Anticholinergic medication use and dementia: latest evidence and clinical implications

Abstract

Use of medications with anticholinergic activity is widespread in older adults. Several studies have highlighted that anticholinergic use may be associated with an increased risk of dementia. The objective of this narrative review is to describe and evaluate studies of anticholinergic medication use and dementia and provide practical suggestions for avoiding use of these medications in older adults. A comprehensive review of the literature, citations from recent reviews and the author’s personal files was conducted. Four studies were found that evaluated anticholinergic use and dementia as the primary outcome. Three studies focused on overall anticholinergic medication use and reported a statistically significantly increased risk of Alzheimer’s disease or dementia. In one study, dementia risk was primarily found with higher cumulative doses; people using anticholinergic medications at the minimum effective dose recommended for older adults for at least 3 years were at highest risk. In contrast, a study conducted in nursing-home residents with depression did not find that paroxetine [a highly anticholinergic selective serotonin reuptake inhibitor antidepressant, (SSRI)] increased risk for dementia compared with other SSRIs (without anticholinergic activity). Further study is needed to understand the mechanism by which anticholinergic medications may increase risk. In conclusion, there is evidence from three observational studies suggesting that anticholinergic medications may increase dementia risk. Given this potential risk and the myriad of other well-known adverse effects (i.e. constipation, blurred vision, urinary retention, and delirium) associated with anticholinergic medications, it is prudent for prescribers and older adults to minimize use of these medications and consider alternatives when possible.

Introduction

The pharmacological activities of naturally occurring anticholinergics have been known since the time of Cleopatra who put belladonna drops in her eyes to dilate her pupils to enhance her beauty. We have learned even more in the past 30 years about drugs with anticholinergic activity. Acetylcholine is a neurotransmitter that can innervate both muscarinic and nicotinic receptors. Drugs typically referred to as anticholinergic can block any one of the five muscarinic (G protein) receptors (M1–M5) found in smooth muscle, motor neurons, the heart, and the central nervous system (CNS) [Karimi et al. 2012; Brown and Laiken, 2011]. Despite this knowledge, there is no ‘gold standard’ by which medications can be classified as anticholinergic. Some rely on serum anticholinergic activity (SAA) as measured by radioreceptor assay, in-vitro muscarinic receptor affinity (pKi), clinical consensus, or a combination of these three approaches [Kersten and Wyller, 2014].

The effects of the blockage of muscarinic receptors have been described in humans to make them: ‘mad as a hatter’ (delirium), ‘blind as a bat’ (mydriasis), ‘red as a beet’ (flushed), ‘dry as a bone’ (xerostomia), and ‘hot as a hare’ (hyperthermia). A well known risk with anticholinergic medications is acute impairment in cognition, which has been demonstrated in single-dose experimental studies [Flicker et al. 1990, 1992; Molchan et al. 1992; Ray et al. 1992]. Older adults are more susceptible to anticholinergic adverse effects for several reasons. One reason is the decrease in cholinergic activity due to decreased levels of acetylcholine synthesis or the number of acetylcholine receptors [Bowie and Slattum, 2007]. This may account for the increased pharmacodynamic sensitivity seen in older adults with anticholinergics [Bowie and Slattum, 2007]. Another reason is increased blood–brain barrier permeability in older adults [Karimi et al. 2012]. Finally, older adults may have lower p-glycoprotein activity, an efflux transporter for the CNS [van Assema et al. 2012]. Thus, it is easier for anticholinergic medications to be transported into the CNS and more difficult to be removed.

