Brief assessment of cognitive function in myotonic dystrophy: multicenter longitudinal study using computer-assisted evaluation

Abstract

Introduction/Aims

Myotonic dystrophy type 1 (DM1) is known to affect cognitive function, but the best methods to assess CNS involvement in multicenter studies have not been determined. This study’s primary aim was to evaluate the potential of computerized cognitive tests to assess cognition in DM1.

Methods

We conducted a prospective, longitudinal, observational study of 113 adults with DM1 at 6 sites. Psychomotor speed, attention, working memory, and executive functioning were assessed at baseline, 3-months and 12-months using computerized cognitive tests. Results were compared with assessments of muscle function and patient reported outcomes (PROs), including the Myotonic Dystrophy Health Index (MDHI) and EQ-5D-5L.

Results

Based on intra-class correlation coefficients (ICCs), computerized cognitive tests had moderate to good reliability for psychomotor speed (0.76), attention (0.82), working memory speed (0.79), working memory accuracy (0.65), and executive functioning (0.87). Performance at baseline was lowest for working memory accuracy (p < 0.0001). Executive function performance improved from baseline to 3-months (p < 0.0001), without further changes over one year. There was a moderate correlation between poorer executive function and larger CTG repeat size (r = −0.433). There were some weak associations between PROs and cognitive performance.

Discussion

Computerized tests of cognition are feasible in multicenter studies of DM1. Poor performance was exhibited in working memory, which may be a useful variable in clinical trials. Learning effects may have contributed to the improvement in executive functioning. The relationship between PROs and cognitive impairment in DM1 requires further study.

1. INTRODUCTION

Myotonic dystrophy type 1 (DM1) is a dominantly inherited neuromuscular disorder characterized by progressive muscle wasting, weakness and myotonia¹. Dysfunction is present across multiple systems including the central nervous system (CNS). Cognitive problems are common in DM1 with findings of impairment in intellectual and executive function, visual spatial skills, attention, and memory in some individuals². Results of longitudinal studies have been mixed with some showing no decline and others showing process-specific declines over long intervals^3–6. The variability in pattern and range of cognitive impairments across studies has been attributed to the heterogenous nature of DM1⁷, even within the congenital, juvenile, and adult onset subtypes⁸.

As molecularly-targeted treatments for DM1 move closer to therapeutic trials there is a need for reliable measures of cognitive impact. Since the CNS is but one of several systems that need to be evaluated, it will be important to develop cognitive assessments that are sensitive to change yet concise and readily applicable across centers in multicenter studies, recognizing time constraints in study visits that also evaluate muscle, cardiac, and gastrointestinal impact. Computerized tests of cognitive function are finding increasing use in multicenter studies and clinical trials. It is unclear, however, how readily these methods can be applied to disorders, such as DM1, that also affect manual dexterity, attention, and social interactions. To that end, we assessed the feasibility of computerized cognitive testing to track the natural history of cognitive function in DM1 over one year, and we tested for associations of cognitive performance with CTG repeats, muscle impairment, and patient reported outcomes (PROs).

2. METHODS

2.1. Participants

Participants were enrolled at six sites of the Myotonic Dystrophy Clinical Research Network (DMCRN): University of Florida, University of Kansas Medical Center, Ohio State University, Stanford University, University of Rochester, and National Institutes of Health (NIH). Participants were recruited from the National Registry for Myotonic Dystrophy and Facioscapulohumeral Dystrophy Patients and Family Members, patient advocacy groups and local registries that included the Myotonic Dystrophy Foundation and the Muscular Dystrophy Association, as well as neuromuscular clinics. Recruitment efforts included postings in bulletins, emails and letters that were sent to potential enrollees in their respective states that met age, mobility and DM1 diagnosis criteria. This study was also publicly registered on clinicaltrials.gov (NCT02308657). All participants consented to participate in this study, which was approved by Institutional Review Boards from all participating centers.

Participants in the study were men and women from 18 to 70 years of age who (1) carried a clinical diagnosis of DM1; (2) had body mass index ≤ 33 kg/m2; (3) had CTG repeat size (if known) > 70; (4) had symptom onset after 10 years of age; and (5) were able to complete the 6-minute walk test without a cane or walker. Detailed exclusion criteria can be found in Supplemental Information.

