Study identification
372 studies were obtained from searching the four databases: 212 from Medline, 100 from PsychInfo, 30 from Embase and 30 from CINHL. After the removal of duplicates with EndNote software, 173 studies remained. More duplicates were identified by Covidence software and removed (including study protocols), leaving 159 studies for the title and abstract screening (Fig. 1).
Characteristics of included studies
The total number of included studies was n=16, and all participants were≥65 years. Of the 16 studies, five were included in the meta-analysis for effects of AB on walking speed [16,19,22,23,26], and four were included in the meta-analysis for effects of AB on IADL scores[16,22,24,26]. The most common type of studies were cross-sectional studies (n=8), followed by cohort studies (n=5)[15,19,20,24,25] and only three were randomised controlled studies (n=3)[13,18,21].The studies used a mixture of statistical tests, but the t-test and p-value were the most commonly reported. A concise summary of study characteristics is presented in Table 1, and a brief description of some of the characteristics follows.
Participant characteristics
Two studies included men only [24,26], one study included women only [27], and the rest included both men and women. The study populations were distributed in four countries, with the USA being the most frequent (n=6), followed by Italy (n=4), Australia (n=2), Turkey (n=1), Netherlands (n=1), Switzerland (n=1) and Germany (n=1) (Table 1). The study participants were mostly living in the community [19,23,25,26], some were outpatients [12,22], whilst others were based in a nursing home [15,18]. The inclusion criteria of other studies were morbidity specific; People with Alzheimer’s [21], Vascular Dementia [20], Hypertension [1], moderate to severe disability [27], as well as those with a history of falls [14].
Exposure (anticholinergic burden)
Except for two [20,21], all studies used AB scales to measure exposure. The Drug Burden Index (n=5) was the most commonly used scale, followed by the anticholinergic cognitive burden scale (n=4). Two studies used more than one AB scale to compare exposure as captured by the scales [12,17].
Outcome measures
The primary outcome measure (mobility) was measured using several tools, including the Barthel Index [29], Timed-Up and Go (TUG)[28], handgrip strength[30], Basic Activities of Daily Living (BADL)[31], Gait assessment [32] etc. The most popular were handgrip strength and gait (n=5). Ten of the studies used more than one mobility measure tool.
Risk of bias in included studies
The Newcastle-Ottawa Scale (NOS)[33] for assessing cohort studies was slightly modified and adapted for this systematic review and employed to assess the risk of bias for cohort studies (Table 2A) and cross-sectional studies (Table 2B). The modified scales are shown in the additional files (Additional file 1 and 2), elaborating on each assessment element and where each evidence can be found in each study. In addition, for the randomised control studies (n=3), the Cochrane risk of bias tool for randomised trials was used (ROB2)[34]. The tool can be found in Additional file 3 and is summarised in Table 2C.
Table 1 Main characteristics of the 19 included studies on anticholinergic burden association with mobility measures
|
Author,
Year
|
Country
|
Method
|
Study Setting
|
Anticholinergic
Burden Scale
Used
|
Mobility measures
|
Sample size
|
Association outcome
|
Mayer, 2017 [12]
|
Germany
|
Cross-sectional
|
Home dwelling
|
ABC, ACB, ADS, AAS, CrAS, ARS, ACL, Martindale, Cancelli
|
Barthel Index
|
2761
|
Yes
|
Kolanowski, 2015 [13]
|
USA
|
Randomized clinical trial
|
Hospital setting
|
ACB
|
Barthel Index
|
99
|
Yes
|
Attoh-Mensah, 2020 [14]
|
Switzerland
|
Cross-sectional
|
Community dwelling
|
ADS
|
Timed-Up and Go (TUG)
|
177
|
Yes
|
Landi, 2014 [15]
|
Italy
|
Prospective (Multicentre)
Cohort
|
Nursing Homes
|
ARS
|
ADL
|
1490
|
Yes
|
Landi, 2006 [16]
