Drugs with anticholinergic effects and cognitive impairment, falls and all-cause mortality in older adults: A systematic review and meta-analysis

Abstract

Aim

The aim was to investigate associations between drugs with anticholinergic effects (DACEs) and cognitive impairment, falls and all-cause mortality in older adults.

Methods

A literature search using CINAHL, Cochrane Library, Embase and PubMed databases was conducted for randomized controlled trials, prospective and retrospective cohort and case-control studies examining the use of DACEs in subjects ≥65 years with outcomes on falls, cognitive impairment and all-cause mortality. Retrieved articles were published on or before June 2013. Anticholinergic exposure was investigated using drug class, DACE scoring systems (anticholinergic cognitive burden scale, ACB; anticholinergic drug scale, ADS; anticholinergic risk scale, ARS; anticholinergic component of the drug burden index, DBI_AC) or assessment of individual DACEs. Meta-analyses were performed to pool the results from individual studies.

Results

Eighteen studies fulfilled the inclusion criteria (total 124 286 participants). Exposure to DACEs as a class was associated with increased odds of cognitive impairment (OR 1.45, 95% CI 1.16, 1.73). Olanzapine and trazodone were associated with increased odds and risk of falls (OR 2.16, 95% CI 1.05, 4.44; RR 1.79, 95% CI 1.60, 1.97, respectively), but amitriptyline, paroxetine and risperidone were not (RR 1.73, 95% CI 0.81, 2.65; RR 1.80, 95% CI 0.81, 2.79; RR 1.39, 95% CI 0.59, 3.26, respectively). A unit increase in the ACB scale was associated with a doubling in odds of all-cause mortality (OR 2.06, 95% CI 1.82, 2.33) but there were no associations with the DBI_AC (OR 0.88, 95% CI 0.55, 1.42) or the ARS (OR 3.56, 95% CI 0.29, 43.27).

Conclusions

Certain individual DACEs or increased overall DACE exposure may increase the risks of cognitive impairment, falls and all-cause mortality in older adults.

Introduction

Drugs with anticholinergic effects (DACEs) are commonly prescribed for the management of different conditions such as depression, psychosis, Parkinson’s disease, muscle spasms, allergy, excessive gastric acid, nausea and vomiting, intestinal motility disorders, overactive bladder and chronic obstructive pulmonary disease 1. Older adults have a relatively high probability of being exposed to DACEs due to their high medical comorbidity and the number of prescribed and over-the-counter medications 1. The wide distribution of muscarinic acetylcholine receptor subtypes (M1–M5) in the central nervous system (CNS) and in the rest of the body largely accounts for the variety of peripheral and CNS adverse effects with DACEs 2–4. Peripheral effects include constipation, dry mouth, dry eyes, tachycardia and urinary retention. CNS effects include agitation, confusion, delirium, falls, hallucinations and cognitive dysfunction.

The neurotransmitter acetylcholine (ACh) is critical for communication between neurons and muscle at the neuromuscular junction for modulating posture and movement, direct neurotransmission in autonomic ganglia, and pathways in the brain that are involved in memory and cognitive function 5,6. In the nucleus basalis, identified basal forebrain cholinergic neurons innervate the cerebral cortex, amygdaloid complex or hippocampus and are necessary for learning and memory formation 6. It has been observed that the use of an anticholinergic drug, such as the antagonist scopolamine, administered to healthy volunteers resulted in impairment of memory function, similar to that seen in Alzheimer dementia 6.

Recent evidence has also demonstrated that DACEs may impair cognitive performance as well as physical function in older adults 7–9. For instance, normal age-related declines in memory could increase with susceptibility to the potential cognitive side effects of DACEs 10. Comorbid conditions in older adults, including Parkinson’s disease and type 2 diabetes, can also predispose to a decline in cognition and amplify the effects of DACEs on cognitive function 10. Cognitive impairment is associated with a large burden of disease in the ageing population restricting daily activities and can result in high care needs 11. Falls are also a major source of hospitalization, long term institutionalization and death in older adults 12. A decline in the ability to ambulate can lead to a decline in ability to perform activities of daily living independently, increased reliance on others for assistance and increased risk of social isolation 13.

