people 48 49 (9). Use of anticholinergics among people with dementia is recognised as inappropriate by 50 51 52 both Beers (10) and STOPP (11) criteria for potentially inappropriate prescribing. Over the 53 54 past decade, prolonged exposure to anticholinergics has been linked to long term cognitive 55 56 decline or dementia incidence among community living cohorts and nursing home residents 57 58 59 60 https://mc.manuscriptcentral.com/bmj Page 5 of 70 BMJ

1 2 3 (12–17). However, these studies have been limited in their ability to determine if the 4 5 increased risk is specific to anticholinergic action itself, and whether or not the association is 6 7 due to the medications themselves or the underlying conditions for which they were 8 9 10 prescribed. 11 Confidential: For Review Only 12 13 Here we present a nested case-control study using the UK’s Clinical Practice Research 14 15 Datalink (CPRD; 18), in which we select patients with a new dementia diagnosis, and 16 17 compare their prescriptions of anticholinergic medication between 4 and 20 years prior to 18 19 20 dementia diagnosis with that of a matched group of control patients without dementia. Our 21 22 objectives were (a) to estimate the association between chronic anticholinergic medication 23 24 use with future dementia incidence while controlling for potential confounders, (b) to 25 26 27 explore whether any observed effect was specific to particular subclasses of medication, 28 29 and (c) to test how the association varied with time to dementia incidence and amount of 30 31 exposure within each class. 32 33 34 Methods 35 36 37 Study design 38 39 40 We performed a nested case-control study using data from CPRD, which includes 41 42 43 anonymised diagnosis, referral and prescribing records for more than 11.3 million patients 44 45 from 674 primary care practices across the UK, and is broadly representative of the UK 46 47 population in terms of age, sex and ethnicity (18). All coded information in the primary 48 49 50 care record for each selected patient is available to researchers. This includes demographic 51 52 detail, lifestyle information, and any diagnoses, symptoms recorded by the general 53 54 practitioner, referrals to other healthcare services and subsequent findings, and treatments 55 56 57 58 59 60 https://mc.manuscriptcentral.com/bmj BMJ Page 6 of 70

1 2 3 initiated or continued in primary care. In the UK a patient’s registered primary care practice 4 5 is the coordinator of the vast majority of their healthcare, and acts as the main gateway for 6 7 access to secondary care. Hence most of a patient’s health care information, except the 8 9 10 detail of secondary care episodes or privately obtained healthcare, where they have not 11 Confidential: For Review Only 12 been communicated to primary care, is held in their primary care record. An ‘up to 13 14 standard’ (UTS) date is recorded for each CPRD practice, which is the date at which the data 15 16 17 recorded by the practice is considered to be of an acceptable research standard where 18 19 diagnoses are assumed to have been recorded correctly. Recording of therapies and 20 21 diagnoses made prior to UTS date are available, but may be incomplete. 22 23 24 25 26 27 Selection of cases and controls 28 29 30 Patients with a recorded diagnosis of dementia made between April 2006 and July 2015, 31 32 33 and aged 65-99 years were eligible to be selected as cases. We also required at least six 34 35 year of UTS data prior to diagnosis to allow for a 4 year lag before dementia diagnosis, as 36 37 well as at least one year drug exposure period (DEP), and one year before the DEP for 38 39 40 ascertainment of covariates. Dementia diagnosis was defined as the presence in the record 41 42 of any Read codes for dementia as a diagnosis, symptom or referral, or prescription of a 43 44 cognitive enhancer (memantine, donepezil, rivastigmine, galantamine or tacrine) if that was 45 46 47 also followed by a dementia diagnosis code within 12 months. Cases were excluded if they 48 49 had been diagnosed with motor neurone disease, HIV/AIDS, multiple sclerosis, Down 50 51 syndrome or alcohol abuse prior to dementia diagnosis. 52 53 54 55 56 57 58 59 60 https://mc.manuscriptcentral.com/bmj Page 7 of 70 BMJ

1 2 3 Using the dementia diagnosis date as the index date, each case was matched to up to seven 4 5 controls who had not been diagnosed with dementia diagnosis up to the index date. Cases 6 7 and controls were matched on sex, year of birth (within three years), years of UTS data 8 9 10 history, and practice-level index of multiple deprivation (IMD) quintile. We used incidence 11 Confidential: For Review Only 12 density sampling to select controls, hence cases were eligible to be selected as controls for 13 14 other cases with earlier index dates (19). 15 16 17 Patients were excluded if their date of dementia diagnosis was ambiguous. That is, if 18 19 20 “dementia annual review’, “H/O: dementia”, “Assessment of psychotic and behavioural 21 22 symptoms of dementia”, or “Antipsychotic drug therapy for dementia” was recorded before 23 24 the index date. 25 26 27 28 29 30 Anticholinergic medication exposure 31 32 33 34 A drug exposure period (DEP) was defined for each case-control group, starting one year 35 36 after UTS data recording and ending four years before the index date. We excluded 37 38 exposures in the four years prior to index date to avoid protopathic bias, whereby the 39 40 medication is given for a sign or symptom of dementia prior to dementia diagnosis (22). 41 42 43 Estimating the anticholinergic effect of individual medications on the human brain is 44 45 46 difficult, and serum anticholinergic activity as measured via receptor bioassay do not 47 48 correlate well with effects on cognition (16,20), and so the anticholinergic effects of 49 50 medications are usually classified using scales developed by expert consensus aided by 51 52 53 literature review (21). In the current study all medication prescribed to each patient during 54 55 the DEP was classified according to the 2012 update of the Anticholinergic Cognitive Burden 56 57 58 59 60 https://mc.manuscriptcentral.com/bmj BMJ Page 8 of 70

1 2 3 (ACB) scale (23). Medications with serum anticholinergic activity or in vitro affinity to 4 5 muscarinic receptors but with no known clinically relevant negative cognitive effects are 6 7 scored 1, while drugs with established and clinically relevant anti-cholinergic effects are 8 9 10 scored 2 based on blood-brain penetration and 3 if also have reported associations with 11 Confidential: For Review Only 12 delirium. All other drugs are scored 0. For the medications available in the UK in the last 30 13 14 years not rated on the ACB scale, we made the following assumptions. We classified all 15 16 17 thiazide diuretics, loop diuretics and antihistamines that had not been scored as ACB1, all 18 19 tricyclic antidepressants not scored as ACB3, and all creams, eye and ear drops as not having 20 21 anticholinergic activity. 22 23 24 Our primary exposures were the number of defined daily doses (DDDs) prescribed within 25 26 27 each category of ACB score during the DEP. The DDD is defined as the assumed average 28 29 maintenance dose per day for a drug based on its main indication in adults, using the DDD 30 31 values assigned by the World Health Organisation’s (WHO) Collaborating Centre for Drug 32 33 Statistics Methodology. We classified exposure as 0, 1-13, 14-89, 90-364, 365-1459 or 34 35 36 more than 1460 DDDs during the DEP. 37 38 39 We further categorised the DDDs within the class of each medication with anticholinergic 40 41 properties based as either analgesic (WHO Anatomical Therapeutic Chemical [ATC] N02*), 42 43 antidepressant (ATC N06A*), antipsychotic (ATC N05A*), cardiovascular (ATC C*), gastro- 44 45 46 intestinal (ATC A*), Parkinsonian (ATC N04*), respiratory (ATC R*), urological (ATC G04*), or 47 48 other. These categories were pre-specified based on common classes of anticholinergic 49 50 medications. 51 52 53 Finally, an average ACB score was calculated as the average across the DEP of the sum of 54 55 56 ACB score each medication being used at any given time. This was then categorised into 57 58 59 60 https://mc.manuscriptcentral.com/bmj Page 9 of 70 BMJ

1 2 3 groups reflecting approximate quintiles of the sample, but with two additional groups 4 5 representing those with particularly high (3-5 and 5 or more) average anticholinergic scores. 6 7 8 9 10 Confidential: For Review Only 11 Covariates 12 13 14 We considered potential confounders as any variable suspected to be linked to dementia 15 16 17 incidence or an indication for any of the medications examined. For a full list of covariates 18 19 and their definition see the Appendix, but in short we included diagnoses of cardiovascular 20 21 disease, other dementia risk factors and correlates, other indications for anticholinergics, 22 23 24 other medication use, sociodemographic variables and records of health-related lifestyle 25 26 where available. 27 28 29 Exposures could occur at any time during an individual’s DEP, and so as with many case- 30 31 control studies (24) it is not clear at what point we should determine the presence or 32 33 34 absence of confounding factors. Our primary analysis measured confounders recorded up to 35 36 the end of the DEP to best capture the medication indications. However, as some variables 37 38 could be consequences of medication exposure we also coded the status of each patient 39 40 with respect to each covariate up to the start of the DEP. 41 42 43 Statistical analysis 44 45 46 47 Patterns of exposures and covariates were described for case and control groups separately. 48 49 50 Primary analysis compared the number of DDDs of medications with ACB score of 1, 2 and 3 51 52 prescribed to cases and controls during the DEP, controlling for covariates recorded at the 53 54 end of each DEP. We used conditional logistic regression to estimate the association 55 56 57 58 59 60 https://mc.manuscriptcentral.com/bmj BMJ Page 10 of 70

1 2 3 between anticholinergic prescriptions and dementia diagnosis, adjusting for covariates 4 5 described above. Adjusted odds ratios (OR) are reported with 95% confidence intervals (CI), 6 7 however p<0.01 was pre-specified as a threshold for statistical significance owing to large 8 9 10 number of subgroups being examined. All analysis was conducted using Stata version 14. 11 Confidential: For Review Only 12 13 Secondary pre-specified analyses included (i) disaggregation of exposure by drug class and 14 15 (ii) testing the effect of exposure 4-10, 10-15 and 15-20 years prior to the index date in 16 17 those with 6 or more, 11 or more and 16 or more years of UTS data history respectively and 18 19 20 (iii) testing the effect of average ACB score over the DEP. 21 22 23 Sensitivity analyses included repeating our primary analysis (i) adjusted for covariates 24 25 measured up to the start rather than the end of the DEP, and the following post-hoc 26 27 analyses of (ii) emulating a new-user design by excluding patients prescribed medications 28 29 30 with an ACB score of 2 and 3 in the 12 months prior to the DEP (25), and (iii) recoding binary 31 32 exposure variables to correspond to 90 or more DDDs prescribed instead of any prescription 33 34 during the DEP, to represent longer term rather than one-off exposure.. 35 36 37 To test the likely impact of missing lifestyle covariate data, we compared findings with and 38 39 40 without adjustment for these lifestyle variables in (i) complete data and (ii) in the full 41 42 dataset with use of a missing category for each variable. 43 44 45 For a final pre-specified sensitivity analysis, we recoded medications according to the 46 47 Anticholinergic Drug Scale (ADS) instead of the ACB scale. ADS classifies the degree of 48 49 50 anticholinergic activity of each medication on a scale from 0 (no anticholinergic effect) to 3 51 52 (marked anticholinergic effect) according to literature review and expert consensus (26). 53 54 Sample size was determined by the maximum available data within CPRD. 55 56 57 58 59 60 https://mc.manuscriptcentral.com/bmj Page 11 of 70 BMJ

1 2 3 4 5 6 Protocol registration 7 8 9 The study protocol was approved by the Independent Scientific Advisory Committee (ISAC) 10 Confidential: For Review Only 11 for CPRD research (protocol number 15_056R) prior to any data being released to the 12 13 14 research team and was made available to journal reviewers. The study was registered on 15 16 the ENCePP e-register of pharmacoepidemiology studies (register number EUPAS9705). 17 18 19 20 21 22 Patient involvement 23 24 25 Four Alzheimer’s Society Research Network Volunteers acted as study monitors and service 26 27 user representatives on our study steering committee. These individuals have all acted as 28 29 30 carers of people with dementia. The monitors contributed to our protocol development by 31 32 sharing their experiences of psychoactive medication use and their view of the balance 33 34 between the benefits of medication use and potential cognitive decline. Monitors met the 35 36 37 study team to discuss study progress every six months. They are assisting us with the 38 39 dissemination of our study, making sure that lay summaries of results are accessible and 40 41 avoid possible misinterpretation. 42 43 44 45 46 47 Results 48 49 50 The source population consisted of 66,136 patients diagnosed with dementia between April 51 52 53 2006 and July 2015. After applying the exclusion criteria, our analysis included 40,770 cases 54 55 and 283,933 controls, with the vast majority of cases being matched to seven controls 56 57 58 59 60 https://mc.manuscriptcentral.com/bmj BMJ Page 12 of 70

1 2 3 (efigure 1). Most patients (n=254,083, 78%) were from general practices in England, with 4 5 also 30,817 (9%) from Scotland, 29,575 (9%) from Wales, and 10,228 (3%) from Northern 6 7 Ireland. The median (IQR) age of patients at index date was 83 (78-87) years and 63% were 8 9 10 female. The median DEP duration was 7.1 (IQR: 4.0-11.3; range: 1-16) years. Diagnoses of 11 Confidential: For Review Only 12 conditions related to dementia and indications for anticholinergics increased during the 13 14 DEP, for example the proportion of dementia cases diagnosed with depression before the 15 16 17 start of the DEP was 12% (n=5,071), rising to 20% (n=8,030) during the DEP (table 1). 18 19 Information on smoking status, harmful alcohol use, and BMI becomes more complete by 20 21 the end of the DEP. As expected, the average BMI decreases during the DEP for dementia 22 23 24 cases, but increases for the control group. 25 26 27 28 29 30 Frequency of anticholinergic medication use 31 32 33 14,453 (35%) cases and 86,403 (30%) controls were prescribed at least one anticholinergic 34 35 classified as ACB3 during their DEP. A total of 1,793,505 ACB3 prescriptions were written 36 37 during the DEP, with the five most common being amitriptyline (29% of ACB3 prescriptions) 38 39 40 dosulepin (16%), paroxetine (8%), oxybutynin (7%) and tolterodine (7%). Only 1429 (3.5%) 41 42 cases and 7909 (2.8%) controls were prescribed an ACB2 medication, with carbamazepine 43 44 accounting for 87% of ACB2 prescriptions. The vast majority of patients received at least 45 46 47 one ACB1 prescription during their DEP (89% of cases and 87% of controls) with 48 49 cardiovascular medication accounting for 63% of these prescriptions. The number of 50 51 prescriptions within each ACB category is further described in etable 1. 52 53 54 55 56 57 58 59 60 https://mc.manuscriptcentral.com/bmj Page 13 of 70 BMJ

1 2 3 Primary analysis 4 5 6 There was a positive and significant association between the prescription of any ACB1, ACB2 7 8 or ACB3 medication and dementia with corresponding odds ratios of 1.10 (95% confidence 9 10 11 intervalConfidential: 1.06 to 1.15), 1.10 (1.03 to 1.16), For and 1.11 Review (1.08 to 1.14), adjusted Only for covariates 12 13 measured at the end of the DEP (table 2). A dose-response effect was evident for ACB2 and 14 15 ACB3 doses prescribed, supporting a causal interpretation However there was a consistent 16 17 OR of approximately 1.1 for those prescribed ACB1 medications, regardless of how many 18 19 20 were prescribed (table 2), and including those with less than 14 DDD exposure, suggesting 21 22 residual confounding rather than a causal effect of ACB1 on dementia incidence. The 23 24 associations are attenuated from the crude rates when adjusting for medication use prior to 25 26 27 DEP, and are further reduced only slightly when adjusting for covariates at the end of the 28 29 DEP. 30 31 32 There was a dose-response effect linking average anticholinergic load, measured by the 33 34 average total ACB score over the DEP, with dementia incidence, although this is only evident 35 36 37 for those with an average score of 3 or more (table 2). 38 39 40 41 42 43 Exposure by medication class 44 45 46 When analysed by class, there was a substantial association between dementia incidence 47 48 and any prescription of ACB3 antidepressants, Parkinsonian drugs and urological 49 50 medications, but no association with ACB3 antispasmodics, antipsychotics, antihistamines, 51 52 53 or other ACB3s (table 3). ACB2 prescriptions were relatively rare, and so results are 54 55 imprecise in this group, but there is some evidence for an association between dementia 56 57 58 59 60 https://mc.manuscriptcentral.com/bmj BMJ Page 14 of 70

1 2 3 incidence and prescription of ACB2 Parkinsonian drugs. For ACB1, associations with 4 5 increased dementia risk are only observed for antidepressants but not with any other ACB1 6 7 medication class. Associations between dementia incidence and number of DDDs by class 8 9 10 is shown in appendix table 2 and are consistent with the findings in table 3. 11 Confidential: For Review Only 12 13 Exposure by time before dementia diagnosis 14 15 16 When examining the effect of exposure in three different time different periods (4-10, 10- 17 18 15, and 15-20 years) prior to the index date, associations with any ACB3 prescriptions were 19 20 consistent across the DEP while associations for ACB1 and ACB2 were more apparent closer 21 22 23 to the index date (table 4). In particular, the prescription of any ACB3 medication 15-20 24 25 years prior to dementia was significantly associated with greater dementia incidence with 26 27 OR of 1.17 (1.10 to 1.24) adjusted for covariates at the start of the DEP. Prescriptions at 15- 28 29 30 20 years prior to dementia for ACB3 antidepressants and urologicals remained consistently 31 32 significantly associated with dementia incidence with OR of 1.19 (1.10 to 1.29) and 1.27 33 34 (1.09 to 1.48). However, for ACB1 antidepressants, the association with dementia increased 35 36 37 for prescriptions given in periods closer to dementia. 38 39 40 41 42 43 Sensitivity analyses 44 45 46 Excluding those with an ACB2 or ACB3 prescription in the 12 months before the DEP, hence 47 48 restricting to ‘new’ ACB2/3 users during the DEP, excluded 5,215 cases and 62,161 controls. 49 50 This led to small reductions in the association between dementia incidence and any ACB3 51 52 53 with OR of 1.07 (1.04 to 1.10) and any ACB2 with OR of 1.12 (1.04 to 1.21) adjusted for 54 55 covariates measured at the end of the DEP (etable 3). No substantial changes were seen in 56 57 58 59 60 https://mc.manuscriptcentral.com/bmj Page 15 of 70 BMJ

1 2 3 the association between dementia incidence and specific classes of medication use when 4 5 restricted to new users. In particular, ACB3 antidepressants and urologicals and ACB1 6 7 antidepressants remained significantly associated with dementia. 8 9 10 11 FindingsConfidential: remained similar when also analysed For by 90+Review DDDs of exposure Only versus less, rather 12 13 than any prescription versus none (results not shown). 14 15 16 Although there were differences in the number of medications rated ACB1 and ACB2 and 17 18 ADS1 and ADS2, the proportion of patients with any prescription of ACB3 or ADS3 19 20 medications were similar (etable 4). There was good concordance between ADS3 and ACB3 21 22 23 (Cohen’s kappa = 0.91). Prescription of any ADS3 medication had a marginally lower 24 25 association with dementia with an odds ratio of 1.08 (1.06 to 1.11) adjusted for covariates 26 27 at the end of the DEP, than ACB3 (odds ratio 1.11 [1.08 to 1.14]). However, when analysed 28 29 30 according to the medication class, the findings were very similar with the ADS3 31 32 antidepressants, Parkinsonian and urologicals scoring and ADS1 antidepressants scoring 33 34 consistently associated with greater dementia incidence (data not shown), but no 35 36 37 association with other ADS classes. 38 39 40 Missing data 41 42 43 There were relatively high proportions with missing data for the BMI, smoking and alcohol 44 45 variables. There was very little difference in any results when comparing estimates with and 46 47 without adjustment for these variables among the main dataset and those with complete 48 49 50 data (etable 5), suggesting that our findings are not sensitive to missing data in these 51 52 covariates. 53 54 55 56 57 58 59 60 https://mc.manuscriptcentral.com/bmj BMJ Page 16 of 70

1 2 3 Discussion 4 5 6 Statement of principal findings 7 8 9 In this case-control study of older adults in the United Kingdom, there was a significant 10 Confidential: For Review Only 11 association between increasing total anticholinergic burden over the previous 4-20 years 12 13 14 and incident dementia diagnosis. However this dose-response effect was not seen for 15 16 cumulative use of medications classed 1 on the ACB scale, and was only evident for certain 17 18 classes of anticholinergics. ACB3 antidepressants (predominantly Amitriptyline, Dosuplepin 19 20 and Paroxetine) and urologicals (predominantly Oxybutynin and Tolterodine) were 21 22 23 consistently associated with incident dementia. These relationships were seen even for 24 25 exposures between 15 and 20 years prior to the dementia diagnosis, suggesting that reverse 26 27 causation or confounding with early dementia symptoms are less likely explanations for the 28 29 30 effect, although depression in midlife may be a risk factor for later dementia. Prescription 31 32 of ACB3 antihistamines and gastrologicals was not associated with dementia in any of our 33 34 analysis. Other antidepressants (those coded as ACB1, predominantly SSRIs) were 35 36 37 associated with dementia, but that association was greater for prescriptions close to 38 39 dementia incidence, suggesting that reverse causation could be a possible explanation for 40 41 this observed association. Other ACB1 medications were not associated with increased 42 43 dementia incidence. 44 45 46 47 48 49 50 Strengths and weaknesses of the study 51 52 53 Our study used a large population representative primary care database that allowed a 54 55 detailed analysis of the association between dementia incidence and prescriptions of many 56 57 58 59 60 https://mc.manuscriptcentral.com/bmj Page 17 of 70 BMJ

1 2 3 different drugs classes up to 20 years prior to dementia diagnosis. This is the first study to 4 5 simultaneously examine classes of anticholinergic medications, hence to have discovered 6 7 differences in association with dementia across these classes. However, there are several 8 9 10 limitations to our approach. 11 Confidential: For Review Only 12 13 Dementia diagnosis codes in CPRD reflect GP diagnoses well (27), however dementia is 14 15 known to be under-diagnosed, with up to 50% of people with dementia not receiving a 16 17 diagnosis (28). It is possible that those more likely to be prescribed anticholinergic 18 19 20 medication are also more likely to be diagnosed with dementia, leading to surveillance bias, 21 22 but this should be apparent for all medication classes and so is unlikely to account for our 23 24 findings. 25 26 27 It is difficult to accurately assess the anticholinergic activity of each medication, and 28 29 30 anticholinergic scales differ with respect to how they classify medications (21). However, 31 32 there was good agreement between the anticholinergic drug scale (ADS) and the 33 34 anticholinergic cognitive burden (ACB) and our findings did not differ when using each scale. 35 36 37 We do not know whether patients were adherent to their prescribed medications. 38 39 Medications obtained over-the-counter (OTC) are not recorded in our data, hence we will 40 41 have likely underestimated the use of particular antihistamines. DDDs can be difficult to 42 43 establish for certain medications yet represent the best available method for comparing the 44 45 46 levels of exposure of different drug classes. Our findings did not change when we instead 47 48 analysed the number of prescriptions to quantify exposure (results not shown). 49 50 51 As in any observational study, unmeasured or residual confounding could underlie positive 52 53 associations, and many causes of dementia are unknown. In particular, primary care 54 55 56 records do not hold detailed lifestyle or demographic information, but do include detailed 57 58 59 60 https://mc.manuscriptcentral.com/bmj BMJ Page 18 of 70

