Sex differences for cognitive decline in progressive supranuclear palsy

Abstract

Introduction:

Cognitive dysfunction is a core clinical feature of progressive supranuclear palsy (PSP), with executive function being most markedly affected. In other neurodegenerative conditions, such as Alzheimer’s and Parkinson’s diseases, there are a growing number of reports demonstrating that cognition is differentially impacted in men and women. In PSP, however, the sex differences in cognitive decline have yet to be fully characterized.

Methods:

Data were obtained from the TAUROS trial for 139 participants with mild-to-moderate PSP (62 women, 77 men). Sex differences in longitudinal change in cognitive performance were evaluated with linear mixed models. Exploratory subgroup analyses assessed whether sex differences varied by baseline executive dysfunction, PSP phenotype, or baseline age.

Results:

In the primary whole group analyses, there were no sex differences for change in cognitive performance. Among participants with normal executive function at baseline, men declined more severely on executive function and language tests. Among the PSP-Parkinsonism subgroup, men declined more severely on category fluency. Across people aged≥65, men had a worse decline on category fluency, whereas across people aged <65, women had a worse decline on DRS construction.

Conclusion:

In people with mild-to-moderate PSP, there are no sex differences in cognitive decline. However, the rate of cognitive decline may differ for women and men based on the level of baseline executive dysfunction, PSP-phenotype and age. Further studies are needed to clarify how sex differences in PSP clinical progression vary by disease stage and to examine the contributions of co-pathology to these observed sex differences.

1. Introduction

Progressive supranuclear palsy (PSP) is a four repeat tauopathy clinically characterized by supranuclear gaze palsy, gait instability, akinesia, and cognitive dysfunction [1]. While a variety of cognitive domains can be affected in PSP, executive function is usually the first and most commonly involved [2]. Executive function is crucial for daily living [3] and correlates with microscopic measures of disease, such as tau burden, in people with PSP [4]. Furthermore, early presence of cognitive changes, especially executive dysfunction, is associated with shorter survival in PSP [5,6]. Thus, it is critical to better understand the changes in executive function in people with PSP.

Sex differences have been an emerging theme in the investigation of neurodegenerative disorders. There is an abundance of data suggesting that women are at greater risk of developing Alzheimer’s disease (AD) [7], whereas men are at greater risk for Parkinson’s disease (PD) [8]. In both, AD and PD, sex has been found to be associated with different rates of cognitive decline [9,10]. Interrogation of sex differences in these diseases has also led to discoveries into new potential disease mechanisms, such as sex-specific genetic and pathological factors [11–15]. In contrast to these more common neurodegenerative conditions, there is a paucity of such studies in PSP and their findings have been inconsistent. One study reported that apart from men having worse tremor, there were no other sex differences in the clinical expression of PSP [16]. Another study reported that men died sooner after PSP diagnosis than women [17]. Lastly, an additional large retrospective study reported several sex differences in PSP clinical expression, including women having worse executive dysfunction and faster rate of cognitive decline [18].

As has been the case for the common neurodegenerative diseases, deeper investigation into sex differences in PSP may reveal new mechanistic insights and sex-specific diagnostics, prognostics, and therapeutics. In this study, we investigated sex differences in longitudinal changes in different cognitive domains in a multinational cohort of PSP patients. In other neurodegenerative disorders, sex differences in cognition can vary by disease stage [19]; accordingly, we conducted subgroup analyses to explore whether sex differences vary by different levels of baseline executive function. Since clinical progression may vary by PSP phenotype [20], we also explored sex differences in two PSP phenotypes, PSP-Richardson syndrome and PSP-Parkinsonism. Lastly, since co-pathologies are common in older age, we performed an age-stratified subgroup analysis to explore the potential contributions of co-pathologies to our findings.