Traditional methods for assessing exposure to DACEs were previously based on a dichotomous yes/no approach or the total number of DACEs taken by the patient 14. However, other characteristics such as the daily dose, binding affinity to the muscarinic receptor(s), permeability of the blood–brain barrier, and serum and tissue concentrations, all influence the risk of anticholinergic effects 1. These characteristics, and the identification of an increasing number of DACEs, have led to the development of several DACE scoring systems, e.g. the Anticholinergic Cognitive Burden (ACB) Scale 15, Anticholinergic Drug Scale (ADS) 16, Anticholinergic Risk Scale (ARS) 17 and the anticholinergic component of the Drug Burden Index (DBI_AC) 18. These scales may provide a more useful way to examine the association between overall exposure to DACEs and adverse outcomes than looking at individual medications or classes of medications separately. However, some DACE scoring systems classify drugs using criteria that are not necessarily based on the affinity for muscarinic acetylcholine receptors. For example, trazodone and pramipexole, listed as DACEs in some scoring systems 15,17,18, have shown no affinity for muscarinic acetylcholine receptors 19,20. Inconsistencies in the different methods used for DACE classification provide opportunities for updating and refining such instruments.

Given the potential for adverse effects of DACEs on a vulnerable population, a comprehensive and critical appraisal of the current literature is required in order to synthesize the available evidence of the increased risk of cognitive impairment, falls and mortality in older people. We conducted a systematic review and meta-analysis of studies examining the association between DACEs and these three endpoints specifically in this population.

Methods

Databases and literature search strategy

Relevant studies published by June 13 2013 (the date of the last search) were identified via CINAHL, Cochrane Library, Embase and PubMed electronic databases and by reviewing reference lists. A combination of keywords and MeSH terms was used to formulate the search strategy in the CINAHL database, with analogous terms used for the other databases. The MeSH terms were anticholinergic drugs, antimuscarinic drugs, cholinergic antagonist, muscarinic antagonist, falls, accidental fall, hip fracture, dementia, cognitive impairment, mortality, death and the list of individual DACEs in Supplementary Table 1 (Appendix S1). An additional search on hip fracture was conducted for the assessment of DACE exposure classification to help identify additional studies for falls. No term limits were applied for language, age or human studies to restrict information bias as adding limitations can result in studies being missed 21. Article titles, abstracts and full text articles were screened by two authors (KR, AAM) for potential relevance.

Study inclusion criteria

The following inclusion criteria were employed: randomized controlled trials, prospective and retrospective cohort and case–control studies that examined the use of DACEs specifically in adults aged 65 years and older with outcomes on either falls, cognitive impairment or all-cause mortality. These outcomes are more fully defined in Supplementary Table 2. Only articles written in English were included and no publication date or publication status restrictions were imposed. Only studies with participants who resided in either a community setting, aged-care facility, institution or hospital were included. Within intervention studies, those that assessed the effects of DACEs either as a class, scoring system (ARS score, ACB, ADS, and DBI_AC) or individual drugs based on the ARS score vs. no DACE exposure or placebo were considered.

Classification of exposure to DACEs

Exposure to DACEs was assessed using three different approaches: 1) as a whole drug class whenever DACE use was defined, 2) the use of DACE scoring systems or 3) use of specific DACEs included in a validated list (Supplementary Table 1; 17).

Drug scoring systems

Anticholinergic Cognitive Burden Scale

The ACB identifies the severity of anticholinergic adverse effects on cognition for a list of prescribed and over-the-counter medications 15. Drugs are quantified for anticholinergic burden by either serum anticholinergic activity (SAA) or the in vitro affinity to muscarinic receptors. Drugs with measurable SAA or in vitro affinity to muscarinic receptors, but with no clinically relevant negative cognitive effects, are given a score of 1. A score of either 2 or 3 is assigned to drugs with recognized and clinically relevant cognitive anticholinergic effects. The total determines the patient’s ACB score.