1 2 3 records of diagnoses and symptoms that have allowed good confounding control. With 4 5 respect to study design: our estimates generalise to those surviving at least four years after 6 7 their initial anticholinergic exposure, and the nested-case control design provides unbiased 8 9 10 estimates of effects when compared to the equivalent cohort study (29). 11 Confidential: For Review Only 12 13 14 15 16 17 Strengths and weaknesses in relation to other studies 18 19 20 Many studies have linked anticholinergic medication use with concurrent or short term 21 22 23 cognitive effects (8). Few studies have examined associations of long term anticholinergic 24 25 exposure with future cognitive decline or dementia incidence, but these tend to report 26 27 positive associations (12,14,15). Our study findings are consistent with these studies. In 28 29 30 particular a US cohort study of 3,434 participants followed over an average of 7 years, 31 32 similar effect sizes were found to our study (12). Anticholinergic antidepressants and other 33 34 anticholinergics were linked to dementia incidence, consistent with our results, however the 35 36 37 authors were not able to test specific classes of non-antidepressants. A US case-control 38 39 study of 141,940 nursing home residents with depression (15), found a greater adjusted 40 41 odds ratio of 1.24 (1.20 to 1.28) for anticholinergic medication use in the 90 days before 42 43 dementia diagnosis based on administrative data. Our analysis builds on these studies using 44 45 46 a longer history in a larger sample of patients enabling a further disaggregation of the 47 48 effects of specific classes in time windows prior to diagnosis. Although set in the UK our 49 50 findings are likely to generalizable to other developed countries. 51 52 53 54 55 56 57 58 59 60 https://mc.manuscriptcentral.com/bmj Page 19 of 70 BMJ

1 2 3 Meaning of the study: possible explanations and implications for clinicians and policymakers 4 5 6 Possible explanations for our findings are that other actions of specific groups of 7 8 anticholinergics may underlie observed effects, or that the medications are markers of 9 10 11 prodromalConfidential: symptoms or dementia risk factors.For Alternatively Review the class-specific Only association 12 13 we have observed may reflect a difference in the ability of different groups of 14 15 anticholinergics to cross the blood-brain barrier. 16 17 18 Mechanistic evidence for a link between anticholingerics and dementia incidence is limited, 19 20 but neuropathological studies in humans and mice do support a role of anticholinergics 21 22 23 affecting neurodegenerative pathology (30,31). Most recently a cross-sectional analysis of 24 25 Neuroimaging Initiative (ADNI) and Indiana Memory and Ageing Study (IMAS) data linked 26 27 anticholinergic use to reduced glucose metabolism and increased brain atrophy, and to 28 29 30 future mild cognitive impairment or Alzheimer’s disease incidence among cognitively 31 32 normal participants (HR=2.47, p=0.01) (32), but did not disaggregate subclasses of 33 34 medications. Evidence from anticholinergic cessation trials have not shown improvements 35 36 37 in cognitive function, but these have been underpowered and focussed on short term 38 39 outcomes (33). 40 41 42 Anticholinergic urologicals, particularly Oxybutynin, have been consistently associated with 43 44 short-term cognitive decline in RCTs (8,34), so a long-term risk of dementia is plausible. 45 46 47 Lower urinary tract symptoms themselves have been linked to future dementia incidence 48 49 (35) and may be a symptom of early neurodegeneration (36). 50 51 52 ACB1 antidepressants (mainly SSRIs) were only associated with dementia close to the time 53 54 of prescription. Conversely the finding of a constant association between dementia and 55 56 57 58 59 60 https://mc.manuscriptcentral.com/bmj BMJ Page 20 of 70

1 2 3 ACB3 antidepressants across the 20 years before dementia incidence, is more likely to 4 5 represent a causal link. Although we adjusted for depression diagnosis, severity, symptoms 6 7 and duration, residual confounding and incomplete coding cannot be excluded. 8 9 10 11 Similarly,Confidential: patients with Parkinson’s disease For have anReview elevated risk of dementia Only (38), yet 12 13 anticholinergic antiparkinsonian drugs have previously been associated with greater 14 15 cognitive decline in a cohort of 235 Parkinson’s disease patients (39). 16 17 18 19 20 Previous research linking all anticholinergics to dementia incidence may have dissuaded 21 22 some patients and clinicians from the use of certain antihistamine or gastrological 23 24 medication for which we see no long-term association with dementia. Our results, although 25 26 27 in need of independent confirmation should reassure these patients. Nevertheless, we do 28 29 not dispute the possible short-term harms of all anticholinergic medication use among 30 31 vulnerable groups, including short-term cognitive impairment, and would advocate current 32 33 guidance of vigilant use or avoidance among frail older people. 34 35 36 37 38 39 Methodological considerations for future research 40 41 42 Adjusting for time-varying confounders is difficult in a matched case-control study (24,40). 43 44 We measured covariates at the start and end of the DEP. Since there was little difference 45 46 47 between these estimates in most cases we are confident that the effect of time-varying 48 49 covariates does not significantly affect our findings. Secondly, to be certain that potential 50 51 confounders preceded medication initiation and not vice versa, we further excluded 52 53 patients who were ‘prevalent’ users at the start of each DEP, again with little difference in 54 55 56 results. We suggest that authors undertaking case-control studies comparing cumulative 57 58 59 60 https://mc.manuscriptcentral.com/bmj Page 21 of 70 BMJ

1 2 3 exposure over a long time period carefully consider how time-varying covariates and 4 5 exposure prior to the period in which they are being directly measured might affect study 6 7 findings. 8 9 10 11 ConclusionConfidential: For Review Only 12 13 14 Many people use anticholinergic medications at some point in their lives, and many are 15 16 prescribed to manage chronic conditions leading to potentially long exposures. There are 17 18 clear robust associations between levels of anticholinergic antidepressants and 19 20 anticholinergic urologicals and risk of dementia diagnosis up to twenty years after exposure. 21 22 23 Future research in this area must consider mechanisms underlying differential effects of 24 25 individual drug classes. Carefully conducted prospective studies in specific patient cohorts 26 27 comparing the long-term cognitive effects and neuropathological correlates of specific drug 28 29 30 classes are now needed. Clinicians should continue to be vigilant with respect to the use of 31 32 anticholinergic medications, and in particular should consider the risk of long term cognitive 33 34 effects, as well as short-term effects, associated with specific classes when weighing their 35 36 37 risks and benefits. 38 39 40 41 42 43 Funding: This research was supported by funding from the Alzheimer’s Society (AS-PG-2013- 44 45 017). The funders had no role is the design of the study or the interpretation of the 46 47 findings. 48 49 50 Data source: This study is based in part on data from the Clinical Practice Research Datalink 51 52 53 obtained under licence from the UK Medicines and Healthcare products Regulatory Agency 54 55 56 57 58 59 60 https://mc.manuscriptcentral.com/bmj BMJ Page 22 of 70

1 2 3 (MHRA). However, the interpretation and conclusions contained in this report are those of 4 5 the authors alone. 6 7 8 Data sharing: Data from the Clinical Practice Research Datalink cannot be shared by the 9 10 11 authorsConfidential: but is available directly from CPRD. For Full code Review lists corresponding Only to the outcome 12 13 and each of the covariates we included are available on request. 14 15 16 Ethical approval: The study was approved by the Independent Scientific Advisory 17 18 Committee for CPRD for Clinical Practice Research Datalink research (protocol No 15_056R). 19 20 No further ethical approval was required for the analysis of the data. The CPRD Group has 21 22 23 obtained ethical approval from a multi-centre research ethics committee for all purely 24 25 observational research using CPRD data. 26 27 28 Contributorship: 29 30 31 KR, CF, GS, PM and IM conceived and developed the initial study. KR and GS drafted the 32 33 34 statistical analysis plan. KR, NS, IM, CF, and NC developed the code lists. KR conducted the 35 36 statistical analysis and is the guarantor. All authors contributed to the study protocol 37 38 development and revision, the interpretation of findings and the revision of the manuscript. 39 40 41 Transparency statement: 42 43 44 KR affirms that this manuscript is an honest, accurate, and transparent account of the study 45 46 47 being reported; that no important aspects of the study have been omitted; and that any 48 49 discrepancies from the study as planned (and, if relevant, registered) have been explained. 50 51 52 Acknowledgements: We would like to thank Ms Tarita Murray-Thomas (MHRA) for 53 54 extracting the CPRD data, Mr Keshav Bajaj (Aston University) for assisting with medication 55 56 57 58 59 60 https://mc.manuscriptcentral.com/bmj Page 23 of 70 BMJ

1 2 3 coding, Prof Ryan Carnahan for providing the ADS scale medications, and Mr Barry 4 5 Plumpton, Mrs Ann McLauchlan, Mrs Barbara di Vita, and Mrs Gloria Swan for providing 6 7 much appreciated assistance in interpretation and oversight as Alzheimer’s Society Research 8 9 10 Network Volunteers. 11 Confidential: For Review Only 12 The Corresponding Author has the right to grant on behalf of all authors and does grant on 13 14 behalf of all authors, a worldwide licence to the Publishers and its licensees in perpetuity, in 15 16 17 all forms, formats and media (whether known now or created in the future), to i) publish, 18 19 reproduce, distribute, display and store the Contribution, ii) translate the Contribution into 20 21 other languages, create adaptations, reprints, include within collections and create 22 23 24 summaries, extracts and/or, abstracts of the Contribution, iii) create any other derivative 25 26 work(s) based on the Contribution, iv) to exploit all subsidiary rights in the Contribution, v) 27 28 the inclusion of electronic links from the Contribution to third party material where-ever it 29 30 may be located; and, vi) licence any third party to do any or all of the above. 31 32 33 Conflict of interest. 34 35 36 All authors have completed the ICMJE uniform disclosure form at www.icmje.org/coi_disclosure.pdf 37 38 and declare: no support from any organisation for the submitted work beyond the Alzheimer's 39 40 Society grant listed below; IM has received personal fees for guest lectures and to support travel 41 42 43 from Astellas Pharmaceuticals; YL reports personal fees from Thame Pharmaceuticals, NC and CF 44 45 have received grants and personal fees from Astellas Pharmaceuticals; no other relationships or 46 47 activities that could appear to have influenced the submitted work. 48 49 50 51 52 53 54 55 56 57 58 59 60 https://mc.manuscriptcentral.com/bmj BMJ Page 24 of 70

1 2 3 References 4 5 1. Deaths registered in England and Wales (Series DR) - Office for National Statistics [Internet]. 6 [cited 2017 Feb 12]. Available from: 7 https://www.ons.gov.uk/peoplepopulationandcommunity/birthsdeathsandmarriages/deaths/b 8 ulletins/deathsregisteredinenglandandwalesseriesdr/2015#dementia-and-alzheimer-disease- 9 leading-cause-of-death-for-both-men-and-women-aged-80-and-over 10 11 2. Confidential: Savva GM, Wharton SB, Ince PG, Forster For G, Matthews Review FE, Brayne C, et al. OnlyAge, Neuropathology, 12 and Dementia. N Engl J Med. 2009 May 28;360(22):2302–9. 13 14 3. White LR, Edland SD, Hemmy LS, Montine KS, Zarow C, Sonnen JA, et al. Neuropathologic 15 comorbidity and cognitive impairment in the Nun and Honolulu-Asia Aging Studies. Neurology. 16 2016 Mar 15;86(11):1000–8. 17 18 4. Matthews FE, Arthur A, Barnes LE, Bond J, Jagger C, Robinson L, et al. A two-decade comparison 19 of prevalence of dementia in individuals aged 65 years and older from three geographical areas 20 of England: results of the Cognitive Function and Ageing Study I and II. The Lancet. 2013 21 Oct;382(9902):1405–12. 22 23 5. Winblad B, Amouyel P, Andrieu S, Ballard C, Brayne C, Brodaty H, et al. Defeating Alzheimer’s 24 disease and other dementias: a priority for European science and society. Lancet Neurol. 2016 25 Apr 1;15(5):455–532. 26 27 6. Gao L, Maidment I, Matthews FE, Robinson L, Brayne C. Medication usage change in older 28 people (65+) in England over 20 years: findings from CFAS I and CFAS II. Age Ageing. :1–6. 29 30 7. Prescriptions Dispensed in the Community, Statistics for England - 2005-2015 - NHS Digital 31 [Internet]. [cited 2017 Oct 16]. Available from: http://digital.nhs.uk/catalogue/PUB20664 32 33 8. Tannenbaum C, Paquette A, Hilmer S, Holroyd-Leduc J, Carnahan R. A Systematic Review of 34 35 Amnestic and Non-Amnestic Mild Cognitive Impairment Induced by Anticholinergic, 36 Antihistamine, GABAergic and Opioid Drugs. Drugs Aging. 2012;29(8):639–58. 37 38 9. National Institute for Health and Care Excellence. Urinary incontinence in women: management. 39 NICE guideline CG171 [Internet]. 2015 [cited 2017 Feb 10]. Available from: 40 https://www.nice.org.uk/guidance/cg171?unlid= 41 42 10. Fick DM, Cooper JW, Wade WE, Waller JL, Maclean JR, Beers MH. Updating the Beers Criteria for 43 Potentially Inappropriate Medication Use in Older Adults: Results of a US Consensus Panel of 44 Experts. Arch Intern Med. 2003 Dec 8;163(22):2716–24. 45 46 11. Gallagher P, O’Mahony D. STOPP (Screening Tool of Older Persons’ potentially inappropriate 47 Prescriptions): application to acutely ill elderly patients and comparison with Beers’ criteria. Age 48 Ageing. 2008 Nov;37(6):673–9. 49 50 12. Gray SL, Anderson ML, Dublin S, Hanlon JT, Hubbard R, Walker R, et al. Cumulative Use of Strong 51 Anticholinergic Medications and Incident Dementia. JAMA Intern Med. 2015 Mar 1;175(3):401– 52 7. 53 54 13. Carriere I, Fourrier-Reglat A, Dartigues J-F, Rouaud O, Pasquier F, Ritchie K, et al. Drugs With 55 Anticholinergic Properties, Cognitive Decline, and Dementia in an Elderly General Population: 56 The 3-City Study. Arch Intern Med. 2009 Jul 27;169(14):1317–24. 57 58 59 60 https://mc.manuscriptcentral.com/bmj Page 25 of 70 BMJ

1 2 3 14. Fox C, Richardson K, Maidment ID, Savva GM, Matthews FE, Smithard D, et al. Anticholinergic 4 Medication Use and Cognitive Impairment in the Older Population: The Medical Research 5 Council Cognitive Function and Ageing Study. J Am Geriatr Soc. 2011 Aug;59(8):1477–83. 6 7 15. Chatterjee S, Bali V, Carnahan RM, Johnson ML, Chen H, Aparasu RR. Anticholinergic Medication 8 Use and Risk of Dementia Among Elderly Nursing Home Residents with Depression. Am J Geriatr 9 Psychiatry Off J Am Assoc Geriatr Psychiatry. 2016 Jun;24(6):485–95. 10 11 16. Confidential:Boustani M, Campbell N, Munger S, Maidment For I, FoxReview C. Impact of anticholinergics Only on the aging 12 brain: a review and practical application. Aging Health. 2008 Jun;4(3):311–20. 13 14 17. Cai X, Campbell N, Khan B, Callahan C, Boustani M. Long-term anticholinergic use and the aging 15 brain. Alzheimers Dement [Internet]. 2012 Nov [cited 2013 Jan 10]; Available from: 16 http://www.alzheimersanddementia.com/article/S1552-5260(12)00081-7/abstract 17 18 18. Herrett E, Gallagher AM, Bhaskaran K, Forbes H, Mathur R, van Staa T, et al. Data Resource 19 Profile: Clinical Practice Research Datalink (CPRD). Int J Epidemiol. 2015 Jun 1;44(3):827–36. 20 21 19. Vandenbroucke JP, Pearce N. Case–control studies: basic concepts. Int J Epidemiol. 2012 Oct 22 1;41(5):1480–9. 23 24 20. Salahudeen MS, Chyou T, Nishtala PS. Serum Anticholinergic Activity and Cognitive and 25 Functional Adverse Outcomes in Older People: A Systematic Review and Meta-Analysis of the 26 Literature. PLOS ONE. 2016 Mar 21;11(3):e0151084. 27 28 21. Salahudeen MS, Hilmer SN, Nishtala PS. Comparison of Anticholinergic Risk Scales and 29 Associations with Adverse Health Outcomes in Older People. J Am Geriatr Soc. 2015;63(1):85– 30 90. 31 32 22. Horwitz RI, Feinstein AR. The problem of ‘protopathic bias’ in case-control studies. Am J Med. 33 1980 Feb;68(2):255–8. 34 35 23. Campbell NL, Maidment I, Fox C, Khan B, Boustani M. The 2012 Update to the Anticholinergic 36 Cognitive Burden Scale. J Am Geriatr Soc. 2013;61:S1–S232. 37 38 24. Schneeweiss S. A basic study design for expedited safety signal evaluation based on electronic 39 healthcare data. Pharmacoepidemiol Drug Saf. 2010 Aug 1;19(8):858–68. 40 41 25. Ray WA. Evaluating Medication Effects Outside of Clinical Trials: New-User Designs. Am J 42 43 Epidemiol. 2003 Nov 1;158(9):915–20. 44 45 26. Carnahan RM, Lund BC, Perry PJ, Pollock BG, Culp KR. The Anticholinergic Drug Scale as a 46 Measure of Drug-Related Anticholinergic Burden: Associations With Serum Anticholinergic 47 Activity. J Clin Pharmacol. 2006;46(12):1481–6. 48 49 27. Imfeld P, Brauchli Pernus YB, Jick SS, Meier CR. Epidemiology, Co-Morbidities, and Medication 50 Use of Patients with Alzheimer’s Disease or Vascular Dementia in the UK. J Alzheimers Dis. 2013 51 Jan 1;35(3):565–73. 52 53 28. Brown A, Kirichek O, Balkwill A, Reeves G, Beral V, Sudlow C, et al. Comparison of dementia 54 recorded in routinely collected hospital admission data in England with dementia recorded in 55 primary care. Emerg Themes Epidemiol. 2016;13:11. 56 57 58 59 60 https://mc.manuscriptcentral.com/bmj BMJ Page 26 of 70

1 2 3 29. Langholz B, Richardson D. Are nested case-control studies biased? Epidemiol Camb Mass. 2009 4 May;20(3):321–9. 5 6 30. Yoshiyama Y, Kojima A, Itoh K, Isose S, Koide M, Hori K, et al. Does Anticholinergic Activity Affect 7 Neuropathology? Implication of Neuroinflammation in Alzheimer’s Disease. Neurodegener Dis. 8 2015;15(3):140–8. 9 10 31. Perry EK, Kilford L, Lees AJ, Burn DJ, Perry RH. Increased Alzheimer pathology in Parkinson’s 11 Confidential:disease related to antimuscarinic drugs. For Ann Neurol. Review 2003;54(2):235–238. Only 12 13 32. Risacher SL, McDonald BC, Tallman EF, West JD, Farlow MR, Unverzagt FW, et al. Association 14 Between Anticholinergic Medication Use and Cognition, Brain Metabolism, and Brain Atrophy in 15 Cognitively Normal Older Adults. JAMA Neurol. 2016 Jun 1;73(6):721–32. 16 17 33. Kersten H, Molden E, Tolo IK, Skovlund E, Engedal K, Wyller TB. Cognitive Effects of Reducing 18 Anticholinergic Drug Burden in a Frail Elderly Population: A Randomized Controlled Trial. J 19 Gerontol A Biol Sci Med Sci. 2013 Mar 1;68(3):271–8. 20 21 34. Kay GG, Ebinger U. Preserving cognitive function for patients with overactive bladder: evidence 22 for a differential effect with darifenacin. Int J Clin Pract. 2008 Nov;62(11):1792–800. 23 24 35. Chiang C-H, Wu M-P, Ho C-H, Weng S-F, Huang C-C, Hsieh W-T, et al. Lower Urinary Tract 25 Symptoms Are Associated with Increased Risk of Dementia among the Elderly: A Nationwide 26 Study. BioMed Res Int. 2015 Jul 28;2015:e187819. 27 28 36. Sakakibara R, Panicker J, Fowler CJ, Tateno F, Kishi M, Tsuyusaki Y, et al. Is overactive bladder a 29 brain disease? The pathophysiological role of cerebral white matter in the elderly. Int J Urol. 30 2014 Jan 1;21(1):33–8. 31 32 37. Bali V, Chatterjee S, Carnahan RM, Chen H, Johnson ML, Aparasu RR. Risk of Dementia Among 33 Elderly Nursing Home Patients Using Paroxetine and Other Selective Serotonin Reuptake 34 Inhibitors. Psychiatr Serv. 2015 Aug 3;66(12):1333–40. 35 36 38. Aarsland D, Andersen K, Larsen JP, Lolk A, Nielsen H, Kragh–Sørensen P. Risk of dementia in 37 38 Parkinson’s disease A community-based, prospective study. Neurology. 2001 Mar 27;56(6):730– 39 6. 40 41 39. Ehrt U, Broich K, Larsen JP, Ballard C, Aarsland D. Use of drugs with anticholinergic effect and 42 impact on cognition in Parkinson’s disease: a cohort study. J Neurol Neurosurg Psychiatry. 2010 43 Feb 1;81(2):160–5. 44 45 40. Rose S, van der Laan M. A Double Robust Approach to Causal Effects in Case-Control Studies. Am 46 J Epidemiol. 2014 Mar 15;179(6):663–9. 47 48 49 50 51 52 53 54 55 56 57 58 59 60 https://mc.manuscriptcentral.com/bmj Page 27 of 70 BMJ