Dementia is a major public health problem for older adults [World Health Organization, 2012]. To date, there have been few potentially modifiable risk factors identified. One potentially important modifiable risk factor is the use of medications (i.e. anticholinergics). Since it is not feasible to examine whether medications increase dementia risk in a randomized controlled clinical trial, we must rely on observational studies. A number of methodological issues need consideration when employing observational designs to examine medications and dementia risk. First, dementia has a long latency period and pathological changes in the brain may begin decades before the onset of symptoms. Therefore, a study that is evaluating anticholinergic use and dementia should take this latency period into consideration when determining the best time window to assess this putative risk factor. From a biological-plausibility standpoint, it is unlikely that a medication started 1 or 2 years prior to the diagnosis would alter the risk of dementia. Ideally, the exposure would be assessed several years prior to the diagnosis and the study design would allow ample follow-up time. Second, some medications with anticholinergic effects (e.g. paroxetine and first-generation antihistamines) are used to treat early (or prodromal) symptoms of dementia, including depression, anxiety or insomnia. Including use of anticholinergic medications to treat early symptoms in the exposure definition could result in reverse-causation bias; these medications are certainly related to the dementia diagnosis, but it is not a causal relationship. Another complication is that it is not clear how long prodromal symptoms occur prior to the diagnosis of dementia and therefore the optimal time window for excluding medication use prior to the diagnosis to avoid this type of bias.

To the best of our knowledge, no recent reviews have examined the relation between anticholinergic medication use and dementia risk. Therefore, the objective of this narrative review is to describe and evaluate studies of anticholinergic medication use and dementia, and provide practical suggestions for avoiding use of these medications in older adults.

Methods

A systematic search of the English-language literature from January 1960 to November 2015 was conducted using PubMed and Google Scholar using a combination of terms: ‘muscarinic receptor antagonists’, ‘anticholinergics’, ‘aged’, ‘elderly’, ‘dementia’, and ‘Alzheimer’s disease’. A manual search of the reference lists from identified articles and the authors’ article files, book chapters, and recent reviews were conducted to identify additional articles. Both authors independently reviewed each article retrieved. Studies were included if the focus was on older adults, a rigorous observational design was used (e.g. cohort or case-control), and dementia was the primary outcome. The critique considered potential threats to internal and external validity using a previously published approach [Hanlon et al. 1996].

Data synthesis

Overall, we identified a total of nine observational studies. Excluded were five studies, in which dementia was not the primary outcome measure [Ancelin et al. 2006; Cai et al. 2013; Campbell et al. 2010; Gnjidic et al. 2012; Whalley et al. 2012]. This left a total of four studies [Carriére et al. 2009; Gray et al. 2015; Jessen et al. 2010a; Bali et al. 2015] summarized in Table 1. In the following, we provide a description and evaluation for each study.

Table 1.

Studies examining the association between anticholinergic medication use and dementia.

Study, site, reference	Study duration/design	Participants	Exposure ascertainment and measurement	Outcome	Results Adjusted HR (95% CI or p value if not available)
3-City Study, France, Carriére et al. [2009]	4 years; Prospective population-based cohort study	n = 7123 (⩾65 years)	Medication inventory; three French pharmacotherapy sources	Dementia as per DSM-IV criteria and adjudication by a panel of neurologists	AC use and dementia1  Continuing2   1.65 (1.00–2.73)  Discontinuing3 1.28 (0.59–2.76) AC use and Alzheimer’s disease  Continuing   1.94 (1.01–3.72)  Discontinuing 1.72 (0.74–3.99)
German Study on Aging, Cognition and Dementia in Primary Care Patients, Jessen et al. [2010a]	54 months; Prospective cohort study	n = 2605 (>75 years) randomly selected from general practice registries	Medication inventory; AC medications as per Chew et al. [2008]	Dementia as per DSM-IV and criteria and clinical panel consensus	Any AC use4   2.08, p < 0.001 Low-activity use 1.80, p < 0.001 Medium-activity use 1.53, p = 0.105 High-activity use  2.58, p = 0.002 Highest-activity use  3.36, p < 0.001
Adult Changes in Thought Study, US, Gray et al. [2015]	Average of 7.3 years; Prospective cohort study	n = 3434 (⩾65 years) enrolled in integrated health care delivery system	Automated pharmacy data; cumulative strong AC medications as per 2012 AGS Beers criteria over 10-year window	Dementia as per DSM-IV and NINCDS criteria and clinical-panel consensus	Days/dose of AC use and dementia5  1 to 90  0.92 (0.74–1.16)  91 to 365 1.19 (0.94–1.51)  366 to 1095 1.23 (0.94–1.62)  >1095   1.54 (1.21–1.96) Days/dose of AC use and Alzheimer’s disease  1 to 90 0.95 (0.74–1.23)  91 to 365 1.15 (0.88–1.51)  366 to 1095 1.30 (0.96–1.76)  >1095  1.63 (1.24–2.14)
US Nursing Homes, Bali et al. [2015]	2 years; Retrospective new user cohort study	n = 19,952 (⩾65 years) nursing home residents with depression	Medicare Part D claims; Paroxetine versus all other SSRIs	Dementia as per ICD-9 codes from outpatient and inpatient claims files	Paroxetine use and dementia6 0.99 (0.79–1.23)