Anticholinergic Drug Scale

The ADS uses a three-level anticholinergic classification system based on 340 medications 14,16. The latter are assigned a rank from 0 (none) to 3 (high) according to clinical experience, the pharmacologic characteristics of each medication and the available ratings for the in vitro anticholinergic activities of the drugs. The individual scores of each drug are summed to determine the patient ADS score.

Anticholinergic Risk Scale

The ARS score was developed following review of 500 drugs within the Veterans Affairs Boston Healthcare System. Each drug is ranked on a scale of 0 (limited or none), 1 (moderate), 2 (strong), and 3 (very strong) based on the dissociation constant for the muscarinic receptor, rates of anticholinergic effects vs. placebo in experimental studies, and a literature review on anticholinergic adverse effects 17. An ARS score is calculated by summing the ARS rankings assigned for each of the prescribed drugs.

Drug Burden Index (anticholinergic component)

The DBI measures the total exposure to both anticholinergic and sedative medications 18. The Total Drug Burden (TDB) is calculated for each patient according to the following equation:

where the B_AC and B_S represent the linear additive burden sum for every anticholinergic (AC) and sedative (S) drug. For each anticholinergic drug the DBI is calculated using the following:

where D is the daily dose and δ is the minimum recommended daily dose approved by the Food and Drugs Administration (FDA).

Data extraction

A comprehensive data extraction form was developed to collect data from eligible studies with the following fields: study design; population and setting, exposure and outcome details, methodological quality and information on analytical models for the statistical analyses. The number of events and non-events for both the exposed and unexposed populations were extracted for each eligible study. Both unadjusted and fully adjusted risk estimates with 95% CIs were extracted for all outcomes of interest and according to each potential DACE exposure classification. One author (KR) extracted the eligible study’s data followed by a second author (AAM) who reviewed the extracted data. Authors from primary studies were contacted when necessary study details were missing and details were considered unattainable if they did not respond despite several reminders.

Quality assessment

Two reviewers (KR, AAM) assessed the risk of bias for each study using the Cochrane risk of bias tool 22 for randomized control trials (RCTs) and the Newcastle-Ottawa scale 23 for non-randomized studies. The Cochrane risk of bias tool assessed six components which included items for sequence generation (randomization), allocation concealment, blinding of participants and personnel, blinding of outcome assessments, incomplete outcome data (attrition and exclusions), selective outcome reporting, and other bias (including topic-specific, design-specific). The Newcastle-Ottawa scale assessed three components: selection of the cohort, comparability of cohorts on the basis of the design or analysis, how the exposure was ascertained and how outcomes of interest were assessed. Studies achieving six or more stars were considered to be of high quality.

Statistical analysis

We performed separate meta-analyses for each of the three outcomes using one or more exposure classifications. Where the raw data for each outcome were not available, the unadjusted risk estimates for each exposure group were combined into either an odds ratio (OR) or rate ratio (RR) with corresponding 95% confidence intervals (CI). An OR or RR above 1 indicated a greater likelihood of the outcome of interest for exposure to DACEs compared with non-exposure. Meta-analyses were performed using a fixed-effects model (Mantel & Haenszel) with inverse variance weighting for each study. However, if significant heterogeneity was observed (Higgin’s I²), a random effects model (DerSimonian & Laird) was also performed. All meta-analyses were performed using STATA software package Version 13 (StataCorp, College Station, TX, USA).

Results

Study characteristics

The search yielded 89 480 citations summarized in Fig.1, according to PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) conventions. Of the 35 341 studies inspected, 35 323 were excluded, leaving 18 studies eligible for the review. Eligible studies comprised a total of 124 286 participants with sample sizes ranging from 50 to 60 746 participants. Study characteristics are summarized in Table 1. Two studies included one patient aged 60 years 24,25. Patients in the remaining studies were ≥65 years. The vast majority of studies (94.4%) included both males and females, with one study only including females (5.6%) 26. Another study included only African-American participants 27. Most studies were conducted in Europe (n = 12), as well as the United States (n = 4), Canada (n = 1) and Australia (n = 1).

