Anticholinergics, executive function, and cognitive/behavioral changes in Down syndrome

Abstract

INTRODUCTION

Anticholinergic (AC) medication use is considered a risk factor for cognitive impairment and deterioration in the general population; yet, this has not been examined in Down syndrome (DS), a disorder with high dementia rates.

METHODS

Family members of 108 young adults with DS (18–39 years) reported on their loved one's medication use, executive function, and changes in cognition and behavior using a dementia screener. Medications were coded for their AC potency using the CRIDECO Anticholinergic Load Scale (CALS).

RESULTS

Forty percent of the sample was taking at least one medication with a CALS AC potency of ≥1. These individuals were reported to have greater executive function difficulties and more changes in cognition/behavior relative to those not taking AC medications.

DISCUSSION

AC medication use may represent a modifiable risk factor for cognitive deterioration in adults with DS; more research on this topic, particularly with older adults with DS, is needed.

Highlights

• Identifying modifiable risk factors for dementia in Down syndrome (DS) is critical.

• A risk factor studied in the general population is anticholinergic (AC) medication.

• This risk factor has not been studied in DS, a high‐risk group.

• AC medication use was associated with everyday cognitive challenges in young adults with DS.

• Longitudinal studies across adulthood, including older adults with DS, are needed.

1. BACKGROUND

In recent years, there has been increasing interest in identifying modifiable risk factors for dementia in both the general population ¹ and populations at elevated risk, such as Down syndrome (DS). ² An emerging line of research suggests that medications that inhibit the activity of acetylcholine, a neurotransmitter important for learning and memory, ³ may represent one such risk factor. ⁴ This is because acetylcholine is thought to be an important neurotransmitter involved with Alzheimer's disease (AD). ⁵ Consequently, it has also been identified as a key target in symptomatic treatment. Specifically, cholinesterase inhibitors (i.e., drugs that increase available acetylcholine at the synapse by inhibiting breakdown) are commonly prescribed medications to treat AD symptoms in the general population ⁶ and individuals with DS. ⁷ , ⁸ , ⁹ , ¹⁰

In contrast, anticholinergic (AC) medications are thought to inhibit or block acetylcholine's activity at the synapse. These prescription and over‐the‐counter medications are used to treat prevalent conditions (e.g., respiratory, gastrointestinal, psychiatric). ¹¹ Over the past two decades, attention has been drawn to the detrimental effects of AC medications, particularly in older adults. ¹² One area of inquiry focuses on associations between AC medications and cognitive functioning. In particular, these medications have been tied to increased risk of dementia ⁴ , ¹³ , ¹⁴ , ¹⁵ and broader cognitive difficulties, including poorer executive function skills (i.e., higher level cognitive/behavioral regulation skills necessary for the completion of complex tasks). ¹⁴ , ¹⁶ , ¹⁷ , ¹⁸

Despite this, research investigating links between AC medications and aspects of cognition and behavior that are relevant to dementia among adults with DS is lacking. This dearth of research is surprising, given that DS has been described as a “genetic form of dementia,” ² and research suggests that adults with intellectual disability (ID) may be exposed to higher rates of AC medications than adults in the general population. ¹⁹ , ²⁰ , ²¹

To our knowledge, just two studies have examined links between AC medications and functional outcomes among individuals with neurodevelopmental conditions. First, a recent investigation ²² documented high rates of AC medication use among autistic adults (without ID) and reported an association between AC medication burden and self‐reported cognitive decline. Second, a study of adults with ID examined associations between AC medications, sedative medications, and instrumental activities of daily living (iADLs). ²¹ After adjusting for covariates, they found that sedative use, but not AC medication use, was linked to reduced independence in iADLs. These researchers did not complete separate analyses with adults with DS; thus, conclusions cannot be made about possible relations between AC medications and the constructs under investigation in DS specifically.

Given (1) the strong ties between DS and AD²; (2) the high rates of AC medication use among adults with heterogeneous ID ¹⁹ , ²⁰ ; (3) research from the general population documenting relations between AC medications and cognitive impairments, including dementia ¹⁵ , ²³ ; and (4) the suggestion that AC medication use should be considered a prevention target for AD, ⁴ the current research sought to characterize AC medication use and its association with reports of cognitive and behavioral changes on the Dementia Screening Questionnaire for Individuals with Intellectual Disabilities (DSQIID) in a sample of young adults with DS. It also sought to examine the association between AC medication use and contemporaneous executive dysfunction measured with the Behavior Rating Inventory of Executive Function‐Adult (BRIEF‐A). The decision to focus on current executive dysfunction alongside cognitive and behavioral changes was driven by the fact that (1) a common early indicator of dementia in DS is worsening executive dysfunction, ²⁴ (2) executive dysfunction is prevalent in DS across the lifespan, ²⁵ and (3) AC use has been tied to poorer executive function among older adults in the general population. ¹⁴ , ¹⁶ , ¹⁷ , ¹⁸ It was hypothesized that AC medication use would be associated with higher scores (denoting greater difficulties) on the DSQIID and the BRIEF‐A.

RESEARCH IN CONTEXT

1. Systematic review: Using traditional review methods (e.g., Web of Science), the literature was searched for articles examining anticholinergic (AC) medication use in Down syndrome (DS). No studies were found that examined relations between AC medication use and changes in cognition and behavior in DS.

2. Interpretation: Using informant questionnaires, the current study found that there was a modest association between taking AC medications and both (1) deterioration in cognition/behavior and (2) greater contemporaneous executive dysfunction challenges in 18‐ to 39‐year‐old adults with DS.

3. Future directions: More work is needed to understand the links between AC medications, executive function, and dementia risk in adults with DS. Prospective, longitudinal research, particularly focused on older adults with DS who have been evaluated for dementia, is needed to elucidate relationships between AC medication use and dementia risk in this vulnerable group.