2.2. Study Design

The study was designed as a prospective, longitudinal investigation of natural history over 1 year. Participants were evaluated at a clinical research center for 1-day sessions at 3 timepoints: screening/baseline (Visit 1), 3 months ±2 weeks (Visit 2), and 12 months ±4 weeks (Visit 3). Research visits also included testing of muscle strength and function (reported separately).

2.3. Cognitive Assessment

Commercially available computerized tests (Cogstate, Melbourne, Australia) have been developed to measure cognition with the advantages of short administration time, and sensitivity to changes in cognitive function of healthy individuals and those with CNS disorders. These have been validated in many disorders including mild traumatic brain injury, schizophrenia, AIDS dementia, and Alzheimer’s disease^9–12. They were designed to not be limited by level of education, culture, or repeat administrations. Selected tests were administered on laptop computers in the same sequence at each time point, in the presence of a study monitor in accordance with procedures recommended by the developers. Study monitors were trained via company webinars. Test sessions were approximately 20 to 30 minutes in duration. Modules selected for this study were Detection (DET), Identification (IDN), One-Back (ONB) and Groton Maze Learning Test (GMLT)^10,13. DET assessed simple reaction time and psychomotor function, IDN assessed choice reaction time and attention, ONB assessed working memory by use of a one-back paradigm, and GMLT assessed executive function. These modules were selected based on previous evidence that DM1 affects processing speed, attention and components of executive functioning⁸. Outcome variables and performance criteria are detailed in Supplemental Information.

For all data, z-scores were calculated based on age matched control subjects¹⁴ in order to determine if participants with DM1 exhibited impairment relative to healthy controls, and to compare changes over time. Z-scores were expressed as the number of standard deviations above (positive values) or below (negative values) the mean. Scores were considered impaired if they fell below one standard deviation of the mean, based on normative data that were previously collected¹⁵.

2.4. Patient Reported Outcomes

Participants completed the Myotonic Dystrophy Health Index (MDHI) at each study visit. The MDHI is a disease-specific patient reported outcome measure developed and validated to measure point-in-time multifactorial DM1 symptomatic burden during clinical trials^16,17. Prior studies have found the MDHI and its subscales to be reliable, correlate to functional state, be preferable over generic patient reported outcome, and have a stronger association with employment status and clinical function compared to generic and semi-generic patient reported outcome measures^18,19. The MDHI consists of 114 symptom questions across 17 subscales that measure self-report of mobility, upper extremity function on a 6-point Likert scale. The ability to perform activities, fatigue, pain, gastrointestinal issues, vision, communication, sleep, emotional issues, cognitive impairment, social satisfaction, social performance, myotonia, breathing, swallowing, and hearing. There is also a total score. Each subscale and the total score ranges from 0 to 100 with a higher score reflecting a greater degree of symptoms. The EQ-5D-5L was also utilized during the study as a generic descriptive profile of health status²⁰. It consists of 5 dimensions, which are mobility, self-care, usual activities, pain/discomfort, and anxiety/depression. There are 5 levels: Level 1 = no problems, Level 2 = slight problems, Level 3 = moderate problems, Level 4 = severe problems and Level 5 = extreme/unable to do.

2.5. Clinical and Genetic Assessment

Clinical muscle assessments were performed by trained physical therapists to determine the Muscular Impairment Rating Scale (MIRS)²¹, a general index of distal to proximal muscle impairment in DM1. CTG repeat size was determined by standard methods using PCR or Southern blot as previously described²².

2.6. Statistical Analysis

All data were entered in REDCap²³ (Research Electronic Data Capture), a secure, HIPAA-compliant database that was accessible to research staff only. Statistics were conducted in the R environment²⁴ or with SPSS Version 26 (IBM, Armonk, NY)²⁵.

We used intraclass correlation coefficients (ICCs) to determine test-retest reliability (2-way mixed, average measures, consistency)²⁶ on computerized cognitive tests from baseline to the 3-month visit. To determine change in computerized cognitive test performance from baseline through 1 year (i.e., longitudinal analyses), linear mixed effects models with a fixed effect for visit and a random intercept for participant were used to model the effect of visit on performance over the three visits. These models take into consideration an individual’s baseline performance, and they are robust and allow for missing data that can occur randomly, such as participants missing sessions, scheduling problems and/or computer issues. Missing data were not imputed. Each model was used to estimate mean outcomes at each visit as well as differences between outcomes from visit to visit.