1 2 3 Table 1 – Characteristics of 40,770 case patients with dementia and 283,933 controls from the Clinical Practice 4 Research Datalink. Values are numbers (percentages) unless stated otherwise 5 At start of drug exposure period At end of drug exposure period 6 Controls Cases Controls 7 Cases (n=40,770) 8 Characteristic (n=283,933) (n=40,770) (n=283,933) Lifestyle 9 Smoker - current* 5,058 12.4 34,652 12.2 4,516 11.1 30,976 10.9 10 Harmful alcoholConfidential: use* 166 For 0.4 Review980 0.3 Only377 0.9 2,316 0.8 11 BMI*† 26.4 6.3 26.5 6.9 26.1 5.2 26.8 5.6 12 Cardio/cerebrovascular and related diagnoses 13 Diabetes 2,734 6.7 16,362 5.8 5,392 13.2 34,015 12.0

14 Diabetes complications 418 1.0 2,206 0.8 1,378 3.4 7,678 2.7

15 Hyperlipidemia 2,740 6.7 16,369 5.8 6,481 15.9 40,796 14.4

16 Hypertension 13,328 32.7 91,737 32.3 22,104 54.2 155,262 54.7

17 Stroke/Transient Ischaemic Attack 1,969 4.8 11,734 4.1 4,747 11.6 25,856 9.1

18 Congestive heart disease 5,279 12.9 33,947 12.0 8,388 20.6 54,615 19.2

19 Heart failure 819 2.0 5,921 2.1 2,140 5.2 15,961 5.6

20 Peripheral arterial disease 897 2.2 5,953 2.1 2,110 5.2 13,818 4.9

Atrial fibrillation 1,303 3.2 8,856 3.1 3,744 9.2 24,591 8.7 21 Angina 3,704 9.1 23,213 8.2 6,006 14.7 38,602 13.6 22 Myocardial infarction 1,820 4.5 12,051 4.2 2,908 7.1 19,602 6.9 23 Coronary artery operations 831 2.0 5,261 1.9 1,711 4.2 11,002 3.9 24 Deep vein thrombosis 552 1.4 4,172 1.5 1,110 2.7 8,101 2.9 25 Mental health diagnoses 26 Depression 5,071 12.4 29,676 10.5 8,030 19.7 44,264 15.6 27 Severe depression 400 1.0 2,175 0.8 804 2.0 4,056 1.4 28 Depression symptoms 1,393 3.4 7,880 2.8 4,577 11.2 24,220 8.5 29 Depression duraon‡, years† 13.8 12.8 14.2 12.8 14.3 13.0 15.2 13.2

30 Severe mental illness 294 0.7 1,558 0.5 431 1.1 2,078 0.7

31 Anxiety 3,641 8.9 22,229 7.8 6,190 15.2 35,598 12.5

32 Anxiety symptoms 1,172 2.9 6,457 2.3 3,763 9.2 20,366 7.2

33 Other diagnoses

34 Parkinson's disease 237 0.6 958 0.3 1,448 3.6 4,283 1.5

35 Epilepsy 542 1.3 2,799 1.0 771 1.9 3,860 1.4

36 Insomnia 2,940 7.2 18,407 6.5 7,351 18.0 45,227 15.9

Fatigue 2,666 6.5 15,825 5.6 8,561 21.0 50,351 17.7 37 Other sleep problems 583 1.4 3,501 1.2 2,772 6.8 15,878 5.6 38 Hemiplegia and paraplegia 131 0.3 816 0.3 235 0.6 1,406 0.5 39 Drug abuse 297 0.7 1,694 0.6 414 1.0 2,357 0.8 40 Migraine 1,320 3.2 8,625 3.0 1,839 4.5 11,999 4.2 41 Headache 2,540 6.2 15,653 5.5 6,102 15.0 37,602 13.2 42 Back/neck pain 12,225 30.0 79,108 27.9 21,123 51.8 138,554 48.8 43 Neuropathy 750 1.8 4,766 1.7 2,030 5.0 13,098 4.6 44 Meniere's disease 340 0.8 2,353 0.8 471 1.2 3,397 1.2 45 Restless legs syndrome 169 0.4 1,088 0.4 610 1.5 3,955 1.4

46 Chronic obstructive pulmonary disease 899 2.2 6,327 2.2 2,534 6.2 18,012 6.3

47 Asthma 3,134 7.7 20,991 7.4 4,375 10.7 29,822 10.5

48 Rhinitis 2,556 6.3 16,617 5.9 5,155 12.6 32,956 11.6

49 GERD/Oesophagitis 2,876 7.1 18,277 6.4 6,141 15.1 39,437 13.9

50 Peptic/gastric ulcer 1,122 2.8 6,779 2.4 2,530 6.2 15,724 5.5

51 Irritable bowel syndrome 210 0.5 1,316 0.5 2,328 5.7 15,277 5.4

52 Inflammatory Bowel disease 1,087 2.7 6,692 2.4 2,451 6.0 15,523 5.5

Intestinal surgery/Colostomy/ileostomy 1,438 3.5 9,515 3.4 3,137 7.7 20,215 7.1 53 Liver disease 93 0.2 650 0.2 207 0.6 1,312 0.5 54 Osteoarthritis 7,925 19.4 50,644 17.8 14,627 35.9 94,681 33.3 55 Rheumatoid arthritis 486 1.2 3,700 1.3 806 2.0 5,799 2.0 56 57 58 59 60 https://mc.manuscriptcentral.com/bmj BMJ Page 28 of 70

1 2 Dermatitis 6,986 17.1 45,613 16.1 13,364 32.8 88,331 31.1 3 Eczema 4,015 9.8 26,696 9.4 8,952 22.0 59,677 21.0 4 Psoriasis 1,059 2.6 6,856 2.4 1,696 4.2 11,200 3.9 5 Urinary incontinence 1,222 3.0 7,186 2.5 3,225 7.9 18,027 6.3 6 Renal disease/ chronic kidney disease 457 1.1 3,138 1.1 5,817 14.3 37,981 13.4 7 Prostatism 1,496 3.7 9,176 3.2 3,490 8.6 21,464 7.6

8 Cancer 3,133 7.7 21,956 7.7 6,307 15.5 44,567 15.7

9 History in 12 months before DEP

10 Any falls 808 2.0 4,391 1.6 808 2.0 4,391 1.6 Confidential: For Review Only 11 Any fractures 596 1.5 3,693 1.3 596 1.5 3,693 1.3

12 Physician consultaons† 5.4 6.3 4.8 5.9 5.4 6.3 4.8 5.9 13 Abbreviations: BMI, body mass index; DEP, drug exposure period; GERD, Gastroesophageal reflux disease 14 15 * Missing smoking data for 9,664 cases and 70,877 controls at the start of the DEP, and 1,465 cases and 13,596 16 controls at the end of the DEP 17 Missing BMI data for 13,086 cases and 94,734 controls at the start of the DEP, and 4,736 cases and 37,526 controls 18 at the end of the DEP 19 Missing alcohol data for 22,940 cases and 159,721 controls at the start of the DEP, and 12,702 cases and 90,129 20 controls at the end of the DEP 21 22 23 † mean (SD) 24 25 ‡ years since first depression symptom or depression diagnosis 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 https://mc.manuscriptcentral.com/bmj

34 of of 29 Page Page CI 95% b Adjusted for covariates forAdjusted covariates 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 measured at end of DEP of DEP atmeasured end CI 95% aOR a Adjusted for covariates forAdjusted covariates 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 measured at start of DEP DEP start atmeasured of BMJ https://mc.manuscriptcentral.com/bmj n n % n % OR CI 95% aOR 65 65 0.2% 214 0.1% 2.15* to 2.83 1.63 1.62* to 2.15 1.22 1.57* to 1.18 2.09 218 218 0.5% 1035 0.4% 1.48* to 1.72 1.28 1.23* to 1.43 1.06 1.20 to 1.03 1.40 266 266 0.7% 1533 0.5% 1.22* to 1.39 1.07 1.08 to 1.23 0.94 1.07 to 0.94 1.23 704 704 1.7% 4301 1.5% 1.15* to 1.25 1.06 1.01 to 1.10 0.93 1.02 to 0.94 1.11 Cases (n=40,770) Cases (n=40,770) Controls (n=283,933) Unadjusted 1052 1052 2.6% 5077 1.8% 1.59* to 1.71 1.49 1.40* to 1.50 1.30 1.31* to 1.22 1.41 2440 2440 6.0% 15154 5.3% 1.23* to 1.28 1.17 1.13* to 1.18 1.08 1.07* to 1.02 1.12 2530 2530 6.2% 17088 6.0% 1.25* to 1.32 1.19 1.15* to 1.21 1.09 1.13* to 1.07 1.19 5663 5663 13.9% 38084 13.4% 1.13* to 1.17 1.10 1.07* to 1.11 1.04 1.04 to 1.01 1.08 3000 3000 7.4% 20530 7.2% 1.24* to 1.30 1.18 1.16* to 1.22 1.11 1.14* to 1.08 1.20 4333 4333 10.6% 36527 12.9% 1.00 9517 9517 23.3% 74445 26.2% 1.00 6084 6084 14.9% 41528 14.6% 1.15* to 1.19 1.11 1.09* 1.05,1.13 1.07* to 1.03 1.11 18105 44.4% 124684 43.9% 1.23* to 1.28 1.19 1.05 to 1.10 1.01 1.05 to 1.01 1.10 26317 64.5% 197530 69.6% 1.00 39341 96.5% 276024 97.2% 1.00 Confidential: For Review Only 90-364 365-1459 1460+ 2786 2512 6.8% 6.2% 15462 12626 5.4% 4.4% 1.37* 1.51* to 1.43 1.31 to 1.58 1.45 1.24* 1.35* to 1.30 1.19 to 1.42 1.29 1.17* 1.27* to 1.12 1.22 to 1.21 1.33 >0-13 >0-13 14-89 0 14-89 >0-13 >0-13 90-364 365-1459 365-1459 1460+ 176 0.4% 826 0.3% 1.50* to 1.76 1.27 1.23 to 1.45 1.04 1.18 to 0.99 1.39 0 90-364 365-1459 1460+ 4253 8549 10.4% 21.0% 28497 56607 10.0% 19.9% 1.26* 1.28* to 1.32 1.21 to 1.33 1.23 1.12* 1.12* to 1.18 1.07 to 1.17 1.07 1.09* 1.10* to 1.04 1.15 to 1.05 1.15 >0-13 >0-13 14-89 0 ACB 3 ACB 2 ACB 1 ACB 1 prescription ACB 1 prescription ACB 2 prescription 36437 ACB 3 prescription 1429 89.4% 14453 3.5% 247406 35.5% 87.1% 7909 86403 1.25* to 1.29 1.21 30.4% 2.8% 1.11* 1.27* 1.27* to 1.15 1.07 to 1.30 1.24 to 1.35 1.20 1.16* 1.10* 1.10* to 1.19 1.13 to 1.17 1.03 to 1.06 1.15 1.11* 1.10* to 1.08 1.14 to 1.03 1.16 No prescriptions No prescriptions 4295 10.5% 36329 12.8% <0.1 <0.1 0.1-<0.5 0.1-<0.5 Number of DDDs Number DDDs of Exposure DEP during Any use Average ACB burden Table 2. Crude and adjusted odds ratios of dementia by prescription of any, DDDs, and total burden of anticholinergics according to the Anticholinergic Anticholinergic the to according anticholinergics burden of total and any, DDDs, of prescription by ofdementia ratios odds adjusted and Crude 2. Table score Burden Cognitive

Page 29 of 70 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 30 of 70

34 of of 30 Page Page BMJ https://mc.manuscriptcentral.com/bmj 9030 9030 22.2% 62348 22.0% 1.14* to 1.17 1.10 1.06* to 1.10 1.03 1.07* to 1.01 1.09 4914 4914 3874 12.1% 9.5% 33113 23954 11.7% 1.17* 8.4% to 1.21 1.13 1.07* 1.28* to 1.33 1.23 to 1.12 1.03 1.13* 1.07* to 1.19 1.08 to 1.02 1.12 1.12* to 1.07 1.18 1570 1570 3.9% 8132 2.9% 1.53* to 1.62 1.44 1.31* to 1.40 1.23 1.28* to 1.19 1.36 5781 5781 14.2% 40413 14.2% 1.12* to 1.16 1.08 1.06* to 1.10 1.02 1.06* to 1.02 1.10 Confidential: For Review Only DEP, drug exposure period; DDD, defined daily dose; OR, odds ratio, aOR, adjusted odds ratio; CI confidence interval; ACB Anticholinergic Anticholinergic ACB interval; CI confidence ratio; odds adjusted aOR, ratio, OR, odds dose; daily defined DDD, period; exposure drug DEP, 0.5-<1 0.5-<1 1-<2 2-<3 2-<3 3-<5 3-<5 5+ the DEP for the following medications not rated as anticholinergic: benzodiazepines, z-drugs, antidepressants, antinausea and antivertigo and antinausea antidepressants, z-drugs, benzodiazepines, anticholinergic: as rated not medications following the for DEP the 1. table in listed diagnoses all and years), >20 15-20, 10-15, 5-10, 0-5, (0, duration depression use, alcohol 0.01 < p-value * Cognitive Burden Burden Cognitive during ofprescriptions number DEP; the prior to months 12 the in consultations GP of and number fractures any falls, any region, age, for Adjusted a antivertigo and antinausea antidepressants, z-drugs, benzodiazepines, anticholinergic: as rated not medications following the for DEP the 1. table in listed diagnoses all and years), >20 10-20, 5-10, 2-5, 0-2, (0, duration depression use, alcohol during prescriptions of number DEP; the to prior months 12 inthe consultations GP of and number fractures any falls, any region, age, for Adjusted b harmful status, smoking BMI, DEP: the of end atthe measured variables following the and drugs; anti-Parkinson and antiepileptics, preparations,

Abbreviations: preparations, antiepileptics, and anti-Parkinson drugs; and the following variables measured at the start of the DEP: BMI, smoking status, harmful harmful status, smoking BMI, DEP: the of start atthe measured variables following the and drugs; anti-Parkinson and antiepileptics, preparations, 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 31 of 70 BMJ

1 2 3 4 5 Table 3. Adjusted odds ratios of dementia by any prescription of an anticholinergic medication by 6 Anticholinergic Cognitive Burden score and drug class 7 8 Adjusted for covariates 9 measured at start of Adjusted for covariates 10 Cases (n=40,770) Controls (n=283,933) DEP measured at end of DEP 11 Confidential: For Review Only a b 12 Drug class n % n % aOR 95% CI aOR 95% CI 13 ACB1 14

15 Analgesic 23,871 58.6% 158,162 55.7% 1.02 1.00 to 1.05 1.02 0.99 to 1.04 16 17 Antidepressant 5,958 14.6% 28,767 10.1% 1.37* 1.32 to 1.42 1.25* 1.20 to 1.30 18 19 Antipsychotic 8,051 19.7% 50,079 17.6% 1.05* 1.02 to 1.08 1.04 1.01 to 1.07 20 21 Cardiovascular 27,926 68.5% 191,895 67.6% 0.97 0.94 to 0.99 0.98 0.95 to 1.01 22 Gastro-intestinal 10,845 26.6% 71,814 25.3% 0.97 0.94 to 0.99 0.96* 0.93 to 0.99 23 24 Respiratory 9,385 23.0% 62,787 22.1% 0.99 0.97 to 1.02 0.99 0.97 to 1.02 25 26 Other 11,521 28.3% 77,345 27.2% 0.95* 0.92 to 0.97 0.95* 0.92 to 0.98 27 28 ACB2 29 Analgesic 385 0.9% 2,337 0.8% 1.03 0.92 to 1.15 1.03 0.92 to 1.16 30 31 Antipsychotic 22 0.1% 69 0.0% 1.44 0.87 to 2.36 1.35 0.82 to 2.23 32 33 Parkinsonian 57 0.1% 141 0.0% 1.55* 1.12 to 2.14 1.32 0.96 to 1.82 34 35 Respiratory 19 0.0% 123 0.0% 0.89 0.55 to 1.45 0.83 0.51 to 1.36 36 37 Other 985 2.4% 5,454 1.9% 1.07 1.00 to 1.15 1.09 1.01 to 1.17 38 ACB3 39 40 Antidepressant 8,823 21.6% 50,817 17.9% 1.13* 1.10 to 1.16 1.11* 1.08 to 1.14 41 42 Antipsychotic 1,036 2.5% 5,140 1.8% 1.09 1.02 to 1.18 1.07 1.00 to 1.16 43 44 Gastro-intestinal 1,817 4.5% 12,057 4.2% 0.94 0.89 to 0.99 0.94 0.89 to 0.99 45 Parkinsonian 270 0.7% 951 0.3% 1.45* 1.25 to 1.68 1.29* 1.11 to 1.50 46 47 Respiratory 4,002 9.8% 25,195 8.9% 1.04 1.00 to 1.08 1.03 1.00 to 1.07 48 49 Urological 3,261 8.0% 16,873 5.9% 1.23* 1.18 to 1.28 1.18* 1.13 to 1.23 50 51 Other 284 0.7% 1,741 0.6% 0.99 0.87 to 1.13 0.99 0.87 to 1.13 52 53 Abbreviations: DEP, drug exposure period; aOR, adjusted odds ratio; CI confidence interval; ACB 54 Anticholinergic cognitive burden 55 56 57 58 Page 31 of 34 59 60 https://mc.manuscriptcentral.com/bmj BMJ Page 32 of 70

1 2 3 a Adjusted for age, region, any falls, any fractures and number of GP consultations in the 12 4 months prior to the DEP; number of prescriptions during the DEP for the following medications 5 not rated as anticholinergic: benzodiazepines, z-drugs, antidepressants, antinausea and 6 antivertigo preparations, antiepileptics, and anti-Parkinson drugs; and the following variables 7 measured at the start of the DEP: BMI, smoking status, harmful alcohol use, depression duration 8 (0, 0-2, 2-5, 5-10, 10-20, >20 years), and all diagnoses listed in table 1. 9 10 11 b AdjustedConfidential: for age, region, any falls, any fracturesFor and Review number of GP consultations Only in the 12 12 months prior to the DEP; number of prescriptions during the DEP for the following medications 13 not rated as anticholinergic: benzodiazepines, z-drugs, antidepressants, antinausea and 14 antivertigo preparations, antiepileptics, and anti-Parkinson drugs; and the following variables 15 measured at the end of the DEP: BMI, smoking status, harmful alcohol use, depression duration 16 (0, 0-5, 5-10, 10-15, 15-20, >20 years), and all diagnoses listed in table 1. 17 18 19 * p-value < 0.01 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 Page 32 of 34 59 60 https://mc.manuscriptcentral.com/bmj

c

34 CI 95% d of of

NA NA NA NA 1.07 1.07 to 1.17 0.99 1.59 1.59 to 2.28 1.11 0.99 0.99 to 1.01 0.96 1.05 1.05 to 1.20 0.91 1.00 1.00 1.03 1.03 to 1.07 1.00 1.00 1.00 0.98 0.98 to 1.01 0.95 33 1.06* 1.06* to 1.09 1.02 1.13* to 1.15 1.10 1.37* 1.37* to 1.42 1.32 1.11* 1.11* to 1.18 1.03 0.95* 0.95* to 0.98 0.92 1.12* to 1.15 1.08 0.95* 0.95* to 0.97 0.92 1.14* 1.14* to 1.23 1.04 0.91* 0.91* to 0.94 0.89 Page Page

No. of of No. controls controls 10 years prior to index date index to prior years 10 - 4 12 12 75 47 47 116 14 14 48 745 745 4,056 254 254 1,518 781 781 3,679 cases cases No. of of No. 5,413 25,566 4,492 38,579 6,392 39,642 1,054 5,734 4,492 38,579 8,404 55,944 7,342 41,826 7,784 52,560 9,792 66,244 21,756 21,756 143,993 35,722 35,722 242,210 12,338 72,335 27,256 27,256 187,816

(N=40,770) (N=40,770) (N=283,933) OR

b

CI 95% d

NA NA NA NA NA NA 1.12 1.12 to 1.26 1.00 0.92 0.92 to 1.14 0.75 1.00 1.00 1.04 1.04 to 1.09 0.99 1.00 1.00 0.98 to 1.02 0.94 0.99 0.99 to 1.02 0.95 1.01 1.01 to 1.06 0.97 0.98 0.98 to 1.10 0.86 0.98 0.98 to 1.02 0.94 1.18* 1.18* to 1.26 1.11 1.05* 1.05* to 1.08 1.02 1.06* 1.06* to 1.10 1.02 1.15* to 1.19 1.10 1.14* 1.14* to 1.26 1.03 1.16* 1.16* to 1.22 1.11

Exposure period Exposure period No. of of No. 15 years prior to index date index to prior years 15 controls controls - 10 8 35 17 17 22 10 10 28 363 363 1,873 102 102 654 493 493 2,556 377 377 2,024 cases cases No. of of No. 3,813 24,345 9,771 62,234 5,602 44,790 5,242 30,303 2,507 15,429 5,602 44,790 4,447 28,709 3,237 17,828 2,913 18,767 1,396 7,091 17,867 17,867 118,973 11,785 11,785 79,119

(N=23,959) (N=23,959) (N=166,735) OR BMJ

a

https://mc.manuscriptcentral.com/bmj 95% CI d

NA NA NA NA NA NA 0.99 0.99 to 1.21 0.80 1.08 1.08 to 1.24 0.93 1.09 1.09 to 1.43 0.83 1.06 1.06 to 1.11 1.01 1.00 1.00 1.05 to 1.10 1.00 1.00 1.07 1.07 to 1.25 0.91 0.99 0.99 to 1.07 0.92 1.02 1.02 to 1.08 0.97 0.96 0.96 to 1.04 0.89 1.20 to 1.42 1.00 0.97 0.97 to 1.03 0.92 0.94 0.94 to 1.02 0.88 1.19* 1.19* to 1.29 1.10 1.17* 1.17* to 1.24 1.10

No. of of No. controls controls 20 years prior to index date index to prior years 20 - 15

8 8 16 <5 <5 26 <5 <5 12 64 64 358 120 120 675 262 262 1,429 193 193 1,057 932 932 6,032 990 990 6,658 189 952 cases cases No.of 3,638 22,914 1,155 6,302 3,638 6,789 27,905 44,564 1,972 11,321 3,638 27,905 3,672 23,851 1,863 12,023 1,230 8,230