|
Italy
|
Cross-sectional cohort
|
Community dwelling
|
SAA
|
SPPB, HGS, IADL, ADL
|
364
|
Yes
|
Pasina, 2013 [17]
|
Italy
|
Cross sectional (prospective)
|
Hospital setting
|
ARS, ACB
|
Barthel Index
|
1380
|
Yes
|
Wilson, 2010 [18]
|
Australia
|
Randomized controlled trial
|
Residential Aged Care Facility
|
DBI
|
HGS, Walking speed, balance
|
602
|
No
|
Wouters, 2020 [19]
|
Netherlands
|
Prospective cohort study
|
Community dwelling
|
DBI
|
Walking test, Cardigan test, Chair Stand test, Balance Test, Functional Independence
|
3107
|
Yes
|
Moretti, 2005 [20]
|
Italy
|
Controlled open-label study(Cohort)
|
Nursing Home
|
Olanzapine
Vs
-Promazine chloridate
-Haloperidol
|
Barthel Index, IADL, Tinetti Scale
|
356
|
Yes
|
Street, 2000 [21]
|
USA
|
Randomised controlled trial (double-blind, placebo)
|
Nursing home
|
Olanzapine
VS
Placebo
|
Gait (Simpson Angus Scale Assessment)
|
206
(7 gait assessment)
|
Yes
|
Soytas, 2021 [22]
|
Turkey
|
Cross-sectional (single centre)
|
Outpatient
|
ACB
|
BADL, HGS
|
256
|
Yes
|
Nebes 2007 [23]
|
USA
|
Cross-sectional
|
Community dwelling
|
SAA
|
Gait, Manual response time
|
90
|
Yes
|
Han, 2008 [24]
|
USA
|
Prospective Cohort
|
Community dwelling
|
CrAS
|
IADL
|
544
|
Yes
|
Hilmer, 2009 [25]
|
USA
|
Prospective cohort
|
Community dwelling
|
DBI
|
SPPB, HGS
|
3075
|
Yes
|
Gnjidic, 2008 [26]
|
Australia
|
Cross-sectional
|
Community dwelling
|
DBI
|
Performance Battery, HGS, IADL
|
1705
|
Yes
|
Cao, 2008 [27]
|
USA
|
Cross-sectional
|
Community dwelling
|
DBI
|
Balance, Gait speed, HGS, ADL, Mobility, Chair stand test
|
932
|
Yes
|
KEY: Cr-AS Clinician-rated Anticholinergic Scale, DBI Drug Burden Index, SAA Serum Anticholinergic Activity, ACB Anticholinergic Cognitive Burden, ARS Anticholinergic Risk Scale, ADS Anticholinergic Drug Scale, ADL Activities of Daily Living, HGS Hand Grip Strength, SPPB Short Physical Performance Battery, IADL Instrumental Activities of Daily Living
|
Association between anticholinergic burden and mobility
Of the 16 studies, 15 demonstrated a statistically significant association between increasing anticholinergic burden and mobility [18]. For example, the paper by Soytas et al. [22] found that AB, as measured by the anticholinergic cognitive burden scale, decreased the handgrip strength in the exposure group compared to non-users (p=0.04)whilst the difference in BADL performance was insignificant between the groups (p=0.232). Nebes et al. initially found significant differences in all the six mobility measures between users and non-users; however, this significance was lost in five of the six measures after adjusting for effects of confounding factors[23], with Hand Grip Strength retaining significance. For conducting a meta-analysis, only two studies[15,16] provided enough information on ADL measurements to conduct a meta-analysis of the association between AB and ADL, whilst one study [14] provided enough information on TUG measurements to conduct a meta-analysis on the association between AB and TUG. Thus the number of studies was too small to conduct a meaningful analysis of these outcomes. The same argument applied to studies with enough information on SPPB scores [16] and BADL [22]. Conversely, five studies provided enough information for a meta-analysis of the association between AB and walking speed [16,19,22,23,26] to be conducted and four studies [16,22,24,26] provided enough information for a meta-analysis of the association between AB and IADL to be conducted, and these form the basis of the two meta-analyses conducted in this study. Although the study by Moretti et al. [20] provided enough information regarding IADL scores, it was not included in the meta-analysis of IADL and AB because instead of having a control group like the other four IADL studies, the study had participants using haloperidol; a less potent anticholinergic drug, and compared the group with participants taking olanzapine; a more potent anticholinergic drug.