To determine if specific cognitive domains were differentially affected by DM1, a repeated measures ANOVA was calculated across performances on computerized cognitive tests at each time point, followed by paired comparisons with Bonferroni corrections. Relationships between computerized cognitive test performances, demographic and clinical measures and PROs were quantified using either Pearson or Spearman rank correlations depending on whether the measure was continuous or ordinal. Correlations were designated as negligible if they ranged from 0.00 to 0.09, weak if they ranged from 0.10 to 0.39, moderate if they ranged from 0.40 to 0.69, strong if they ranged from 0.70 to 0.89 and very strong if they ranged from 0.90 to 1.00²⁷.

3. RESULTS

3.1. Cognitive

We screened 119 participants for eligibility and enrolled 113 adults with DM1 into the study (Table 1). The number of participants assessed at baseline, 3 months and 1 year are presented in Figure 1. There were no differences in demographic information between participants who completed the study and those who withdrew. Percentages ranged from 92.6% to 99% of evaluations across all tests for all participants who met completion criteria as established by Cogstate. Intra-class correlations across all tests ranged from 0.65 to 0.87 (Table 2).

Table 1.

Participant characteristics at baseline

Characteristic
Participant	n = 113
Age (years + SD)	43 (11.9)
Sex (% female)	59.3
Education (years ± SD)	15 (2.4)
Employed Full Time (%)	28.6
Age at 1st Symptom (years ± SD)	25 (11.6)
Age at Diagnosis (years ± SD)	33 (11.8)
Disease Duration (years ± SD)	18 (10.6)
CTG Repeats (± SD)	570 (353)

Figure 1.

Participant enrollment

Table 2.

Test-Retest results of computerized cognitive tests from baseline to 3 months

Test	ICC	Confidence Interval 95%	df1	df2	F Test	p-value
DET	0.76	(0.67, 0.82)	112	224	4.1	< 0.00001
IDN	0.82	(0.76, 0.87)	112	224	5.7	< 0.00001
ONBS	0.79	(0.71, 0.84)	112	224	4.7	< 0.00001
ONBA	0.65	(0.53, 0.75)	112	224	2.9	< 0.00001
GMLT	0.87	(0.82, 0.90)	112	224	7.5	< 0.00001

Abbreviations: DET = Detection, IDN = Identification, ONBS = One-back Speed, ONBA = One-back Accuracy, GMLT = Groton Maze Learning Test

Results of intra-class correlations (ICC) using 2-way mixed-effects, average measures, consistency ICC (3, k).

Baseline performance in our cohort trended lower than healthy controls but generally fell within one standard deviation of the mean (Figure 2). The one exception was working memory (ONB task) which showed impairment at baseline, (z-score = −1.60), 95% CI [−1.95, −1.25]. Accuracy on the working memory test remained one standard deviation below normative data at 1 year follow-up (z-score = −1.15), 95% CI [−1.15, −0.74]. The results of repeated measures ANOVA confirmed that there were significant differences in level of impairment across different computerized cognitive tests at baseline [F (4, 97) = 14.525, p < 0.0001], which persisted at follow-up [F (4, 92) = 23.314, p < 0.0001] at 3 months, and [F (4, 69) = 10.510, p < 0.0001] at 1 year. Pairwise comparisons with Bonferroni correction for each cognitive test at baseline, 3 months and 1 year are shown in Supplemental Table 1. Notably, the accuracy measure on the working memory test was significantly lower than for the simple reaction time test (p < 0.0001), choice reaction time test (p < 0.0001), or executive functioning test (p < 0.0001), or the speed of the working memory test (p < 0.0001), suggesting that reduced performance on working memory was not explained primarily by motor impairment.

Figure 2.

Performance on computerized cognitive tests at baseline. Z-scores represent the difference from healthy controls. Error bars represent standard errors. Z-scores between 0 and −1.0 are considered within normal limits and are depicted within the area between the two dotted blue lines.

*Repeated measures ANOVA with post-hoc paired comparisons with Bonferroni correction indicated that performance on the ONB Accuracy test was more impaired than the other computerized cognitive tests (p < 0.0001).

Abbreviations: DET = Detection, IDN = Identification, ONB = One-back, GMLT = Groton Maze Learning Test

Longitudinal analyses showed significant improvement of the executive function test (GMLT) (p < 0.0001) in the interval from baseline to 3 months (p < 0.0001) and was then stable from 3 months to 1 year (p = 0.618) (Table 3). There was no significant change over time for the other computerized cognitive tests. Of note, performance levels at baseline did not differ among study completers as compared to participants who withdrew or failed to complete testing at 1 year.

