Alcohol Use, Mental Health, and Functional Capacity as Predictors of Workplace Disability in a Cohort With Manifest Huntington’s Disease

Enroll-HD (ClinicalTrials.gov identifier NCT01574053) is a global clinical research platform designed to facilitate clinical research in HD. As of July 1, 2019, there were 18,845 active participants enrolled at 174 sites in 19 countries. Individuals with premanifest HD or manifest HD, genotype negative or unknown, family members, and community controls participate in the Enroll-HD study. The diagnosis of HD is based on two factors: investigator-determined status, including clinical signs and symptoms, and genotyping status. Participants are classified as manifest HD if they show evidence of clinically significant motor symptoms (diagnostic confidence level of 4 from the Unified Huntington’s Disease Rating Scale [UHDRS], which is associated with motor abnormalities that are considered unequivocal signs of HD, scored by the investigator).

All enrolled participants undergo an annual assessment using a core data set covering motor, cognitive, and behavioral symptoms, with assessments also exploring work productivity measures, quality of life, and physical functioning. All sites are required to obtain and maintain local Ethics Committee approvals, and the project conforms to the provisions of the Declaration of Helsinki. Written informed consent was obtained from all participants. Data are monitored for quality and accuracy by using a risk-based monitoring approach, and all participants are deidentified.

For this study, we analyzed the third Enroll-HD baseline data set, extracted from the database on October 31, 2016, and made available on December 15, 2016. The sample consisted of 316 participants who were classified as manifest HD, having a CAG repeat length range between 39 and 60, and currently engaged in paid full- or part-time employment, who provided data (at least one subscale total) about work disability outcomes as measured by the Work Productivity and Activity Impairment-Specific Health Problem (WPAI-SHP) questionnaire.

Data were collected on the following measures from included participants.

Because cell sizes were small for certain categories of predictor variables, some of the predictor variable categories were combined prior to analysis. Region was combined into three categories: Australasia, Europe, and the Americas (North and Latin America). Years of education was combined into three categories: below diploma (International Standard Classification of Education [ISCED] levels 0, 1, 2, and 3), diploma-certificate (ISCED level 4), and tertiary studies (ISCED levels 5 and 6). Dichotomized variables were marital status (single versus married or in a partnership); residence (town or city versus rural or village), and smoking and drinking status (alcohol drinker versus nondrinker, smoker versus nonsmoker). Employment status was categorized as full-time, part-time, and self-employed. To address study aims (and because the WPAI-SHP data for our sample were zero inflated), the WPAI-SHP variables were dichotomized into presence or absence of any impairment for analysis by using binary logistic regression models, following the approach suggested in the literature (27).

In the data set, values were missing for 11 of the 22 variables, with the proportions ranging from 0.3% (residence and TMS) to 32.6% (overall work impairment). Of the 316 individuals who constituted the study cohort, 46.5% had one or more missing values, with the missing values in total accounting for 5.2% of the total values. Before the transformation of the four WPAI-SHP variables mentioned above and formal data analysis, we performed multiple imputation to address the missing data based on the assumption that data were missing at random (28–30). Specifically, we included all baseline variables of interest in the imputation model, selected “Automatic” imputation method, set the number of imputations as 35 (slightly larger than the highest proportion of missing values [32.6%] for the variable overall work impairment), and set constraints for all continuous variables to ensure that the imputed values would be within the normal range of the variables (29, 31). After creating 35 imputed data sets, we used standard statistical methods (logistical regression) to fit the model of interest to each of the imputed data sets. Estimated associations in each imputed data set were averaged together to give overall estimated associations between the predictor variables and outcome variables (28, 29).

We calculated mean values and SDs as well as frequencies and percentages for sociodemographic, lifestyle, clinical, and neuropsychological measures. To identify significant predictors of the four WPAI-SHP outcomes, we first conducted a series of univariate logistic regression analyses (“Enter” method) for each outcome variable using the imputed data sets. Variables with a p value <0.25 were included in a multivariable logistic regression analysis (“Enter” method) for each WPAI-SHP outcome (32). We also conducted the same regression analyses using the data set with missing data and found some different significant factors. Given that missing data, if not specified, can lead to the loss of power and potentially biased results, we reported the results of the imputed data sets (31).

Collinearity of all predictor variables were tested by using the linear regression procedure. Statistical significance was assessed using a p value <0.05 and a 95% confidence interval in logistic regression models. Odds ratios were used to quantify the association between independent and outcome variables. All analyses were performed using SPSS 22 (IBM Corporation, Armonk, N.Y.).