Association between anticholinergic burden and walking speed
Using the DerSimonian and Laird method [42], the random-effects model was used to meta-analyse the data to establish the overall change in the mean difference in walking speed scores between participants exposed to anticholinergic medication and those not exposed to anticholinergic medication. Figure 3 below is a funnel plot of the five studies used in the walking speed meta-analysis. The small number of studies makes visual inspection for funnel plot asymmetry inconclusive. However, as all but one study showed a positive association of AB with walking speed, publication bias is implied. This contrasts with the statistically non-significant Eggers Regression test for funnel plot asymmetry (P=0.494). However, when the number of studies is low, as is the case here, the statistical power of Egger’s test may not be high enough to detect real asymmetry. Therefore, Stern et al. [47] recommend at least n≥10 studies for Eggers regression to be relevant. Figure 2 is a forest plot of the meta-analysis between AB and walking speed. The studies in the walking speed meta-analysis show a high degree of heterogeneity (I2=99% p<0.01). The instrument for walking speed was the same in all studies, so the difference in means was used as the effect estimate. The pooled estimated effect of AB on mobility was a general reduction in physical function as measured by the walking speed instrument. The pooled estimated effect of AB on walking speed was a decrease of 0.079m/s ±0.035 (MD±SE 95% CI: 0.010 to 0.149) p=0.026.
Anticholinergic burden and IADL
The random-effects model was also used to meta-analyse data on the association between AB and IADL scores. Two of the four studies in the meta-analysis used the IADL [20,26] as developed by Lawton et al. [46], while the other two used modified versions of the IADL score [16,24]. Ordinarily, higher IADL scores indicate high functioning and, therefore, greater independence, while low IADL scores indicate poor physical functioning and greater dependence. However, two studies [16,26] used an inverted IADL scale where low IADL scores represent high functioning and vice versa for high IADL scores. Hence, the direction of the effect was used to pool the results for meta-analysis. The mean differences obtained showed that mobility was impaired in all four studies. Landi et al. [16] assessed IADL using the Minimum Data Set for Home Care (MDS-HC) instrument [48], whilst Han et al. [24] used the IADL scale of the Older American Resources and Services (OARS) instrument [49]. The mean differences were standardised because of the different modifications in the IADL instrument used to measure the underlying singular construct [50]. The studies in this meta-analysis for IADL and AB show a high degree of heterogeneity (I2=99% p<0.001, Fig. 4). The pooled estimated effect of AB on IADL was a decrease of 0.27±0.12 (SMD±SE 95% CI: 0.03 to 0.52), p=0.027. Thus mobility was negatively impaired in both outcomes (IADL scores and walking speed).
Studies with a focus on specific anticholinergic drugs
Two studies did not employ an anticholinergic rating scale but compared taking a single anticholinergic drug with a placebo group[21] or having two groups stratified by the anticholinergic drug they were taking[20]. The study by Moretti et al. [20] had an equal number of participants in group A taking olanzapine n=178 and group B with participants taking either haloperidol or promazine chloridate, both known to have anticholinergic effects [8]. Longitudinal progression in both groups was analysed at baseline and a follow-up after 12 months. The group taking olanzapine showed only moderate worsening of mobility that was not significant p>0.05 in gait, balance and equilibrium (as stated by the Tinetti scale)when following up results were compared with baseline. However, group B, with participants taking either haloperidol or promazine, registered significantly worsened scores for physical performance in follow-up to baseline (p<0.05 and p<0.01, ΔBI and ΔTinetti respectively)However, the study by Street et al. [21] demonstrated olanzapine is strongly associated with gait. The placebo group did not receive any anticholinergic medication, whilst three intervention groups had participants taking either 5mg/d, 10mg/d or 15mg/d of the anticholinergic drug olanzapine. All three intervention groups had higher odds of abnormal gait than the placebo group (OR: 11.2, 7.5, 9.4 for 5,10,15mg/d respectively).
Ceiling and floor effects
Three studies discussed the possibility of analysis being affected negatively by a “ceiling” or “floor” effect. To address this concern, Pasina et al. [17] used BI scores to discriminate participation, with those scoring less than 20 at baseline excluded from the study altogether. This was done to exclude patients with the highest degree of physical impairment, in whom it might be more difficult to detect any potential effects of anticholinergic drugs on physical performance. The study by Hilmer et al. [25] also introduced exclusion criteria to avoid the floor effect by requiring that subjects not report difficulty walking 0.25 miles, climbing ten steps, or performing activities of daily living as measured at baseline. On the other hand, the paper by Wilson et al. [18], the only study to not suggest an association between AB and poor physical performance, did not extend the exclusion criterion to filter out participants with already poor physical performance. However, the study suggested that a ceiling effect might have been introduced by a “decline in cholinergic receptors present in the very old populations with high rates of cognitive impairment that is independent of external exposure to anticholinergic medications.”