2. METHODS

Following guidelines set forth in the Declaration of Helsinki and approval by the Drexel University Institutional Review Board, parents or adult siblings of young adults with DS completed an online eligibility screening questionnaire, informed consent procedures, and study questionnaires evaluating the health and behavior of the individual with DS via secure online platforms (REDCap [Research Electronic Data Capture] ²⁶ , ²⁷ and test publishers’ websites). Following questionnaire completion, participants received a $30 e‐gift card.

2.1. Participants

2.1.1. Recruitment, inclusion criteria, and demographic characteristics

Close family members (parents: n = 106; siblings: n = 2) of 108 individuals with DS completed questionnaires about their loved one. These participants were recruited via the National Institutes of Health participant registry, DS‐Connect, and via partnerships with DS clinics around the United States, the National Down Syndrome Society, local DS family support groups, and in‐person events, such as Buddy Walks and the National Down Syndrome Society Adult Summit and Advocacy Conference. To be included, participants were required to be the close family member (e.g., parent or sibling) of a young adult with DS, be 18 years of age or older, and be able to read English well enough to report on their family member's health and behavior. The family member with DS was required to have a genetically‐confirmed diagnosis of DS due to trisomy 21 or translocation per family member report, be between 18 and 39 years of age, and not have a history of the following: an acquired head injury involving a loss of consciousness, hydrocephalus, or encephalopathy. In addition, to qualify for inclusion in the current report, participants were required to have complete medication data (i.e., adequate details to permit coding medications by class and AC burden) as well as data on the DSQIID.

The 108 individuals with DS comprising the final sample had a mean age of 27.58 (SD = 4.99) and included 53 female (49.1%) and 55 male (50.9%) participants. The sample included 2 (1.9%) Hispanic participants and 105 (97.2%) non‐Hispanic participants. (Ethnicity information was missing for 1 [0.9%] participant.) The racial breakdown was as follows: Asian, n = 5 (4.6%); Black, n = 1 (0.9%); more than one race, n = 8 (7.4%), White, n = 93 (86.1%), and missing race data, n = 1 (0.9%).

Of these 108 individuals, 89 (82.4%) lived with the rater who completed study procedures, whereas 19 (17.6%) did not. To rule out concerns that raters who did not live with their family member with DS may have impacted study findings, primary study analyses that compared those who were and were not taking AC medications on the DSQIID and BRIEF‐A were re‐run in the sub‐sample of 89 individuals who resided in the same household as the rater. Results of these analyses were nearly identical to the results focused on the complete sample of 108 participants (see Supporting Information for a side‐by‐side comparison). Given this, the primary study findings (which are reported in the main text) are based on the complete sample of 108 individuals.

2.1.2. Data validity checks

As this was an online study, numerous validity checks were put into place at different stages of the research process to ensure data validity and exclude fraudulent data. Many of the approaches used in the current investigation have been used effectively in other online, survey‐based studies of DS (e.g., ²⁸ ). These checks were as follows. At the consent stage, (1) comprehension checks were included on consent forms and (2) other bot‐defeating measures were in place (consistent with recommendations from other research). ²⁹ Once consent and contact information were obtained, responses were probed for (1) duplicate email addresses and phone numbers; (2) suspicious names (e.g., Peter Pan) and incomplete or suspicious addresses (e.g., addresses that were not residences); and (3) questionable recruitment sources (e.g., multiple sign‐ups listing the same recruitment site that was spelled incorrectly in each one). Following these preliminary screening procedures, study surveys were sent out to participants.

Figure 1 outlines data inclusion and exclusion at different stages of the data validity check/quality control process once survey links were sent out to the participants. Of the 155 personalized survey links emailed to participants (Figure 1, Box a), responses were received from 126 individuals (Figure 1, Box b). Survey responses were then scrutinized for validity as follows: (1) confirming that the participant's reported mailing address was indeed a residence using Google Maps street or satellite view functions; (2) reviewing birth dates reported for the individual with DS and the respondent on the initial contact information form (completed at the consent stage) and ages for these individuals reported in study measures (completed days, if not weeks later) for consistency (e.g., we confirmed that that the birth date reported for the person with DS on the contact information form, which was completed prior to study questionnaire links being sent out was consistent with the age reported for the person with DS in the study questionnaires); (3) examining questionnaire completion times; and (4) reviewing the content of responses to ensure the absence of fraudulent responses (e.g., receiving a DS diagnosis other than at birth or during the prenatal period, except under rare circumstances, such as international adoption). Any concerns about consistency and data validity resulted in study exclusion. Of the 118 participants who passed data validity checks (Figure 1, Box c), a further requirement for inclusion in the current report, as described above, was having data on the two primary domains of interest–medication usage and cognitive and behavioral changes reported on the DSQIID. This resulted in the final sample: n = 108 participants (Figure 1, Box d).

FIGURE 1.

Flow of participants through different study inclusion steps and validation procedures.

2.2. Measures

2.2.1. Background questionnaire assessing demographic and medication information

Participants reported on their family member's demographic characteristics, health history, and medication use on a questionnaire developed for the current study. Family members were asked to provide information about their loved one's medication usage as follows: “Is the individual currently taking any medication?” If they responded “yes,” they were prompted with the following instructions and completed an open text field: “Please list any medications the individual is taking and the condition or symptoms it is used to treat. For example, Zoloft (depression); Synthroid (hypothyroid); Zyrtec (allergies).”

2.2.2. Dementia Screening Questionnaire for Individuals with Intellectual Disability (DSQIID)

Family members of the individual with DS completed the DSQIID ³⁰ electronically via the REDCap platform. ²⁶ , ²⁷ The DSQIID was developed to be completed by the caregivers of individuals with ID to screen for cognitive and behavioral changes associated with dementia risk. It has three parts. Part 1 queries raters about the individual's “best ability level.” Parts 2 and 3 focus on changes in cognition and behavior and are the focus of the current investigation. Specifically, Part 2 includes 43 items and asks about changes in cognition and behavior that are commonly reported for individuals with ID experiencing dementia (e.g., “can't find words,” “can't recognize familiar persons”). Part 3 of the DSQIID includes 10 items that contain comparative statements to again capture changes in cognition and behavior (e.g., “speaks (signs) less”). Response options include “does not apply,” “always been the case,” “always but worse,” or “new symptom.” Per the DSQIID authors, questions responded to with “always been the case” or “does not apply” receive scores of 0; questions responded to with “always but worse” or “new symptom” receive scores of 1. The sum of the 53 items included in Parts 2 and 3 represents the individual's score on the instrument and was the primary variable of interest in the current study. Higher scores on this measure denote greater concerns about changes in cognition and behavior.