2. Methods

2.1. Study participants

Data was obtained from the Tau Restoration on PSP (TAUROS) trial (ClinicalTrials.gov Identifier NCT01049399). The TAUROS trial was a phase 2 randomized, double-blind, placebo-controlled, multi-center, and multi-national clinical trial evaluating the efficacy of NP03111240, a GSK-3 inhibitor, in the treatment of PSP [21]. The trial did not meet its primary or secondary endpoints. Inclusion criteria have been described previously [21]. Briefly, the trial included individuals aged 40 to 85 with a diagnosis of possible or probable PSP according to the National Institute of Neurological Disorders and Stroke-Society for PSP criteria [22], and a disease severity of mild-to-moderate stage [23]. The clinical trial protocol was approved by the associated local ethics committee of each of the clinical trial sites. Prior to participating in the study, signed informed consent was obtained from participants.

2.2. Demographic, clinical, and cognitive measures

As part of the clinical trial protocol, participants were evaluated at baseline, 4, 8, 26, 40, 52 and 60 weeks. Educational attainment was available as a categorical variable, with the following possible levels: able to read and write, less than elementary school, elementary school, less than high school, high school, more than high school, university degree, and postgraduate degree. In addition to these categories, we also binarized educational attainment as less than high school completion and at least high school completion. This was done in lieu of converting educational level to a continuous years-of-education variable because such a conversion is not straightforward in a multinational trial in which different countries have different educational systems.

From the clinical evaluation, we used disease duration, number of concomitant medical conditions, and the PSP Rating Scale (PSPRS) data for our analyses [23]. From the neuropsychological evaluation, we utilized the Dementia Rating Scale (DRS) total and domain-specific scores, Frontal Assessment Battery (FAB), lexical fluency (average score between A and S) and semantic fluency (average score between Animals and Supermarket) scores. A full clinical and neuropsychological evaluation was completed at the baseline visit. Follow-up PSPRS, DRS, and FAB scores were subsequently collected at 8, 26, 40, and 52 weeks. Lexical and semantic fluency were collected at all follow-up visits. In addition to considering FAB score as a continuous variable, it was also binarized as normal FAB versus abnormal FAB to estimate the absence/presence of executive dysfunction at baseline. Abnormal FAB was defined as having an age- and education-corrected FAB score below 13.48 [24].

2.3. Statistical analysis

R version 4.0.3 and Python 3.8.3 were used for statistical analyses. Baseline sex differences in demographic and cognitive variables were assessed with Mann-Whitney U-tests for continuous variables and χ² tests for categorical variables. Sex differences in longitudinal change in cognitive performance were assessed using linear mixed effects models. These models were fit using the lme4 and lmerTest packages in R [25,26], and can handle missing data using maximum likelihood estimation. Each model included the raw cognitive score as the outcome and age, sex, education, baseline raw cognitive score, disease duration, and interactions of these variables with time as predictors. A subject-specific random intercept and random slope were included in the model. Alpha-level was set at 0.05 for statistical significance and FDR correction was used to adjust for multiple comparisons.

For exploratory purposes, we also conducted subgroup analyses. In the first subgroup analysis, we split the sample by FAB performance, with a normal FAB subgroup and an abnormal FAB subgroup. Within each subgroup, we performed the linear mixed effects models to assess for sex differences in cognitive decline. In a second subgroup analysis, we split the sample by PSP-phenotype (PSP-Richardson syndrome and PSP-Parkinsonism; a single subject with PSP-pure akinesia was excluded) and again, linear mixed effects models were performed in each subgroup. We were also interested in the potential contributions of aging-related cognitive decline observed in our study and performed an age-stratified subgroup analysis. We split the sample into Younger (age<65) and Older (age ≥ 65) subgroups and conducted linear mixed effects models within each subgroup. The threshold of 65 years was chosen as this is typically considered the lower age cutoff for late-onset AD, the most common co-pathology [27], and dementias before this age are rare [28]. Each of the linear models included the raw cognitive score as the outcome and sex, education, age, baseline raw cognitive score, disease duration, and interactions of these variables with time as predictors. Given that these subgroup analyses were exploratory, p values were reported as uncorrected.

