Using Downgaze Palsy Progression Rate to Model Survival in Progressive Supranuclear Palsy–Richardson Syndrome

Rapid development of downgaze palsy, the most specific symptom of progressive supranuclear palsy (PSP), has been associated with shorter survival in small studies.


Introduction
For neurodegenerative disorders without specific treatment such as progressive supranuclear palsy (PSP), progression patterns are very important in clinical practice and trial design but are inadequately studied, possibly because large, multiyear, longitudinal data sets are lacking.
Downgaze palsy, the most specific symptom of PSP, 1 results from damage to the oculomotor nerve nuclei, with supranuclear input from the rostral interstitial nuclei. 2 The four-repeat tau pathology of PSP, best studied in PSP-Richardson syndrome and PSP-parkinsonism, spreads here from the globus pallidus interna, subthalamic nucleus, and substantia nigra before continuing rostrally to the frontal and parietal cortices and caudally to the pontine and cerebellar nuclei. 3,4 This pathogenetic pattern suggests that more rapid development of downgaze palsy could serve as a marker for more rapid overall disease progression and shorter survival duration.
A retrospective study on pathologically verified PSP cases supported this hypothesis, with earlier onset of the downgaze palsy associated with shorter survival. 5 However, in practice, detecting subtle downgaze palsy and determining its date of onset is difficult. Using the rate of progression of the downgaze palsy, as determined by quantitative assessment of severity when first detected divided by the disease duration at that point, may provide a useful marker for survival. Our recent prospective study on clinically diagnosed patients with PSP-Richardson syndrome and PSP-parkinsonism accomplished this by adding PSP onset age to the statistical model. 6 However, both studies were limited by the small sample size and by the failure to control for other variables potentially contributing to survival. 5,6 Such data on other clinical variables comprise the PSP Rating Scale (PSPRS), with 28 clinical items covering the common features of the disease. 7 It is widely used in clinical trials [8][9][10][11] and has recently been modified and validated for teleneurology use. 12 A large, longitudinal data set including PSPRS scores is needed to take best advantage of the potential value of the timing of downgaze loss and other variables in providing overall prognostic guidance. Therefore, for this analysis, we used a large, longitudinal PSPRS data set that has been previously used to estimate survival and acquisition of disease milestones but without specific reference to downgaze palsy. 13 The PSPRS rates downgaze palsy only visually and semiquantitatively, but we hypothesized that a longitudinal data set of sufficient size could provide the statistical power needed to relate the downgaze item score to the overall disease course. Specifically, we hypothesized that the rate of progression in downgaze palsy and the age at PSP symptom onset could predict survival after controlling for various individual PSPRS items alone and in combination. 7,13 Given the absence of a practical prognostic biomarker in PSP, this study could have significant implications for clinical practice and both observational and interventional research. [8][9][10][11]14

Methods and Patients
Each of the 28 items on the PSPRS is rated ordinally from 0 (unaffected) to 2 (six items) or 0 to 4 (22 items, including downgaze palsy in oculomotor assessment). The total score, with a maximum of 100, progresses by about 11 points per year. 7 This single-center cohort was collected by author L.I.G. from 1994 to 2020 and included 462 patients with probable PSP-Richardson syndrome according to the Movement Disorder Society (MDS)-PSP criteria (2017). 15 For the patients who enrolled before 2017, we used the National Institute of Neurological Disorders and Stroke and Society for Progressive Supranuclear Palsy criteria (1996) for PSP 16 and included them when they met the MDS-PSP criteria (2017) for probable PSP. 15 We confined our analysis to 414 patients with a score of at least 1 on the downgaze item, with upgaze palsy at least as severe. The 414 patients had a total of 1411 visits, with a median visit interval of 5 months and a median of three visits per patient. We used data on birth year, symptom onset month and year using a previously published operational definition of PSP onset, 7 death month and year, visit month and year, and individual scores for the 28 items. The downgaze item is scored as: 0, not slow or hypometric, 86% to 100% of normal amplitude; 1, slow or hypometric, 86% to 100% of normal amplitude; 2, 51% to 85% of normal amplitude; 3, 16% to 50% of normal amplitude; and 4, 15% of normal amplitude or worse. Gaze was tested as a verbal command to "look down," not as a command to seek a specific target, nor as a pursuit task. Some patients were able to generate saccades when providing their history, but not when gaze was being formally examined, became anxious, and exhibited "frozen" or "apraxic" gaze. In such cases, the examiner provided reassurance and moved his face away from the patient's axis of primary gaze. This allowed the patient to generate evaluable saccades.