The DSQIID has adequate psychometric properties, with strong test–retest reliability (r = 0.95) and interrater reliability (intraclass correlation coefficient: r = 0.9). ³⁰ Within our sample, internal consistency was evaluated and deemed to be excellent: Cronbach's α = 0.93. To maximize sample size, participants with data missing on less than 10% of the items from Parts 2 and 3 were included in analyses. To address missing data, the mean score on the items in Part 2 was used to replace any missing item‐level data in Part 2. Similarly, the mean score on the items in Part 3 was used to replace any missing item‐level data in Part 3.

According to the DSQIID guidelines, a score of ≥20 (the screening cutoff score) suggests concern for dementia‐related symptomatology, and follow‐up with a health care provider for evaluation is recommended. Just two participants in the sample screened positive. Both participants were taking AC medications. One participant was 29 years of age; the other participant was 30 years of age. Because the DSQIID is used as a questionnaire for research purposes, no formal follow‐up evaluation was completed as a part of the research procedures. However, families were notified of the positive screening and encouraged to discuss the findings with their loved one's physician. A review of the medical history survey obtained as a part of study procedures provided the following additional information about these two participants. The 29‐year‐old participant was reported to have experienced regression, with a likely diagnosis of Down syndrome regression disorder (DSRD) prior to the age of 20. It was reported that although symptoms had improved, the family was still seeking further information about the possibility of DSRD and other treatment options. The 30‐year‐old participant was also reported to have experienced regression, including catatonia. For this participant, the regression occurred in their late 20s. This individual had also experienced the following changes shortly before the completion of the DSQIID: a recent illness and a medication change as well as an illness of a close family member. Thus, consistent with studies that indicate that the average age at onset for dementia in people with DS is in their mid‐50s and that cases are generally not reported prior to age 35, ³¹ , ³² these young individuals appear to have screened positive on the DSQIID not due to dementia onset, but rather due to a combination of medical and psychosocial factors, most notably regressive symptoms suggestive of DSRD, illness, and challenging life circumstances.

2.2.3. Behavior Rating Inventory of Executive Function–Adult (BRIEF‐A) ³³ Informant report

The BRIEF‐A was also completed by family members (via the PariConnect website). Because this measure was completed after the DSQIID and required participants to go to a different website to complete the measure, there was modest participant attrition. Thus, BRIEF‐A data were not available for all study participants with DSQIID data. Instead, these data were available for 83 participants.

The BRIEF‐A is a standardized questionnaire designed to capture executive functioning behaviors in real‐world settings, as reported by someone who knows the participant well. It includes 75 items rated on a 3‐point Likert scale (1 = never, 2 = sometimes, and 3 = often), reflecting the frequency of executive functioning difficulties. The measure yields two index scores—the Behavioral Regulation Index (BRI) and the Metacognition Index (MI)—as well as a Global Executive Composite (GEC) score. The BRI includes the Inhibit, Shift, Emotional Control, and Self‐Monitor subscales, whereas the MI includes the Initiate, Working Memory, Plan/Organize, Task Monitor, and Organization of Materials subscales, all of which contribute to the overall GEC score. Per the user manual, the BRIEF‐A demonstrates strong internal consistency, test–retest reliability, and validity. Within the current sample, internal consistency was noted to be excellent: Cronbach's α = 0.96. The GEC raw score was used as the variable of interest in the current study. Higher GEC raw scores indicate greater executive functioning difficulties.

2.2.4. Vineland Adaptive Behavior Scales – Third Edition (VABS‐3) parent/caregiver form

The VABS‐3 ³⁴ parent/caregiver domain‐level form was also completed by family members of the individual with DS electronically via Q‐Global. This computer‐adapted version of the VABS‐3 asks questions about adaptive behavior skills in three overarching domains: communication, social, and daily living skills. The VABS‐3 is normed for individuals ages 3 years and older, has been used extensively with samples with ID, and has strong internal consistency (Cronbach's α = 0.98). For the current investigation, the VABS‐3 adaptive behavior composite (ABC) was used to compare adaptive behavior among those who were and were not taking AC medications.

2.3. Coding, medication inclusion criteria, and group assignment

2.3.1. Medication coding

The Anatomical Therapeutic Chemical (ATC) code was assigned for each medication using the World Health Organization (WHO) Collaborating Center for Drug Statistics Methodology ATC classification system (https://atcddd.fhi.no/). This was done to characterize the most common medications taken by sample participants. After applying ATC codes, a medication count was generated for each participant. This count included all prescription and over‐the‐counter medications the participant was reported to be taking. Medications were coded for their AC potency using the CRIDECO Anticholinergic Load Scale (CALS). ¹¹ To create the scale, the CALS's authors completed a systematic literature review to identify existing scales that characterized the AC properties of different medications. These included scales from countries in Europe and Asia as well as North and South America. Using these different scales, the authors of the CALS created a comprehensive tool that characterizes the AC potency of 217 medications using a scale from 1 to 3, where 1 corresponds to “low potency” and 3 to “high potency.” For participants in the current sample, each medication reported was coded using the CALS and assigned one of the following scores: 1 = low AC potency; 2 = medium AC potency; 3 = high AC potency; 0 = no AC potency. These scores were then summed to obtain a total AC score.