3. Results

3.1. Baseline characteristics

Our analysis included 139 participants (77 men, 62 women) from the TAUROS trial. At baseline, there were no significant sex differences in age at first study visit, education, age of disease onset, disease duration, number of medical comorbidities, PSP-phenotype, or PSPRS total score (Table 1). There were no significant sex differences in any of the baseline cognitive scores (Table 1).

Table 1.

Baseline demographics and cognitive scores.

	Men (N = 77)	Women (N = 62)	Comparison
Age at baseline visit, years	69.1 (7.8) [51.4–85.8]	68.4 (5.6) [54.0–84.3]	U = 2206.0, p = 0.78
Age at PSP onset, years	66.4 (8.3) [48.7–84.3]	65.3 (5.7) [51.5–82.4]	U = 2162.5, p = 0.78
Education, n			χ2 = 11.1, p = 0.62
Able to read/write	1	0
Less than elementary	1	2
Elementary	21	20
High school	20	25
More than high school	11	8
University degree	17	3
Post-graduate degree	6	4
Education (binarized), n
< High school	23	22	χ2 = 0.49, p = 0.79
≥ High school	54	40
Number Conditions	5.4 (3.8) [0.0–16.0]	5.6 (3.6) [1.0–17.0]	U = 2281.0, p = 0.78
Phenotype, n			χ2 = 0.90, p = 0.85
PSP-Parkinsonism	30	26
PSP-Richardson	46	36
Pure akinesia with freezing gait	1	0
PSPRS total	39.2 (13.5) [9.0–67.0]	39.3 (10.1) [10.0–64.0]	U = 2277.0, p = 0.78
Disease duration, years	2.8 (3.2) [0.03–22.8]	3.3 (2.5) [0.12–12.3]	U = 1890.0, p = 0.28
FAB
Total score	11.0 (4.2) [0–18]	11.6 (4.0) [3–18]	U = 2185.5, p = 0.78
Corrected FAB <13.48, n	46	41	χ2 = 0.5987, p = 0.78
DRS
Memory	20.5(5.3) [5–25]	21.0 (4.1) [12–25]	U = 2253.0, p = 0.78
Attention	32.9(4.8) [15–37]	33.2 (4.2) [19–37]	U = 2333.5, p = 0.78
Construction	4.1 (2.0) [0–6]	4.0 (2.3) [0–6]	U = 2321.0, p = 0.78
Initiation/Perseveration	24.2 (7.0) [4–37]	25.2 (7.3) [8–37]	U = 2205.0, p = 0.78
Conceptualization	32.2 (6.8) [5–39]	31.6 (6.0) [14–39]	U = 2165.0, p = 0.78
Total	114.0 (21.0) [33–144]	115.0 (19.4) [66–142]	U = 2312.0, p = 0.78
Fluency ^
Category	9.7 (4.5) [1.5–25.0]	10.2 (4.5) [1.5–22.0]	U = 2241.0, p = 0.78
Lexical	4.3 (3.2) [0.0–14.5]	5.3 (3.9) [1.5–22.0]	U = 2045.0, p = 0.16

Findings for continuous variables are reported as mean (SD), [minimum-maximum]. Findings for categorical variables are reported as counts. P-values are from Mann-Whitney test for continuous variables and from χ² test for categorical variables, and are FDR-corrected. PSP: Progressive Supranuclear Palsy, PSPRS: Progressive Supranuclear Palsy Rating Scale, FAB: Frontal Assessment Battery, DRS: Dementia Rating Scale.

^#Sum across disorientation, bradyphrenia, emotional incontinence, and grasping/imitative/utilizing behavior from the PSPRS.

^{^}Fluency values are reported as mean across the categories “animals” and “supermarket” for category fluency, and mean across the letters “A” and “S” for lexical fluency.

3.2. Sex differences for cognitive change over time

The median (SD) time-delay from baseline to the final follow-up visit was 1.15 (0.36) years for the overall cohort. Women and men had a similar change in the overall disease severity, as measured by the PSPRS, over the follow-up period. In terms of cognition, there were no sex differences for changes in DRS, FAB and letter fluency scores. There was a nominally significant interaction between sex and time for category fluency (uncorrected p = 0.007); men had a worse decline within a year compared to women (Fig. 1). However, this interaction did not survive multiple comparisons correction (FDR-corrected p = 0.22).

