FormalPara Key Points

Use of drugs with anticholinergic effects is common in older hospitalized patients, and hospitalization is associated with an increase in the overall anticholinergic burden.

The Anticholinergic Cognitive Burden (ACB) and Anticholinergic Risk Scale (ARS), two different tools for the evaluation of the anticholinergic burden, have only a moderate degree of correlation.

Higher anticholinergic burden is associated with a steeper decline in cognitive performance in the year after hospital discharge.

Higher anticholinergic burden is associated with higher risk of incident disability in the year after hospital discharge.

1 Introduction

Medications with anticholinergic properties, widely used in older people, are associated with several adverse events at both the peripheral (urinary retention, decreased gastrointestinal motility, alterations of glandular secretions) and central level (alterations of attention and consciousness and increased risk of delirium) [1,2,3,4,5]. Older subjects are more susceptible to side effects for various reasons, including polypharmacy [6], physiological reduction of acetylcholine reserve and increased risk of neurodegenerative diseases involving cholinergic neurons [7, 8], increased blood–brain barrier permeability [9], and pharmacokinetic and pharmacodynamic changes [10].

In order to predict the possible adverse effects, especially in the context of polypharmacy, several measures of the anticholinergic burden, the cumulative effect of drugs with anticholinergic activity [11], have been developed for use in clinical practice. These tools are fast, simple to administer, objective and reproducible, but they tend to oversimplify the pharmacological complexity of drug interaction, and there is discrepancy in the inclusion of drugs and in the attribution of the anticholinergic weight. Durán et al. conducted a systematic review of seven anticholinergic scales in order to synthesize the data and to develop a single list of 100 drugs with clinically relevant anticholinergic activity [12]. Although it represents a noteworthy attempt to create a unique classification tool, this list has not been validated. Consequently, there is no agreement on what is the best classification model to quantify anticholinergic exposure in clinical practice [12, 13]. Of the seven instruments currently available, the Anticholinergic Cognitive Burden (ACB) [14] and the Anticholinergic Risk Scale (ARS) [15] are the two most frequently validated by expert opinion and more frequently used in clinical research [16].

Several studies have investigated the relationship between use of drugs with anticholinergic activity and cognitive and functional performance [17,18,19,20,21,22,23]. Overall, the results of these studies suggest a significant association between use of drugs with anticholinergic effects and cognitive and functional decline. Of them, however, only one cross-sectional study compared the predictive value of both the ACB and ARS scales on change in cognitive and functional status [17]. None of these studies investigated if hospitalization for an acute event is associated with change in anticholinergic burden.

The aim of our study was therefore to analyze and compare the relationship between global anticholinergic burden estimated using the ACB and ARS scales and cognitive and functional impairment in a sample of Italian older hospitalized patients. The secondary aims of the study were to investigate the effects of hospitalization on change in anticholinergic burden and to evaluate the degree of correlation between the two different tools used for the evaluation of anticholinergic burden.

2 Methods

2.1 Study Population

We analyzed data from 1123 subjects enrolled in the CRIteria to Assess Appropriate Medication Use among Elderly Complex Patients (CRIME) project, an observational study performed in geriatric and internal medicine acute care wards of seven Italian hospitals (Gemelli Hospital, Catholic University of Sacred Heart of Rome; University of Perugia; University of Ferrara; National Institute of Hospitalization and Care—INRCA at Ancona, Cosenza, Fermo and Rome). The methodology of the CRIME project has been described elsewhere [24, 25].

Briefly, all patients consecutively admitted to participating wards during the enrolment period (June 2010–May 2011) entered the study. The only two exclusion criteria were age < 65 years old and unwillingness to take part in the study. Ethical approval for the study was obtained in all participating centers. Hospitalized patients were invited to take part in the study and were free to decline participation. Written consent was obtained with assurance of data confidentiality.

Data collection was conducted through a dedicated questionnaire, administered at hospital admission and at discharge by study researchers. Patients were then reassessed at 3, 6 and 12 months after discharge through either ambulatory visits or telephone contacts (if the patient’s clinical condition did not allow the visit). During each follow-up control, participants’ information was obtained by interviewing the patient or their relatives.

2.2 Sample Size

At baseline study population was composed of 1123 patients. Functional status was evaluated in the whole sample, while the cognitive performance was available for 824 patients (73.4%). For the longitudinal study of cognitive status, we used the data of subjects examined at least once during the follow-up (442, 53.6%). For the longitudinal analysis of functional status, we considered patients independent at baseline (581, 51.7%) and revalued at least once in the follow-up (493, 84.9%).