2.3.2. Medication inclusion criteria

For the purposes of the current report, participants were included in analyses only if the information provided about their medications could be unambiguously coded for their AC properties. Thus, if participants were said to be taking medications, but no details were provided, they were excluded. If participants were said to be taking a particular class of medication but the specific medication was not provided (e.g., the individual takes an allergy medication), they were also excluded. However, there were two exceptions to this. First, if the individual was reported to be taking an oral contraceptive or a statin (for cholesterol) but the specific medication was not provided, these participants were included, as these medication classes are not known to have AC properties and thus these medications could be assigned a CALS score of 0. Second, because the primary variable of interest in the current report (as detailed below) was whether the person was taking any AC medications, a person with an ambiguous medication but who was also taking a codable AC medication was included in the AC medication use positive group (see below). In contrast, the opposite scenario—that is, a person taking one ambiguous medication but not taking any other AC medications— resulted in exclusion from the study. Following the aforementioned guidelines for inclusion, four participants were excluded due to ambiguity in their medication use.

2.3.3. Anticholinergic medication use (AMU) group assignment

Medication potency scores for all medications were summed to create an AC total score. These scores ranged from 0 to 7. However, the majority of participants (i.e., 27 of 44) who were taking AC medications were taking only one low potency AC medication, which resulted in a total AC score of 1. Thus, participants were stratified into two groups. Participants were assigned to the “anticholinergic medication use positive” (AMU+) group if they were taking any medications with an AC potency of 1 or higher. Participants who had total AC scores of 0, indicating that none of the medications they were taking were noted to have AC properties, were assigned to the “anticholinergic medication use negative” (AMU−) group. Individuals in the sample who were not taking any medications were assigned to the AMU− group by default. In total, 44 participants were included in the AMU+ group and 64 participants were included in AMU− group.

2.4. Analytic Plan

Analyses were completed in four parts. Part 1 analyses focused on characterizing medication use in general (i.e., all prescription and over‐the‐counter medications regardless of their AC properties) and AC medication use specifically. Part 2 focused on examining differences in participant characteristics as a function of AC medication use, with a particular emphasis on group differences in changes in cognition/behavior as measured via the DSQIID and current executive function using the BRIEF‐A. Part 3 focused on examining relationships among medications, demographic characteristics, and both DSQIID and BRIEF‐A scores dimensionally using Spearman rank‐order correlations. Part 4 focused on sensitivity analyses. Analyses associated with these four parts are detailed below.

In Part 1, information about medication use in the sample (regardless of AC potency) was first summarized to provide a description of medication use more broadly. Then medications with different AC potencies were considered and participants were stratified into groups based on whether they were taking at least one medication with an AC potency of 1 or higher. To provide a description of the types of medications participants were taking, medication counts were presented as a function of the ATC first level (anatomic) classification.

In Part 2, relations between AC medication use and participant characteristics were examined. First, group differences in demographic characteristics and total medication use were examined to determine if covariates needed to be included in the primary analyses that focused on group differences on the DSQIID and BRIEF‐A. The VABS‐3 was also included here to characterize the sample's adaptive function skills overall and to provide another measure of behavioral functioning to demonstrate the discriminant validity of the findings involving the DSQIID and BRIEF‐A (considered next). Second, group differences in DSQIID and BRIEF‐A scores were considered. Prior to conducting these analyses, data were inspected for normality.

Because the DSQIID is a screening questionnaire, scores were not expected to be normally distributed. Although scores ranged from 0 to 32 on the measure (with 20 being the screening cutoff for recommended dementia evaluation), the majority of participants received a score of 0 on the questionnaire. Due to the distribution of these data (skew >3; kurtosis >16; Kolmogorov–Smirnov and Shapiro–Wilks tests, p’s < .01), non‐parametric tests (Mann–Whitney U) were used to compare DSQIID scores as a function of AMU group assignment. In contrast, the BRIEF‐A GEC score data were normally distributed (skew and kurtosis <1; Kolmogorov–Smirnov and Shapiro–Wilks tests were non‐significant). As a result, t‐tests were used to compare group differences on the BRIEF‐A.

In Part 3, associations between demographic, medication, and DSQIID data were explored using Spearman rank‐order non‐parametric correlations. Finally, in Part 4, two sets of sensitivity analyses were completed to evaluate alternative hypotheses about what could be driving study findings. First, as presented below in the results section, there was an association between total medication count (regardless of AC potency) and DSQIID scores; this association was not observed with BRIEF‐A scores. To ensure that AC medications, specifically, not total medication count, were uniquely associated with higher DSQIID scores, partial Spearman rank‐order non‐parametric correlations were completed examining the association between total AC potency scores and DSQIID scores controlling for total medication count (and vice versa). Second, to exclude the possibility that heightened DSQIID scores in the AMU+ group were driven by participants in that group who were only taking medications with AC potency scores of ≥1 to treat nervous system disorders, AC potency total scores were re‐calculated excluding medications falling into the ATC nervous system anatomic group. Participants whose AC potency total score dropped to 0 were excluded from analyses. Then, the Mann–Whitney U test was re‐run with the smaller AMU+ group (n = 26) relative to the complete AMU− group (n = 64) for the DSQIID. Similarly, the t‐test was re‐run for the smaller AMU+ group (n = 19) relative to the complete AMU− group (n = 49) for the BRIEF‐A. In addition, Spearman rank‐order correlations were completed examining associations between both the DSQIID and BRIEF‐A scores and AC total scores that excluded nervous system medications. Finally, an additional, very conservative set of sensitivity analyses were completed in which participants taking any nervous system medication(s) with AC potency scores of ≥1 were excluded (regardless of whether they would have been included in the AMU+ group due to non‐nervous system medications with AC potencies of ≥1). This conservative, follow‐up set of sensitivity analyses contrasts with the one above in the following way. In the analyses above, AC potency total scores were recalculated with nervous system medications removed. Thus, if a participant was taking a nervous system medication with an AC potency of ≥1 but was also taking a non‐nervous system medication with an AC potency of ≥1, they were included in the AMU+ group in the analyses above because their total AC score was 1 or greater. However, in this very conservative follow‐up set of sensitivity analyses, this participant was excluded completely. As a result, the AMU+ group was very small for this follow‐up analysis (DSQIID: AMU+ group n = 17; BRIEF‐A: AMU+ group n = 13). Due to space limitations and the very small sample size, the results of this set of follow‐up sensitivity analyses are mentioned briefly in the text of the main document but are detailed in the Supporting Information.