(N=10,684) (N=10,684) (N=74,145) OR Confidential: For Review Only Other Other Respiratory Respiratory Antidepressant Antidepressant Analgesic Analgesic Analgesic Parkinsonian Antidepressant Antipsychotic Antipsychotic Antipsychotic Antipsychotic Cardiovascular Cardiovascular Respiratory Respiratory Antipsychotic Gastro-intestinal Gastro-intestinal Other Other No prescriptions prescriptions No prescription 1 ACB prescription 2 ACB prescription 3 ACB prescription No ACB1 ACB2 ACB3 Table 4. Adjusted odds ratios of dementia by prescription of anticholinergic medications, by drug class, and by time period before dementia diagnosis diagnosis dementia before timeperiod and by drug class, by medications, anticholinergic of prescription by dementia of ratios odds Adjusted 4. Table Any use by ACB score and class class ACB and Any use score by class Drug ACB Any use score by Page 33 of 70 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 34 of 70

34 of of 1.03 1.03 to 1.08 0.99 1.02 1.02 to 1.20 0.86 34 0.91* 0.91* to 0.97 0.85 1.56* 1.56* to 1.83 1.33 1.23* 1.23* to 1.29 1.18 Page Page 230 230 786 176 176 1,051 1,283 8,721 2,839 17,807 2,870 14,654

0.93 0.93 to 1.02 0.84 1.04 1.04 to 1.10 0.97 0.97 0.97 to 1.21 0.78 1.54* 1.54* to 1.96 1.22 1.22* 1.22* to 1.32 1.13 93 93 569 522 522 3,453 104 104 379 810 810 4,139 1,228 7,615 BMJ

https://mc.manuscriptcentral.com/bmj 1.11 1.11 to 1.28 0.97 1.29 1.29 to 2.01 0.83 1.07 1.07 to 1.18 0.97 0.97 0.97 to 1.34 0.70 1.27* 1.27* to 1.48 1.09 29 29 146 43 43 281 273 273 1,568 524 524 3,247 207 207 995

Confidential: For Review Only

OR, odds ratio, aOR, adjusted odds ratio; CI confidence interval; ACB Anticholinergic cognitive burden burden cognitive Anticholinergic ACB interval; CIconfidence ratio; odds adjusted aOR, ratio, odds OR, Gastro-intestinal Gastro-intestinal Parkinsonian Parkinsonian Respiratory Respiratory Urological Urological Other Other

Abbreviations: index before date history the ofUTS data years with ≥16 patients Including a date the index before data history ≥11 of UTS years with patients Includingb index date history the before of UTS ≥6 data years Including with patients c anticholinergic: as rated not medications following for the period the exposure during number of prescriptions region, for age, d Adjusted following the and drugs; anti-Parkinson and antiepileptics, antivertigoand preparations, antinausea antidepressants, z-drugs, benzodiazepines, months priorthe 12 in the to numberandof consultations fractures GP any any falls, exposure period: of the start measured the variables at table in listed diagnoses all and >20 years), 10-20, 0-2,2-5,duration 5-10, (0, use, depression alcohol smoking harmful BMI, period: status, exposure 1. 0.01p-value < * 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 35 of 70 BMJ

Supplementary Material 1 2 3 4 Appendix 5 CPRD medcodes, corresponding Read codes and Read terms used to define dementia 6 7 8 medcode read_code read_term 9 10 1350 E00..12 Senile/presenile dementia 11 1916 Confidential:E00..11 Senile dementia For Review Only 12 1917 F110.00 Alzheimer's disease 13 2882 E00z.00 Senile or presenile psychoses NOS 14 4357 Eu02z14 [X] Senile dementia NOS 15 16 4693 Eu02z00 [X] Unspecified dementia 17 6578 Eu01.00 [X]Vascular dementia 18 7323 E000.00 Uncomplicated senile dementia 19 7572 F116.00 Lewy body disease 20 21 7664 Eu00.00 [X]Dementia in Alzheimer's disease 22 8195 Eu00z11 [X]Alzheimer's dementia unspec 23 8634 E004.11 Multi infarct dementia 24 8934 Eu01200 [X]Subcortical vascular dementia 25 9509 Eu02300 [X]Dementia in Parkinson's disease 26 27 9565 Eu01.11 [X]Arteriosclerotic dementia 28 11136 F111.00 Pick's disease 29 11175 Eu01100 [X]Multi-infarct dementia 30 11379 Eu00112 [X]Senile dementia,Alzheimer's type 31 12621 Eu02.00 [X]Dementia in other diseases classified elsewhere 32 33 15165 E001.00 Presenile dementia 34 15249 E00y.00 Other senile and presenile organic psychoses 35 16797 F110000 Alzheimer's disease with early onset 36 18386 E002000 Senile dementia with paranoia 37 38 19393 Eu01z00 [X]Vascular dementia, unspecified 39 19477 E004.00 Arteriosclerotic dementia 40 21887 E002100 Senile dementia with depression 41 25386 E041.00 Dementia in conditions EC 42 25704 Eu00011 [X]Presenile dementia,Alzheimer's type 43 44 26270 Eu02500 [X]Lewy body dementia 45 27677 E001300 Presenile dementia with depression 46 27759 Eu02z16 [X] Senile dementia, depressed or paranoid type 47 27935 Eu02z15 [X] Senile psychosis NOS 48 28402 Eu02000 [X]Dementia in Pick's disease 49 50 29386 Eu00z00 [X]Dementia in Alzheimer's disease, unspecified 51 29512 F112.00 Senile degeneration of brain 52 30032 E001200 Presenile dementia with paranoia 53 30706 Eu00200 [X]Dementia in Alzheimer's dis, atypical or mixed type 54 55 31016 Eu01300 [X]Mixed cortical and subcortical vascular dementia 56 32057 F110100 Alzheimer's disease with late onset 57 33707 E00..00 Senile and presenile organic psychotic conditions 58 34944 Eu02z13 [X] Primary degenerative dementia NOS 59 60 https://mc.manuscriptcentral.com/bmj BMJ Page 36 of 70

37014 Eu02200 [X]Dementia in Huntington's disease 1 37015 E003.00 Senile dementia with delirium 2 3 38438 E001z00 Presenile dementia NOS 4 38678 Eu00100 [X]Dementia in Alzheimer's disease with late onset 5 41089 E002z00 Senile dementia with depressive or paranoid features NOS 6 42279 E004z00 Arteriosclerotic dementia NOS 7 42602 E001000 Uncomplicated presenile dementia 8 9 43089 E004000 Uncomplicated arteriosclerotic dementia 10 43292 E004300 Arteriosclerotic dementia with depression 11 43346 Confidential:Eu00113 [X]Primary degen dementia For of Alzheimer's Review type, senile Only onset 12 44674 E002.00 Senile dementia with depressive or paranoid features 13 46488 Eu01000 [X]Vascular dementia of acute onset 14 15 46762 Eu00111 [X]Alzheimer's disease type 1 16 47619 Eu02z12 [X] Presenile psychosis NOS 17 48501 Eu02z11 [X] Presenile dementia NOS 18 49263 Eu00000 [X]Dementia in Alzheimer's disease with early onset 19 20 49513 E001100 Presenile dementia with delirium 21 51494 E00y.11 Presbyophrenic psychosis 22 53446 Eu04100 [X]Delirium superimposed on dementia 23 54106 Eu02100 [X]Dementia in Creutzfeldt-Jakob disease 24 55313 Eu01y00 [X]Other vascular dementia 25 26 55467 E004200 Arteriosclerotic dementia with paranoia 27 55838 Eu01111 [X]Predominantly cortical dementia 28 56912 E004100 Arteriosclerotic dementia with delirium 29 59122 Fyu3000 [X]Other Alzheimer's disease 30 60059 Eu00012 [X]Primary degen dementia, Alzheimer's type, presenile onset 31 32 61528 Eu00013 [X]Alzheimer's disease type 2 33 62132 E02y100 Drug-induced dementia 34 64267 Eu02y00 [X]Dementia in other specified diseases classif elsewhere 35 109288 A411000 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 https://mc.manuscriptcentral.com/bmj Page 37 of 70 BMJ

Definition of covariates 1 2 3 We extracted and coded the following covariates at the start and end of each patient’s Drug Exposure 4 5 Period (DEP), unless stated otherwise, from the patient’s record: 6 7 8 Diagnosis of cardiovascular and related diseases: any previous diagnosis of diabetes, diabetes 9 10 complications, hyperlipidemia, hypertension, stroke/ Transient Ischaemic Attack, congestive heart disease, 11 Confidential: For Review Only 12 13 heart failure, peripheral arterial disease, atrial fibrillation, angina, myocardial infarction, coronary artery 14 15 operations, deep vein thrombosis. 16 17 18 Other potential dementia risk factors or correlates: depression (diagnosis, maximum severity, symptoms 19 20 and duration), anxiety, anxiety symptoms, Parkinson's disease, severe mental illness, epilepsy, drug abuse, 21 22 23 cancer, insomnia, sleep problems, migraine, headache, pain, neuropathy, Meniere's disease, restless leg 24 25 syndrome 26 27 28 Other indications for anticholinergics: Chronic obstructive pulmonary disease, asthma, rhinitis, 29 30 Gastroesophageal reflux disease, peptic/gastric ulcer, Irritable bowel syndrome, Inflammatory Bowel 31 32 disease, intestinal surgery, liver disease, arthritis, dermatitis, eczema, psoriasis, incontinence, chronic 33 34 35 kidney disease, or prostatism. 36 37 38 Other factors: GP practice region (Scotland, Northern Ireland, Wales and 10 health regions of England), 39 40 smoking status (never, former, current or unknown), body mass index (<20, 20-24.9, 25-29.9, 30+ kg/m2, 41 42 or missing), and problematic alcohol use (categorised as more than 49 units per week for men and more 43 44 than 35 units per week for women, or missing). In the 12 months prior to the DEP: any falls, any fractures, 45 46 47 and number of GP consultations. 48 49 50 Other medication: Number of prescriptions during the DEP of other medications rated without 51 52 anticholinergic activity from the following classes: benzodiazepines (ATC N05BA, N05CD, N03AE), z-drugs 53 54 (ATC N05CF), antidepressants (ATC N06A), antinausea and antivertigo preparations (ATC N07C, A03F), 55 56 57 antiepileptics (ATC N03), and anti-Parkinson drugs (ATC N04). 58 59 60 https://mc.manuscriptcentral.com/bmj BMJ Page 38 of 70

eFigure1 – Selection of 40,770 patients with dementia and 283,933 controls from the Clinical Practice Research 1 Datalink. 2 3 4 5 6 7 8 9 10 11 Confidential: For Review Only 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 https://mc.manuscriptcentral.com/bmj

8% 1% 1% 2% 1% 1% 1% 2% 1% 2% 2% 7% 7% 2% 1% 2% 2% 4% 1% 1% 29% 16%

% % ACB3 (1,793,505 prescriptions) prescriptions) (1,793,505 ACB3

Drug Drug Amitriptyline Dosulepin Paroxetine Lofepramine Clomipramine Imipramine Trimipramine Nortriptyline Hyoscine Dicycloverine Atropine Chlorphenamine Hydroxyzine Promethazine Trifluoperazine Olanzapine Chlorpromazine Thioridazine Procyclidine Tolterodine Oxybutynin Solifenacin

6% 2% 2% 1%

% % BMJ ACB2 (132,731 prescriptions) prescriptions) (132,731 ACB2

Drug Drug Zuclopenthixol Amantadine Nefopam Pethidine https://mc.manuscriptcentral.com/bmj

1% 6% 6% 4% 1% 5% 1% 3% 2% 4% 11% 13%

% %

ACB1 (25,122,902 prescriptions) prescriptions) (25,122,902 ACB1 Confidential: For Review Only Drug Drug

Citalopram Citalopram Ranitidine Prochlorperazine CODEINE Dextropropoxyphene Dihydrocodeine Tramadol Bendroflumethiazide Atenolol Furosemide Ramipril Lisinopril

Antidepressant Respiratory Antipsychotic Cardiovascular Class Class Gastrointestinal Parkinsonian Urological Analgesic eTable 1. Frequency* of prescriptions during the drug exposure period according to Anticholinergic Cognitive Burden score and drug class class drug and score Burden Cognitive Anticholinergic to according period exposure drug the during prescriptions of Frequency* 1. eTable Page 39 of 70 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 40 of 70

87%

BMJ

Carbamazepine https://mc.manuscriptcentral.com/bmj

2% 4% 1% 2% 2% 4% 2% 3% 2% 3% 2%

Confidential: For Review Only Isosorbide Isosorbide Warfarin Nifedipine Doxazosin Digoxin Enalapril Perindopril Amiloride Trinitrate Glyceryl Prednisolone Diazepam

Other * Percentage of total prescriptions in the drug exposure period within each Anticholinergic Cognitive Burden score category with prevalence of 1% or more more or 1% of prevalence with category score Burden Cognitive Anticholinergic each within period exposure drug the in prescriptions total of Percentage * ACB, Anticholinergic Cognitive Burden Burden Cognitive Anticholinergic ACB, 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

CI 95% b b

Adjusted for covariates 1.00 1.00 measured at end of DEP of atmeasured end

BMJ CI 95% aOR a

aOR Adjusted for covariates measured at start of DEP of DEP start atmeasured https://mc.manuscriptcentral.com/bmj

7.1% 1.04 to 1.00 1.09 3.9% 1.06 2.5% 2.4% 1.29* to 1.01 1.11 1.36* to 1.23 1.36 1.47* to 1.28 1.44 1.19* to 1.39 1.57 1.23* 2.2% to 1.13 1.25 1.34* to 1.16 1.31 to 1.26 1.43 1.10* to 1.02 1.17 1.09 4.2% to 1.01 1.16 6.1% 1.03 1.01 to 0.98 1.09 to 0.96 1.06 1.03 1.01 to 0.97 1.09 to 0.96 1.06 0.0% 1.04 to 0.68 1.59 1.03 to 0.67 1.57 1.2% 1.12 to 1.02 1.23 1.16* 1.1% to 1.05 1.27 1.55* to 1.42 1.69 0.7% 1.47* to 1.34 1.60 1.09 to 0.97 1.23 1.6% 1.07 1.00 to 0.95 1.20 to 0.92 1.09 0.99 to 0.91 1.07 0.3% 1.05 to 0.86 1.28 0.99 to 0.81 1.21 0.2% 1.11 to 0.90 1.38 1.09 to 0.88 1.36 44.3% 1.00 89.9% 1.00 1.00 82.4% 1.00 1.00 1.00 74.7% 1.00 32.4% 1.00 1.00 39.4% 0.93* to 0.90 0.96 0.95* to 0.91 0.99 15.9% 19.7% 11.8% 1.03 1.00 1.03 to 1.00 1.07 to 0.97 1.03 1.02 to 0.99 1.07 1.00 to 0.99 1.05 1.04 to 0.97 1.03 to 1.00 1.08 14.5% 1.04 to 1.01 1.07 1.03 to 1.00 1.06 16.3% 0.98 to 0.94 1.01 0.98 to 0.95 1.02

131 131 713 713 542 7,222 7,222 6,883 6,195 3,325 3,325 2,982 1,977 4,550 92,038 92,038 45,258 45,258 55,924 33,483 20,172 10,967 41,234 11,912 17,242 46,285 125,771 125,771 255,166 233,854 212,119 111,906 111,906

8.1% 5.2% 3.6% 3.8% 2.7% 4.6% 6.4% 0.1% 1.5% 0.3% 1.7% 0.9% 0.3% 1.7% 41.4% 41.4% 85.4% 80.3% 31.5% 73.4% 39.1% 39.1% 16.1% 16.1% 20.0% 12.8% 15.8% 16.7%

n n % n % 29 29 Cases (n=40,770) Cases (n=40,770) Controls (283,933) 615 615 114 708 359 108 713 6,581 8,166 5,216 3,293 2,103 1,481 1,552 6,453 1,102 1,863 2,625 6,802 16,899 16,899 34,812 32,719 12,844 29,925 15,923 15,923 Confidential: For Review Only

1459 - 89 364 - - 0 0 0 0 0 365 14 1460+ 90 >0-13 >0-13 14-89 90-364 1460+ >0-13 90-364 365-1459 1460+ >0-13 14-89 90-364 365-1459 >0-13 14-89 365-1459 1460+ Analgesic Analgesic Gastro-intestinal Gastro-intestinal Cardiovascular Cardiovascular Antidepressant Antidepressant Antipsychotic eTable 2. Adjusted odds ratios of dementia by number of DDDs of anticholinergic medications by drug class and ACB score score ACBand class drug by medications anticholinergic of DDDs of number by ofdementia ratios odds Adjusted 2. eTable class Drug ACB1 Page 41 of 70 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 42 of 70

1.00 1.00 1.00 1.00 1.00 1.00

BMJ

https://mc.manuscriptcentral.com/bmj

6.8% 8.5% 5.3% 1.03 3.7% 0.95* 0.92* to 0.98 1.07 to 0.91 0.99 0.96 1.01 to 0.88 0.97 0.94* 3.6% 0.92* to 0.91 1.01 to 0.97 1.06 to 0.90 0.98 0.96 3.6% to 0.87 0.96 1.01 2.6% to 0.90 1.01 0.95 to 0.95 1.07 0.97 1.02 to 0.89 1.00 6.1% to 0.91 1.04 to 0.96 1.08 0.95 0.98 5.2% 0.98 to 0.89 1.00 3.3% to 0.92 1.05 0.93* to 0.93 1.03 0.92* 0.97 to 0.88 0.98 to 0.86 0.97 0.94 to 0.93 1.02 0.91* to 0.89 0.99 to 0.86 0.97 0.0% 1.62 0.0% 0.0% to 0.90 2.92 0.0% 1.43 1.61 1.80 1.48 to 0.85 2.40 to 0.89 2.91 to 1.02 3.17 1.33 to 0.84 2.62 1.47 to 0.79 2.23 1.18 to 0.84 2.59 to 0.67 2.08 0.1% 0.0% 1.13 1.21 to 0.79 1.59 to 0.76 1.93 1.12 1.21 to 0.79 1.59 to 0.76 1.93 1.0% 0.94 to 0.85 1.05 0.95 1.4% to 0.86 1.06 0.94 to 0.86 1.04 0.96 0.8% to 0.87 1.05 0.86 to 0.76 0.97 0.4% 0.86 1.10 to 0.76 0.98 to 0.94 1.27 1.10 to 0.95 1.28 0.3% 0.89 to 0.73 1.08 0.89 to 0.73 1.09 0.0% 1.11 to 0.45 2.75 0.92 to 0.36 2.33 77.9% 1.00 72.8% 1.00 99.2% 1.00 1.00 10.9% 1.01 to 0.98 1.05 1.00 11.8% to 0.97 1.04 0.95* to 0.92 0.99 0.95 to 0.92 0.99 100.0% 100.0% 1.00 100.0% 1.00

45 45 63 37 41 24 24 208 208 108 832 832 7,363 7,363 9,441 2,815 2,815 3,867 2,179 1,189 19,201 19,201 24,251 15,138 10,409 10,163 31,037 10,357 17,447 33,590 14,688 221,146 221,146 206,588 281,596 283,864 283,792

7.3% 8.8% 5.4% 3.9% 3.7% 3.7% 2.7% 6.5% 5.4% 3.3% 0.1% 0.1% 0.0% 0.0% 0.0% 0.0% 1.1% 1.4% 0.8% 0.3% 0.5% 0.0% 77.0% 77.0% 71.7% 99.1% 99.9% 99.9% 11.5% 11.5% 12.3%

6 39 39 22 16 20 19 18 435 435 566 311 116 208 2,995 3,592 2,216 1,607 1,507 4,702 1,515 1,095 2,653 5,013 2,182 1,362 31,385 31,385 29,249 40,385 40,748 40,713 Confidential: For Review Only

1459 - 364 13 13 89 - - - - 0 0 0 0 0 90 >0 365 >0 >0 >0-13 >0-13 14-89 365-1459 1460+ 14-89 90-364 365-1459 1460+ >0-13 14-89 90-364 1460+ 14-89 90-364 365+ >0-89 90+ 90-364 365+ Analgesic Analgesic Parkinsonian Parkinsonian Respiratory Antipsychotic Antipsychotic Respiratory Respiratory Other ACB2 ACB2 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

1.00 1.00 1.00 1.00

BMJ

https://mc.manuscriptcentral.com/bmj

7.5% 3.2% 3.4% 1.08* 2.7% 1.07 to 1.04 1.13 1.16* 1.24* 1.07* to 1.01 1.14 to 1.10 1.23 1.05 to 1.02 1.11 to 1.17 1.31 1.14* 1.22* to 0.99 1.12 to 1.08 1.21 to 1.14 1.29 2.8% 0.95 to 0.90 1.02 0.95 to 0.89 1.01 0.0% 0.89 to 0.49 1.61 0.85 to 0.47 1.53 0.1% 1.58* to 1.14 2.20 1.56* to 1.12 2.17 0.1% 1.31 to 0.96 1.80 1.32 0.1% to 0.96 1.82 1.09 0.0% 0.1% to 0.82 1.46 0.1% 1.29 1.12 1.40 1.44 to 0.95 1.74 to 0.84 1.49 to 0.94 2.07 1.19 to 1.08 1.92 1.27 to 0.88 1.61 1.27 to 0.86 1.89 to 0.95 1.70 0.9% 0.5% 1.02 0.97 to 0.92 1.14 to 0.84 1.12 1.05 0.99 to 0.94 1.17 to 0.85 1.15 1.1% 1.30* 0.8% 0.4% to 1.18 1.42 1.29* 0.98 1.11 to 1.18 1.41 to 0.88 1.09 to 0.96 1.28 0.96 1.10 to 0.86 1.07 1.0% to 0.95 1.27 0.91 to 0.82 1.00 0.91 to 0.82 1.01 0.3% 0.2% 1.17 1.26 to 0.98 1.39 to 1.05 1.52 1.17 1.23 to 0.97 1.40 to 1.01 1.48 0.3% 0.2% 1.31* 1.20 to 1.12 1.54 1.29* to 0.99 1.46 1.16 0.3% to 1.09 1.52 to 0.95 1.42 0.80 to 0.65 0.99 0.82 to 0.66 1.02 0.0% 0.89 to 0.37 2.11 0.80 to 0.34 1.93 98.1% 1.00 82.1% 1.00 1.00 98.2% 1.00 1.00 95.8% 1.00 1.00 99.7% 1.00 1.00 100.0% 100.0% 1.00