Figure 1.

PRISMA study selection flow diagram

The study designs were comprised of prospective cohorts (n = 10), retrospective cohorts (n = 4), case–control studies (n = 2), and randomized control trials (n = 2). Of the two case–control studies 28,29, one was a nested case–control study 29. The follow-up duration ranged from 1.25 to 6 years for cognitive impairment, 1 month to 4 years for falls and 1 month to 3.3 years for all-cause mortality. Sixteen studies used no DACE exposure as the control group while two studies 26,30 used placebo.

The results from the eligible studies are summarized in Table 2. Of the six studies examining falls, one examined DACE exposure as a class 31 while the others assessed the effects of nine individual DACEs 26,30–34. Of the three studies examining cognitive impairment, all considered exposure to DACEs as a class 27,35,36. Of the 10 studies examining all-cause mortality, three investigated exposure as a DACE class 37–39, three examined a total of eight individual DACEs 28,29,32 and four examined DACE exposure through drug scoring systems 24,25,40,41.

Risk of bias

The risk of bias is shown in Supplementary Table 3 and Supplementary Fig.1. The Cochrane Risk of Bias tool for the two RCTs 26,30 demonstrated an overall low risk of bias. Since one of these trials was presented only as an abstract 30 there was insufficient information to permit judgement of bias specific to randomization and blinding procedures. The other RCT trial 26 demonstrated a low risk of bias for randomization, blinding of participants and personnel and completeness of outcome data, whilst risk from allocation concealment and assessor blinding was deemed as ‘unclear’.

The Newcastle-Ottawa scale indicated moderate to high methodological quality for non-RCTs. Both of the two case–control studies had adequate definitions for cases and controls, matched cases and controls for at least two factors with the exception of one drug 29, and used the same methods to ascertain exposure 28,29. All cohort studies (n = 14) selected their non-exposed cohort from the same community as the exposed cohort and controlled for at least one factor with adequate follow-up periods for the outcome of interest. Limitations included recruiting a selective group of participants (n = 4) 33,34,38,41, self-report on the outcome of interest (n = 1) 31, and the presence of a medical condition at the beginning of a study for mortality (n = 4) 32,38,39,41.

Systematic review data

Cognitive impairment

One study 35 showed an increased risk of cognitive impairment in patients prescribed DACEs (HR 1.65, 95% CI 1.00, 2.73). By contrast, two studies 27,36 showed no significant associations between DACEs and cognitive impairment (OR 1.43, 95% CI 0.98, 2.07; HR 0.67, 95% CI 0.40, 1.15, respectively). Differences in the results may be due to study population size, which was larger in Carriere et al. 35 than the other two studies.

Falls

Drug class

Only one study 31 met the criteria for DACE class exposure and falls, categorized as either one or less falls and two or more falls. DACE class was not associated with the risk of falling (adjusted RR 1.18, 95% CI 0.96, 1.47). When no falls was compared with one or more falls, there was a significant increase in the risk of falls for participants exposed to DACEs (RR 1.19, 95% CI 1.06, 1.34).

Individual medications

Of the nine individual medications assessed for their association with falls, imipramine (RR 2.2, 95% CI 1.8, 2.6), mirtazapine (HR 1.19, 95% CI 1.05, 1.36), nortriptyline (RR 2.0, 95% CI 1.8, 2.3), amitriptyline (HR 1.32, 95% CI 1.22, 1.42; RR 1.9, 95% CI 1.7, 2.1), paroxetine (HR 1.45, 95% CI 1.31, 1.59; RR 1.7, 95% CI 1.5, 1.9) and trazodone (HR 1.55, 95% CI 1.29, 1.87; RR 1.2, 95% CI 1.0, 1.4) were each associated with an increased risk of falling 32,34, whereas risperidone (HR 1.32, 95% CI 0.57, 3.06; OR 1.48, 95% CI 0.57, 3.87) 30,33 and oxybutynin (OR 0.44, 95% CI 0.04, 5.19) 26 were not. Olanzapine also showed an increased risk of a fall (HR 1.74, 95% CI 1.04, 2.90) in one study 33 but not in another (OR 1.86, 95% CI 0.73, 4.74) 30.