3. RESULTS

3.1. Medication use: totals of all medications and stratification by AC potency

In total, 90 participants were taking at least one prescription or over‐the‐counter medication; 18 were not (Figure 2a). Total medication use per participant ranged from 0 to 15 medications (Figure 2b). These medications had AC potency scores ranging from 0 to 3. As stated in the Methods, AC potency scores across all medications were summed for each participant. If participants had a total AC potency score of 0 (which included those who were taking no medications), they were placed in the “anticholinergic medication use negative” (AMU−) group; n = 64. If they had an AC potency score of 1 or greater, they were placed in the “anticholinergic medication use positive” (AMU+) group; n = 44. See Figure 2c for AMU‐ and AMU+ breakdown. AC potency total scores ranged from 0 to 7. See Figure 2d for the count of participants with different AC potency total scores. Finally, ATC groupings were considered. Participants were taking medications from all of the ATC first‐level groups (corresponding to the main anatomic group) except “P‐antiparasitic products, insecticides and repellents” and “V‐various.” A breakdown of the number of participants taking at least one medication in each of the ATC first‐level groups according to the AC potency (AC = 0 or AC ≥1) is provided in Figure 2e. As can be seen, the greatest counts of participants taking medications with an AC ≥1 were in the “N‐Nervous system” and “R‐Respiratory system” groups. The numbers of participants taking at least one medication in the different nervous system pharmacological subgroups (third‐level ATC grouping) were as follows: antidepressants, n = 20; antipsychotics, n = 5; anxiolytics, n = 3; antiepileptics, n = 2; dopaminergics, n = 1; anticholinergics, n = 1. Regarding respiratory system medications, the breakdown of participants taking medications from different pharmacological subgroups (third‐level ATC grouping) was as follows: antihistamines, n = 18; expectorants, n = 1.

FIGURE 2.

Medication use in the sample. (A) Participants were stratified by whether they took any medications. (B) Counts of participants taking different quantities of medications. (C) Participants were stratified by whether they took at least one medication with an anticholinergic (AC) potency of ≥1 (AC medication use positive or AMU+ group) or <1 (AC medication use negative or AMU− group). (D) Counts of participants with different AC potency total scores (i.e., the sum of AC potency scores across all medications taken, with those not taking any medications receiving an AC potency total score of 0 by default). (E) Counts of participants taking at least one medication in the different Anatomical Therapeutic Chemical (ATC) first‐level (anatomic) groups as a function of AC potency. Note: AC total potency score may be an underestimate for three participants in the AMU+ group. These participants were taking at least one medication with an AC potency ≥1; however, their loved one reported that they were taking another mediation that could not be coded for AC potency due to insufficient details.

Given this study's focus on AC medications, it is useful to note that only one participant in the sample was taking an acetylcholinesterase inhibitor—that is, a medication used to treat dementia that increases available acetylcholine at the synapse (and thus, is not an AC medication). This participant was in the AMU− group and did not screen positive on the DSQIID. The parent of the participant noted that this medication was prescribed to support memory and verbal expression without elaborating as to whether these skills had declined prior to the onset of the medication.

3.2. Examining differences in participant characteristics as a function of AC medication use

Next, differences in participant characteristics for those in the AMU+ and AMU− groups were considered. First, differences in background characteristics were considered to identify variables that may need to be included as covariates in analyses. Second, primary study analyses were completed, in which group differences (AMU+ and AMU−) on the DSQIID and the BRIEF‐A were probed. Finally, relations among all variables of interest (demographic, medication use, AC potency scores, and DSQIID scores) were examined dimensionally using Spearman rank‐order non‐parametric correlations.

3.2.1. AMU+ and AMU− group differences in background characteristics

Table 1 displays the demographic characteristics of the sample. It also includes VABS‐3 scores for a subset of participants who had these data available (n = 83). These data were included to characterize the sample's adaptive function skills overall and to provide another measure of behavioral functioning to demonstrate the discriminant validity of the relationship between AMU group membership and DSQIID/BRIEF‐A scores (described in the next section).

TABLE 1.

Demographics and descriptive information by anticholinergic medication use (AMU) group.

	AMU− (n = 64)		AMU+ (n = 44)		Statistical significance
	M	SD	M	SD	t
Age	27.08	5.13	28.30	4.73	t (106) = −1.25, p > 0.20
Vineland ABC SS1	74.24	13.72	71.68	9.73	t (81) = 0.94, p > 0.30

	Md	IQR	Md	IQR	U
Medication total	1	0‐2	3	2‐5	U = 2493.00, p < 0.001

	N	%	n	%	Χ2
Sex					Χ2 (1) = 0.05, p > 0.80
Female	32	50.00	21	47.73
Male	32	50.00	23	52.27
Race					Χ2 (1) = 0.05, p > 0.80
White	56	87.50	37	84.09
Other	8	12.50	6	13.64
Missing			1	2.27
Ethnicity					Χ2 (1) = 0.08, p > 0.70
Hispanic	1	1.56	1	2.27
Non‐Hispanic	63	98.44	42	95.45
Missing			1	2.27
Rater relationship					Χ2 (1) = 1.40, p > 0.20
Parent	62	96.88	44	100.00
Sibling	2	3.13	0	0.00
Rater sex					Χ2 (1) = 0.10, p > 0.70
Female	57	89.06	40	90.91
Male	7	10.94	4	9.09

Notes: ¹Vineland ABC: AMU− group: n = 49; AMU+ group: n = 34.

Abbreviations: AMU, anticholinergic medication use; DSQIID, Dementia Screening Questionnaire for Individuals with Intellectual Disabilities; Vineland ABC SS, Vineland Adaptive Behavior Scales – Third Edition Parent/Caregiver Report Adaptive Behavior Composite Standard Score.

As can be seen, the groups did not differ statistically on age, sex assigned at birth, ethno‐racial background, or VABS‐3 adaptive behavior composite scores. However, the groups did differ on the median number of medications taken. Not surprisingly, those in the AMU+ group took a greater number of medications than those in the AMU− group.