Table 3.

Results of computerized cognitive tests z-scores (means and 95% confidence intervals) at each study visit

Test	Baseline	3-Months	1-Year	p-value
DET	−0.62 (−0.84, −0.39)	−0.65 (−0.88, −0.42)	−0.87 (−1.12,−0.62)	0.062
IDN	−0.49 (−0.71, −0.28)	−0.45 (−0.67, −0.24)	−0.29 (−0.52,−0.06)	0.112
ONBS	−0.93 (−1.25, −0.71)	−0.91 (−1.13, −0.68)	−0.75 (−1.00, −0.51)	0.332
ONBA	−1.60 (−1.95, −1.25)	−1.33 (−1.70, −0.97)	−1.15 (−1.56, −0.74)	0.127
GMLT	−0.58 (−0.78, −0.37)	−0.18 (−0.39, −0.03)	−0.28 (−0.49, −0.06)	<0.0001

Abbreviations: DET = Detection, IDN = Identification, ONBS = One-back Speed, ONBA = One-back Accuracy, GMLT = Groton Maze Learning Test

The p-value represents the statistical significance of the linear mixed effects model for changes over time, adjusted with Bonferroni correction for multiple comparisons.

3.2. Cognitive Associations with Demographic Information, Patient-Reported Outcomes and Clinical Muscle Measures

There were no significant associations between gender or employment and performance on normalized cognitive tests at baseline. There were weak, but statistically significant correlations between education and choice reaction time (r = 0.207, p = 0.036) and education and executive functioning at baseline (r = 0.241, p = 0.014). A higher education level was associated with better performances on these tests. We also tested for associations among cognitive tests and age at first symptom, age at diagnosis, and disease duration. The only significant correlation was between age at first symptom and choice reaction time (ODN) at baseline (r = −0.214, p < 0.05), but this was also weak. Older age of onset at first symptom was associated with worse performance at baseline, but not at subsequent visits. There was not a significant association between muscle involvement and cognitive test performance as determined by the MIRS.

There was a statistically significant, but weak correlation, between size of the expanded CTG repeat and the one-back accuracy measure at baseline. Additionally, there was a moderate correlation between size of the expanded CTG repeat and executive functioning at baseline (Figures 3a and 3b). Larger CTG repeats were associated with lower performance. There were twelve participants with CTG repeats > 1000; 1^st symptom at age 10, 11, 12 (2 participants), 14, 16.5, 17, 19, >20 (2 participants), 29 and 46.

Figure 3.

Correlations between CTG repeat size and cognition

(A) Pearson correlation between CTG repeats and One-Back Speed, r = −0.299, p < 0.01

(B) Pearson correlation between CTG repeats and Groton Maze Learning Test, r = −0.433, p < 0.01

At baseline, the scores for 4 out of 5 cognitive tests were weakly correlated with MDHI total scores, and 3 of 5 cognitive tests were weakly correlated with scores of the MDHI’s cognitive subscale (Supplemental Table 2). All of the cognitive tests were weakly correlated with the MDHI’s activity participation subscale. No EQ-5D-5L subscales correlated with all of the cognitive tests (Supplemental Table 3).

4. DISCUSSION

In an unselected cohort of ambulatory patients with DM1 we found that nearly all participants were able to complete a brief battery of computerized cognitive tests, during study visits that also included muscle and cardiac assessment, and that results showed moderate to good test-retest reliability over 3 months. These findings attest to the feasibility of using computerized cognitive tests in adults with DM1.

Our study also showed a modest trend for adults with DM1 to display lower performance than age-matched controls, but scores were generally within one standard deviation of the mean values of controls. Many examples of cognitive tasks that trend lower than controls but not in a range that denotes clear cognitive impairment have been previously reported in DM1 adults, and there is wide agreement that attention, executive functioning, and visual spatial abilities are more impaired than other abilities^28–30. In a recent meta-analysis of cognition in DM1, visual spatial perception had the largest effect size³¹ indicating that this area of cognition was the most challenging for adults with DM1. The computerized cognitive test battery in this current study included attention and executive functioning measures but there was not a test for visual spatial perception. Visual spatial processing may be informative in future studies.