82 82 41 41 170 170 208 311 125 207 220 773 773 674 800 601 717 8,976 8,976 9,682 7,732 8,052 2,499 2,499 1,338 3,192 2,356 1,175 2,977 21,235 21,235 278,479 278,479 233,116 278,793 271,876 282,982 283,810 283,810

8.3% 3.7% 4.4% 3.7% 3.0% 0.0% 0.1% 0.1% 0.1% 0.1% 0.1% 0.2% 1.0% 0.5% 0.4% 0.4% 1.6% 1.0% 0.6% 0.5% 0.3% 1.1% 0.2% 0.0% 97.6% 97.6% 78.4% 97.5% 95.5% 99.3% 100.0% 100.0%

6 13 13 48 52 57 30 60 65 424 424 218 152 143 656 413 243 194 134 454 101 3,389 1,505 1,775 1,498 1,205 40,751 40,751 39,785 31,947 39,734 38,953 40,500 Confidential: For Review Only

364 364 13 - - - 0 0 0 0 0 0 14+ 90 1460+ 90 >0 >0-13 >0-13 >0-13 14-89 365-1459 1460+ >0-13 14-89 90-364 365-1459 >0-13 14-89 365-1459 1460+ 14-89 90-364 365+ >0-13 14-89 90-364 Antidepressant Antidepressant Other Other Gastro-intestinal Gastro-intestinal Parkinsonian Antipsychotic Antipsychotic ACB3 ACB3 Page 43 of 70 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 44 of 70

BMJ

https://mc.manuscriptcentral.com/bmj

4.6% 3.3% 1.02 1.05 to 0.97 1.08 to 1.00 1.12 1.02 1.05 to 0.97 1.07 2.3% to 0.99 1.11 1.15* to 1.08 1.23 1.10* to 1.03 1.17 0.0% 1.61 to 1.04 2.50 1.39 to 0.89 2.18 0.1% 1.56* to 1.22 2.01 1.35 to 1.05 1.74 0.1% 1.01 to 0.69 1.47 1.01 to 0.70 1.48 0.2% 0.1% 1.06 1.52 to 0.84 1.35 1.04 to 1.05 2.20 to 0.82 1.32 1.55 to 1.07 2.24 0.6% 1.02 to 0.90 1.15 0.6% 1.01 1.4% 1.06 to 0.89 1.15 1.3% 0.4% 1.27* to 0.94 1.20 1.41* 1.02 to 1.18 1.38 1.28* to 1.30 1.53 1.21* to 0.90 1.15 0.4% to 1.11 1.48 1.35* to 1.12 1.31 1.24* to 1.24 1.46 0.89 to 1.07 1.44 to 0.75 1.05 0.89 to 0.76 1.06 0.3% 1.18 to 1.00 1.40 1.18 to 1.00 1.40 91.1% 1.00 94.1% 1.00 1.00 99.4% 1.00 1.00 1.00

98 98 301 301 206 488 144 893 893 9,321 9,321 6,514 1,809 1,809 1,810 3,891 3,550 1,108 1,109 12,966 12,966 258,738 258,738 267,060 282,179

4.9% 3.7% 2.9% 0.2% 0.1% 0.1% 0.2% 0.1% 0.7% 0.4% 0.7% 1.9% 2.0% 0.6% 0.4%

90.2% 90.2% 92.0% 99.3%

88 88 27 32 83 36 293 293 169 295 773 800 226 165 1,997 1,511 1,167 36,768 36,768 Confidential: 37,509 For 40,483 Review Only

1459 - 89 - 0 0 0 365 14 365-1459 365-1459 1460+ >0-13 14-89 90-364 1460+ >0-13 90-364 365-1459 1460+ >0-13 14-89 90-364 Respiratory Respiratory Urological Urological Other Other * p<0.01 p<0.01 * burden cognitive Anticholinergic ACB interval; CI confidence ratio; odds adjusted aOR, ratio, odds OR, dose; daily defined DDD, period; exposure drug DEP, 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 DEP CI 95% b NA NA Measured at end of of Measured at end 1.00 1.00 1.00 1.00 BMJ DEP CI 95% aOR a Measured at startMeasured of https://mc.manuscriptcentral.com/bmj

n n % n % aOR 6 0.0% 12 0.0% NA 13 0.0% 77 0.0% 0.89 to 1.61 0.49 0.85 to 1.54 0.47 91 0.3% 283 0.1% 1.32 to 1.69 1.03 1.16 to 1.49 0.90 39 0.1% 83 0.0% 1.55 to 2.30 1.04 1.32 to 1.96 0.88 912 2.6% 4594 2.1% 1.11* to 1.20 1.03 1.12* to 1.21 1.04 220 0.6% 1169 0.5% 1.03 to 1.19 0.89 1.03 to 1.20 0.89 591 1.7% 3019 1.4% 1.07 to 1.18 0.98 1.10 to 1.21 1.00 286 0.8% 1510 0.7% 1.08 to 1.23 0.95 1.09 to 1.24 0.96 503 1.4% 2230 1.0% 1.14 to 1.27 1.03 1.12 to 1.25 1.01 Cases (n=35,555) Cases (n=35,555) Controls (n=221,772) 9470 9470 26.6% 57341 25.9% 0.95* to 0.98 0.92 0.99 to 1.02 0.96 7853 7853 22.1% 47517 21.4% 0.99 to 1.02 0.96 0.95* to 0.98 0.92 3114 3114 8.8% 17752 8.0% 1.04 to 1.09 1.00 1.04 to 1.08 0.99 4468 4468 12.6% 19010 8.6% 1.39* to 1.45 1.34 1.26* to 1.32 1.20 5708 5708 16.1% 30035 13.5% 1.10* to 1.14 1.06 1.08* to 1.11 1.04 2239 2239 6.3% 10630 4.8% 1.22* to 1.28 1.16 1.17* to 1.23 1.11 3791 3791 10.7% 28866 13.0% 1.00 3791 3791 10.7% 28866 13.0% 1.00 6534 6534 18.4% 36702 16.5% 1.05* to 1.09 1.02 1.04 to 1.07 1.01 10163 10163 28.6% 55534 25.0% 1.13* to 1.16 1.09 1.07* to 1.10 1.04 31463 31463 88.5% 191183 86.2% 1.12* to 1.16 1.08 1.11* to 1.16 1.07 20056 20056 56.4% 119509 53.9% 1.02 to 1.05 1.00 1.02 to 1.05 0.99 Confidential: 24092 67.8% 148297 For 66.9% 0.97 Review to 1.00 0.94 0.98 to 1.02 0.95 Only Cardiovascular Cardiovascular Gastro-intestinal Gastro-intestinal Respiratory 8870 24.9% 53297 24.0% 0.96* to 0.99 0.93 0.95* to 0.98 0.93 Antidepressant Antidepressant Antipsychotic Other Analgesic Analgesic Other Other Analgesic Analgesic Antipsychotic Antidepressant Antipsychotic Gastro-intestinal Parkinsonian Respiratory Urological 1296 3.6% 7854 3.5% 0.93 to 0.99 0.88 0.93 to 0.99 0.87 Parkinsonian Parkinsonian Respiratory Other New ACB2 New ACB3 No prescription No prescription Any ACB1 Any ACB1 New ACB 3 3 New ACB New ACB 2 2 New ACB Any ACB 1 Any ACB 1 No prescription No prescription Any use by ACB score and class class ACB and Any use score by eTable 3. Adjusted odds ratios of dementia by any new prescription of a probably or potently anticholinergic medication and by drug class drug byand medication anticholinergic potently or probably a of prescription new any by ofdementia ratios odds Adjusted 3. eTable Any use DEP during ACBAny use score by burden cognitive Anticholinergic ACB interval; confidence ratio; CI odds adjusted aOR, period; exposure drug DEP, period the of start the at measured forcovariates adjusted a p<0.01 * Page 45 of 70 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 46 of 70 CI 95% b 1.00 1.00 1.00 95% CI aOR BMJ a 0.6% 1.12 to 1.26 1.00 1.10 to1.24 0.97 1.1% 1.02 to 1.12 0.93 1.01 to1.10 0.92 2.4% 1.10* to 1.18 1.04 1.07 to1.14 1.00 9.6% 1.02 to 1.06 0.98 1.02 to1.06 0.98 2.2% 0.90* to 0.97 0.84 0.91 to0.98 0.85 1.9% 0.96 to 1.04 0.89 0.96 to1.04 0.89 0.3% 1.48* to 1.72 1.28 1.28* to1.49 1.10 7.5% 1.01 to 1.05 0.97 1.00 to1.04 0.96 9.5% 1.02 to 1.06 0.98 1.01 to1.05 0.98 5.9% 1.24* to 1.29 1.19 1.17* to1.23 1.12 27.9% 1.13* to 1.16 1.10 1.08* to1.11 1.06 16.0% 0.98 to 1.01 0.95 0.96 to0.99 0.93 47.6% 1.07* to 1.09 1.04 1.03 to1.06 1.01 40.4% 1.00 40.4% 1.00 11.8% 0.96 to 1.00 0.93 0.97 to0.00 0.93 16.0% 1.10* to 1.14 1.07 1.09* to1.12 1.06 10.4% 1.34* to 1.39 1.30 1.22* to1.27 1.18 13.8% 0.94* to 0.97 0.91 0.93* to0.96 0.90 16.9% 1.03 to 1.06 1.00 1.02 to1.05 0.99 17.5% 0.94* to 0.97 0.91 0.94* to0.97 0.92 https://mc.manuscriptcentral.com/bmj

951 951 1,707 3,247 6,688 6,176 5,513 79,114 45,301 33,364 45,295 29,393 39,251 47,934 27,358 21,246 26,921 49,734 16,750 135,248 135,248 114,814 114,814 114,814 0.9% 1.6% 3.2% 2.3% 2.0% 0.7% 7.9% 7.9% 32.4% 17.3% 51.7% 35.8% 35.8% 12.9% 19.1% 14.9% 14.3% 18.6% 10.6% 10.5% 17.8%

n n % n % aOR 367 367 649 649 921 921 832 832 270 270 Cases (n=40,770) Cases (n=40,770) Controls (n=283,933) DEP at start Measured of DEP of Measured at end 7,038 5,275 7,778 6,061 5,820 7,602 1,314 4,320 3,206 4,281 7,240 3,236 13,203 21,074 14,585 14,585

Confidential: For Review Only Analgesic Analgesic Antipsychotic Antidepressant Antidepressant Antidepressant Antipsychotic Gastro-intestinal Gastro-intestinal Antipsychotic Antipsychotic Other Other Gastro-intestinal Gastro-intestinal Cardiovascular Cardiovascular Gastro-intestinal Gastro-intestinal Parkinsonian Parkinsonian Respiratory Respiratory Respiratory Respiratory Other Other Urological Urological ADS3 ADS3 ADS2 ADS2 ADS1 ADS1 ADS1 No prescription No prescription No prescription ADS3 ADS3 ADS2 ADS2 Any use by ADS score and class class ADSscore and Any use by eTable 4. Adjusted odds ratios of dementia by prescription of an anticholinergic medication as measured by the Anticholinergic drug scale and by drug class class drug by and scale drug Anticholinergic by the asmeasured medication anticholinergic an of prescription by ofdementia ratios odds Adjusted 4. eTable class Drug ADSscore Any use by scale drug ADSAnticholinergic interval; confidence CI ratio; odds adjusted aOR, ratio, odds OR, period; exposure drug DEP, the of period start the at measured for covariates adjusted a p<0.01 * 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

95% CI CI 95% 0.90 to 0.99 0.99 to 0.90 0.94 to 1.03 1.03 to 0.94 1.04 to 0.86 0.96 to 1.05 1.05 to 0.96 1.25 to 0.82 1.17 to 1.35 1.35 to 1.17 1.00 to 1.15 1.15 to 1.00 1.26 to 1.01 0.99 to 1.28 1.28 to 0.99 1.19 to 0.77 0.93 to 1.03 1.03 to 0.93 1.86 to 0.31 1.10 to 0.96 1.11 to 1.21 1.21 to 1.11 0.40 to 3.15 3.15 to 0.40 2.29 to 0.69 1.17 to 0.88 1.34 to 1.51 1.51 to 1.34 1.13 to 1.02 1.24 to 1.12 0.88 to 0.97 0.97 to 0.88 1.09 to 1.95 1.95 to 1.09

b 0.94 0.94 0.98 0.98 0.94 1.01 1.01 1.01 1.07 1.07 1.13 1.13 1.13 0.96 0.98 0.98 0.76 1.03 1.12 1.12 1.26 1.01 1.00 1.00 1.00 1.00 aOR BMI smoking, abuse, alcohol for Unadjusted 95% CI CI 95% BMJ

a Patients without missing data at start of DEP (15,214 cases, 48,486 controls) controls) cases, 48,486 (15,214 DEP of start at missingdata without Patients 1.00 1.00 1.00 1.00 aOR https://mc.manuscriptcentral.com/bmj 95% CI CI 95% smoking, BMI BMI smoking, Adjusted for covariates measured at the start of the drug exposure period period drug exposure the of thestart at measured covariates for Adjusted

b 1.00 1.00 1.00 1.00 Unadjusted for alcohol abuse, abuse, alcohol for Unadjusted

95% CI CI 95% aOR All patients (40,770 cases, 283,933 controls) controls) 283,933 cases, (40,770 patients All

Confidential:a For Review Only 1.07 1.07 1.15 to 1.00 1.07 1.15 to 1.00 0.99 1.13 1.13 to 0.87 0.99 1.28 to 0.99 1.12 to 0.87 0.96 1.19 to 0.77 1.02 1.02 1.05 to 1.00 1.02 1.04 to 0.99 1.03 1.02 1.15 to 0.92 1.03 1.06 to 0.97 1.15 to 0.92 1.02 1.26 to 0.82 1.00 1.00 0.95* 0.95* 0.97 to 0.92 0.95* 0.97 to 0.92 0.94 0.99 to 0.90 1.23* 1.23* 1.28 to 1.18 1.23* 1.28 to 1.18 1.26* 1.36 to 1.17 1.26*

Analgesic Analgesic Antidepressant Antipsychotic Cardiovascular 1.37* Gastro-intestinal 1.05* Respiratory 1.42 to 1.32 0.97 0.97 Other 1.08 to 1.02 1.37* 0.99 to 0.94 0.99 to 0.94 Analgesic 1.05* 0.99 Antipsychotic 1.42 to 1.32 0.96 0.97 Parkinsonian 1.02 to 0.97 1.08 to 1.02 Respiratory 1.44 Other 0.99 to 0.94 1.43* 1.00 0.99 to 0.94 1.55* 1.08* 2.36 to 0.87 Antidepressant 0.89 2.14 to 1.12 1.52 1.02 to 1.34 to 0.97 Antipsychotic 0.94 0.98 1.43 Gastro-intestinal 1.13* 1.56* 1.45 to 0.55 1.14 to 1.02 Parkinsonian 1.09 0.94 0.97 Respiratory 1.16 to 1.10 0.99 to 0.90 0.90 1.42* 2.35 to 0.87 1.03 to 0.94 2.15 to 1.13 Urological 1.08* 1.13* 1.45* 1.18 to 1.02 0.99 to 0.89 Other 1.46 to 1.02 0.55 to 0.93 1.04 0.93* 1.13 1.09 1.68 to 1.25 0.94 1.23 1.16 to 1.10 1.44* 1.08 to 1.00 0.74 1.18 to 1.023.17 to 0.40 0.99 to 0.89 2.25 to 0.68 1.18* 1.04 1.68 to 1.24 1.82 to 0.30 1.01 0.94 1.24 1.08 to 1.12 to 1.00 1.47* 1.17 to 0.88 1.03 to 0.86 1.03 1.18* 1.97 to 1.10 1.11 to 0.97 1.46* No prescription prescription No No prescriptions prescriptions No 1.00 ACB1 ACB2 ACB3 ACB 1 prescription prescription 1 ACB prescription 2 ACB prescription 3 ACB 1.11* 1.10* 1.15 to 1.07 1.16* 1.17 to 1.03 1.11* 1.19 to 1.13 1.10* 1.16* 1.15 to 1.07 1.16 to 1.03 1.19 to 1.13 1.08 1.13 1.16* 1.16 to 1.01 1.26 to 1.01 1.21 to 1.11 1.16* Any use by ACB score and class class scoreand ACB by use Any

etable 5a. Adjusted odds ratios of dementia by prescription of a potent anticholinergic by drug class, by adjustment for BMI, smoking and alcohol abuse and by patientsby and abuse alcohol and smoking BMI, for adjustment by drug byclass, anticholinergic potent aof prescription by ofdementia ratios odds Adjusted 5a. etable class drug and score ACB score ACB by use Any aOR index massbody BMI, score; Burden Cognitive Anticholinergic ACB, interval; confidence CI, ratio; odds adjusted aOR, ratio, odds OR, period; exposure drug DEP, p<0.01 * with complete data. Adjustments made for covariates as measured at the start of the drug exposure period. period. exposure drug the start of theat measured as forcovariates made Adjustments data. complete with Page 47 of 70 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 48 of 70

95% CI 95% CI 0.89 to 1.02 1.02 to 0.89 0.97 to 2.21 2.21 to 0.97 1.09 to 1.22 1.22 to 1.09 0.96 to 1.02 1.02 to 0.96 1.17 to 0.89 1.00 to 1.11 1.11 to 1.00 1.01 to 1.17 1.17 to 1.01 0.97 to 1.04 1.04 to 0.97 1.45 to 0.46 1.17 to 0.97 0.99 to 1.08 1.08 to 0.99 0.86 to 1.16 1.16 to 0.86 1.07 to 1.15 1.15 to 1.07 1.16 to 1.28 1.28 to 1.16 1.05 to 1.12 1.12 to 1.05 0.94 to 1.01 1.01 to 0.94 0.91 to 0.98 0.98 to 0.91 0.91 to 1.10 1.10 to 0.91 0.80 to 3.12 3.12 to 0.80 0.91 to 0.98 0.98 to 0.91 1.08 to 1.59 1.59 to 1.08 1.02 to 1.09 1.09 to 1.02

d 1.00 1.00 1.00 0.95 0.95 1.47 1.47 0.99 0.99 1.02 1.02 1.05 1.05 1.09 1.09 1.00 1.00 0.82 1.07 1.03 1.03 1.00 1.00 0.97 0.97 1.00 1.00 0.94 0.94 1.58 1.58 aOR

BMI smoking, abuse, alcohol for Unadjusted 95% CI CI 95% BMJ

c Patients without missing data at end of DEP (26,642 cases, 126,075 controls) controls) 126,075 cases, DEP (26,642 of end at missingdata without Patients 1.00 1.00 1.00 aOR

https://mc.manuscriptcentral.com/bmj smoking, BMI BMI smoking, CI 95% d Adjusted for covariates measured at the end of the drug exposure period period exposure drug the of theend at measured covariates for Adjusted 1.00 1.00 1.00 aOR Unadjusted for alcohol abuse, abuse, alcohol for Unadjusted

95% CI CI 95% All patients (40,770 cases, 283,933 controls) controls) 283,933 cases, (40,770 patients All

c 1.09 1.09 1.17 to 1.01 1.09 1.17 to 1.01 1.07 1.18 to 0.98 0.99 0.99 1.13 to 0.87 0.98 1.12 to 0.86 1.01 1.17 to 0.87 0.83 0.83 1.36 to 0.51 0.88 1.44 to 0.54 0.75 1.33 to 0.42 1.03 1.03 1.07 to 1.00 1.03 1.07 to 0.99 1.04 1.08 to 0.99 0.99 0.99 1.02 to 0.97 1.00 1.02 to 0.97 1.00 1.04 to 0.97 1.03 1.03 1.16 to 0.92 1.02 1.14 to 0.91 1.04 1.19 to 0.91 1.00 1.00 0.98 0.98 1.01 to 0.95 0.97 1.00 to 0.94 0.97 1.01 to 0.93 1.00 1.00 1.32 1.32 1.82 to 0.96 1.37 1.89 to 1.00 1.40 2.10 to 0.93 1.07 1.07 1.16 to 1.00 1.07 1.15 to 0.99 1.01 1.11 to 0.91 1.35 1.35 2.23 to 0.82 1.32 2.18 to 0.80 1.65 3.27 to 0.84 1.04 1.04 1.07 to 1.01 1.04 1.07 to 1.01 1.05* 1.09 to 1.02 1.05* 0.95* 0.95* 0.98 to 0.92 0.95* 0.97 to 0.92 0.95* 0.98 to 0.91 0.94* 1.18* 1.18* 1.23 to 1.13 1.17* 1.22 to 1.12 1.16* 1.22 to 1.10 1.16* 1.10* 1.10* 1.16 to 1.03 1.09* 1.16 to 1.02 1.09 1.18 to 1.01 1.10* 1.10* 1.15 to 1.06 1.12* 1.16 to 1.07 1.07 1.13 to 1.01 1.11* 1.11* 1.14 to 1.08 1.10* 1.13 to 1.08 1.09* 1.13 to 1.06 1.09* Confidential: For Review Only 1.29* 1.50 to 1.11 1.29* 1.50 to 1.11 1.31* 1.59 to 1.07 1.31* Gastro-intestinal Gastro-intestinal 0.94 0.99 to 0.89 0.94 0.99 to 0.89 0.95 1.01 to 0.89 Parkinsonian Parkinsonian Other Other Other Antipsychotic Antidepressant Antidepressant 1.11* 1.14 to 1.08 1.11* 1.14 to 1.07 1.11* 1.16 to 1.07 1.11* Parkinsonian Parkinsonian Respiratory Respiratory Respiratory Urological Other Other Analgesic Analgesic Cardiovascular 1.02 1.04 to 0.99 1.00 1.03 to 0.98 1.01 1.04 to 0.98 Analgesic Analgesic Antidepressant Antidepressant Gastro-intestinal 1.25* Respiratory 1.30 to 1.20 0.96* 0.99 to 0.93 1.25* Antipsychotic 1.30 to 1.20 0.96 1.22* 0.99 to 0.94 1.28 to 1.16 0.97 1.22* 1.01 to 0.94 Antipsychotic Antipsychotic No prescription prescription No ACB3 ACB3 ACB 2 prescription prescription 2 ACB No prescriptions prescriptions No prescription 1 ACB ACB1 ACB2 ACB2 ACB 3 prescription prescription 3 ACB Any use by ACB score and class class scoreand ACB by use Any etable 5b. Adjusted odds ratios of dementia by prescription of a potent anticholinergic by drug class, by adjustment for BMI, smoking and alcohol abuse and by patients patients by and abuse alcohol and smoking BMI, for adjustment by drug class, by anticholinergic potent a of prescription by dementia of ratios odds Adjusted 5b. etable period. exposure drug the end of theat measured as forcovariates made Adjustments data. complete with class drug and score ACB score ACB by use Any aOR index massbody BMI, score; Burden Cognitive Anticholinergic ACB, interval; confidence CI, ratio; odds adjusted aOR, ratio, odds OR, period; exposure drug DEP, p<0.01 *