3.2.2. AMU+ and AMU− group differences in DSQIID scores

Next, the DSQIID distributions for AMU+ and AMU− groups were compared using a Mann–Whitney U test (given the non‐normal nature of the data; see 2.4, Analytic Plan, for details). The Mann–Whitney U test indicated significantly different distributions for the two groups (U = 1940.50; p < 0.001; effect size, r _pb = 0.4 [medium]). Specifically, the mean rank for the DSQIID scores of the AMU+ group was significantly higher (denoting more reported changes in cognitive and behavioral functioning) than for the AMU− group. In addition to the higher mean rank of DSQIID scores in the AMU+ group, this group also had a wider range of scores (variance = 44.04) than the AMU− group (variance = 3.19). Although both groups had a minimum score of 0, the maximum score in the AMU− group was 8, whereas the maximum score in the AMU+ group was 32. The distribution of scores, along with the median and interquartile range (IQR) for each of the groups, are presented in Figure 3.

FIGURE 3.

Box and whisker plot of Dementia Screening Questionnaire for Individuals with Intellectual Disabilities (DSQIID) scores as a function of anticholinergic medication use group membership. Note that higher scores on the DSQIID denote greater declines in cognition/behavior as reported by family members. Also note that the horizontal lines in the boxes represent the group medians (Anticholinergic medication use negative or AMU− group: Md = 0; Anticholinergic medication use positive or AMU+ group: Md = 1) and the box represents the interquartile range (AMU− group: IQR = 0−1; AMU+ group: IQR = 0−6.5). The range of scores for each of the groups was as follows: AMU−: 0 to 8. AMU+: 0 to 32. Finally, the variance of the AMU− group was 3.19 and the variance for the AMU+ group was 44.04, suggesting greater variance in scores in the AMU+ than in the AMU− group.

3.2.3. AMU+ and AMU− group differences in BRIEF‐A scores

Next, the AMU+ and AMU− groups’ BRIEF‐A GEC scores were compared using a t‐test. Consistent with the DSQIID findings, the AMU+ group received a higher mean score on the BRIEF‐A than the AMU− group, denoting greater everyday executive function challenges (t[81] = 3.73, p < 0.001; effect size, Cohen's d 0.8 [large]); see Figure 4.

FIGURE 4.

Bar graph presenting Behavior Rating Inventory of Executive Function – Adult (BRIEF–A) Global Executive Composite (GEC) raw scores as a function of anticholinergic medication use group membership. Note that higher scores the BRIEF‐A denote greater executive function challenges. Also note that the error bars correspond to the standard error of the mean. For the anticholinergic medication negative or AMU− group, the mean (SD) and range of scores were as follows, respectively: mean = 108.84 (SD = 20.54); range: 76−171. For the anticholinergic medication use positive or AMU+ group, the mean (SD) and range of scores were as follows, respectively: mean = 126.88 (SD = 23.23); range: 78−189.

3.3. Relations among variables of interest

Next Spearman rank‐order non‐parametric correlations were run to examine associations among total AC potency scores, total medication use, DSQIID and BRIEF‐A GEC scores, and other participant characteristics; these are summarized in Table 2. As can be seen, higher DSQIID scores were associated with higher total AC potency scores (Spearman ρ = 0.36 [effect size = medium]) and higher total medication use (Spearman ρ = 0.26 [effect size = small to medium]). As a result, analyses were undertaken in the following section to control for the possible role that total medications may play in the association between DSQIID scores and AC potency.

TABLE 2.

Spearman rank‐order correlations.

	AC potency total	Total medication count	DSQIID score	BRIEF‐A GEC
Age	0.13	0.27 **	0.01	0.00
Sex (F1, M2)	0.04	0.01	−0.06	0.23 *
VABS‐3 ABC	−0.15	−0.14	0.29 **	−0.59 ***
AC potency total	–	–	–	–
Total medication count	0.71 ***	–	–	–
DSQIID score	0.36 ***	0.26 **	–	–
BRIEF‐A GEC	0.41 ***	0.17	0.27 *	–

Notes: (1) n = 108 for all analyses except for those involving the VABS‐3 ABC and BRIEF‐A GEC raw score in which n = 83. The analysis for the VABS‐3 ABC and BRIEF‐A GEC raw score included 76 participants. (2) AC total potency score may be an underestimate for three participants in the AMU+ group. These participants were taking at least one medication with an AC potency ≥1; however, their loved one reported that they were taking another mediation that could not be coded for AC potency due to insufficient details.

Abbreviations: AC, anticholinergic; BRIEF‐A GEC, Behavior Rating Inventory of Executive Function–Adult version Global Executive Composite raw score; DSQIID, Dementia Screening Questionnaire for Individuals with Intellectual disabilities; VABS‐3 ABC, Vineland Adaptive Behavior Scales, Third Edition−Adaptive Behavior Composite standard score.

*p < 0.05; **p < 0.01; ***p < 0.001.

BRIEF‐A GEC scores were also significantly associated with total AC potency (Spearman ρ = 0.41 [effect size = medium]). However, they were not significantly correlated with total medications (Spearman ρ = 0.17 [effect size = small]). Finally, consistent with the group level comparisons summarized in Table 1, AC potency score was not significantly related to age, sex, or VABS‐3 scores.

3.4. Sensitivity analyses to rule out alternative hypothesis

3.4.1. Alternative Hypothesis 1. It is the total number of medications of any kind, not AC medications, that is driving the association with DSQIID scores

In order to exclude the possibility that the total number of medications and not the use of medications with an AC potency of 1 or higher was driving the DSQIID findings, partial Spearman rank‐order correlations were completed in which DSQIID scores were correlated with (1) AC total potency scores controlling for medication total and (2) medication total controlling for AC total potency scores. Consistent with primary analyses, DSQIID scores were significantly associated with total AC potency scores when total medication values were controlled (Spearman ρ _p = 0.26, p < 0.01). In contrast, when DSQIID scores were correlated with total medications and AC total potency scores were controlled, significant associations were not found (Spearman ρ _p = 0.01, p > 0.80).