DSM-IV, criteria used for diagnosing dementia; NINCDS, National Institute of Neurological and Communicative Disorders and Stroke and Alzheimer’s Disease and Related Disorders Association; ApoE4, apolipoprotein E4; SSRI, selective serotonin reuptake inhibitor; CI, confidence interval; HR, hazard ratio; AC, anticholinergic.

¹Models adjusted for: center, age, sex, education, body mass index, alcohol use, tobacco use, caffeine intake, mobility, hypercholesterolemia, ApoE4 status, diabetes mellitus, asthma, depression, ischemic diseases, Parkinson disease, and hypertension.

²Continuing users defined as participants using at baseline and year 2.

³Discontinuing users defined as participants using at baseline only

⁴Models adjusted for: age, sex, education, depression and ApoE4 status.

⁵Models adjusted for: for ACT cohort, age, sex, education, body mass index, current smoking, regular exercise, self-rated health, hypertension, diabetes, stroke, coronary heart disease, Parkinson’s disease, history of depressive symptoms, and current benzodiazepine use.

⁶Treatment groups were matched on propensity-score calculation based on more than 70 covariates (e.g. comorbid conditions, sociodemographic characteristics, co-medications). Medications included other anticholinergics such as antihistamines, antipsychotics, and genitourinary products.

The first study suggesting an association between anticholinergic medications and dementia risk was published by Carriére and colleagues, in 2009 [Carriére et al. 2009]. These authors conducted a population-based cohort study using data from the Three-City Study in France, which recruited 7123 community-dwelling older adults, 65 years and older between 1999 and 2001. Trained research assistants performed cognitive testing and the diagnosis of incident dementia was established by an expert-panel adjudication at 2, and 4 years after baseline, using standard methods. Participants self-reported medication use and for analyses, were categorized as ‘continuing users’ of anticholinergic medications (use at baseline and 2-year follow up; n = 319), ‘discontinuing users’ (baseline use only; n = 175) or nonusers. The main classes of anticholinergic medications taken by at least 1.0% of the population were antidepressants (1.9%), gastrointestinal antispasmodics (1.6%), bladder antispasmodics (1.3%), and first-generation antihistamines (1.0%). Over the 4-year study period, 221 people were diagnosed with incident dementia. Although the risk of dementia was increased for both continuing [hazard ratio (HR), 1.65; 95% CI, 1.00–2.73] and discontinuing (HR, 1.28; 95% CI, 0.59–2.76) users of anticholinergics, neither result was statistically significant. Likewise, the risk for Alzheimer’s disease was increased for both continuing (HR, 1.94; 95% CI, 1.01–3.72) and discontinuing (HR, 1.72; 95% CI, 0.74–3.99) users, with only continued use found to be statistically significant.