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 49 of 70 BMJ

1 ISAC APPLICATION FORM 2 PROTOCOLS FOR RESEARCH USING THE CLINICAL PRACTICE RESEARCH DATALINK (CPRD) 3 4 ISAC use only: IMPORTANT 5 Protocol Number ...... If you have any queries, please contact ISAC Secretariat: 6 Date submitted ...... [email protected] 7 8 1. Study Title 9 Benzodiazepine, non benzodiazepine derivatives, and anticholinergic use and incident dementia 10 2. Principal Investigator (full name, job title, organisation & e-mail address for correspondence regarding this 11 Confidential:protocol) For Review Only 12 George Savva, Senior Lecturer, University of East Anglia, [email protected] 13 3. Affiliation (full address) 14 School of Health Sciences, University of East Anglia, Norwich Research Park, Norwich, NR4 7TJ 15 4. Protocol’s Author (if different from the principal investigator) 16 Kathryn Richardson, University of East Anglia, [email protected] 17 5. List of all investigators/collaborators (please list the names, affiliations and e-mail addresses* of all collaborators, 18 other than the principal investigator) 19 Kathryn Richardson, University of East Anglia, [email protected] 20 Chris Fox, University of East Anglia, [email protected] 21 Ian Maidment, Aston University, [email protected] 22 Antony Arthur, University of East Anglia, [email protected] Yoon Loke, University of East Anglia, [email protected] 23 Nicholas Steel, University of East Anglia, [email protected] 24 Kathleen Bennett, Trinity College Dublin, [email protected] 25 Carol Brayne, University of Cambridge, [email protected] 26 Noll Campbell, Purdue University, [email protected] 27 Phyo Myint, University of Aberdeen, [email protected] 28 Malaz Boustani, University of Indiana, [email protected] Carlota Grossi, University of East Anglia, [email protected] 29 30 *Please note that your ISAC application form and protocol must be copied to all e-mail addresses listed above at the time of 31 submission of your application to the ISAC mailbox. Failure to do so will result in delays in the processing of your application. 32 33 6. Type of Institution (please tick one box below)

34 Academia Research Service Provider Pharmaceutical Industry 35 NHS Government Departments Others 36 7. Financial Sponsor of study 37 38 Pharmaceutical Industry (please specify) Academia(please specify) 39 Government / NHS (please specify) None 40 Other (please specify) Alzheimer’s Society 41 8. Data source (please tick one box below) 42 Sponsor has on-line access Purchase of ad hoc dataset 43 Commissioned study 44 Other (please specify) 45 9. Has this protocol been peer reviewed by another Committee? 46 47 Yes* No 48 49 * Please state in your protocol the name of the reviewing Committee(s) and provide an outline of the review process and outcome. 50 10. Type of Study (please tick all the relevant boxes which apply) 51 52 Adverse Drug Reaction/Drug Safety Drug Use Disease Epidemiology 53 Drug Effectiveness Pharmacoeconomic Other 54 11. This study is intended for: 55 56 Publication in peer reviewed journals Presentation at scientific conference Presentation at company/institutional meetings Other Public 57 events /educational meetings in primary and secondary care 58 1 59 60 https://mc.manuscriptcentral.com/bmj BMJ Page 50 of 70

1 2 12. Does this protocol also seek access to data held under the CPRD Data Linkage Scheme? 3 4 Yes No 5 6 7 13. If you are seeking access to data held under the CPRD Data Linkage Scheme*, please select the 8 source(s) of linked data being requested.

9 Hospital Episode Statistics Cancer Registry Data** 10 Confidential: MINAP For ONS Mortality Review Data Only 11 Index of Multiple Deprivation/ Townsend Score 12 Mother Baby Link Other: (please specify) 13 14 * As part of the ISAC review of linkages, the protocol may be shared - in confidence - with a 15 representative of the requested linked data set(s) and summary details may be shared - in confidence - 16 with the Confidentiality Advisory Group of the Health Research Authority.

17 **Please note that applicants seeking access to cancer registry data must provide consent for 18 publication of their study title and study institution on the UK Cancer Registry website. Please contact 19 the CPRD Research Team on +44 (20) 3080 6383 or email [email protected] to discuss this requirement 20 further. 21 22 14. If you are seeking access to data held under the CPRD Data Linkage Scheme, have you already 23 discussed your request with a member of the Research team? 24 25 Yes No* 26 27 *Please contact the CPRD Research Team on +44 (20) 3080 6383 or email [email protected] to discuss 28 your requirements before submitting your application.

29 Please list below the name of the person/s at the CPRD with whom you have discussed your request. 30 Antonis Kousoulis 31 32 15. If you are seeking access to data held under the CPRD Data Linkage Scheme, please provide the 33 following information: 34 35 The number of linked datasets requested: 1 36 A synopsis of the purpose(s) for which the linkages are required: 37 Index of Multiple Deprivation as we are going to match on this, and 38 39 Townsend Score as we are going to adjust for this in a sensitivity analysis. 40 41 Is linkage to a local dataset with <1 million patients being requested? 42 43 Yes* No 44 45 * If yes, please provide further details: 46 47 48 49 16. If you have requested linked data sets, please indicate whether the Principal Investigator or any of 50 the collaborators listed in response to question 5 above, have access to any of the linked datasets 51 in a patient identifiable form, or associated with a patient index. 52 Yes* No 53 54 * If yes, please provide further details: 55 56

57 58 2 59 60 https://mc.manuscriptcentral.com/bmj Page 51 of 70 BMJ

1 2 17. Does this protocol involve requesting any additional information from GPs? 3 4 Yes* No 5 6 * Please indicate what will be required: ψ 7 Completion of questionnaires by the GP Yes No 8 Provision of anonymised records (e.g. hospital discharge summaries) Yes No 9 Other (please describe) 10 ψ Confidential:Any questionnaire for completion by GPs or Forother health Reviewcare professional must be Only approved by ISAC 11 before circulation for completion. 12 18. Does this protocol describe a purely observational study using CPRD data (this may include the 13 review of anonymised free text)? 14 15 Yes* No** 16 * Yes: If you will be using data obtained from the CPRD Group, this study does not require separate 17 ethics approval from an NHS Research Ethics Committee. 18 ** No: You may need to seek separate ethics approval from an NHS Research Ethics Committee for this 19 study. The ISAC will provide advice on whether this may be needed. 20 19. Does this study involve linking to patient identifiable data from other sources? 21 22 Yes No

23 20. Does this study require contact with patients in order for them to complete a questionnaire? 24 25 Yes No 26 N.B. Any questionnaire for completion by patients must be approved by ISAC before circulation for 27 completion. 28 21. Does this study require contact with patients in order to collect a sample?

29 Yes* No 30 31 * Please state what will be collected 32 22. Experience/expertise available 33 34 Please complete the following questions to indicate the experience/expertise available within the team of researchers 35 actively involved in the proposed research, including analysis of data and interpretation of results 36 Previous GPRD/CPRD Studies Publications using GPRD/CPRD data 37 None 38 1-3 39 > 3 40 41 Yes No Is statistical expertise available within the research team? 42 If yes, please outline level of experience KR, GS, and KB are experienced 43 statisticians. KB is an Assoc Prof in Pharmacoepidemiology and co-investigator on a current CPRD ovarian cancer study. 44 KR has experience analysing CPRD data (with paper on childhood asthma in press) and attended the LSHTM Practical 45 Pharmacoepidemiology course (which included CPRD training). 46 47 Is experience of handling large data sets (>1 million records) available within the research team? 48 If yes, please outline level of experience KR has used Canada’s British 49 Columbia linked primary care data (>4 million patients) and has 5 publications using this. KB regularly works with Irish 50 pharmacy dispensing data and has many publications on this. AA has 3 publications using QResearch data. 51 52 Is UK primary care experience available within the research team? If yes, please outline level of experience NS has significant research 53 experience and previous UK primary care clinical experience, and has supervised a PhD student using CPRD data. 54 55 23. References relating to your study 56 57 See reference list below in section 15. 58 3 59 60 https://mc.manuscriptcentral.com/bmj BMJ Page 52 of 70

1 PROTOCOL CONTENT CHECKLIST 2 In order to help ensure that protocols submitted for review contain adequate information for protocol 3 evaluation, ISAC have produced instructions on the content of protocols for research using CPRD data. 4 These instructions are available on the CPRD website (www.cprd.com/ISAC). All protocols using CPRD 5 data which are submitted for review by ISAC must contain information on the areas detailed in the 6 instructions. IF you do not feel that a specific area required by ISAC is relevant for your protocol, you will 7 need to justify this decision to ISAC.

8 Applicants must complete the checklist below to confirm that the protocol being submitted includes all the 9 areas required by ISAC, or to provide justification where a required area is not considered to be relevant 10 for a specific protocol. Protocols will not be circulated to ISAC for review until the checklist has been 11 completedConfidential: by the applicant. For Review Only 12 13 Please note, your protocol will be returned to you if you do not complete this checklist, or if you answer ‘no’ and fail to include justification for the omission of any required area. 14 15 Included in 16 protocol? 17 Required area Yes No If no, reason for 18 omission 19 Lay Summary (max.200 words) 20 Background 21 Objective, specific aims and rationale 22 Study Type 23 Descriptive 24 Hypothesis Generating 25 Hypothesis Testing 26 Study Design 27 Sample size/power calculation 28 (Please provide justification of 29 sample size in the protocol) 30 Study population 31 (including estimate of expected number of relevant patients in the CPRD) 32 Selection of comparison group(s) or controls 33 34 Exposures, outcomes and covariates Exposures are clearly described 35 Outcomes are clearly described 36 Use of linked data

37 (if applicable) 38 Data/ Statistical Analysis Plan 39 There is plan for addressing confounding

40 There is a plan for addressing missing data † 41 Patient/ user group involvement 42 Limitations of the study design, data sources 43 and analytic methods 44 Plans for disseminating and communicating study results 45

46 † It is expected that many studies will benefit from the involvement of patient or user groups 47 in their planning and refinement, and/or in the interpretation of the results and plans for 48 further work. This is particularly, but not exclusively true of studies with interests in the 49 impact on quality of life. Please indicate whether or not you intend to engage patients in any 50 of the ways mentioned above.

51 Voluntary registration of ISAC approved studies: 52 Epidemiological studies are increasingly being included in registries of research around the world, including 53 those primarily set up for clinical trials. To increase awareness amongst researchers of ongoing research, 54 ISAC encourages voluntary registration of epidemiological research conducted using MHRA databases. This 55 will not replace information on ISAC approved protocols that may be published in its summary minutes or 56 annual report. It is for the applicant to determine the most appropriate registry for their study. Please inform the ISAC secretariat that you have registered a protocol and provide the location. 57 58 4 59 60 https://mc.manuscriptcentral.com/bmj Page 53 of 70 BMJ

1 2 PROTOCOL 3 4 5 6 1 Lay Summary 7 8 There is no effective cure for dementia and so identifying modifiable risk factors is essential. 9 The long term use of medications known to affect the brain for short durations may cause 10 permanent harm. Some of these medications are widely used among older people, and so 11 evenConfidential: small increases in individual risks could For lead to Review large increases in the numberOnly of people 12 with dementia. In this project we will examine whether the use of certain medicines that affect 13 cognition increase the risk of dementia. We will also examine whether the risk persists or is 14 reduced once the medication is stopped. 15 16 Clinical trials rarely examine the long term cognitive harms of medications, and so 17 ‘observational’ studies are needed. We will identify CPRD patients who, since April 2004, were 18 diagnosed with dementia aged 65-99 years. We will compare the medication history of each 19 patient with dementia to that of 7 people of the same age and sex who had no dementia (on 20 the same date). We will compare the use and dosage of medications prescribed during the 21 period beyond 4 years before dementia diagnosis, whilst controlling for other health and social 22 differences. 23 24 Our findings will contribute to prescribing guidance and practice for management of many 25 common long term conditions and have the potential to reduce the future risks of dementia. 26 27 28 29 2 Objectives, specific aims and rationale 30 Our broad aim is to test whether the short term cognitive effects of commonly used 31 medications translate into an increased risk of dementia incidence with longer term use. 32 33 As we intend this project to be hypothesis testing (rather than hypothesis generating), we 34 focus on examining medications with adverse cognitive effects for which there has been (1) 35 consistent evidence of short-term effects, (2) published evidence of potential long-term 36 effects, and (3) are commonly prescribed rather than obtained over-the-counter (OTC). Hence 37 we aim to examine the use of benzodiazepines, non-benzodiazepine derivatives, and 38 39 medications with anticholinergic activity. 40 41 2.1 Objectives 42 We will conduct a nested case-control study among CPRD patients to: 43 44 1. determine whether the use of benzodiazepines (BZD), non benzodiazepine derivatives (Z- 45 drugs), or medications with anticholinergic activity increases the risk of dementia, and 46 whether risk increases with greater duration, dosage or concurrent use. 47 2. determine whether these effects persist beyond medication cessation. 48 49 2.2 Aims 50 All hypotheses were generated a priori based on published studies (see background section). 51 52 For (1) benzodiazepines, (2) Z-drugs, and (3) medications with anticholinergic activity we will: 53 54 1. test whether older people with dementia (aged 65-99 and diagnosed since April 2004) 55 were more likely to have been prescribed a greater quantity of medication than people not 56 diagnosed with dementia in the period beyond 4 years before the index date (defined by 57 the dementia diagnosis date), whilst adjusting for confounding factors. 58 5 59 60 https://mc.manuscriptcentral.com/bmj BMJ Page 54 of 70

1 2 2. asses the evidence for a dose-response relationship by: 3 a. examining duration and quantity prescribed (i.e. defined daily doses [DDDs]); 4 b. and for anticholinergics - examine whether dementia risk increases with greater 5 anticholinergic burden 6 3. examine the above associations in subgroups of patients with depression, anxiety or 7 insomnia. 8 4. test whether this association for exposures beyond 4 years pre diagnosis persists once 9 adjusted for the exposures occurring 0-4 years pre diagnosis. 10 5.Confidential: test if the association remains once medicationFor Reviewuse is discontinued. Only 11 6. test for how many years since discontinuation the association remains within categories of 12 previous use (light, medium, heavy user). 13 14 15 2.3 Rationale 16 • Aim 1 will address objective 1. Drug exposure will not be examined in the 4 years 17 before index date to reduce issues of protopathic bias and surveillance bias, and 18 because the drug exposure would be expected to cause dementia over a longer period 19 20 of time. 21 • Aim 2 will provide evidence of a dose-response to further support objective 1, hence 22 strengthen the case for a causal effect if one exists [1]. 23 • Aim 3 also supports objective 1 by additionally reducing the impact of confounding by 24 indication and protopathic bias. 25 • Aim 4 will examine a potential mechanism by addressing whether exposure beyond 4 26 years before index is just a marker for 0-4 year exposure and the effect is only short 27 lived. Associations remaining for exposures beyond 4 years before index once adjusted 28 for 0-4 year exposure would indicate evidence of a cumulative effect. 29 • Aim 5 will address objective 2. 30 • Aim 6 will further examine objective 2 by looking at whether the effect of stopping 31 depends on prior cumulative use. 32 33 34 35 36 3 Background 37 Dementia is a syndrome of progressive decline in cognitive and daily function causing 38 significant patient, carer and societal burden. There are estimated to be between 670,000 and 39 850,000 people with dementia in the UK [2,3]. The number of dementia cases worldwide are 40 expected to double over the next 20 years [4]. Much is still not understood about the causes 41 and risk factors for dementia. There is no cure for dementia and so the identification of 42 potentially modifiable risk factors is essential. 43 44 Many middle aged and older people take multiple medications for the prevention or 45 management of chronic diseases. Yet there is concern over the impact of some medication use 46 on cognitive health. A recent systematic review of randomised controlled trials examined the 47 risks of short term cognitive impairment due to benzodiazepines, anticholinergics, 48 antihistamines, antipsychotics and opioid use [5]. Consistent amnestic and nonamnestic 49 cognitive impairments were found for benzodiazepine use, but evidence was less consistent 50 for nonbenzodiazepine derivatives. Antihistamines and anticholinergic tricyclic antidepressants 51 induced non-amnestic deficits in attention and information processing. There was insufficient 52 evidence surrounding anticholinergic bladder relaxant drugs, opioids and antipsychotics to 53 draw conclusions. 54 55 Whether these potential short-term cognitive impairments lead to an increased risk in 56 dementia is unknown. Benzodiazepine use has been associated with increased dementia 57 58 6 59 60 https://mc.manuscriptcentral.com/bmj Page 55 of 70 BMJ

1 2 incidence in longer term observational studies, but these studies have been limited by their 3 size, adjustment for confounding, or type of database used [6–9]. They rarely adjust for the 4 use of other classes of medications with adverse cognitive effects. A recent Canadian study 5 found a greater incidence of dementia in those taking more than 90 DDDs of benzodiazepines 6 in the period 5-10 years before dementia diagnosis [7]. Data from Taiwanese medical records 7 also suggest that elevated risks persist for at least three years after benzodiazepine use 8 cessation, depending on the previous usage [8]. 9 10 Benzodiazepines are indicated to treat anxiety and insomnia and are commonly used by older 11 people.Confidential: Benzodiazepines are only indicated For for short-term Review use (2 to 4 weeks) Only because they are 12 highly addictive with serious withdrawal reactions, however studies show that many people 13 take benzodiazepines for many years [10]. In the NHS England in 2013, 15.4 million 14 benzodiazepine prescriptions were dispensed, costing £73 million [11]. Benzodiazepines are 15 known to affect memory, concentration and coordination, but whether this continues after 16 cessation is controversial. More recently, non-benzodiazepines called ‘Z-drugs’ (zolpidem, 17 zopiclone, and zaleplon) have been developed to have an improved safety profile than 18 benzodiazepines in the treatment of insomnia, but their cognitive effects are not so well 19 understood, and the improved safety profile requires further confirmation. 20 21 Many of the medications with short term cognitive effects have anticholinergic activity. Longer 22 term observational studies have also linked medications with anticholinergic activity with 23 cognitive decline and dementia [12,13]. Many commonly used medicines including some used 24 for urinary incontinence, depression, asthma, and heart problems block the neurotransmitter 25 acetylcholine which affects many body systems and is important for forming memories. 26 Anticholinergic medications can cause dry mouth, constipation, urinary retention, low blood 27 pressure and also confusion, difficulty concentrating, agitation and memory problems. These 28 side effects have been assumed to be temporary, but recent studies suggest that long term 29 use of anticholinergics may cause cognitive decline and dementia. 30 31 Estimates of the prevalence of definite anticholinergic use in the elderly ranges from 4%-27%, 32 with around 42%-47% regularly using medications that possibly possess anticholinergic activity 33 [14–16]. Taking a few medications that are only possibly or potentially anticholinergic for a 34 longer duration may pose long-term cognitive risks, as our project team members have found 35 that people taking at least 3 possible anticholinergics for 90 days or longer had a higher risk of 36 37 mild cognitive impairment, but not dementia [17]. However, an American study found an 38 increased risk of dementia after 90 DDDs of strong anticholinergics [13]. A study in France of a 39 small sample of 372 people aged 60 and over found a higher risk of mild cognitive impairment 40 among people taking anticholinergic medications; there was no identified increased risk of 41 dementia during an eight year follow-up period, but the study lacked statistical power [18]. In 42 2011 we used data from 13,004 participants of the MRC Cognitive Function and Ageing Study 43 to show that people taking anticholinergic medication not only had worse cognition, but also 44 additional cognitive decline over two years compared to people who did not take such 45 medications [14]. So while anticholinergics are known to bring short term cognitive harms, 46 evidence of their long term effects is conflicted. 47 48 Antihistamine use was also shown to have short-term cognitive effects [5], but many 49 antihistamines have anticholinergic activity. However, as most antihistamines are purchased 50 OTC rather than prescribed, we will not be able to examine this medication class specifically, 51 but will adjust for non-anticholinergic antihistamine use in our analyses. 52 53 The main limitations to observational studies in this area are confounding, in particular 54 confounding by indication, and protopathic bias. For example depression is an indication for 55 the anticholinergic antidepressants, yet also associated with increased dementia risk [19]. 56 Protopathic bias, by which the use of these medications is a marker of prodromal symptoms of 57 58 7 59 60 https://mc.manuscriptcentral.com/bmj BMJ Page 56 of 70

1 2 dementia, is a concern. Studies have suggested that symptoms of depression and anxiety may 3 occur in the period before dementia diagnosis [20]. 4 Hence we focus our hypothesis testing study on the use of benzodiazepines, Z-drugs, and 5 medications with anticholinergic activity as these have the greatest evidence of short- and 6 7 long-term cognitive effects. We will adjust for the other medication classes mentioned in the 8 literature as confounders and will aim to address confounding and protopathic bias. This 9 project aims to refute or confirm the effects of anticholinergic medications, benzodiazepines 10 and Z-drugs on dementia incidence. 11 Confidential: For Review Only 12 13 14 4 Study Type - Hypothesis testing 15 16 17 5 Study Design - Nested case-control study 18 19 6 Sample size 20 We plan to use data from all available people with dementia meeting our definition to 21 22 maximise precision. The following power calculations are based on the estimated 34,125 cases 23 (with 238,875 matched controls) identified during the CPRD feasibility check. This number is 24 substantially larger than any currently published study on this topic. 25 The prevalence of medication use and health conditions were estimated from the CPRD GOLD 26 27 sample dataset v1.5 in 631 people aged over 65 years. Over a 6 year period, 6%, 3%, and 11% 28 were given prescriptions for benzodiazepines, Z-drugs and definite anticholinergics lasting 90 29 days or longer. A diagnosis of depression, insomnia, and anxiety was received by 19%, 17%, 30 and 18% of patients, respectively. 31 32 Previous studies generally suggest an increased risk of dementia or cognitive decline after 90 33 days or 90 DDDs of exposure, rather than after any use of a benzodiazepine or anticholinergic 34 medication [7,13,17]. We have estimated the smallest detectable effect sizes for detecting a 35 difference in the proportion of cases and controls with 90 days of prescription exposure 36 assuming a 2-sided test with alpha as 0.01. The smallest odds ratios (OR) detectable are 1.08, 37 1.12, and 1.06 for 90 days of BZD, Z-drug, and definite anticholinergic exposure at 80% power. 38 39 40 41 7 Use of linked data (if applicable) 42 Index of Multiple Deprivation will be used in the matching criteria, and we will adjust for 43 44 Townsend Score in a sensitivity analysis. 45 46 8 Study population 47 First patients will be excluded if they ever have diagnoses of: 48 49 • Motor neuron disease (read code F152%), HIV/AIDS (A788%, A789%), Multiple 50 sclerosis (F20%), Down’s syndrome (PJ0%), Alcoholism or related diseases (see 51 appendix for read codes). 52 o Rationale: Conditions associated with a greater risk of dementia through 53 different pathologies, and differential use of BZDs or anticholinergics. 54 55 Dementia cases definition criteria (with rationale for decisions below): 56 57 58 8 59 60 https://mc.manuscriptcentral.com/bmj Page 57 of 70 BMJ