Fig. 1. Longitudinal change in cognitive test performance.

X-axis represents time (in years) elapsed since the baseline visit. Y-axis represents change in respective cognitive scores since baseline. The line indicates the linear regression model fit with time since baseline visit as the independent variable and change in cognitive score as the dependent variable.

3.3. FAB stratified subgroup analyses

In the FAB stratified subgroup analyses, there were 52 participants (31 men, 21 women) with normal FAB scores implicating normal executive function and 87 participants (46 men, 31, women) with abnormal FAB scores implicating abnormal executive function. Among the normal FAB subgroup, men were younger at baseline (uncorrected p = 0.03) and had a lower baseline DRS total score (uncorrected p = 0.03) (Supplementary Table 1). In the longitudinal analysis within this group, we found a significant interaction between sex and time for category fluency (uncorrected p = 1.63e-6), such that men had a more severe decline than women (Fig. 2). For lexical fluency (uncorrected p = 0.012), DRS Total (uncorrected p = 0.032), and DRS Initiation/Perseveration (uncorrected p = 0.0024), men also had a more severe decline than women (Fig. 2).

Fig. 2. Longitudinal change in cognitive test performance within normal FAB participants.

X-axis represents time (in years) elapsed since the baseline visit. Y-axis represents change in respective cognitive scores since baseline. The line indicates the linear regression model fit with time since baseline visit as the independent variable and change in cognitive score as the dependent variable.

In the abnormal FAB subgroup, women were younger at first visit (uncorrected p = 0.0028) and at disease onset (uncorrected p = 0.0026) (Supplementary Table 2). Compared to men, women had a higher baseline lexical fluency performance (uncorrected p = 0.03). Otherwise, there were no further baseline or longitudinal sex differences for cognition in this abnormal FAB subgroup.

3.4. PSP phenotype subgroup analyses

Among the PSP-Richardson subgroup there were 46 men and 36 women. Within this group, there were no sex differences in demographics or baseline cognitive performance (Supplementary Table 3). We also found no longitudinal change differences for this group in any of the cognitive tests we examined.

Among the PSP-Parkinsonism subgroup there were 30 men and 26 women. There were no baseline sex differences in demographics within this subgroup (Supplementary Table 4). Women had a higher baseline performance in DRS Construction (uncorrected p = 0.01), DRS Initiation/Perseveration (uncorrected p = 0.04), and lexical fluency (uncorrected p = 0.03). For longitudinal analyses, we found an interaction between sex and time only for category fluency (uncorrected p = 0.02, Fig. 3); men had a worse decline compared to women.

Fig. 3. Change in cognitive test performance of participants with PSP-Parkinsonism.

X-axis represents time (in years) elapsed since the baseline visit. Y-axis represents change in respective cognitive scores since baseline. The line indicates the linear regression model fit with time since baseline visit as the independent variable and change in cognitive score as the dependent variable.

3.5. Age-stratified subgroup analysis

There were 38 participants (23 men, 15 women) in the Younger subgroup (Supplementary Table 5) and 101 participants (54 men, 47 women) in the Older subgroup (Supplementary Table 6). Within the Younger subgroup, women had higher PSPRS (uncorrected p = 0.02) than men at baseline. In the longitudinal analysis for this subgroup, there was an interaction between sex and time only for DRS construction (uncorrected p = 0.04); men had slower decline than women.

In the Older subgroup, at baseline, women were younger had an earlier age at diagnosis (uncorrected p = 0.007) and had a longer disease duration (uncorrected p = 0.03). Women performed better at baseline than men on lexical fluency (uncorrected p = 0.03). In the longitudinal analysis, we found a sex and time interaction for category fluency (uncorrected p = 0.04); women had a slower decline than men.

