Selected autonomic signs and symptoms as risk markers for phenoconversion and functional dependence in prodromal Parkinson’s disease

Abstract

Purpose

To determine whether dysautonomia can stratify individuals with other prodromal markers of Parkinson’s disease (PD) for risk of phenoconversion and functional decline, which may help identify subpopulations appropriate for experimental studies.

Methods

Data were obtained from Parkinson’s Progression Markers Initiative. Cohorts without PD but with at-risk features were included (hyposmia and/or rapid-eye-movement-sleep behavior disorder, LRRK2 gene mutation, GBA gene mutation). Dysautonomia measures included Scales-for-Outcomes-in-Parkinson’s-Disease Autonomic (SCOPA-AUT), seven SCOPA-AUT subscales, and cardiovascular dysfunction (supine hypertension, low pulse pressure, neurogenic orthostatic hypotension). Outcome measures were phenoconversion and Schwab-and-England Activities-of-Daily-Living (SE-ADL) ≤ 70, which indicates functional dependence. Cox proportional-hazards regression was used to evaluate survival to phenoconversion/SE-ADL ≤ 70 for each dysautonomia measure. If a significant association was identified, a likelihood-ratio test was employed to evaluate whether a significant interaction existed between the measure and cohort. If so, regression analysis was repeated stratified by cohort.

Results

Median follow-up was 30 months. On multivariable analysis, gastrointestinal and female sexual dysfunction subscales were associated with increased risk of phenoconversion, while the cardiovascular subscale and neurogenic orthostatic hypotension were associated with increased risk of SE-ADL ≤ 70; respective hazard ratios (95% confidence intervals) were 1.13 (1.01–1.27), 3.26 (1.39–7.61), 1.87 (1.16–2.99), 5.45 (1.40–21.25). Only the association between the cardiovascular subscale and SE-ADL ≤ 70 was modified by cohort.

Conclusions

Symptoms of gastrointestinal and female sexual dysfunction predict phenoconversion in individuals with other risk markers for PD, while signs and symptoms of cardiovascular dysfunction may be associated with functional decline.

Introduction

Years prior to Parkinson’s disease (PD) diagnosis based on motor signs, non-motor prodromal signs are often present including hyposmia, sleep abnormalities, and autonomic dysfunction, or dysautonomia. A growing research focus has been identifying individuals at risk for PD, who therefore may have prodromal PD (PPD), with maximal accuracy by examining relationships between PD risk markers, prodromal symptoms, and onset of clinical features that indicate presence of a parkinsonian disorder (i.e., phenoconversion). Accurate risk stratification will assist in identifying prodromal individuals for clinical trials, who have less neurodegeneration and may be more responsive to putative disease-modifying therapeutics than individuals with overt PD [1]. Research criteria for probable PPD incorporating independent associations between risk markers and PD incidence have been published and validated [2–8], but they have low sensitivity in the general older adult population [9]. Further studies on the predictive value of particular risk marker combinations are critical for optimizing accuracy of algorithms to detect those at highest risk for PD and related disorders (PDRD), which also comprises dementia with Lewy bodies (DLB) and multiple system atrophy (MSA).

Signs of dysautonomia are among the earliest features in the prodrome of PDRD [10]. Studies have already identified associations between signs of dysautonomia including constipation, urinary dysfunction, and erectile dysfunction and later PD diagnosis in large populations nonenriched for other risk markers [11–14]. Patients with pure autonomic failure (PAF), defined clinically by neurogenic orthostatic hypotension without evidence of an alternative etiology or significant central neurodegeneration, are also at risk for PDRD [15–17]. While there is literature suggesting that certain autonomic signs increase the risk of phenoconversion in cohorts with RBD [18], and that conversely presence of RBD increases risk of PDRD in individuals with PAF [15], further research is needed to explore whether subpopulations with specific autonomic signs in combination with other risk markers are at higher risk for later PDRD relative to other prodromal subpopulations.

The Parkinson’s Progression Markers Initiative (PPMI) is an observational study that initially evaluated individuals with early PD, but subsequently recruited individuals without PD along a spectrum of PD risk, ranging from asymptomatic gene mutation carriers to individuals with hyposmia and rapid eye movement sleep behavior disorder (RBD) with abnormal dopamine transporter imaging who are thought to be at the highest short-term risk of phenoconversion [19–21]. Here, we examine these at-risk populations for specific features of dysautonomia in association with other risk markers and time to two outcomes: a diagnosis of PDRD (i.e., phenoconversion) and progression to functional dependence as a clinically relevant outcome that may develop in prodromal individuals [22].

Methods

Study design

Data were obtained from PPMI (ppmi-info.org), an observational study that performs longitudinal clinical assessments and collects biomarkers with the goal of verifying PD progression markers. The study involves nearly 50 international sites and includes cohorts with prodromal signs including hyposmia and RBD as well as genetic markers including LRRK2 and GBA gene mutations. Further details of methodology are published [19]. Data for this study were downloaded on October 13, 2020.