1 2 1. Patients with first dementia event recorded between 01/04/2006 and the practice 3 “Last Collection Date” (this first date defines the “Index Date”). 4 o Rationale: Date chosen to capture dementia coded since the start of the 5 Quality Outcomes Framework (QOF) and more recent medication use. 6 2. First dementia event defined as the first of a dementia read code (see appendix) or a 7 prescription for a dementia drug (memantine, donepezil, rivastigmine, or galantamine) 8 which is followed by a dementia code within a year. 9 o Rationale: Dementia read codes from QOF excluding alcohol induced 10 Confidential:dementia. Dementia drugs For added dueReview to occasional suspected Only delay in 11 diagnosis coding (i.e. from specialists), and to be in line with other validated 12 definitions [21,22]. 13 3. Patients aged between 65 and 99 years at the index date 14 o Rationale: Age criteria due to the vast majority of dementias occurring at this 15 age, more homogeneous dementia, and consistency with other publications. 16 4. Patients have at least 6 years of continuous follow-up with Up-To-Standard (UTS) data 17 (and acceptable records) before the index date. 18 o Rationale: This is to enable at least 1 year of drug exposure prior to 4 years 19 before diagnosis and to enable drug exposure to occur after at least one year 20 of UTS data, which gives sufficient time for confounder and indication 21 22 estimation, and recording to have stabilised for patients new to the practice 23 [23]. 24 Exclusion criteria: 25 26 • A read code for “dementia annual review” before the index date (read code 6AB.00). 27 o Rationale: Due to dementia diagnosis date being missing. 28 29 CPRD feasibility estimates returned 34,125 potential dementia patients in CPRD (using the 30 June 2015 version of the database) before the health conditions above are excluded. 31

32 33 34 8.1 Selection of controls 35 All cases will be categorised according to the number of years of acceptable, UTS, continuously 36 registered data available before index date. The number of years before four years before the 37 index date (excluding the first year of UTS follow-up) will define the drug exposure period. See 38 figure 1 which summarises the study design. 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 Figure 1. Study design 58 9 59 60 https://mc.manuscriptcentral.com/bmj BMJ Page 58 of 70

1 2 7 controls per case will be selected (without replacement) using incidence density sampling 3 [24,25] from all patients (including cases pre diagnosis) in CPRD with the following matching 4 criteria on the index date: 5 1. Same sex 6 7 2. Year of birth +/- 3 years o 8 Rationale: Age and sex matches to select control patients similar to dementia 9 patients. 10 3. No dementia diagnosis or dementia medication on or before the index date (see 11 Confidential:appendix). For Review Only 12 o Rationale: To ensure no dementia events exist before the index date for 13 controls. 14 4. The same number of years of continuous UTS follow-up before the index date as the 15 case. 16 o Rationale: To enable the same duration and calendar time for the exposure 17 period. The controls will be assigned the same drug exposure period as the 18 case. 19 5. Same quintile of practice level Index of Multiple Deprivation (IMD). 20 o Rationale: To enable matching on area level deprivation which correlates with 21 comorbidity, health behaviours and dementia risk. 22 23 24 The ratio of 7 controls per case was chosen to maximise controls (10 controls per case 25 previously found to give as precise estimates as use of the full cohort [26]) whilst being below 26 the total 300,000 CPRD patient limit. When incidence density sampling is used to select 27 controls, the estimated odds ratios are unbiased estimates of incidence rate ratios [24]. 28 29 30 31 32 9 Exposures 33 The drug exposure period is defined by the number of years before the index date excluding 34 the four years before index and the first year of UTS follow-up (see figure 1). The main 35 exposures of interest are BZDs, Z-drugs, and medications with anticholinergic activity. We will 36 extract all prescriptions for patients occurring during their exposure period. Extracted 37 information will include the date of each prescription, the drug name, the number of daily 38 doses instructed, the dose and quantity prescribed. 39 40 Our primary exposure variable is the number of defined daily doses (DDD) and we hypothesise 41 a relationship with dementia risk above 90 DDDs of exposure (based on published studies). We 42 will calculate the total dose of each prescription by multiplying the dose of each tablet by the 43 number of tablets prescribed. To enable comparison of doses across drug classes, we will 44 convert the total dose for each medication to a number of DDDs, defined as the assumed 45 average maintenance dose per day for a drug based on its main indication in adults, using the 46 DDD values assigned by the World Health Organisation’s Collaborating Centre for Drug 47 Statistics Methodology (www.whocc.no/atc_ddd_index). 48 49 We will also estimate the duration of each prescription in days by dividing the number of 50 tablets prescribed by the dosing instructions (e.g. number of tablets to be taken per day). 51 When the dosing instructions are missing, we will assume an estimate based on the median 52 tablets to be taken per day for the same drug in patients with dosing instructions. 53 54 Benzodiazepines will be defined as those in WHO-Anatomical Therapeutic Chemical (ATC) 55 categories N05BA or N05CD and the Z-drugs defined as Zopiclone, Zaleplon, or Zolpidem (ATC 56 N05CF). 57 58 10 59 60 https://mc.manuscriptcentral.com/bmj Page 59 of 70 BMJ

1 2 The anticholinergic properties of prescribed medications will be assessed using the 3 Anticholinergic Cognitive Burden (ACB) scale (www.agingbraincare.org/tools/abc- 4 anticholinergic-cognitive-burden-scale/) [27,28]. This is a frequently updated scale that 5 classifies, through expert consensus and literature review, the evidence for the anticholinergic 6 activity of medications. Medications with serum anticholinergic activity or in vitro affinity to 7 muscarinic receptors but with no known clinically relevant negative cognitive effects are 8 scored 1 ‘possible’ , while drugs with established and clinically relevant anti-cholinergic effects 9 are scored 2 ‘probable’ based on blood-brain penetration and 3 ‘definite’ if also have reported 10 associationsConfidential: with delirium. We will classify For separately Review medication exposure Only according to 11 exposure to (1) possible, (2) probable, and (3) definite anticholinergics. 12 13 For a patient at a given point in time, total anti-cholinergic burden is defined as the sum of ACB 14 scores for all medications taken. We will estimate the cumulative anticholinergic burden as the 15 mean ACB score over the exposure period (by summing each medication’s duration multiplied 16 by its ACB score, and dividing by the length of the exposure period) and the duration taking at 17 least 3 possible anticholinergics simultaneously. 18 19 Anticholinergic medication exposure will also be assessed within medication classes of 20 antidepressants, first-generation antipsychotics, second-generation antipsychotics, urologicals, 21 respiratory, and cardiovascular drugs. Exposures will also be assessed for the 5 most 22 commonly prescribed BZDs or Z-drugs. 23 24 As a sensitivity analysis, other anticholinergic medication scales will be used instead of the 25 ACB. We will code the medications according to both the US based Anticholinergic Drug Scale 26 (ADS) and the Anticholinergic Risk Scales (ARS), which also quantify the degree of anti- 27 cholinergic activity exhibited by a medication into scores of 0, 1, 2, or 3 according to literature 28 review and expert consensus [29,30]. We will also quantify the anticholinergic component of 29 the Drug Burden Index over the exposure period, which estimates anticholinergic burden using 30 the daily dose and the minimum efficacious dose [31]. 31 32 As a secondary analysis, by considering all medication exposures occurring during the exposure 33 period and the 4 years prior to the index date, we will quantify time since discontinuation for 34 (1) BZD/Z-drugs and (2) definite anticholinergics. We will categorise never users, current users 35 and the time since discontinuation in groups. We will also classify cumulative use by DDDs pre 36 discontinuation (in 3 groups of light, medium and heavy user). 37 38 39 40 41 10 Covariates 42 Patients will already be matched on sex, birth year (+/- 3 years), calendar time and IMD 43 quintile. It is important to capture to the greatest extent possible the health status of each 44 patient (particularly all indications for the medications and conditions that have a greater risk 45 of dementia). We measure covariates at 4 years prior to dementia rather than at the index 46 date, as this is when the drug exposure period ends and we want to limit the amount of events 47 captured that are consequences of taking the medications. 48 49 The following additional covariates will be adjusted for in regression analyses. 50 51 10.1 Covariates measured at 4 years prior to the index date: 52 • Age (years) 53 • Body Mass Index, calculated as weight (in kg) divided by height2 (in m) using the latest 54 records and categorised [32,33]. 55 • Smoking status (non-smoker, ex-smoker, current smoker) using most recent record 56 [34]. 57 58 11 59 60 https://mc.manuscriptcentral.com/bmj BMJ Page 60 of 70

1 2 • Previous diagnoses (at any point in records up to 4 years prior to the index date): 3 o Insomnia, depression and anxiety (see appendix for codes). Duration with 4 insomnia, depression and anxiety (time since first diagnosis, categorised), 5 depression severity (mild, moderate, severe), phobia, drug abuse, symptoms 6 of depression, symptoms of anxiety, time since first depression symptom 7 (categorised), time since first anxiety symptom (categorised) [35,36]. 8 9 o Family history of dementia recorded (where recorded). 10 o Severe Mental Illnesses (SMI) - schizophrenia, bipolar disorder, delusional 11 Confidential:disorder, depressive psychosis, For schizoaffective Review disorder, brief Only psychoses, and 12 psychoses not otherwise specified [37,38]. 13 o Parkinson’s disease, Parkinson’s disease severity - prescription for drugs 14 containing levodopa (ATC N04BA), MAO-B inhibitor prescription (N04BD), 15 COMT-inhibitor prescription (N04BX01, N04BX02). 16 o Diabetes, Diabetes with complications, dyslipidaemia, Mood affective 17 disorders, Meniere’s disease, restless leg syndrome, Epilepsy, Hemiplegia and 18 paraplegia, Neuropathic pain, migraine, Hypertension, stroke, transient 19 ischaemic attack, angina, Cerebrovascular disease, Congestive heart disease, 20 Myocardial infarction, Peripheral vascular disease, Other atherosclerotic 21 disease, Atrial fibrillation, Deep vein thrombosis, Acute coronary syndrome, 22 Coronary artery bypass graft, Percutaneous coronary intervention, Other 23 cardiovascular surgery, Osteoarthritis, rheumatoid arthritis, other connective 24 tissue disease, chronic back pain, Chronic obstructive pulmonary disease, 25 26 asthma, rhinitis, other allergic disease, GERD, reflux oesophagitis, Peptic ulcer, 27 gastric ulcer, Irritable bowel syndrome, Inflammatory Bowel disease, Chronic 28 Diarrhoea, Gastric Bypass Surgery, intestinal surgery, Mild liver disease, 29 Moderate or severe liver disease, Dermatitis, eczema, psoriasis, Urinary 30 incontinence/overactive bladder, Prostatism, Renal disease, Cancer, 31 Metastatic tumour [39–41]. 32 33 10.2 Covariates measured during the exposure period and the year prior: 34 • Recent stress/bereavement [42–44]. 35 • Recent agitation 36 37 10.3 Covariates measured during the exposure period: 38 • Mean alcohol consumption (categorised by units per week) [45]. 39 Blood pressure (categorised) - average systolic blood pressure (in mmHg) [46,47]. 40 • 41 • Mean annual frequency of GP visits (categorised). 42 • Categorised DDDs and duration of exposure to non-anticholinergic medications with 43 evidence of cognitive effects: 44 o Other all CNS active medications by class: 45
Hypnotics and anxiolytics (BNF 4.1) 46
First-generation antipsychotics, second-generation antipsychotics (BNF 47 4.2) 48
Antidepressants (BNF 4.3) 49
Drugs used in Nausea and vertigo (BNF 4.6) 50
Analgesics - Opioids (BNF 4.7.2), non-opioids (rest of BNF 4.7) 51
Antiepileptic drugs (BNF 4.9) 52
Drugs used in Parkinson’s disease (BNF 4.9) 53 o Antihistamines (BNF 3.4.1) not listed on the anticholinergic scale. 54 • Any statin prescription, antiplatelet agent prescription, ACE inhibitor prescription. 55 • Average number of all other cardiovascular and non-cardiovascular medications 56 57 prescribed (excluding topical medications and eye drops). 58 12 59 60 https://mc.manuscriptcentral.com/bmj Page 61 of 70 BMJ

1 2 3 4 5 11 Data/ Statistical Analysis 6 Stata (version 14) will be used for data management and statistical analysis. All reported P 7 values reported will be two tailed, with P<0.01 defining significance. Alpha of 0.01 was chosen 8 as being equivalent to a Benforroni correction for the five primary exposure outcomes. 9 10 First,Confidential: we will explore the internal validity For of the dementia Review diagnosis in the cases.Only In line with 11 other dementia studies [21], we will examine the proportion of cases with a second 12 (confirmatory) dementia-related code, e.g. additional dementia medication prescriptions, 13 additional dementia diagnosis codes, a dementia annual review, a specific dementia test prior 14 to diagnosis (e.g. Mini Mental State Examination, Clock Drawing Test, or Abbreviated Mental 15 Test [7-Minute Screen]), a referral to a specialist (e.g. neurologist, geriatrician or psycho- 16 17 geriatrician), or an assessment based on neuroimaging technique (e.g. magnetic resonance 18 imaging, computed tomography, or single photon emission computed tomography), or 19 dementia symptoms (e.g. memory impairment, aphasia, apraxia, or agnosia). Dementia cases 20 will be also be described according to their first coded sub-type. 21 22 Second, the characteristics 4 years prior to the index date of the cases and controls will be 23 described (percentages, means (SD), median (IQR) where relevant). We will summarise the 24 usage patterns of medications with adverse cognitive effects over the exposure period. 25

26 27 11.1.1 Primary analysis 28 29 We will assess the association between benzodiazepine exposure, anticholinergic exposure 30 and dementia using conditional logistic regression. Odds ratios (and 95% confidence intervals) 31 will be provided unadjusted and adjusted for exposure to each class of medications with 32 adverse cognitive effects and the covariates listed above. Where covariates are categorised 33 they will be grouped into sensible categories according to the distribution of the data, i.e. 34 tertiles or quintiles for continuous measures. 35 36 The association between dementia and the primary exposure of DDDs will be examined, for (i) 37 benzodiazepines, (ii) Z-drugs, and medications with (iii) possible, (iv) probable and (v) definite 38 anticholinergic activity. The model will adjust for the covariates as well as the number of DDDs 39 of the other medications with potential adverse cognitive effects. 40 41 The association between dementia and the following secondary exposures will be separately 42 examined: 43 • Duration of prescriptions. Adjusting for the covariates as well as the duration of 44 prescriptions of the other medications with adverse cognitive effects. 45 • Any prescription. Adjusting for the covariates as well as any prescriptions for the other 46 medications with adverse cognitive effects. 47 • For anticholinergics - average ACB sum (categorised) and duration taking 3 or more 48 possible anticholinergics concurrently (categorised). 49

50 51 11.1.1.1 Sensitivity analyses 52 53 1) For the primary analysis above, the interactions between medication use and both age and 54 sex will be tested, with subgroup results reported if p<0.05. 55 2) For a potential anticholinergic synergistic effect we will test for an interaction between 56 definite and possible anticholinergic use (defined as the DDDs of simultaneous use of the 57 definite and possible anticholinergics). 58 13 59 60 https://mc.manuscriptcentral.com/bmj BMJ Page 62 of 70

1 2 3) We will repeat the primary analysis: 3 a) In cases (and their matched controls) with a second dementia-related code or 4 medication as a subgroup with ‘confirmed’ dementia due to the potential for cases 5 with only one code to be misclassified. 6 b) In cases (and their matched controls) with the index date defined as the first diagnosis 7 code for dementia. 8 9 c) In three subgroups: those with a diagnosis of insomnia, anxiety and depression. 10 d) Excluding patients with Parkinson’s disease, Severe Mental Illness, or drug abuse. 11 Confidential: For Review Only 12 e) Using a covariate selection procedure, i.e. retaining those that affect the odds ratio by 13 10% or more. 14 f) Adjusting for patient level deprivation in those with data linkage, using the Townsend 15 deprivation score in deciles [48]. 16 g) Adjusting only for confounders measured up to the start of the drug exposure period, 17 so to not adjust for variables potentially on the causal pathway during the drug 18 exposure period. However, this analysis has the limitation of potentially under 19 20 adjusting for the medication indications. 21 4) The anticholinergic analyses will be repeated with the ADS and ARS scales instead of the 22 ACB. As the Drug Burden Index does not categorise anticholinergic burden, we will only 23 compare the average score of the anticholinergic component over the exposure period. 24 5) To examine whether associations observed for exposures beyond 4 years pre index are 25 only due to an increased propensity to continue taking those medications closer to 26 diagnosis, we will also adjust for exposures in the 0-4 years before the index date, and 27 28 shall also split exposures in the period beyond 4 years into several 2-year periods. 29 6) Anticholinergic medications will also be assessed within medication classes of 30 antidepressants, first-generation antipsychotics, second-generation antipsychotics, 31 urologicals, respiratory, and cardiovascular drugs. The effect of the 5 most commonly 32 prescribed BZDs or Z-drugs will also be assessed. The Benjamini-Hochberg procedure will 33 be applied to these analyses to control the false discovery rate at 5% to take into account 34 the multiple tests being conducted [49]. 35 36 37 11.1.2 Secondary analysis 38 Finally, we will examine the association between medication cessation and dementia 39 incidence. Here we will consider all medication exposures occurring during the exposure 40 period and in the 4 years before the index date. The association between time since 41 discontinuation (categorised) and dementia will be examined separately using conditional 42 logistic regression for (1) BZD/Z-drugs and (2) definite anticholinergics, adjusting for 43 confounders up to 4 years prior to the index date. Associations will also be provided according 44 to cumulative use pre discontinuation (in three groups of light, medium and heavy user, 45 46 defined by DDDs). 47 In sensitivity analyses we will adjust for confounders measured up to the index date (this will 48 include adjusting for new drug use in that period), due to being unable to adjust for 49 confounders at the exact date of medication cessation. 50 51 11.1.3 Missing data 52 53 In the primary analyses, patients with missing covariate data will be coded in a missing data 54 category. For covariates with at least 10% of patients with missing data, we will summarise the 55 characteristics of those with and without missing data, and examine sensitivity analyses using 56 (1) multiple imputations (5 sets) based on reasonable assumptions and subject to 57 computational feasibility, (2) single imputation, and (3) restricting to those with complete data. 58 14 59 60 https://mc.manuscriptcentral.com/bmj Page 63 of 70 BMJ

1 2 3 4 11.2 Limitations of the study design, data sources and analytic methods 5 11.2.1 Study design limitations 6 • The nested case-control study design includes prevalent medication users rather than 7 only new users of the medications so may over-represent those who tolerate the drugs 8 [50]. However, the nested case-control study design is best placed to deal with 9 complex time-varying and longer term medication exposure [51]. We also include 10 indications in a wide period including at least one year before the prescriptions, and 11 Confidential: For Review Only preliminary data analysis reveals that this is more likely to capture the medication 12 13 indications than through a new user design. 14 • Our results may suffer from protopathic bias, in which the use of these medications is 15 a marker of prodromal symptoms of dementia [20]. Examining medication use many 16 years before dementia diagnosis will help to limit this bias as well as subgroup analyses 17 of those who have depression, anxiety and insomnia whilst also adjusting for time 18 since diagnosis. 19 20 11.2.2 Analytic methods limitations 21 • Our dementia cases and controls could be misclassified. However, we will use read 22 codes included in the QOF business rules for dementia and only include dementia 23 diagnosed since this was introduced in the QOF, thus improving the quality of 24 dementia reporting. We also used an algorithm similar to one validated by GPs 25 confirming dementia in 95% of cases [21]. Using older CPRD data, another study found 26 a sensitivity and specificity of 100% in 150 cases and 50 controls sampled and sent GP 27 letters [22]. We will also assess what proportion of our cases have more than one 28 dementia related code including dementia diagnoses, symptoms, medications, 29 specialist referrals, neuroimaging assessments, or cognitive tests. We will also perform 30 two sensitivity analysis with more specific dementia definitions. 31 32 • We cannot be sure of the timing of dementia onset. By definition diagnosis in primary 33 care records is delayed as the patient will have had symptoms for some time before. 34 As a slowly developing degenerating disease, the disease onset is not a precise point in 35 time. Not examining medication exposures in the period 4 years prior to diagnosis for 36 our primary analysis reduces the impact of any errors in the recording of this timing. 37 • Dementia is also under-diagnosed in primary care [3], hence we could misclassify some 38 early dementias as controls which would have the effect of reducing the strength of 39 our associations. 40 • Our analysis will be subject to confounding by indication. We will address this by (1) 41 adjusting for many indications of the medications, (2) analysing subgroups with just 42 the main indication, and (3) additionally adjusting for many diseases and related 43 symptoms to best possibly capture the reasons for prescription, as many medications 44 will be prescribed outside of their guidelines [38]. 45 • The ACB scale used to evaluate anticholinergic activity of medications has not been 46 validated against in vitro measures of anticholinergic activity, although it is uncertain 47 whether assays reflect anti-cholinergic activity in the brain, and not all relevant drugs 48 49 have been assayed in vitro [52]. The ACB scale was chosen as it is commonly use and 50 has been recently updated with respect to mediations currently used in the UK [27]. 51 Due to some medications potentially being misclassified using this scale, we will also 52 perform sensitivity analyses using different anticholinergic scales. 53 54 11.2.3 Data source limitations 55 • We cannot be sure the patients take the prescribed medications, and adherence 56 would be difficult to ascertain. This would be more of a concern if our analysis was 57 examining patients prescribed one medication, as they may not have taken it, but as 58 15 59 60 https://mc.manuscriptcentral.com/bmj BMJ Page 64 of 70