From these raw data, we calculated each patient's PSP onset age and for each visit, the number of months since PSP onset, the total PSPRS score, 7 and the PSP stage. 13 Most important, for each patient, we calculated the rate of progression on the downgaze palsy item by dividing its score at the first visit when it exceeded 0 by the number of months from PSP onset to that visit. We calculated the rate of progression for each of the other 24 nonocular items by dividing its score at the visit where the downgaze score first exceeded 0 by the number of months from PSP onset to that visit.
For the disease progression pattern analysis, we binned the visit dates into quarters of the year and compared the total PSPRS scores and downgaze palsy scores across multiple years of serial visits.
We calculated the PSP stage, as previously described, 13 from the sum of the following four PSPRS items: dysphagia for solids by history, dysphagia for liquids by exam, gait by exam, and ability to return to the seat by exam. The worst possible total score was 16. We referred to this as the "four-item score." We then binned these as described into six groups, referred to as the "PSP stage." The binning scheme was based on points of a possible 16: stage 0, 0 points; stage 1, 1 to 4 points; stage 2, 5 to 8 points; stage 3, 9 to PSP onset date was, per Golbe and Ohman-Strickland, 7 defined as the month/year of onset of the first symptom that was both likely related to PSP and that subsequently progressed in concert with other PSP symptoms. For cases when the records, the patient, or the caregiver were only able to provide the year of onset, July was recorded as the onset month. When "early" or "late" in a year was provided, March or October, respectively, was recorded as the onset month.
Of the 414 patients, 309 were known to be deceased as of the date of censoring of the database, June 30, 2020. Brain autopsy was performed in 24 (8%) of these, all by the brain bank at Mayo Clinic Jacksonville.

Statistical Analysis
We summarized all variables in the study with descriptive statistics using mean AE standard deviations or median with interquartile range for continuous variables based on the data distribution and frequencies (percentages) for categorical variables. We used the independent-samples t test or Mann-Whitney U test for continuous variables and Pearson χ 2 or Fisher's exact test for categorical variables after evaluating the data distributions for normality. For overall survival, the univariate Cox proportional hazards model was used to estimate the hazard ratio (HR) and the corresponding 95% confidence interval (CI) for each potential risk factor. After fitting a Cox model for each risk parameter, the proportional hazards assumption was checked based on Schoenfeld residuals. To carry out multivariate Cox proportional hazards analyses, we first examined the multicollinearity of the established clinical importance for overall survival or potential risk variables with a P value less than 0.15 in the univariate Cox model by the correlation coefficient matrix and the variance inflation factor. Then, a series of multivariate models were used to identify independent prognostic risk factors for survival. Gonen and Heller's K assessment (ranging from 0 to 1) provided a measure of overall predictive accuracy for Cox proportional hazards models. To test the hypothesis that PSP progression was approximately linear during the entire disease course, we studied the pattern of changes in PSPRS scores and downgaze palsy scores over time in patients with survival durations of 1 year within the median survival time to represent typical PSP-Richardson syndrome patients with usual natural history of survival. A two-tailed P value of less than 0.05 was considered statistically significant.
All analyses were performed under the ethical standards of the institutional review board and the Declaration of Helsinki. Table 1 shows demographics for our 414 patients with probable PSP-Richardson syndrome with median progression rate of individual and complex items when downgaze palsy was first detected. For 309 patients, a month/year of death were known. The average disease  duration from onset to the date of appearance of downgaze palsy with a score of >0 was 3.3 years. The median overall survival duration of the patients who were known to have died was 6.58 years. The HR and 95% CI for the survival risk in each item based on the univariate Cox proportional hazards model are also shown in Table 1.

Results
Of the 24 autopsied brains, 22 had PSP pathology (usually with copathology), and two had non-PSP pathology (one had corticobasal degeneration with argyrophilic grain disease, and the other had frontotemporal lobar degeneration with transactive response DNA-binding protein-43 immunoreactive pathology (FTLD-TDP)).
For the four ocular motor symptoms/signs, multicollinearity was strong, with Spearman's ρ values ranging from 0.548 to 0.829, all highly significant with P values <0.001. We therefore focused solely on downgaze palsy given its specificity to PSP and the rationale for the current study. Table 2 shows the PSPRS items with significant results for the multivariate analysis. Here, downgaze palsy had an HR of 1.72. Other items with HR >1 and P < 0.05 were age at onset, dysphagia for liquids, sitting down, gait, sleep, and grasping/imitative/utilizing behavior. HR <1 was found for tremor, falls, disorientation, and emotional incontinence, although all but tremor were mild. Therefore, we created a second multivariate Cox survival model based on all axial symptoms and signs with known likely implications for mortality, including arising from a chair, gait, postural stability, sitting down, falls, dysphagia for liquids, and dysphagia for solids along with the rate of downgaze palsy and onset age. Significant results appear in Table 3. We again found that age of onset, the progression rate of downgaze palsy, dysphagia for liquids, and impaired sitting down were significant risk factors for shorter survival in PSP.