Strengths of this study included the population-based sample, adjustment for many important confounders and the use of standard methods for dementia ascertainment. It is also notable that the use of nonprescription medications was captured. Some potential concerns were that medication exposure was confined to those collected cross-sectionally at two time points and exposure included medications not consistently agreed upon as being highly anticholinergic (e.g. anxiolytics and antiepileptics) [Durán et al. 2013]. Moreover, neither a dose- nor a duration-response relationship was reported. The potential for reverse-causation bias was not addressed. Lastly, this study also had a short follow-up time and a relatively small number of dementia cases, which likely reduced the precision of the study as evidenced by the wide confidence intervals. This has important implications for interpreting nonsignificant findings, as was the case for those who had discontinued anticholinergics. For example, people who discontinued anticholinergic medications had an HR of 1.72 for Alzheimer’s disease, and because of the wide confidence intervals, was a nonsignificant finding, however, the true HR may be as high as 3.99.

In 2010, Jessen and colleagues published data from the German Study on Aging, Cognition and Dementia in Primary Care Patients cohort study which included 2605 randomly selected individuals from the GP registries above the age of 75 without dementia [Jessen et al. 2010a]. Participants were followed longitudinally every 18 months for assessment of clinical, medication and neuropsychological measures. Anticholinergic medications were defined as those included on a list developed by Chew and colleagues, which is based on serum anticholinergic activity [Chew et al. 2008]. Anticholinergic medications were included if reported prior to the interview that triggered the dementia diagnosis. The dementia diagnosis was based on the DSM-IV criteria using an interdisciplinary consensus conference [Jessen et al. 2010b]. The anticholinergic medications were mainly cardiovascular agents, analgesics or anti-inflammatories, antidiabetics, and antidepressants. Over the follow up, 220 cases of dementia were diagnosed. After adjusting for sex, age, level of education, presence of depression and apolipoprotein E4 status, any anticholinergic medication use was associated with a statistically significant (p < 0.001) increased risk for dementia (adjusted HR, 2.08). The authors also reported higher dementia risk for medications classified as having the strongest anticholinergic activity.

A strength of this study was the use of standard methods for determining dementia diagnosis. Of potential concern is the lack of detail provided in the methods and results sections. Again, exposure included medications not consistently agreed upon as being highly anticholinergic (e.g. antidiabetics, nonsteroidal anti-inflammatory drugs) [Durán et al. 2013]. The influence of duration of anticholinergic use or dose was not reported. The findings could be confounded by the lack of control for potential indications for anticholinergic medication use. The potential for reverse-causation bias was not addressed. Lastly, while the results are generalizable to older adults who receive primary care, they may not be applicable more broadly.

Gray and colleagues published a prospective population-based cohort study using data from the Adult Changes in Thought (ACT) study in Group Health, an integrated health care delivery system in Seattle, Washington [Gray et al. 2015]. The study included 3434 participants, 65 years or older, with no dementia at study entry. Participants were assessed at baseline and every 2 years for cognition, demographic and health information. Computerized pharmacy-dispensing data were used to ascertain highly anticholinergic medications as per the 2012 American Geriatrics Society Beers criteria [American Geriatrics Society Beers Criteria Update Expert, 2012]. For purposes of analyses, exposure dose and duration were defined as the total standardized daily doses [(TSDDs): daily dose divided by minimum effective geriatric daily dose] dispensed in the past 10 years. The most recent 12 months of use were excluded to avoid use related to prodromal symptoms (e.g. reverse causation). Cumulative exposure was updated as participants were followed over time. Standard research criteria were used to ascertain dementia and Alzheimer’s disease using a multidisciplinary consensus conference. The most common anticholinergic medication classes used were antidepressants, first-generation antihistamines, and bladder antimuscarinics, accounting for 63%, 17%, and 11% of overall exposure, respectively. During a mean follow up of 7.3 years, 797 participants (23.2%) developed dementia. A 10-year cumulative dose-response relationship was observed for dementia and Alzheimer’s disease (test for trend, p < .001). In particular, participants in the highest exposure category (>1095 days or 3 years of daily use of a minimum effective geriatric dose) had a statistically significant increased risk for dementia (adjusted HR, 1.54; 95% CI, 1.21–1.96) or Alzheimer’s disease (HR, 1.63; 95% CI, 1.24–2.14) compared with those with no use. Results were robust in secondary, sensitivity, and post hoc analyses.