All-cause mortality

Drug class

Two of the three studies that assessed the association between DACEs as a class and all-cause mortality showed no association (HR 1.12, 95% CI 0.75, 1.68; HR 1.57, 95% CI 0.78, 3.15) 37,39. Another assessed DACE exposure with all-cause mortality by gender 38. Twenty-seven potent DACEs were covered in the study using ATC codes. Although potent DACEs showed no association of all-cause mortality in women, there was a significant association seen for men at 3 months (HR 2.52, 95% CI 1.05, 6.03) and at 1 year (HR 1.99, 95% CI 1.09, 3.63) after adjustment for confounders.

Individual medications

Eight individual DACEs investigated all-cause mortality. Amongst cohort studies, amitriptyline (HR 1.10, 95% CI 1.03, 1.18), mirtazapine (HR 1.76, 95% CI 1.62, 1.91), olanzapine (OR 6.70, 95% CI 1.40, 32.10), paroxetine (HR 1.24, 95% CI 1.14, 1.35) and trazodone (HR 1.82, 95% CI 1.59, 2.08) each showed an increased risk of all-cause mortality 29,32. In case–control studies, quetiapine was associated with increased all-cause mortality (OR 1.80, 95% CI 1.10, 3.00) 28. Clozapine and risperidone were the only two DACEs that were not associated with all-cause mortality (OR 1.80, 95% CI 0.30, 11.20; OR 1.70, 95% CI 0.90, 3.40), respectively 29.

Drug scoring systems

Of the four drug scoring systems, one study included all four and assessed associations at both 3 months and 1 year 41. No associations with the risk of all-cause mortality were observed per each unit increase in the ACB scale at 3 months (HR 1.10, 95% CI 0.70, 1.80) and 1 year (HR 1.10, 95% CI 0.70, 1.80) 41. However, the ACB was associated with increased risk of all-cause mortality at 2 years in another study (HR 1.26, 95% CI 1.20, 1.32) 40. For each unit increase in the ARS score, there was a 60% increased risk in all-cause mortality at 3 months (HR 1.60, 95% CI 1.20, 2.20) and 1 year (HR 1.40, 95% CI 1.10, 1.80) in one study 41 but not in another investigating in-hospital all-cause mortality (HR 1.04, 95% CI 0.67, 1.62) 24. Similar findings were observed for the anticholinergic component of the DBI, with an increased risk in all-cause mortality at 3 months (HR 4.50, 95% CI 1.20, 16.70) and 1 year (HR 3.20, 95% CI 1.10, 9.40) 41, but not in another study at 6.5 months (HR 1.10, 95% CI 0.44, 2.74) 25. The ADS at both 3 months (HR 1.30, 95% CI 0.90, 1.90) and 1 year (HR 1.20, 95% CI 0.90, 1.60) were also not associated with an increased risk in all-cause mortality 41.

Meta-analyses

Of the 18 studies included in the review, only 11 studies were suitable for inclusion in a meta-analysis based on having equivalent exposure classifications 24,25,27,30,32–36,40,41. Figs.4 summarize the meta-analyses results for the 11 studies.

Figure 4.

Forest plot of odds ratios of all-cause mortality with drugs with anticholinergic effect scoring systems (dashed line indicates line of no effect). OR: odds ratio; CI: confidence interval; a: random effects analysis; ACB Scale: Anticholinergic Cognitive Burden Scale; ARS Score: Anticholinergic Risk Scale Score; DBI Score: Drug Burden Index (anticholinergic component) Score. The odds ratio refers to the odds of an event in those exposed to drugs with anticholinergic effects compared with the control exposure

Figure 2.