The task that was the most difficult for this group was the one-back test, measuring working memory, a component of executive functioning. In advance, there was concern that a simple one-back paradigm would have limited capability to detect disease impact. However, DM1 participants had reduced performance on this task across all three time points, particularly in terms of accuracy. In 53% of subjects, the results at one year were below the baseline score, even after performing the test on three occasions. Thus, while this test did not show evidence of progression, it also did not improve through practice alone, and could emerge as a useful measure in clinical studies. Speed on this measure tended to be lower than the simple and choice reaction time measures, suggesting the possibility of slower working memory processing. Working memory supports the active maintenance of task-relevant information during the performance of a cognitive task, and one of working memory’s main attributes is to update and support purposeful behavior³². Problems in working memory can lead to an inability to sequence steps and complete tasks, which is a concern in patients with DM1. Additionally, strong relationships between working memory and white matter integrity, as measured by diffusion tensor imaging (DTI), were previously reported in DM1³³.

Measures of cognitive performance in our study did not significantly decline over one year. This finding is not surprising in light of the slowly progressive nature of the disease and fits with results of previous longitudinal studies spanning 5, 9, 11 and 12 years^5,6,34,35. The 12-year study of 16 DM1 patients did not show significant longitudinal changes³⁵. In a much larger longitudinal study comparing DM1 participants and controls, significant changes occurred after 11 years, but only on measures that assessed visual spatial/construction abilities and visual memory⁶, whereas performance on measures of attention, reaction time, verbal abilities, verbal memory, and executive functioning did not significantly change. In fact, the only significant change over time in our study was apparent improvement of executive function from baseline to 3 months, as determined by the Groton Maze Learning Test, which suggested a learning or practice effect. To mitigate this effect in future studies, it may be advisable to administer a practice module prior to the baseline session. Interestingly, similar practice effects were not observed for the other tasks, consistent with previous studies using the same computerized cognitive tests^36–38.

Cognitive performance in our study was weakly associated with self-reported (MDHI instrument) assessment of cognitive function on three tasks. There was also a weak correlation with the MDHI subscale that measures activity participation. The EQ-5D-5L, a non-DM-specific PRO, demonstrated no consistent associations with performance on computerized cognitive tests. The MDHI was designed specifically for individuals with myotonic dystrophy, and prior studies have shown moderate to strong associations between MDHI subscales and strength and physical functioning, however, cognition had not been a focus of these studies^18,19. The weak correlations pertaining to cognition and self-report measures in DM1 may be because motor symptoms are more salient but another explanation could be lack of awareness of deficits. One study showed that approximately 51% of adults with DM1 had anosognosia³⁹, which would limit accuracy in self-report measures. Another study indicated similar findings, in that self-report of cognitive impairment was not strongly associated with objective test performances and was mediated by low mood⁴⁰. Therefore, adding family- and/or caregiver-report questionnaires in clinical trials as outcome measures should be considered.

It is also noteworthy that clinical signs of muscle impairment (MIRS) were not associated with cognitive performance. CTG expansion size was moderately associated with poorer executive function and weakly associated with poorer working memory, which has been previously reported^29,40,41.

There were limitations in this study. We enrolled an unselected group of ambulatory patients with DM1. If we had selected participants according to cognitive symptoms, the differences in cognition may have been more apparent. Since we performed multiple correlations to test for associations among clinical measures, self-report measures, and test performance, the possibility of Type 1 errors is increased. Accordingly, confirmatory studies are needed. And, as noted above, the absence of run-in or practice sessions to mitigate learning effects may have limited our ability to detect longitudinal decline.

In conclusion, computerized tests are feasible for assessing cognition in individuals with DM1. Compared to other cognitive measures of psychomotor speed, reaction time and executive functioning, we found that a one-back working memory paradigm captured an area of greater impairment in this population. Further studies are needed to determine if computerized tests of visual spatial function could be added to assess cognition in DM1. Family- and/or caregiver-report questionnaires are suggested to supplement PROs in clinical trials.

ABBREVIATIONS

CNS

Central Nervous System

DET

Detection

DM1

Myotonic Dystrophy Type-1

GMLT

Groton Maze Learning Test

ICCs

Intraclass Correlations

IDN

Identification

MDHI

Myotonic Dystrophy Health Index

MIRS

Muscle Impairment Rating Scale

ONB

One-back

PROs

Patient Reported Outcomes