Strengths of this study were the large number of dementia cases, the long follow-up period, and use of computerized pharmacy data to ascertain medication use 10-years before study entry and throughout follow up to capture detailed cumulative anticholinergic exposure. Furthermore, these authors were able to examine whether risk varies according to extent of cumulative exposure and exclude use that may have been for prodromal symptoms. It is possible that nonprescription anticholinergic medication use was not completely captured. The sample population was primarily Caucasian and therefore results may not be generalizable to other ethnic or racial groups. Lastly, while the results are generalizable to older adults who received care at this integrated healthcare system, they may not be applicable more broadly.

In 2015, Bail and colleagues published the results of a retrospective cohort study of 19,952 elderly nursing-home patients with depression who were new users of selective serotonin reuptake inhibitors (SSRIs) [Bali et al. 2015]. Paroxetine, the only SSRI with strong anticholinergic effects, was compared for dementia risk with other SSRIs. The study used 2007 to 2010 Minimum Data Set-linked Medicare data. Exposure was ascertained from Medicare Part D claims. Patients were followed for a maximum of 2 years after SSRI initiation. The dementia diagnosis was based on ICD-9 codes obtained from administrative claims data. The unadjusted incidence of dementia in the propensity-matched cohort (3796 patients) was 7.5% for users of paroxetine and 8.6% for users of other SSRIs. Use of paroxetine was not associated with a greater risk of dementia compared with other SSRIs. (HR, 0.99; 95% CI, 0.79–1.23).

Strengths of this study are the new-user-design cohort study, the use of other SSRIs as an active comparator, and a large national sample. This study had a short follow-up time (2 years) and from a biological-plausibility consideration, this timeframe might not be sufficient for a medication to alter risk of dementia that has a long latency period. Moreover, neither dose nor duration was considered. The use of administrative data to identify dementia diagnoses is problematic, as incident cases may be underestimated and detection may be delayed compared with routine surveillance, as used in the prior three studies. Finally, these results are only generalizable to nursing home residents with depression.

Discussion

Four studies were found that assessed the risk of anticholinergic medications with the primary outcome of dementia or Alzheimer’s disease. Three studies that focused on overall anticholinergic medication use reported a statistically significantly increased risk of Alzheimer’s disease [Carriére et al. 2009; Gray et al. 2015] or dementia [Gray et al. 2015; Jessen et al. 2010a]. In contrast, one study that focused on comparing paroxetine (a highly anticholinergic SSRI) to other SSRIs in older nursing-home residents with depression, did not find an association with dementia [Bali et al. 2015]. As with all observational studies, there are strengths and weaknesses that potentially threaten the internal/external validity of the findings.

A key question is whether the association between anticholinergic medications and dementia is causal. Although causality cannot be determined definitively through observational studies, general criteria adapted from Bradford-Hill can be useful for determining whether the association is likely to be causal: (1) strength of the association; (2) consistency of the association; (3) temporal relationship between exposure and outcome; (4) dose-response gradient; and (5) biological plausibility [Stolley, 1990]. We discuss each of these criterion below, as well as areas for future study. By conventional standards, an effect size (e.g. HR in these studies) of 2.0 or less is considered a weak association because associations of this magnitude may be explained by subtle bias (e.g. confounding). The associations found in these studies ranged from 1.5–2.1, suggesting a small increased risk. Associations of this magnitude may still be important and valid, especially if it is a modifiable risk factor such as medication use. There is consistency found with a total of three studies, although additional studies in different populations using other study designs are needed. The exposure preceded the diagnosis of dementia in all studies, however, given the long latency period of the disease, it is difficult to determine the critical exposure window to evaluate. Even exposures occurring 1 to 2 years prior to the diagnosis may not be relevant and any noted association could reflect reverse causation. One of four studies examined a dose response and found that risk was associated with the highest use. Future studies should more fully examine the dose- and duration-response nature of risk. The underlying biological mechanisms by which anticholinergic medications may increase dementia risk are unknown. However, limited evidence suggests that anticholinergics may be associated with increased Alzheimer’s disease-related pathology [Perry et al. 2003], increased brain atrophy and reduced brain-glucose metabolism [Risacher et al. 2016]. Neurodegeneration may be caused by decreased number of brain synapses through increased amyloid-ß deposition or decreased brain levels of phosphatidylcholine [Wurtman, 2015]. Additional study is needed to elucidate the underlying biological mechanisms in animal or cellular models or by examining duration of anticholinergic medication use in relation to neuropathology outcomes in autopsy samples.