Forest plot of odds ratios of cognitive impairment with the use of drugs with anticholinergic effects as a class (dashed line indicates line of no effect). OR: odds ratio of an event in those exposed to drugs with anticholinergic effects compared to the control exposure; CI confidence interval

Figure 3.

Forest plot of risk estimates (odds ratio or relative risk) of falls with the use of individual drugs with anticholinergic effects (dashed line indicates line of no effect). CI: confidence interval; a: random effects analysis; b: analysis using odds ratio (OR). The odds ratio or relative risk refers to the odds or risk of an event in those exposed to drugs with anticholinergic effects compared with the control exposure

Cognitive impairment

Of the three included studies, DACE exposure was associated with increased odds of cognitive impairment (OR 1.45, 95% CI 1.16, 1.73). In a fixed effect analysis there was no evidence of heterogeneity being present (I² = 0%).

Falls

Olanzapine and trazodone were each associated with an increased odds or risk of falling (OR 2.16, 95% CI 1.05, 4.44 and RR 1.79, 95% CI 1.60, 1.97 respectively). No heterogeneity was present for olanzapine (I² = 0%). However, some heterogeneity was present in the trazodone analysis (I² = 28.2%) in fixed effect analyses. Exposure to amitriptyline, paroxetine, and risperidone was not associated with an increased risk of falling (RR 1.73, 95% CI 0.81, 2.65, RR 1.80, 95% CI 0.81, 2.79 and RR 1.39, 95% CI 0.59, 3.26 respectively). In the fixed effect analysis risperidone displayed no significant heterogeneity (I² = 0%) whilst amitriptyline and paroxetine were assessed with random effects models due to substantial heterogeneity (I² = 97.5% and 97.7%, respectively).

All-cause mortality

Out of the three different DACE scoring systems, only the ACB scale showed a significant association with all-cause mortality (OR 2.06, 95% CI 1.82, 2.33), while the DBI (OR 0.88, 95% CI 0.55, 1.42) and ARS score (OR 3.56, 95% CI 0.29, 43.27) did not. The ACB scale and DBI analyses showed no heterogeneity (I² = 0%), however, the ARS scale showed moderate heterogeneity (I² = 87.4%) and a random effect analyses was therefore conducted.

Discussion

This is the first comprehensive systematic review and meta-analysis that quantifies the effect of DACEs as either a drug class, drug scoring system or as individual drugs on clinically relevant outcomes, i.e. falls, cognitive impairment and all-cause mortality specifically in older adults. There was evidence of increased risk for each of the three outcomes. Exposure to DACEs as a class was associated with a 45% increase in the odds of cognitive impairment, the individual medications olanzapine and trazodone were associated with an approximate doubling in the risk of falls and the ACB scale was associated with an approximate doubling in the odds of all-cause mortality. Another systematic review has recently investigated associations between DACEs, cognitive function and mortality 42. However, there were substantial differences in the search criteria, e.g. inclusion of participants <65 years, publication time and reporting of specific DACE scoring systems (total vs. anticholinergic DBI) 42.

Other reviews have examined associations that included younger participants and those with cognitive impairment 43 and normal cognition 44. Amongst those with cognitive impairment 43 participants were aged between 56 to 99 years, and the study duration ranged from 8 weeks to 6 months. A significant decline in the Mini Mental State Examination (MMSE) was observed with risperidone (weighted mean difference (WMD) 0.69, 95% CI 0.07, 1.31) but not with olanzapine (WMD 0.64, 95% CI, –0.09, 1.35) and quetiapine (WMD 0.68, 95% CI –0.26, 1.62). Amongst subjects aged 60 years or older with normal cognition there were deleterious amnestic and non-amnestic effects for oxybutynin, but no clear impairments for amitriptyline and imipramine across varying drug dosages over exposure periods of up to 3 weeks 44.