Surprisingly, the progression on the falls item was a mild protective factor for survival. We suspect that experiencing early falls (at the time when downgaze palsy was first noted) made the patient/caregiver more careful in preventing further falls and reducing the risk of death. However, given its uncertain role in PSP survival, we removed the factor of falls and remodeled the rest of the factors. We again found that the progression rate of the downgaze palsy, dysphagia to liquids, and age at onset along with the difficulty in arising from a chair were risk factors for shorter survival (data not shown).
Given the multicollinearity of the individual items with the complex items (total PSPRS score, the six domain scores, the four-item staging score, and the PSP stage), we created an additional Cox proportional hazards model using these variables along with symptom onset age. We found that the onset age and progression rate of the four-item staging score and the PSP stage were risk factors for shorter survival (Table 4).
We considered the 82 patients with survival duration within 1 year of the median as representative on that variable. They displayed an approximately linear increase in PSPRS scores (Fig. 1A) and downgaze palsy scores (Fig. 1B) across years 2 to 6 of the disease course.

Discussion
This study used a large, longitudinal, single-center PSPRS data set to confirm the hypothesis arising from a previous study, 6 using a different methodology, that the rate of downgaze palsy progression and PSP age at symptom onset are associated with survival duration in patients with PSP-Richardson syndrome. We defined the rate of downgaze palsy progression as the score on  that PSPRS item at the time it initially reached one or more divided by the time since initial PSP symptom onset. We used the progression rate of the other 24 nonocular PSPRS items as other input variables. We defined the progression rate of each of those as the item score at the visit where a downgaze impairment was first detected divided by time since PSP onset. This innovative approach, using downgaze palsy as a referent anchor, showed that shorter survival was associated with age at onset and the rate of progression of dysphagia for liquids. This result was consistent across two different multivariate Cox survival models. We also used PSPRS-derived stages and the six PSPRS subdomains ("complex" risk factors) as input variables with similar results. Our methodology relies on the observation that downgaze impairment is one of the features central to the pathobiology of PSP along with other axial/midline deficits. In fact, these features have been found to correlate with PSP survival. 6,[17][18][19][20][21][22] The observation that downgaze is controlled by fibers along the neural pathway where PSP pathology is thought to form and spread is further justification for our approach. [2][3][4] The consideration of progression rates rather than cross-sectional scores could allow us to assess the effect of a specific deficit on survival regardless of when the patient's eye movements were first evaluated. This accommodates the constraints of ordinary clinical practice in that formal vertical oculomotor evaluation may not take place until several years after symptom onset. We found a smoothly linear increase in the PSPRS score and downgaze palsy score in the first 2 to 6 years of the disease course in PSP-Richardson syndrome patients who had a typical survival duration of within 1 year of a median of 6.58 years in this cohort. This is similar to the mean survival duration in another large size study of autopsy-confirmed PSP-Richardson syndrome cases. 23 Our current study will have significant implications for disease prognostication, end-of-life planning, clinical trial enrollment, and outcome assessment for PSP-Richardson syndrome patients, the most frequent form of PSP. 23 Our multivariate Cox proportional hazards model found that the onset age and progression rates of downgaze palsy, dysphagia for liquids, gait, returning to the seat, sleep, and grasping behavior were each independently associated with shorter survival. Counterintuitively, we also found protective effects, albeit mild and with marginal statistical significance even with so large a sample size, for disorientation, emotional incontinence, and falls. The last is puzzling but may be explained by the possibility that the early appearance of falls could have prompted the patient or caregivers to institute fall precautions at a point in the disease when the patient was able to accept, practice, and habituate them. This may have reduced injuries and stress and may have extended survival.
To probe our data in a rational neuro-anatomical context, we repeated the analysis using only onset age, downgaze palsy, and the axial motor items arising from a chair, gait, postural stability, sitting down, falls, dysphagia for liquids, and dysphagia for solids. We again found that the progression of the downgaze palsy, age at onset, dysphagia for liquids, and difficulty in sitting down were significant independent risk factors for shorter survival. The falls item was again a mild protective factor.