3.4.2. Alternative Hypothesis 2: DSQIID and BRIEF‐A GEC scores are higher in the AMU+ group due to a high number of participants taking nervous system medications with AC potencies of ≥1

Cross‐sectional analyses such as the ones included in this report prevent interpretations about the directional relationships between constructs. Given the high number of medications with an AC potency >0 that are nervous system medications, an alternative explanation for the findings here is that individuals taking nervous system medications are prescribed these due to cognitive difficulties that would result in elevated DSQIID and BRIEF‐A scores. We attempted to rule out the possibility that these participants were driving the findings by removing nervous system medications from the AC potency total score. Participants who were included in the AMU+ group due to the exclusive use of medications in the “N‐Nervous system” ATC group were excluded from analyses (because by removing these medications from AC potency total score, their scores were reduced to 0).

For the DSQIID, this resulted in an AMU+ group that included 26 participants (because 18 of the 44 participants in the AMU+ group were taking only AC medications that belonged to the “N‐Nervous system” ATC group). The smaller AMU+ group was then compared to the AMU− group using a Mann−Whitney U test. Results were consistent with those when the entire AMU+ group was considered—the DSQIID distributions differed for the groups and the mean rank for the DSQIID was higher in the (now smaller) AMU+ group compared to the AMU− group (U = 1147.50, p < 0.01; effect size, r_pb = 0.3 [medium]). (See Supporting Information for the median, interquartile range, minimum, maximum, and effect size for the AMU− group and this smaller AMU+ group.) As a point of comparison, these values are also provided for the complete sample of 108 participants.

To complement these group‐level comparisons, Spearman correlations were re‐run to examine relations between the DSQIID total score and AC potency total score with nervous system medications removed. This analysis excluded those who were in the AMU+ group due to exclusive use of nervous system medications with an AC potency ≥1. In this sub‐sample, the relationship between AC total and DSQIID scores remained (Spearman ρ = 0.36, p < 0.05; n = 90).

These sensitivity analyses, removing those who were in the AMU− group due to exclusive use of nervous system medications with AC properties, were completed for the BRIEF‐A as well. This resulted in an AMU+ group that included 19 participants (because 15 of the 34 participants in the AMU+ group who had available BRIEF‐A data were taking only AC medications that belong to the “N‐Nervous system” ATC group). When the t‐test comparing these groups was re‐run, results were maintained, with the AMU+ group receiving higher BRIEF‐A GEC raw scores (denoting greater challenges) than the AMU− group (t[66] = 2.18, p < 0.04; Cohen's d = 0.6; medium effect). See Supporting Information for the mean, SD, minimum, maximum, and effect size results for the AMU− group and this smaller AMU+ group. Again, as a point of comparison, these values are also provided for the complete sample of 83 participants with BRIEF‐A data.

Again, to complement group‐level comparisons, Spearman correlations were re‐run between the AC potency total score and BRIEF‐A GEC raw score with this sub‐sample. Although the magnitude of Spearman ρ was reduced (ρ = 0.29, p < 0.02; n = 68), a significant correlation persisted.

Finally, to be very conservative, the DSQIID and BRIEF‐A group comparisons described were re‐run with an even smaller sub‐sample that excluded any participant taking nervous system medications that had an AC potency of 1 or higher. (Stated another way, unlike the analysis above in which participants remained in the AMU+ group if they were taking at least one non‐nervous system medication with an AC potency of ≥1, in this analysis, any participant taking a nervous system medication with an AC potency of ≥1 was excluded regardless of the AC potency of the other medications they took.) Effect sizes associated with these conservative follow‐up analyses were as follows. For the DSQIID, a small effect, indicating (non‐significantly) greater impairment in AMU+ group (n = 17) relative to the AMU− group (n = 64) was observed. For the BRIEF‐A, a medium effect was observed, indicating (non‐significantly) greater impairment in AMU+ group (n = 13) relative to the AMU− group (n = 49). (See Supporting Information for details.)

4. DISCUSSION

In the first study of its kind, the current research documented AC medication use and its links to cognition/behavior (as measured via a screener for cognitive and behavioral changes associated with dementia risk and a questionnaire assessing current executive functioning) in young adults with DS. Consistent with heterogeneity in the DS phenotype, ³⁵ considerable variability was observed in both total medication use and AC medication use specifically. Although 83% of the sample was taking ≥1 medication, total medication counts varied from 0 to 15. For AC medications, 41% of the sample was taking at least one medication with an AC potency of ≥1. Total AC potency scores ¹¹ ranged from 0 to 7 (with 61% of those in the AMU+ group obtaining a score of 1). This AC potency appears to be relatively low compared to adults with heterogeneous ID ³⁶ and to autistic adults without ID. ²² Nevertheless, the modest AC medication use observed in this sample related to greater changes in cognition/behavior on a dementia screener and greater current informant‐reported executive function challenges.

Specifically, individuals taking AC medications had higher scores on the DSQIID, indicating more deterioration in cognition/behavior than in those not taking AC medications. Similarly, they presented with higher levels of current everyday executive dysfunction on the BRIEF‐A. Discriminant validity for these associations was observed, as AC medication use was not significantly associated with contemporaneous adaptive behavior skills (on the VABS‐3).