Sociodemographic, lifestyle, clinical, and neuropsychological characteristics are summarized in Table 1. Most participants were male (52.5%), were married or in a partnership (61.7%), worked full-time (64.2%), and lived in a town or city (79.7%). The mean age of participants was 47.0 years (SD=10.5, range 22–76), and most were from Europe (64.2%), followed by the Americas (30.7%) and Australasia (5.1%).

Mean CAG repeat length was 44.5 (SD=3.4, range=39–60). Participants had a mean TMS score of 24.4 (SD=12.9, range=0–87), which suggested a wide variability in motor symptoms in the participants classified as manifest HD (the maximum possible total TMS score is 124). Nevertheless, all included participants had been classified clinically as having manifest HD by the Enroll-HD study investigators on the basis of the criteria described above. Higher scores on the TMS indicate more impaired motor function, and individuals with manifest HD are classified in Enroll-HD based on TMS scores >5 (as consistent with previous literature [33]). In our sample, 308 participants had a TMS score >5 (97.7% of our cohort).

Participants had a mean MMSE score of 27.2 out of a possible 30 (SD=2.4, range=20–30), indicating minimal cognitive impairment. The mean TFC score was close to the maximum level of functioning of 13 (11.5, SD=1.7, range=4–13). The mean total scores on both the HADS anxiety (HADS-A) and HADS depression (HADS-D) subscales were below the cutoff score of 7, indicating absence of clinically relevant symptoms of anxiety and depression (HADS-A: 6.8, SD=4.2, range=0–21; HADS-D: 6.0, SD=4.0, range=0–18). Likewise, participants fell significantly below the cutoff score of 14 points for irritability measured by the Snaith Irritability scale (6.6, SD=4.5, range=0–20). On the SF-12, the mean physical composite score (PCS) (50.7, SD=8.6) and the mean mental composite score (MCS) (44.2, SD=10.8) indicated an “average” level of physical and mental health in this sample, according to the normative mean of 50.

As shown in Table 1, the proportion of participants who reported having missed work due to HD symptoms was 20.3% (N=64). The proportion who reported experiencing impairment while working due to HD was 60.1% (N=190). In addition, 79.1% (N=250) reported having work-related activity impairment due to HD, and 60.8% (N=192) reported experiencing impairment in overall work productivity due to HD.

Based on the results of univariate logistic regression analyses using the significance level of 0.25, the potential significant predictors for each outcome were determined. For absenteeism, they were education (p=0.15), current alcohol use (p<0.001), TMS score (p=0.08), SF-12 MCS (p=0.01), TFC score (p<0.001), MMSE score (p=0.03), HADS-D score (p=0.01), irritability score (p=0.02), and SF-12 PCS (p=0.01). For presenteeism, they were region (p<0.01), current alcohol use (p=0.06), current smoker (p=0.06), TFC score (p=0.02), HADS-D score (p=0.11), irritability score (p=0.08), and SF-12 MCS (p=0.01). For activity impairment, they were current alcohol use (p=0.20), current smoker (p=0.15), TMS score (p=0.08), TFC score (p=0.06), HADS-A score (p<0.001), HADS-D score (p<0.001), irritability score (p<0.001), SF-12 PCS (p=0.026), and SF-12 MCS (p<0.001). For overall work impairment, they were region (p=0.12), current alcohol use (p=0.18), current smoker (p=0.12), TFC score (p=0.04), HADS-A score (p=0.24), irritability score (p=0.09), and SF-12 MCS (p=0.02),

On the basis of multivariable logistic regression analyses of the imputed data sets, we identified significant (p<0.05) factors associated with two outcomes (Table 2). For absenteeism, significant factors included current alcohol use (odds ratio=2.86, 95% CI=1.62, 5.03) and TFC score (odds ratio=0.76, 95% CI=0.64, 0.91). Significant factors for activity impairment included SF-12 MCS (odds ratio=0.95, 95% CI=0.897, 0.995). In contrast, no significant factors were identified for the other two outcomes—presenteeism and work productivity loss–overall work impairment.

No collinearity within the independent variables was identified based on the collinearity statistics (tolerance value above 0.2). The Hosmer-Lemeshow test was used to assess the goodness of fit, with a significant p value indicating poor model fit. Using this criterion, all regression models were assessed as having a good model fit.