What are the clinical implications of the findings from these four observational studies? First, clinicians should keep in mind that there may be a cumulative dosage and duration effect with anticholinergic medication use. This is important, as a recent study of community-dwelling older adults that ascertained the use of both prescription and over-the-counter anticholinergic drugs reported that as many as 17% took one or more and that use did not decrease over the decade studied [Felton et al. 2015]. Second, effective alternative nonpharmacological and pharmacological options are available for most chronic conditions that might be treated with an anticholinergic medication. An excellent review on this topic provides alternatives for the use of the most commonly reported classes of anticholinergic medications (i.e. first-generation antihistamines, tricyclic antidepressants, certain skeletal-muscle relaxants) [Hanlon et al. 2015]. The management of urge-urinary incontinence or overactive bladder often includes the use of anticholinergics, which is considered first-line pharmacological therapy. For this condition, behavioral therapies are considered first-line management and should be maximized. In 2015, the American Urological Association updated the guidelines for the management of overactive bladder to expand the first-line pharmacological therapy to also include the ß₃ adrenoreceptor agonist mirabegron [Gormley et al. 2015]. It is important to note, however, that this is a relatively newly marketed agent and its effectiveness and safety in frail older adults is not completely known at this time. Finally, it is important to note that anticholinergic medications have adverse effects beyond the potential risk of dementia such as constipation, urinary retention, especially in men with benign prostatic hyperplasia or lower urinary-tract symptoms, delirium, and perhaps increased health-service use and mortality [Karimi et al. 2012]. Although we are not aware of evidence-based systematic approaches to reduce the anticholinergic burden that can be widely adopted [Kersten et al. 2013; Salahudeen et al. 2014], it is still prudent for prescribers to avoid starting any new anticholinergic medications [American Geriatrics Society Beers Criteria Update Expert Panel, 2015]. Older adults should also be counseled to avoid using anticholinergics that are available over the counter, including drugs used for allergies or for sleeping. Healthcare professionals should review the medication regimen of older adults after querying for over-the-counter anti-cholinergic drug use on a regular basis and determine if any of these could be discontinued or the dose reduced. In the scenario that an anticholinergic is considered an effective therapeutic option for an individual patient, the healthcare provider should engage the patient in shared decision-making and discuss the risks versus benefits of therapy.

There are several limitations to this narrative review. It is possible that we may have missed negative studies not published, published papers not identified by the databases searched, or studies not published in English. We did, however, look in our personal files; manually search the reference lists of the identified articles and recent reviews and articles cited by these to potentially identify studies for inclusion. We also did not perform a formal evaluation of the quality of the four observational studies using a standardized approach. It is important to note that there is currently no gold standard for evaluating the quality of pharmacoepidemiological studies of drug safety [Neyarapally et al. 2012].

Conclusion

There is evidence from three observational studies suggesting that anticholinergic medications may increase dementia risk. Further studies are needed to confirm these findings and understand the biological mechanism by which medications with anticholinergic effects may increase risk. Nonetheless, given this potential risk and the myriad of other well known adverse effects (i.e. constipation, blurred vision, urinary retention, and delirium) associated with anticholinergic medications, it is prudent for prescribers and older adults to minimize use of these medications and consider alternatives when possible.