Meta-analyses results for falls showed a 116% and 79% increase in the odds of a fall occurring with the use of olanzapine and trazodone, respectively. Amitriptyline, paroxetine and risperidone showed no significant associations with falls whereas olanzapine and trazodone did and there was also a trend towards an increased risk with amitriptyline, paroxetine and risperidone. Significant heterogeneity was observed in several of the five analyses across two studies 32,34, which may have resulted from the different population characteristics. Whilst one study included participants newly diagnosed with depression 32 the other included participants with depression, insomnia, cognitive impairment and anxiety 34. Depression is a known risk factor for deconditioning and an associated decline in muscle mass 45, particularly in the older population, has been linked to falls, functional decline, increased frailty and immobility. One study also included only nursing home residents 34 which indicated that this population was on average older with a mean age of 82 years, compared with 75 years in the other study 32, and included subjects who were frail and highly impaired. Advancing age might also indicate greater variability in other clinical and demographic characteristics as well as medication prescribing and this might have, at least partly, contributed to the inconsistent results across the analyses.

Results from the meta-analyses for all-cause mortality showed that exposure to DACEs was associated with a 106% increase in mortality rates for each additional point scored on the ACB scale but there was no observed association with the ARS. Possible differences in the associations between DACE scoring scales and all-cause mortality include inherent differences in DACE definition within each scoring system, the relatively small number of studies analyzed and differences in the baseline characteristics of the study populations. Significant heterogeneity was observed in the ARS score analysis. This may be due to the differences between the two studies population characteristics with one study including only patients with hip fracture and awaiting surgery as well as having a higher proportion of females. Imprecision of results with wide confidence intervals and no overlapping can be due to studies including smaller numbers of participants with a few events, such as mortality rates, occurring. Other reviews, including patients younger than 65 years, have examined all-cause mortality with individual DACE exposure. Results from such studies investigating associations between inhaled DACEs and all-cause mortality have been ambiguous. Two meta-analyses that investigated the adverse effects of tiotropium in six 46 and 11 47 RCTs amongst COPD patients, respectively, observed no significant associations between DACE exposure and all-cause mortality (OR 0.96, 95% CI 0.63, 1.47 46; OR 0.94, 95% CI 0.82, 1.08 47). Another two studies in COPD patients observed no significant associations between the use of inhaled DACEs and all-cause mortality (RR 0.80, 95% CI 0.50, 1.20 48 and RR 1.26, 95% CI 0.99, 1.61 49). However, two more recently published RCT reviews observed that tiotropium was associated with an approximate 50% increased risk of all-cause mortality compared with placebo (OR 1.51, 95% CI 1.06, 2.19 50 and RR 1.52, 95% CI 1.06, 2.16 51), respectively.

Within Australia, medication related hospital admissions make up approximately 2–3% of total hospital admissions of which around 50% are potentially preventable 52. In the USA medication related hospital admissions are slightly higher at 4.7% in 2008 53. Cognitive impairment, falls and mortality have serious economic consequences for the health care system and the older population. For instance, serious injuries, including hip fracture, due to falls in older adults result in lengthy periods of hospitalization and are associated with substantial morbidity and mortality. Falls themselves are also risk factors for future falls. The estimated financial cost of fall-related injuries in Australia and the USA, including indirect costs such as costs covered by family or the community, easily exceed $1 billion 54 and $30 billion per year 55, respectively. Not only is the cost to the health care system from fall-related injuries considerable but also it has a large impact on the individual and others who surround them. A decline in the ability to ambulate can lead to a decline in ability to perform activities of daily living independently, increased reliance on others for assistance and increased risk of social isolation 13. Cognitive impairment is associated with a large burden of disease in the ageing population with most of this associated with residential aged care facilities. Cognitive impairment restricts daily activities and, in the long term, can result in high care needs 11. Therefore people requiring residential care are usually frailer than in the community, with substantial physical and behavioural needs and multiple comorbidities.