2.3 Anticholinergic Burden

Anticholinergic burden was evaluated using two different scores: ACB [15] and ARS [16]. Both scales are based on a systematic review of literature and on expert opinion, but they include a different number of drugs with anticholinergic activity (88 for ACB and 49 for ARS). As shown in Electronic Supplementary Material Figure S1, anticholinergic burden was estimated for home therapy, during hospital stay, at discharge, and at 3-month follow-up. The lists of drugs included in the ACB and ARS scales and the distribution of these drugs prescribed at discharge in the study population are shown in Electronic Supplementary Material Tables S1 and S2.

2.4 Outcomes

Cognitive status was assessed using the Mini Mental State Examination (MMSE) [26], administered at hospital discharge and during the follow-up visits. Functional status was evaluated at hospital admission, at discharge and over the follow-up using five basic activities of daily living (ADLs) [27] (bathing, dressing, toileting, transfer and eating). We considered as dependent, patients who needed assistance in at least one activity. Disability at follow-up was defined as presence of dependency in two consecutive follow-up visits or presence of dependency in a follow-up visit followed by loss to follow-up or death. The time points at which outcomes were evaluated are shown in Electronic Supplementary Material Figure S1.

2.5 Covariates

Socio-demographic characteristics of the study population included age, sex, education and marital status. The home therapy was ascertained from the baseline interview. The prevalence of specific medical conditions was established using standardized criteria that combined information from baseline interview, medical records, physical examination and blood test results. Disease categories considered were dementia, Parkinson disease, hypertension, coronary heart disease, diabetes mellitus, infectious disease, renal failure and anemia. To assess the presence of depressive symptoms, we used the Geriatric Depression Scale (GDS) [28].

2.6 Statistical Analysis

Selected socio-demographic and clinical characteristics of the study population were compared using a χ2 test and analysis of variance (ANOVA) model for categorical and continuous variables, respectively, according to ARS and ACB classes of therapy prescribed at discharge. Anticholinergic burden at discharge was also the primary variable in the longitudinal analyses described below. We then analyzed the degree of correlation between the two anticholinergic burden scales using the Spearman’s rank correlation coefficient.

The retrospective association between anticholinergic burden and cognitive status was analyzed using a linear multiple regression analysis predicting MMSE score at hospital discharge as a function of the anticholinergic burden of the home therapy. Change in cognitive status over the follow-up period as a function of anticholinergic burden at hospital discharge was then estimated using mixed linear regression models for repeated measures, combining ACB and ARS scores ≥ 1 in the same category due to sample size. Regression models were adjusted for several potential confounding factors of the association between use of drugs with anticholinergic activities and cognitive impairment (variables clinically significant and variables found to be associated in univariate analysis). In order to take into account any change in anticholinergic burden after discharge due to primary care physician intervention, the model was adjusted for anticholinergic burden evaluated at the first follow-up visit (3-month follow-up).

To calculate the likelihood of prevalent disability (cross-sectional analysis) according to anticholinergic burden of home therapy and the risk of developing new disability during follow-up according to anticholinergic burden of the therapy prescribed at discharge (longitudinal analysis excluding patients with disability at hospital discharge and combining ACB and ARS scores ≥ 1 in the same category due to sample size), we estimated the odds ratio (OR) and the 95% confidence interval (CI) using multivariable logistic regression models. Logistic models were adjusted for several potential confounders, including change in cognitive status over the follow-up. All analyses were performed using Stata 13.0 for Windows (Stata Corporation; College Station, TX).

3 Results

Mean age of study participants was 81 ± 7.5 years, and 56% were female. Patients had a mean of 5.2 ± 2.6 diseases and took at home a mean of 7.4 ± 2.8 drugs.

Table 1 presents selected socio-demographic and clinical characteristics of the study population according to ARS and ACB classes of the therapy prescribed at discharge. Comparing the two ARS groups, subjects with higher scores were significantly more likely to be female, smoke less, and have a greater prevalence of dementia, Parkinson disease and depressive symptoms. Patients with higher ACB scores were older and had a greater prevalence of dementia, coronary heart disease, infectious diseases, renal failure and anemia. With both scales, patients with higher anticholinergic burden had lower MMSE scores and a greater prevalence of disability in ADLs.