To our knowledge, this is the first study to report a link between AC medications and informant report of either cognitive/behavioral deterioration or executive dysfunction in adults with DS. Regarding the DSQIID findings, it is important to note that although this research represents a potential lead in identifying modifiable risk factors for dementia in this vulnerable population, findings should be interpreted within the context of the DSQIID's scope of measurement, the age of participants (i.e., 18−39 years), and the modest changes in cognition/behavior noted for most of the young adults with DS. Although the DSQIID was developed to screen for changes in cognition/behavior that may be indicative of dementia, the changes indexed may also be markers of other clinical entities, such as DSRD, ³⁷ depression, ³⁸ and untreated or undetected sleep apnea or hypothyroidism. Moreover, elevated scores may be associated with stressful life circumstances, such as moves and personal losses, as well as a lack of cognitive stimulation in daily activities. This is particularly important to consider given the young age of the sample and the low rates of DSQIID symptom endorsement overall. Specifically, more than 50% of the participants received a DSQIID score of 0, indicating no declines in cognition/behavior. Moreover, only two participants scored over the dementia risk screening threshold on the DSQIID, and these participants had reportedly experienced regression that was suggestive of DSRD. Because research indicates that dementia diagnoses are exceptionally rare among individuals with DS younger than age 40, ³¹ , ³² it is important to emphasize that we are not equating the functional declines measured by the DSQIID in the current report with dementia. Rather, we are describing the association between a continuous measure of cognitive/behavioral decline and AC medication use.

It is important to note that the current study's correlational study design precludes drawing conclusions about the directionality of the relationship between AC medications and changes in cognition/behavior, particularly given that the most frequent ATC anatomic grouping associated with AC potency was “N‐Nervous system” medications. Consequently, an alternative explanation for the findings is that individuals taking nervous system medications are prescribed these due to changes in cognition/behavior that would contribute to elevated DSQIID or BRIEF‐A scores. We attempted to rule out the possibility that these participants were driving the findings in follow‐up analyses by excluding nervous system medications from AC total scores and re‐running analyses three ways. First, we removed individuals who were in the AMU+ group who were only taking AC nervous system medications and continued to find higher DSQIID and BRIEF‐A scores relative to the AMU−group. Second, we examined associations between dimensional DSQIID and BRIEF‐A scores and total AC scores that excluded nervous system medications and found that although the magnitude of the association was reduced (in the smaller sub‐sample), a statistically significant association between total AC scores and both DSQIID and BRIEF‐A scores remained. Finally, in a very conservative set of follow‐up sensitivity analyses (presented primarily in Supporting Information), all participants taking nervous system medications with an AC potency of ≥1 were excluded from analyses. Although the magnitude of the group difference was reduced to a small effect for the DSQIID, the effect size of the group differences for the BRIEF‐A was medium. Taken together, the association between AC medication and everyday cognitive difficulties reported here does not appear to be driven exclusively by AC nervous system medications. (This appears to be especially true for executive dysfunction, in which medium effect sizes were observed in sensitivity analyses.)

As stated earlier, compared to reports of AC medication use in other groups, such as autistic adults, AC potency total scores were relatively “low” for the current sample with DS. For example, <7% of our sample had a total AC score of ≥3 (high potency per the CALS) compared to over 26% in a study of autistic adults. ²² Thus, finding links between AC medication use and deterioration in cognition/behavior in such a young sample with a relatively low AC burden is quite remarkable. One possible explanation for this could relate to differences in how individuals with DS process AC medications. For example, it has been suggested that individuals with DS may absorb medications differently than their non‐DS peers. ³⁹ This difference in pharmacodynamics could alter their vulnerability to the adverse effects of AC medications. Indirect support for this possibility comes from research that suggests that older adults may be particularly susceptible to the adverse effects of AC medications ⁴⁰ due to age‐related changes in how the body processes these medications ⁴¹ (e.g., changes in blood−brain barrier permeability ⁴² ). Given that research suggests that individuals with DS experience premature aging ⁴³ and may be subject to cellular senescence at earlier ages than individuals without DS, ⁴⁴ even low levels of AC medication exposure may result in adverse consequences. Further support for differential sensitivity comes from research conducted in the 1980s, suggesting that individuals with DS respond differently to AC medications than their peers without DS. ⁴⁵ Future investigations are needed on this topic not only in people with DS but also animal models of the syndrome, where the ability to experimentally control exposure to AC medications and examine the cognitive/behavioral sequelae of this exposure is possible.

This study has limitations. First, the study relied upon informant report of cognition/behavior at one point in time. Given the study's inclusion of families from around the country, direct cognitive testing was not feasible, and we do not yet have longitudinal data from this cohort to track changes in cognition/behavior over time. Second, family members reported current medication use by the individual with DS. Thus, we do not have information about how long individuals were taking these medications, the dose, or past AC medication use. Moreover, informant report can result in medication reporting inaccuracies. However, research suggests that electronic health records, which can be an alternative source of information about medication use, are also prone to inaccuracies. ⁴⁶ Thus, although relying on informant report may be a limitation of the current study, it appears that other methods of studying medication use may also have their limitations. Thus, it is important to consider findings within the context of how the data were acquired. Third, the current study's sample comprised individuals with DS who were primarily White and non‐Hispanic, which limits generalizability of findings to more diverse groups of young adults with DS.

Acknowledging these limitations, we also note study strengths. First, by using informant report tools, we were able to capture changes in cognition/behavior as well as current executive functioning among individuals with DS with a wide range of ability levels, something that can be difficult when using direct testing tools with this group. Second, by collecting data online, we were able to capture responses from families from a large geographic region, increasing generalizability to a broader group of individuals with DS.

In closing, the current study provides the first report of its kind linking AC medications to deteriorations in cognition/behavior as well as greater levels of executive dysfunction in DS. The linkage between AC medications and less favorable ratings of cognition and behavior is remarkable, given the young age of the sample and the relatively low levels of AC medication use. These findings suggest that associations between AC medications and cognition in DS may be even more impactful later in adulthood. Thus, investigations into this topic in older groups with DS, particularly using longitudinal study designs and a combination of informant report and direct cognitive testing, are warranted. Should the current findings be replicated with longitudinal investigations, AC medication use may represent a modifiable risk factor for dementia in this already vulnerable population.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest. Author disclosures are available in the supporting information.

CONSENT STATEMENT

All participants provided informed consent prior to completing study procedures.

Supporting information

Supporting Information

Supporting Information

Lee NR, McQuaid GA, Jiddou H, et al. Anticholinergics, executive function, and cognitive/behavioral changes in Down syndrome. Alzheimer's Dement. 2025;21:e70649. 10.1002/alz.70649