4. Discussion

In this study, we investigated sex differences for changes across different cognitive domains using data from a clinical trial with a large multinational cohort of participants clinically diagnosed with PSP. In our primary analyses, women and men had similar cognitive performances at baseline and had similar rates of decline during the follow-up period. However, our exploratory subgroup analyses showed that among participants with a normal executive function at baseline, men had a more severe decline in cognition than women. Sex differences were also noted in the PSP-Parkinsonism group with men declining more than women on category fluency. These findings may suggest subtle sex differences in PSP based on level of executive dysfunction and phenotypic groups that need further investigations.

Despite the burgeoning number of studies dedicated to examining sex differences in other neurodegenerative conditions, there have been fewer such studies in PSP and, to date, these have produced ambiguous findings [16–18]. In this first prospective study, we attempted to clarify these associations. The primary analysis of our entire PSP cohort yielded a nominally significant association for category fluency. A prior finding of men dying sooner after diagnosis than women [17] could lend some credence to the faster progression in men that we detected in this primary analysis. The same study [17], though, found that men were also diagnosed later, which could offset the sex difference in time to mortality. Another study, with a larger cohort and longer follow-up, reported worse baseline executive dysfunction and faster time to severe cognitive impairment in women [18]. However, this association was also nominally significant and the study was retrospective. The study did not correct for multiple comparisons and their associations would not have met multiple comparisons correction. Taken together, among people with a clinical diagnosis of PSP, there does not appear to be convincing evidence for sex differences in cognitive decline in PSP.

While sex differences may not exist across people with PSP as a whole, our exploratory analyses revealed there may be sex differences in specific subgroups of this population. In the FAB-stratified analysis, we found an interaction between sex and time for category fluency in the normal FAB subgroup, but not the abnormal FAB subgroup. In healthy aging, women tend to excel at tasks that involve verbal and language skills [29–32]. This premorbid advantage may afford women the ability to maintain a level of cognitive performance despite harboring some amount of PSP pathology (i.e., cognitive reserve). In the context of AD, for example, women have preserved verbal memory abilities compared to men in the early disease stages [19,33]. Our findings may suggest that early-on in the course of PSP-related cognitive decline, men may decline faster than women, similar to what has been observed in AD. More to the point of AD, we unfortunately did not have AD biomarker characterization in this sample. With our age-stratified subgroup analysis, however, we attempted to interrogate the contributions of age-related co-pathologies to our findings. In the Older subgroup, the subgroup that is more likely to harbor age-related co-pathologies such as AD, we found that men declined faster than women on category fluency. This association was not present in the Younger subgroup. This finding, coupled with prior observations discussed above that women have preserved verbal test performance early in the course of AD suggests that the sex differences we are detecting may in part be due to age-related co-pathologies. However, without biomarker or pathological confirmation, we cannot ascertain the contributions of these other pathologies to our findings. Prior studies have reported that while co-pathologies are common in patients diagnosed with PSP, they are not associated with the clinical expression in PSP [27,34]. Overall, future studies into the potential sex differences for clinicopathological correlations in PSP can benefit from cohorts with more complete biomarker and pathological characterization through detailed history, genetic, biofluid, neuro-imaging and pathology data.

There is marked heterogeneity in the clinical presentation of PSP, with several recognized clinical phenotypes [35]. Prior work has demonstrated that the cognitive trajectories between phenotypes may differ, such that people with the PSP-Richardson phenotype deteriorate more rapidly [20,36], a phenomenon that has been demonstrated within this TAUROS cohort [37]. With regards to sex differences, there is some suggestion that there is a male predominance in the PSP-Richardson subtype; prior studies have reported 64% [20] and 59% [36] of their sample’s PSP-Richardson participants to be male. The sex distribution in PSP-Parkinsonism is thought to be even [20]. Given that the clinical picture may differ substantially between them, we stratified our sample by phenotype and examined sex differences within each. In accordance with prior studies, there were nominally more men (56%) than women within our PSP-Richardson subgroup. However, the phenotype distribution did not significantly vary by sex. In the PSP-Parkinsonism group, women had better executive functioning at baseline as suggested by higher scores on DRS Initiation/Perseveration and lexical fluency. Men had a worse decline in language, indicated by category fluency scores, than women. These findings may suggest subtle sex differences based on the PSP phenotype. However, as is the case for the FAB-stratified analyses, these phenotype-stratified analyses were also exploratory in nature and, thus, we take caution in the interpretation of our findings.