We also found in multivariate models comprising progression rate of downgaze palsy that onset age and either the complex four-item staging score or the PSP stage were independently associated with survival. This supplements the analysis of Golbe and colleagues as validation for the PSP staging system they propose. 13 It also confirms the onset age as a survival risk factor. 6 Prior 12-month longitudinal data from the placebo group in a clinical drug trial showed a linear Golbe et al, 2020 13 Velocity of total PSPRS score from onset to first visit Jabbari et al, 2021 31 rs2242367 polymorphism at SLC2A13/LRRK2 locus progression in PSPRS scores during that limited timespan. 8 To extend this analysis to a longer time period, we used our longer term data set to tabulate the median and 95% CI total PSPRS scores and downgaze palsy scores through the seventh year of disease. We focused on the patients whose survival was within 1 year of the median to better understand the natural history of the disease without the influence of outliers.
We found a steady increase in the PSPRS scores and downgaze palsy scores from years 2 to 6 with a possible ceiling effect thereafter, when patients had reached a median score of about 70/100 and nearly 4/4, respectively, representing a very severe level of disability. We started at year 2 from the onset of the PSP symptoms because our data set started in most cases only after the patients had been referred to a tertiary center. The ceiling effect is probably explained by the fact that not all PSPRS items progress to a maximally advanced score in every patient and that some items such as tremor, dystonia, or aggressive behavior never occur at all in some patients. This is the first documentation and quantification of a long-term, continued increase in the PSPRS score and downgaze palsy score and their ceiling effects. This suggests that treatment trials using slowing of PSPRS progression as the primary outcome measure should complete their observations before year 7 for PSP-Richardson syndrome to avoid possible masking of disease-modifying effects due to this ceiling effect. Our observation of the linearity of PSPRS progression during a 5-year period (years 2 to 6 of symptoms) validates the assumption of the other part of the current study and other studies that used a linear model in the early or middle phase of PSP when the downgaze palsy was first detected. 5,6 It also provides justification for interventional trials to assume that the PSPRS progresses for each patient at a constant rate regardless of whether the patient enters the trial at year 2 or year 5 after symptom onset.
Survival in PSP-Richardson syndrome can be predicted to some extent by the PSPRS total score and its progression rate early in the disease course. 7,13 Table 5 summarizes previous studies of factors associated with shorter survival in PSP. 1,5,6,7,13,21,[25][26][27][28][29][30][31] The present study does the same using a different methodology and further confirms the predicting value of the progression rate of downgaze palsy in survival.
The major strengths of the study include its large size, which permitted a robust multivariate analysis; the experience and presumed internal consistency of a single expert examiner; the uniform method of ascertainment of the symptom onset date; the use of a pathologically relevant event, downgaze palsy, as a base for the analysis of the other input variables; the use of progression rate rather than cross-sectional severity of PSP features to reduce the limitations related to the timing of the clinical evaluation; and the consistent identification of independent risk factors across our multivariate models.
There are some limitations to this study. First, the PSPRS estimates downgaze palsy semiquantitatively by the examiner's visual inspection because automated oculomotor recordings were not available at the time for the routine care visits that contributed to our database. Second, medical comorbidities and causes of death were not recorded or considered in our analysis. Third, postmortem examination was obtained in only 24 patients (8%), but PSP pathology proved to be present in 22 (92%) of these, similar to the approximately 90% positive predictive value for PSP clinical diagnostic criteria. 32,33 The number of autopsies was insufficient for a subanalysis of pathologically confirmed patients. However, our intent was not to study the biology of PSP, but to model the clinical environment where living patients diagnosed by imperfect clinical criteria seek prognostic advice or consideration for an interventional study. Fourth, the use of only a single examiner may limit the generalizability of our data despite that examiner's experience and internal consistency in applying the PSPRS. Fifth, the onset month/year of PSP symptoms may not be precisely ascertainable in many cases.
In summary, our analysis of this large, longitudinal data set shows that faster rates of progression of downgaze palsy, dysphagia for liquids, and difficulty in returning to a chair and the stage of the disease along with older age at onset are associated with shorter survival in PSP-Richardson syndrome. This study also found a steady increase, approximating arithmetic linearity, in PSPRS scores over years 2 through 6 of the disease course. These results have actionable implications for PSP-Richardson syndrome with regard to clinical prognostication, patient-centered management, end-oflife planning, and interventional trial design and enrollment. Further studies of more quantitative and detailed, technology-based, longitudinal measurement of downgaze palsy as a predictor of survival in PSP-Richardson syndrome are under way.