Table 1 Socio-demographic and clinical characteristics of study population according to ARS and ACB classes at discharge
Full size table

Figure 1 displays the distribution of ACB and ARS scores of the therapies before hospital admission, during hospitalization, at discharge, and at the first follow-up visit. ACB classification identified a lower percentage of patients without anticholinergic burden, whereas ARS identified a lower percentage of subjects with positive anticholinergic burden. With both scales we found an increase in anticholinergic burden during hospitalization. Indeed, comparing home and hospital therapies, the number of subjects who did not use drugs with anticholinergic properties reduced from 40.6 to 27.5% according to ACB and from 82.1 to 75.6% according to ARS. At discharge there was a reduction of the anticholinergic burden, without reaching pre-admission levels (ACB = 0: 35.1%; ARS = 0: 81.5% of patients). A further reduction of the anticholinergic burden was observed at the first follow-up visit: 54.9% and 87.2%, respectively. Comparing ACB and ARS scores, we found a moderate concordance between these two classification systems, with correlation coefficients ranging from 0.39 to 0.43. At hospital discharge, 13.4% of the patients were treated with antipsychotic drugs, 17.9% with antidepressants and 16.7% with anxiolytics including benzodiazepines.

Fig. 1
figure 1

Distribution of a ACB and b ARS scores before hospital admission, during hospitalization, at discharge, and at the first follow-up visit. ACB Anticholinergic Cognitive Burden, ARS Anticholinergic Risk Scale, F/U follow-up

Full size image

In the retrospective analysis (n = 824, 73.4% of the sample), we found a significant association between the home anticholinergic burden (ACB ≥ 2: β: − 1.34, p = 0.017; ARS ≥ 1: β: − 2.47, p < 0.001) and in-hospital MMSE score, even after adjustment for potential confounders (age, sex, education, smoke, hypertension, coronary heart disease, renal failure, anemia, infectious diseases, and variation of ACB/ARS during hospital stay) (data not shown). Then we analyzed the relationship between anticholinergic burden at discharge and decline in MMSE score over time, estimating the monthly variation in MMSE score in patients with in-hospital cognitive evaluation and examined at least once during the follow-up (n = 442, 53.6%) (Table 2). We found that patients of the two ACB classes had similar cognitive trajectories, whereas subjects with higher anticholinergic burden according to ARS score had a significantly steeper decline in MMSE score (p = 0.04), even after adjustment for potential confounders, including ARS score at the first follow-up (83% of the patients still had an ARS score of > 1). Particularly, in the fully adjusted model, we found that patients with ARS = 0 did not have a significant decline in MMSE (− 0.02/month), whereas subjects with ARS ≥ 1 presented a significant monthly MMSE score variation (− 0.15/month).

Table 2 Multiple regression analysis for repeated measures predicting MMSE score over the follow-up as a function of anticholinergic burden of the therapy prescribed at discharge
Full size table

Similarly, for functional status, we first performed a cross-sectional analysis to explore the likelihood of disability according to anticholinergic burden of the home therapy. At multivariate logistic regression analysis, the risk of disability was more than twofold in subjects with ACB ≥ 2 (OR 2.29, 95% CI 1.42–3.72) compared to ACB = 0 and 3.5-fold in subjects with ARS ≥ 1 (OR 3.50, 95% CI 2.13–5.75) compared to ARS = 0 after adjustment for potential confounding variables (age, sex, education, smoke, MMSE score, hypertension, coronary heart disease, renal failure, anemia, infectious diseases, and variation of ACB/ARS during hospital stay) (data not shown). Then we analyzed the likelihood of developing new disability during follow-up through a logistic regression analysis performed on patients without disability at hospital discharge and revalued at least once in the follow-up (n = 493, 84.9% of the entire sample) (Table 3). After adjustment for potential confounders, including ACB/ARS score at the first follow-up, patients with ACB ≥ 1 had almost a threefold increased risk of developing disability (OR 2.77, 95% CI 1.39–5.54) compared to the ACB = 0 group. Patients with ARS ≥ 1 also presented an increased risk of disability in the partially adjusted model (OR 2.19, 95% CI 1.09–4.40), but after full adjustment the association was no longer significant (OR 1.49, 95% CI 0.60–3.70).