Our study is not without limitations. First, our sample comes from a clinical trial which limits the generalizability of our findings. For example, health conditions that are common in older adults of the general population, such as cerebrovascular disease and clinically significant heart disease [38], were part of the clinical trial exclusion criteria [21]. Thus, our findings may not necessarily be applicable to all people with PSP. The use of an investigational drug can also have impact on the disease course, and this impact can differ based on sex, as shown in other neurodegenerative diseases [39,40]. However, there was no treatment effect on change in the primary outcome (PSPRS) or any of the secondary cognitive outcomes in the TAUROS trial [21]. Moreover, the investigators did not report any sex- and gender-specific effects in this trial. Next, one could argue that the lack of sex differences could simply be due to a follow-up interval too short to detect meaningful cognitive decline. While the average follow-up of about one year is indeed short when considering other common neurodegenerative conditions, PSP typically progresses faster than these other disorders. In the context of other degenerative conditions, sex differences, if existent, are discernible even with cross-sectional data [33,41]. Thus, it is reasonable to expect that if there are meaningful sex differences in PSP, it would be detectable within our dataset. Another limitation of this study, as touched upon above, is the lack of pathological confirmation to confirm underlying PSP pathology and determine the contribution of co-pathologies to the clinical profile. Because of the well-known predilection of PSP for affecting executive function, the clinical trial preferentially collected data from neuropsychological tools testing this domain. There may be sex differences in other domains that were not detected as they were not probed as thoroughly. With further regards to the limitations in cognitive testing, practice effects may also be contributing. In our dataset, study participants were exposed to a test multiple times during a trial period in which the time between the first and last visit was on average 1.15 years. For verbal fluency, repeated administration at one month can lead to a statistically significant, but modest practice effect [42]. In mild PD, repeat administration of cognitive tests at 17 months resulted in significant practice effects, but with small effect size [43]. These data suggest that while practice effects may be affecting our findings, their contributions are minor and would have affected both sexes. Moreover, it has been shown that there is worsening in these parameters over time in this PSP cohort [37].

Given the nature of the data collection in the clinical trial, our cohort did not include a healthy control group. PSP is a disease of older age and the literature in healthy cognitive aging is rife with data showing sex-specific advantages in certain domains. Cognitively healthy women, for example, perform better on tests that involve verbal and language skills, while men tend to excel in tests of visuospatial function [29–32]. These sex differences persist throughout life; while cognition, on average, declines with age, this decline occurs in parallel between men and women [44,45]. The advantage for women in verbal tasks is relevant to our study as it may contribute to cognitive reserve for women, where they can maintain preservation in fluency performance compared to men early in the disease course. Similarly, the advantage for visuo-spatial skills for men may also be relevant to our study. In the age-stratified analysis, men had slower decline in DRS Construction than women in the Younger subgroup. Their premorbid advantage in visuo-spatial skills [29] may contribute to cognitive reserve for men in this domain. Interestingly, given the known pre-morbid advantages in certain domains, one might have expected baseline sex differences in performance of cognitive tests in our study. Across the entire sample, we did not find any cognitive sex differences at baseline. PSP may override the previously determined patterns of sex differences in healthy aging and the lack of sex difference in our study may, in fact, indicate a sex difference for PSP compared to healthy aging.

To conclude, among people diagnosed with PSP, there does not appear to be marked sex differences in cognitive decline. However, among subgroups of PSP, subtle sex differences may exist. Specifically, in people with normal FAB score, which may represent a state of yet unaffected executive function, men may decline faster in tests of executive function than women. These findings warrant further investigation into how sex differences vary by disease stage, phenotype, and the role of co-pathologies in contributing to these differences.

Data statement

The data used in these analyses are available upon reasonable request.