1 2 our primary hypotheses concentrate on those receiving multiple prescriptions (i.e. >90 3 days) we can be more sure the patient has taken at least some of them [53,54]. Having 4 prescription data also enables detailed and accurate information on dosing and 5 prescribing instructions. 6 • We may under-estimate medication use as we will lack data on medications purchased 7 over-the-counter and illegally, and those prescribed in secondary care. However, in 8 most cases of prescriptions given in secondary care the responsibility of longer term 9 prescribing is transferred to primary care via shared care arrangements, or equivalent. 10 Confidential:Also, the majority of the medications For we are Review interested in are prescription Only only and 11 the bulk of prescribing should be in primary care. We may have limited ability to 12 extrapolate the risks of OTC medication use with anticholinergic activity. 13 • Some confounder data will be missing for some patients (e.g. smoking, BMI, alcohol 14 use, blood pressure), however we will examine appropriate techniques to address this. 15 • There is always the possibility of residual confounding. We are missing information on 16 certain disease severities (e.g. depression scores), genetics (e.g. APOE E4 status), and 17 cognitive scores. However, CPRD contains information on all of the most important 18 19 confounding variables in these hypotheses. 20 • There could be issues over the quality of data recording in primary care records, but 21 we will use UTS data only and examine medication exposures after at least one year of 22 UTS to enable recording to have stabilised [23]. The CPRD has also been well validated 23 for a range of conditions [40,55], and we will use QOF definitions or validated read 24 code lists where possible. 25 26 27 28 12 Patient/ user group involvement 29 The funding for this study was awarded by the Alzheimer’s Society grant advisory board which 30 has substantial lay representation. 31 32 As the funder of this study, The Alzheimer’s Society has appointed three Research Network 33 Volunteers to act as study monitors and service user representatives on our study steering 34 committee. These individuals have all acted as carers of people with dementia. 35 36 The monitors contribute to our protocol development by sharing their experiences of 37 psychoactive medication use and their view of the balance between the benefits of medication 38 use and potential cognitive decline. Monitors meet the study team to discuss study progress 39 every six months. They will be particularly concerned with the dissemination phase of our 40 study, making sure that lay summaries of results are accessible and avoid possible 41 misinterpretation. 42 43 44 45 46 13 Plans for disseminating and communicating study results 47 Our communication plan is designed to gain maximum impact for our work among clinicians, 48 patient groups and researchers, and the Alzheimer’s Society are partners to ensure the public 49 and patient message can be developed. Specific dissemination activities are: 50 51 Academic dissemination. We will disseminate our findings though one or two peer-reviewed 52 academic papers in high impact medical journals, as well as including the findings in a planned 53 and funded systematic review and discussion articles to link our work to that of others at the 54 end of the project. Academic conference presentations will be made to clinical and public 55 health audiences. We aim to analyse and publish our findings within one year of receiving the 56 57 58 16 59 60 https://mc.manuscriptcentral.com/bmj Page 65 of 70 BMJ

1 2 CPRD data. Funding for this project currently ends in July 2017, and although we will also apply 3 for an Alzheimer’s Society dissemination grant, activity will be restricted beyond this point. 4 Medical practitioners, planners and policy makers. These groups have already supported our 5 study and asked to be kept informed of progress. We will write articles for professional 6 7 periodicals and give updates about the study on relevant websites (e.g. that of the British 8 Geriatrics Society, NHS Primary Care Commissioning website, and NHS Evidence) and via our 9 regional and professional networks. We will contribute study findings to the professional 10 training programmes at our Universities. Our previous experience of Department of Health 11 consultationsConfidential: will help us to contact key Forpolicy groups. Review Only 12 13 For patient and public groups. We will contribute to newsletters including those for carers and 14 the voluntary sector (e.g. Living with Dementia - Alzheimer's Society newsletter) and websites 15 (e.g. Dementia UK, Alzheimer’s Society). We will also present to lay groups advocating for 16 dementia and older people’s issues. We have comprehensive service user involvement and 17 through this we will explore all opportunities to disseminate study information and 18 implications to relevant people without dementia, people with dementia and their carers. 19 20 Website/social media. We have a project website at www.uea.ac.uk/drug-safety-and- 21 dementia/ to draw attention to the study and provide progress updates. We will use personal, 22 academic, institutional and study specific Twitter accounts to provide details of the 23 publication(s). 24 25 International dissemination. We have established links to groups such as the American 26 Geriatrics Society, the American Association of Geriatric Psychiatry, the European Delirium 27 Association and international public health bodies and we will produce briefings for these. 28 29 30 31 14 Protocol peer review 32 33 34 This exact protocol has not been peer reviewed by a committee; however an earlier simplified 35 version was peer reviewed by our study funder, the Alzheimer’s Society, when submitted as 36 part of a larger project. Our research grant application underwent scientific peer review and 37 lay review by their Research Network volunteers. We provided replies and the application was 38 then considered by a Grant Advisory Board. The scientific peer reviewers provided some 39 specific comments regarding the CPRD analysis, such as that not all dementia patients will 40 have a diagnosis, concerns about protopathic bias, and that anticholinergic burden scales 41 differ. The project was successful, particularly at the Grant Advisory Board stage, and was 42 43 funded in full. 44

45 46 47 48 49 50 51 52 53 54 55 56 57 58 17 59 60 https://mc.manuscriptcentral.com/bmj BMJ Page 66 of 70

1 2 15 References 3 4 5 1 Howick J, Glasziou P, Aronson JK. The evolution of evidence hierarchies: what can Bradford 6 Hill’s ‘guidelines for causation’ contribute? JRSM 2009;102:186–94. 7 doi:10.1258/jrsm.2009.090020 8 9 2 Alzheimer’s Society. Dementia UK: Update [Internet]. Available from: 10 Confidential:http://www.alzheimers.org.uk/site/scripts/download_info.php?fileID=2323 For Review Only (accessed 22 11 Jan2015). 12 13 3 Matthews FE, Arthur A, Barnes LE, et al. A two-decade comparison of prevalence of 14 dementia in individuals aged 65 years and older from three geographical areas of England: 15 results of the Cognitive Function and Ageing Study I and II. The Lancet 2013;382:1405–12. 16 doi:10.1016/S0140-6736(13)61570-6 17 18 4 Ferri CP, Prince M, Brayne C, et al. Global prevalence of dementia: a Delphi consensus 19 study. The Lancet 2006;366:2112–7. 20 21 5 Tannenbaum C, Paquette A, Hilmer S, et al. A Systematic Review of Amnestic and Non- 22 23 Amnestic Mild Cognitive Impairment Induced by Anticholinergic, Antihistamine, GABAergic 24 and Opioid Drugs. Drugs Aging 2012;29:639–58. 25 26 6 Billioti de Gage S, Begaud B, Bazin F, et al. Benzodiazepine use and risk of dementia: 27 prospective population based study. BMJ 2012;345:e6231–e6231. doi:10.1136/bmj.e6231 28 29 7 Billioti de Gage S, Moride Y, Ducruet T, et al. Benzodiazepine use and risk of Alzheimer’s 30 disease: case-control study. BMJ 2014;349:g5205–g5205. doi:10.1136/bmj.g5205 31 32 8 Wu C-S, Ting T-T, Wang S-C, et al. Effect of Benzodiazepine Discontinuation on Dementia 33 Risk. Am J Geriatr Psychiatry 2011;19:151–9. doi:10.1097/JGP.0b013e3181e049ca 34 35 9 Lagnaoui R, Bégaud B, Moore N, et al. Benzodiazepine use and risk of dementia: A nested 36 case–control study. J Clin Epidemiol 2002;55:314–8. doi:10.1016/S0895-4356(01)00453-X 37 38 10 Bartlett G, Abrahamowicz M, Tamblyn R, et al. Longitudinal patterns of new 39 Benzodiazepine use in the elderly. Pharmacoepidemiol Drug Saf 2004;13:669–82. 40 doi:10.1002/pds.908 41 42 11 Prescribing and Primary Care team, Health and Social Care Information Centre. 43 Prescription Cost Analysis England 2013. 2014. 44 45 12 Fox C, Smith TO, Maidment I, et al. An Updated Systematic Review of the Impact of 46 Medications with Anti-Cholinergic Activity on Cognitive function and Mortality. In: 47 Alzheimer’s Association International Conference. Boston, USA: 2013. 48 http://www.rcpsych.ac.uk/pdf/Posters%20Final.pdf (accessed 26 Jul2013). 49 50 13 Gray SL, Anderson ML, Dublin S, et al. Cumulative use of strong anticholinergics and 51 incident dementia: A prospective cohort study. JAMA Intern Med Published Online First: 26 52 January 2015. doi:10.1001/jamainternmed.2014.7663 53 54 14 Fox C, Richardson K, Maidment ID, et al. Anticholinergic Medication Use and Cognitive 55 Impairment in the Older Population: The Medical Research Council Cognitive Function and 56 Ageing Study. J Am Geriatr Soc 2011;59:1477–83. doi:10.1111/j.1532-5415.2011.03491.x 57 58 18 59 60 https://mc.manuscriptcentral.com/bmj Page 67 of 70 BMJ

1 2 15 Richardson K, Bennett K, Maidment I, et al. Use of medications with anti-cholinergic 3 activity and self-reported injurious falls in the older aged community-dwelling population. J 4 Am Geriatr Soc 2015;(In Press). 5 6 16 Ness J, Hoth A, Barnett MJ, et al. Anticholinergic medications in community-dwelling 7 older veterans: Prevalence of anticholinergic symptoms, symptom burden, and adverse 8 drug events. Am J Geriatr Pharmacother 2006;4:42–51. 9 doi:10.1016/j.amjopharm.2006.03.008 10 11 17Confidential: Cai X, Campbell N, Khan B, et al. ForLong-term anticholinergicReview use and Only the aging brain. 12 Alzheimers Dement Published Online First: November 2012. doi:10.1016/j.jalz.2012.02.005 13 14 18 Ancelin ML, Artero S, Portet F, et al. Non-degenerative mild cognitive impairment in 15 elderly people and use of anticholinergic drugs: longitudinal cohort study. BMJ 16 2006;332:455–9. doi:10.1136/bmj.38740.439664.DE 17 18 19 Ownby RL, Crocco E, Acevedo A, et al. Depression and risk for Alzheimer disease: 19 systematic review, meta-analysis, and metaregression analysis. Arch Gen Psychiatry 20 2006;63:530–8. doi:10.1001/archpsyc.63.5.530 21 22 20 Masters MC, Morris JC, Roe CM. ‘Noncognitive’ symptoms of early Alzheimer disease A 23 longitudinal analysis. Neurology 2015;:10.1212/WNL.0000000000001238. 24 doi:10.1212/WNL.0000000000001238 25 26 21 Imfeld P, Brauchli Pernus YB, Jick SS, et al. Epidemiology, Co-Morbidities, and 27 Medication Use of Patients with Alzheimer’s Disease or Vascular Dementia in the UK. J 28 Alzheimers Dis 2013;35:565–73. doi:10.3233/JAD-121819 29 30 22 Dunn N, Holmes C, Mullee M. Does lithium therapy protect against the onset of 31 dementia? Alzheimer Dis Assoc Disord 2005;19:20–2. 32 33 23 Lewis JD, Bilker WB, Weinstein RB, et al. The relationship between time since 34 registration and measured incidence rates in the General Practice Research Database. 35 Pharmacoepidemiol Drug Saf 2005;14:443–51. doi:10.1002/pds.1115 36 37 24 Flanders WD, Louv WC. The Exposure Odds Ratio in Nested Case-Control Studies with 38 Competing Risks. Am J Epidemiol 1986;124:684–92. 39 40 25 Beaumont JJ, Steenland K, Minton A, et al. A Computer Program for Incidence Density 41 Sampling of Controls in Case-Control Studies Nested Within Occupational Cohort Studies. 42 43 Am J Epidemiol 1989;129:212–9. 44 26 Steenland K, Deddens JA. Increased Precision Using Countermatching in Nested Case- 45 46 Control Studies. Epidemiology 1997;8:238–42. 47 48 27 Campbell NL, Maidment I, Fox C, et al. The 2012 Update to the Anticholinergic 49 Cognitive Burden Scale. J Am Geriatr Soc 2013;61:S1–232. doi:10.1111/jgs.12263 50 51 28 Boustani M, Campbell N, Munger S, et al. Impact of anticholinergics on the aging brain: 52 a review and practical application. Aging Health 2008;4:311–20. 53 doi:10.2217/1745509X.4.3.311 54 55 29 Rudolph JL, Salow MJ, Angelini MC, et al. The anticholinergic risk scale and 56 anticholinergic adverse effects in older persons. Arch Intern Med 2008;168:508–13. 57 doi:10.1001/archinternmed.2007.106 58 19 59 60 https://mc.manuscriptcentral.com/bmj BMJ Page 68 of 70

1 2 30 Carnahan RM, Lund BC, Perry PJ, et al. The Anticholinergic Drug Scale as a Measure of 3 Drug-Related Anticholinergic Burden: Associations With Serum Anticholinergic Activity. J 4 Clin Pharmacol 2006;46:1481–6. doi:10.1177/0091270006292126 5 6 31 Narayan SW, Hilmer SN, Horsburgh S, et al. Anticholinergic Component of the Drug 7 Burden Index and the Anticholinergic Drug Scale as Measures of Anticholinergic Exposure in 8 Older People in New Zealand: A Population-Level Study. Drugs Aging 2013;30:927–34. 9 doi:10.1007/s40266-013-0111-y 10 11 32Confidential: Bhaskaran K, Forbes HJ, Douglas ForI, et al. Representativeness Review and optimal Only use of body 12 mass index (BMI) in the UK Clinical Practice Research Datalink (CPRD). BMJ Open 13 2013;3:e003389. doi:10.1136/bmjopen-2013-003389 14 15 33 Bhaskaran K, Douglas I, Forbes H, et al. Body-mass index and risk of 22 specific cancers: 16 a population-based cohort study of 5·24 million UK adults. The Lancet 2014;384:755–65. 17 doi:10.1016/S0140-6736(14)60892-8 18 19 34 Lewis JD, Brensinger C. Agreement between GPRD smoking data: a survey of general 20 practitioners and a population-based survey. Pharmacoepidemiol Drug Saf 2004;13:437–41. 21 doi:10.1002/pds.902 22 23 35 Martinez C, Rietbrock S, Wise L, et al. Antidepressant treatment and the risk of fatal 24 and non-fatal self harm in first episode depression: nested case-control study. BMJ 25 2005;330:389. doi:10.1136/bmj.330.7488.389 26 27 36 Coupland C, Dhiman P, Morriss R, et al. Antidepressant use and risk of adverse 28 outcomes in older people: population based cohort study. BMJ 2011;343:d4551. 29 30 37 Osborn DPJ, Levy G, Nazareth I, et al. Relative risk of cardiovascular and cancer 31 mortality in people with severe mental illness from the United Kingdom’s General Practice 32 Rsearch Database. Arch Gen Psychiatry 2007;64:242–9. doi:10.1001/archpsyc.64.2.242 33 34 38 Marston L, Nazareth I, Petersen I, et al. Prescribing of antipsychotics in UK primary 35 care: a cohort study. BMJ Open 2014;4:e006135. doi:10.1136/bmjopen-2014-006135 36 37 39 Khan NF, Perera R, Harper S, et al. Adaptation and validation of the Charlson Index for 38 Read/OXMIS coded databases. BMC Fam Pract 2010;11:1. doi:10.1186/1471-2296-11-1 39 40 40 Khan NF, Harrison SE, Rose PW. Validity of diagnostic coding within the General 41 Practice Research Database: a systematic review. Br J Gen Pract 2010;60:e128–36. 42 43 doi:10.3399/bjgp10X483562 44 41 Martinez C, Assimes TL, Mines D, et al. Use of venlafaxine compared with other 45 46 antidepressants and the risk of sudden cardiac death or near death: a nested case-control 47 study. BMJ 2010;340:c249. 48 49 42 Martín-Merino E, Ruigómez A, Wallander M-A, et al. Prevalence, incidence, morbidity 50 and treatment patterns in a cohort of patients diagnosed with anxiety in UK primary care. 51 Fam Pract 2009;:cmp071. doi:10.1093/fampra/cmp071 52 53 43 Shah SM, Carey IM, Harris T, et al. Initiation of Psychotropic Medication after Partner 54 Bereavement: A Matched Cohort Study. PLoS ONE 2013;8:e77734. 55 doi:10.1371/journal.pone.0077734 56 57 58 20 59 60 https://mc.manuscriptcentral.com/bmj Page 69 of 70 BMJ

1 2 44 Yang C, Jick SS, Jick H. LIpid-lowering drugs and the risk of depression and suicidal 3 behavior. Arch Intern Med 2003;163:1926–32. doi:10.1001/archinte.163.16.1926 4 5 45 Khadjesari Z, Marston L, Petersen I, et al. Alcohol consumption screening of newly- 6 registered patients in primary care: a cross-sectional analysis. Br J Gen Pract 2013;63:e706– 7 12. doi:10.3399/bjgp13X673720 8 9 46 Delaney JAC, Moodie EEM, Suissa S. Validating the effects of drug treatment on blood 10 pressure in the General Practice Research Database. Pharmacoepidemiol Drug Saf 11 Confidential:2008;17:535–45. doi:10.1002/pds.1553 For Review Only 12 13 47 Davies NM, Kehoe PG, Ben-Shlomo Y, et al. Associations of Anti-Hypertensive 14 Treatments with Alzheimer’s Disease, Vascular Dementia, and Other Dementias. J 15 Alzheimers Dis 2011;26:699–708. doi:10.3233/JAD-2011-110347 16 17 48 Townsend P, Phillimore P, Beattie A, et al. Health and deprivation: inequality and the 18 North. Croom Helm London 1988. 19 20 49 Benjamini Y, Hochberg Y. Controlling the False Discovery Rate: A Practical and 21 Powerful Approach to Multiple Testing. J R Stat Soc Ser B Methodol 1995;57:289–300. 22 doi:10.2307/2346101 23 24 50 Ray WA. Evaluating Medication Effects Outside of Clinical Trials: New-User Designs. Am 25 J Epidemiol 2003;158:915–20. doi:10.1093/aje/kwg231 26 27 51 Essebag V, Genest J, Suissa S, et al. The nested case-control study in cardiology. Am 28 Heart J 2003;146:581–90. doi:10.1016/S0002-8703(03)00512-X 29 30 52 Lampela P, Lavikainen P, Garcia-Horsman JA, et al. Anticholinergic Drug Use, Serum 31 Anticholinergic Activity, and Adverse Drug Events Among Older People: A Population-Based 32 Study. Drugs Aging 2013;30:321–30. doi:10.1007/s40266-013-0063-2 33 34 53 Fischer MA, Stedman MR, Lii J, et al. Primary Medication Non-Adherence: Analysis of 35 195,930 Electronic Prescriptions. J Gen Intern Med 2010;25:284–90. doi:10.1007/s11606- 36 010-1253-9 37 38 54 Weich S, Pearce HL, Croft P, et al. Effect of anxiolytic and hypnotic drug prescriptions 39 on mortality hazards: retrospective cohort study. BMJ 2014;348:g1996–g1996. 40 doi:10.1136/bmj.g1996 41 42 55 Herrett E, Thomas SL, Schoonen WM, et al. Validation and validity of diagnoses in the 43 General Practice Research Database: a systematic review. Br J Clin Pharmacol 2010;69:4–14. 44 doi:10.1111/j.1365-2125.2009.03537.x 45 46 56 Wallander M-A, Johansson S, Ruigómez A, et al. Morbidity Associated With Sleep 47 48 Disorders in Primary Care: A Longitudinal Cohort Study. Prim Care Companion J Clin 49 Psychiatry 2007;9:338–45. 50 51 52

53 54 55 56 57 58 21 59 60 https://mc.manuscriptcentral.com/bmj BMJ Page 70 of 70

1 2 16 Appendix 3 4 5 16.1 Code lists 6 7 16.1.1 Alcoholism or alcohol related disease 8 The alcoholism code lists includes alcoholism, alcoholic dementia, alcoholic liver disease, and 9 alcoholic cirrhosis. 10 Confidential: For Review Only 11 Alcoholism read codes: 1462, E01, E23..00, E23..11, E23z.00, E230, E231, Eu102, Eu103, Eu104, 12 Eu1051, Eu10611, Eu10711, F11x000, F11x011, F144000, F375.00, F394100, G555.00, 13 G852300, J153.00, J610.00, J611.00, J612.00, J612000, J613.00, J613000, J617.00, J617000, 14 15 J671000, ZV11300 16 17 16.1.2 Dementia 18 The dementia code list is the QOF definition excluding alcohol induced dementia, and is very 19 similar to lists published and validated [21]. 20

21 22 Dementia diagnosis read codes: 23 E00, E02y1, E041, Eu00, Eu01, Eu02, Eu041, F110, F111, F112, F116, Fyu3000 24 25 16.1.3 Depression 26 27 The depression code list is the QOF definition excluding depression in dementia, but also 28 including ‘Depressive psychoses’. 29 30 Depression read codes: 31 E11..12, E112, E113, E118., E11y2, E11z2, E130., E135., E2003, E291., E2B.., E2B1, Eu204, 32 Eu251, Eu32, Eu33, Eu341, Eu412 33

34 35 16.1.4 Anxiety 36 The anxiety codes are essentially the ones used in a published and validated study, but 37 excluding anxiety symptoms, interventions and phobias [42]. 38 39 Anxiety read codes: 1466, E200, E292000, Eu054, Eu34114, Eu41, Eu51511, Eu60600. 40 41 42 16.1.5 Insomnia 43 The insomnia/sleep disturbance codes are essentially the ones used in a published study, but 44 including related sleep disorders/disturbance and excluding sleep apnoea [56]. 45

46 47 Insomnia read codes: 1B1B (excluding 1B1B.12), 1B1Q, E274, Eu51, Fy0..00, Fy00, Fy01, Fy02, 48 R005 (excluding R0051, R0053). 49 50 51 52 53 54 55 56 57 58 22 59 60 https://mc.manuscriptcentral.com/bmj