Table 3 Logistic regression analysis for likelihood of disability over time according to anticholinergic burden of the therapy prescribed at discharge
Full size table

4 Discussion

Our study demonstrated a significant association between anticholinergic burden, as evaluated by either the ACB or ARS scale (two rapid assessment scales that take only a few minutes to complete, and even less time with use of existing online calculators), and cognitive and functional status in a sample of hospitalized older subjects. Results were consistent in retrospective and longitudinal analyses with patients exposed to higher anticholinergic burden being at higher risk of cognitive decline and new disability over time. Furthermore, we demonstrated a poor agreement between ACB and ARS, suggesting that these two tools cannot be used interchangeably in research and clinical practice.

In particular, we found a significant inverse relationship between the home anticholinergic burden (ACB ≥ 2 and ARS ≥ 1) and in hospital MMSE score, regardless of the variations of anticholinergic burden during hospitalization. Longitudinal analysis showed that compared to patients with ARS = 0, those with ARS ≥ 1 for therapy prescribed at discharge experienced a steeper decline in MMSE score during the follow-up, regardless of anticholinergic burden variation during the first follow-up.

Similarly, patients with ACB ≥ 2 and ARS ≥ 1 for home therapy had a significantly greater likelihood of disability at hospital admission compared to subjects with lower scores. We also demonstrated that patients with ACB ≥ 1 at discharge have an almost threefold increased risk of developing disability during follow-up compared to the ACB = 0 group. Interestingly, this relationship is independent from the patient’s cognitive status, an important risk factor for disability, suggesting other pathogenic mechanisms underlying this association, including a direct peripheral effect of anticholinergic drugs on muscle performance or a central effect on processes involved in motor skills not assessed by MMSE.

Another interesting aspect of our study is the analysis of correlation between ACB and ARS, which showed a poor correlation. Analyzing the distribution of different scores in the study population, we can note that the two scales give different information, with the ACB demonstrating greater sensitivity and the ARS greater specificity. A possible explanation for the greater sensitivity of the ARS compared with the ACB is that these scales consider a different number of drugs and sometimes assign different scores to the same drug. These findings agree with the results of a single previous study published in 2013 that compared the two scales in a sample of elderly hospitalized patients [17]. The modest concordance between the two scales might also explain the lack of consistency between ACB and ARS as predictors of cognitive decline and incident disability. It is noteworthy that with both scales we found an increase in anticholinergic burden during hospitalization, probably due to additional treatments for intercurrent acute events. This observation supports the role of hospitalization as a risk factor for increased chronic use of drugs with anticholinergic properties, as already reported in the literature [29].

Our work confirms and expands the results of previous studies, analyzing the association between anticholinergic burden and cognitive and functional status with a 1-year prospective study with repeated assessment of anticholinergic burden over the follow-up. Despite several strengths of our study (detailed clinical and pharmacological characterization of participants, multiple assessments of anticholinergic burden, statistical analysis adjustment for several potential confounders), some limitations should be considered. The first is that only 53.6% of the patients with baseline cognitive assessment were included in the longitudinal analysis because of the lack of MMSE assessment during the follow-up. This aspect might have reduced the statistical power of the study and might represent a selection bias since people with poor health status are more likely to be lost over the follow-up. Secondly, combining patients with ACB or ARS scores of ≥ 1 and the small proportion of patients with high anticholinergic burden (≥ 3) did not allow us to determine whether a dose relationship was present. Nevertheless, based on data obtained with the three classes of ACB, we can suppose that the impairment is dose related, and we were able to demonstrate that in the elderly, even a moderate burden is significantly associated with cognitive and functional decline, emphasizing its important clinical impact. Finally, because of the observational design, we cannot rule out the presence of residual confounding, in particular, the so-called confounding by indication phenomenon (i.e., the negative outcome is not related to the drug itself but to the clinical indication for which the drug is prescribed).

The results of our study stress the need for caution in the prescription of drugs with anticholinergic properties, especially in patients with complex polypharmacy, and the need for a standardized model for the evaluation of anticholinergic burden. Our findings could also represent a starting point for further studies focused on the dose-related effect of these drugs, in order to identify a dosage of drugs with anticholinergic properties that could be considered “safe” in elderly patients.

5 Conclusions

In conclusion, we found a moderate correlation between ACB and ARS, a different predictive value of these scores and an increase in anticholinergic burden during hospital stay with both scales. Moreover, we demonstrated an independent association between use of drugs with anticholinergic properties and cognitive and functional decline in a sample of hospitalized geriatric patients.