The definitions of the specific outcomes that we assessed including falls and cognitive impairment were not the same across studies. The different diagnostic tests, criteria and clinical assessments used for the diagnosis of cognitive impairment or dementia in each study may have resulted in either an over or under diagnosis of cognitive impairment amongst the participants. Changes in cognition may also be caused by an intercurrent illness. Second, there may have been bias due to differential measurement error amongst participants. For example, those susceptible to cognitive decline and with poor recall may not have recorded falls reliably. Third, some studies assessing mortality relied on databases to ascertain the subject’s vital status and may not have recorded deaths for subjects who moved to areas outside of the databases jurisdiction. Fourth, each of the meta-analyses included a relatively small number of studies and was therefore potentially underpowered to detect some associations which suggested that a trend was present. Similarly, publication bias was not assessed using funnel plots for the same reason and there were too few studies to establish asymmetry visually. Fifth, the presence of significant between-study heterogeneity was observed in a number of meta-analyses that may have arisen from methodological diversity or differences in outcome assessments. This could have been due to the combination of different study designs, and in particular differences in the length of follow-up and variability between studies in subject inclusion and exclusion criteria. The duration of exposure to DACEs differed from as short a period as 1 month to as long as 6 years which would have likely contributed to at least some of the differences for each of the assessed outcomes. Another potential issue is that the DACE scoring systems considered in our study, i.e. ARS, ADS, ACB and DBI_AC, are largely based on drugs marketed in the USA 15–18. This raises the possibility that DACEs available in other countries, but not in USA (e.g. biperiden, a drug used in many countries for counteracting extrapyramidal side effects of antipsychotic medications), were missed in the literature search unless they were generically included within a whole drug class whenever DACE use was defined (see Methods). Some of the outcomes assessed, particularly falls and cognitive impairment, might be associated not only with exposure to DACEs, e.g. some anti-Parkinsonian drugs, but also with medical conditions for which these drugs were prescribed, e.g. Parkinson’s disease. Therefore, there is the possibility of confounding by indication. Finally, several DACEs, e.g. olanzapine, can bind to different non-muscarinic receptors 56. This can make the specific attribution of adverse outcomes to anticholinergic effects challenging 57.

A major strength of the study was the use of published standards for the reporting of observational studies and RCTs 58,59. Search limitations of animal and non-English studies were performed manually rather than relying on automatic limitations in order to reduce retrieval loss. The methodological quality of included studies in the review ranged from moderate to high indicating that the results of the individual studies were each reliable. Finally, our inclusion criteria for the exposure period being more than 365 days for studies assessing cognitive impairment was determined to ensure that any subjects classified with cognitive impairment may have plausibly resulted from an exposure to medications.

Our findings raise concerns about the risks of prescribing of DACEs and have potential clinical implications. Health professionals prescribing DACEs should be aware of the potential adverse effects associated with these drugs as well as their benefits in the older population. However, our conclusions require confirmation in larger trials or cohort studies which also include adequate and appropriate follow-up periods, direct comparisons of the predictive capacity between different DACE scoring systems and sub-group analyses stratified for different age groups, e.g. <80 vs. ≥80 years. Future studies that assess DACE scores are required for studies assessing falls and cognitive impairment which were not available for this review. Studies might also consider assessing DACE prescribing patterns since one study included in this review suggested that discontinuing DACEs was associated with a decreased risk in cognitive impairment 35. In summary although our analysis was confined to a relatively limited number of studies, it provides preliminary evidence for negative effects of DACEs on cognitive impairment, falls and mortality.

Competing Interests

All authors have completed the Unified Competing Interest form at www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author) and declare no support from any organization for the submitted work, no financial relationships with any organizations that might have an interest in the submitted work in the previous 3 years and no other relationships or activities that could appear to have influenced the submitted work.

Supporting Information

Appendix S1 Search Strategy

Table S1 Individual anticholinergic drugs used in search strategy

Table S2 Outcome definitions

Table S3 Assessing bias in individual studies using Newcastle-Ottawa Scale

Figure S1