A recent systematic review noted that early identification of persons with dementia in the workplace, reasonable adjustments for people with dementia, and provision of information to raise awareness and facilitate informed choice were key factors in good employee management (37). Several key predictors of work disability outcomes identified in our study (current alcohol use and mental well-being) are modifiable and are potentially responsive to interventions. For example, there are many effective pharmacological and nonpharmacological treatments for mental health disorders, particularly for depression and anxiety. Based on our results, in order to minimize and delay workplace disability for persons with HD, health promotion should be a target for workplace health interventions, even if individuals report that they are functioning well. These health promotion interventions should strategically target mental health and well-being. Psychoeducation and teaching effective coping strategies could be incorporated into these interventions, which may potentially reduce alcohol use as a coping mechanism and may also positively affect functional capacity. These workplace interventions would be useful for all employees (regardless of diagnosis) and in all workplaces. Appropriate workplace support and modifications for persons with HD would engender inclusivity and diversity of employees, retain experienced workers, and keep persons with HD functioning within society for longer. Then, if and when they choose to leave the workforce, it can be done with dignity and choice.

We were limited in our choice of variables to those collected by the Enroll-HD study, and thus we were unable to explore other potential significant factors associated with workplace disability, such as job-related stress, job demands, and type of occupation. Exploring differences between occupations, such as between those with a greater or lesser degree of physical, cognitive, or emotional demands, may highlight important differences and identify occupation-specific predictors of workplace disability. In addition, it should be acknowledged that the Enroll-HD study relies on self-reported data for most outcomes. Because participants in this study had manifest HD, difficulties with capacity and insight may have affected the quality of the data collected. It is also well known that alcohol consumption is underreported in self-reports (38). Collection of collateral information from family members and employers may help to overcome this issue in future.

We also transformed some variables and combined some variable categories, leading to the loss of information. In addition, the size of the study cohort meant that it was not feasible to examine the interactions of variables associated with the outcomes. In future, we recommend longitudinal studies that examine time-dependent predictor variables to identify determinants of work disability outcomes in individuals with manifest HD and studies with larger samples so that interactions of variables can be investigated.

There is evidence that apathy is one of the most disabling behavioral symptoms for patients and caregivers, affecting up to 90% of individuals with HD (39). Greater levels of apathy have also been found to be related to less independence, increased motor impairment, and more clinician-rated behavioral problems (i.e., anger, irritability, and depression) (40). In addition, a recently published systematic review found that the most common significant factors associated with decreased functional capacity were depression and apathy (41). Given the prevalence of apathy among persons with HD reported in the literature and a recent study finding that apathy was an independent determinant of unemployment in HD mutation carriers (17), future research should include a measure of apathy in the methodology and analyses. This is particularly pertinent because our most significant predictors were related to absenteeism from work, which may be related to apathy manifesting as a reduction in the goal-directed behavior of attending work.

This study dichotomized participants into drinkers and nondrinkers because of small cell sizes, but it is also worth exploring the amount and type of alcohol consumption. There is a need to explore the extent of mental health problems in relation to alcohol consumption in order to guide interventions that decrease risk and improve psychological well-being and workplace ability of people with HD, particularly because substance abuse (including alcohol abuse) has been found to have a strong effect on the age at onset of motor symptoms in patients with HD in the Enroll-HD study (42). In addition, there are reports of positive aspects of leaving work, such as the sense of relief (43), which should be explored.

We used raw MMSE scores in our analysis, because we wished to directly compare findings in a premanifest and manifest cohort (Table 3), and previous studies have also found problems in the interpretation of transformed MMSE scores in specific populations (44). When we conducted an analysis after transforming MMSE scores to Z scores using established normative data (published 26 years ago) (45) to ascertain the potential influence of this approach to MMSE score interpretation on the study results, we observed two differences in the results of the multivariate regression following this analysis. Significant predictors of work productivity loss–overall work impairment included TFC score (odds ratio=−0.51, 95% CI=0.37, 0.96). In addition, where participants resided in the world (region) was a significant predictor of presenteeism (when workers come to work when they are unwell or attend work but work at levels that are less than optimal). The North and South American region had an odds ratio of 2.46 (95% CI=2.6, 77.8), and the European region had an odds ratio of 1.60 (95% CI=1.06, 23.3). All other results were the same. There is evidence to suggest that that the use of Z scores results in a loss of the meaningfulness of the raw score and standard deviation and may magnify certain effects and diminish others (46). Furthermore, we do not feel it is justified to transform MMSE raw scores into Z scores because age and education may influence not only MMSE scores but also other predictor variables employed in our study. Therefore, we elected to present the study results using raw MMSE scores. In addition, the participants represented a global cohort, with participants from Europe (64.2%), North and South America (30.7%), and Australasia (5.1%).

Although this study focused on worldwide predictors of workplace disability, it is known that the context of people’s lives determines their health. Future research should further investigate the social and economic determinants of employment and of health in HD, since the current literature is scarce in this area.