Study population

Participants with hyposmia were recruited through mail and online. These individuals then completed the University of Pennsylvania Smell Identification Test (UPSIT); anyone scoring ≤ 10th percentile for age/gender were invited for screening. Participants with isolated RBD (without PD or other neurodegenerative Parkinsonian disorders) were recruited if they had a recent polysomnogram consistent with RBD [23]. Other inclusion criteria for individuals with hyposmia and/or RBD (H/RBD) included age ≥ 60 years old, while exclusion criteria included diagnosed PD or dementia. Additionally, H/RBD participants were selected so that approximately 80% had dopamine transporter binding < 80% expected for age on dopamine transporter singlephoton emission computerized tomography (DAT-SPECT) similar to the early PD cohort originally enrolled in PPMI. Participants were invited for LRRK2/GBA genotyping if they belonged to an ethnic/geographic population with high mutation prevalence (i.e., populations of Ashkenazi Jewish, Basque descent) and had at least one relative with PD, or if they had a first-degree relative with a mutation. Mutation screening included G2019S, R1441G for LRRK2 and N370S, L483P, L444P, IVS2 + 1, 84GG for GBA. Those with positive test results were eligible for inclusion. Other inclusion criteria included an age ≥ 45 years old, while exclusion criteria included diagnosed PD or dementia. Further details of inclusion/exclusion criteria have been published [19–21].

We included all otherwise-qualifying participants with a baseline Scales-for-Outcomes-in-Parkinson’s-Disease Autonomic (SCOPA-AUT) and/or orthostatic vital sign testing, as well as a Movement-Disorders-Society Unified-Parkinson’s-Disease-Rating-Scale Part-III (MDS-UPDRS-III) and Montreal Cognitive Assessment (MoCA). Additionally, all participants needed to have at least one follow-up visit with a primary diagnosis assessment and Schwab-and-England Activities-of-Daily-Living (SE-ADL). Since we were interested in evaluating for effect modification of individual risk markers in all Cox models associated with a significant hazard ratio (see Statistical Analysis below), 13 participants with a LRRK2/GBA double mutation identified on initial genotyping were excluded to evaluate outcomes associated with each gene separately. Five participants initially enrolled in the LRRK2/GBA cohorts were later determined to have PD at baseline and were also excluded. Lastly, most participants in all cohorts underwent hyposmia testing after enrollment as well as additional genetic testing to identify pathogenic variants, which included some combination of genome-wide association testing, whole-genome sequencing, and RNA sequencing, depending on the individual. A flowchart of participant inclusion/exclusion employed by our study as well as results of additional testing after enrollment is depicted in Fig. 1.

Fig. 1.

Inclusion and exclusion of participants. H/RBD hyposmia and/or REM sleep behavior disorder participants, LRRK2 LRRK2 mutation carrier participants, GBA GBA mutation carrier participants, H ± presence and absence of hyposmia, respectively, on post-enrollment testing, LRRK2 ± presence and absence of a LRRK2 mutation, respectively, on post-enrollment testing, GBA ± presence and absence of a GBA mutation, respectively, on post-enrollment testing

Assessments

Patient-Reported Autonomic Symptoms:

The SCOPA-AUT is a 23-item patient-reported questionnaire validated in PD [24]. Each item asks about frequency of an autonomic symptom in the past 1–6 months and is scored from 0–3 (i.e., never to often). We incorporated the total scale at baseline and created seven subscales based on system for the purpose of this study: gastrointestinal (SCOPA-GI; seven items), urinary (SCOPA-U; six items), cardiovascular (SCOPA-CV; three items), thermoregulatory (SCOPA-TR; four items), pupillomotor (SCOPA-PM; one item), female sexual dysfunction (SCOPA-FS; two items), male sexual dysfunction (SCOPA-MS; two items). Any participant with a catheter was excluded from analysis involving the SCOPA-AUT or SCOPA-U. Any participant who answered not applicable for SCOPA-FS or SCOPA-MS items was not included in analysis involving the SCOPA-AUT, SCOPA-FS, or SCOPA-MS. All scales were treated as continuous.

Vital Sign Testing:

Participants undergo orthostatic vital sign testing with systolic/diastolic blood pressure and heart rate measured supine and standing at baseline. These data were used to generate the variable neurogenic orthostatic hypotension (supine minus standing systolic blood pressure ≥ 20 mmHg and/or supine minus standing diastolic blood pressure ≥ 10 mmHg, with standing minus supine heart rate < 20 beats-per-minute), which has been associated with PD risk [25]. We also evaluated hypertension while lying (systolic blood pressure ≥ 140 mmHg and/or diastolic blood pressure ≥ 90 mmHg) separately as a risk marker since one study found that prevalence of orthostatic hypotension did not differ between patients with and without supine hypertension, despite its association with orthostatic hypotension in PD [26]. Lastly, although not a described feature of autonomic dysfunction, these data were also used to generate a variable for low pulse pressure (supine and/or standing systolic minus standing blood pressure ≤ 40 mmHg) as a potential risk marker since pulse pressure may be lower in patients with PD compared to the general population [27].

Outcome Measures:

Each outcome measure was assessed at each 6-month follow-up visit. The first was phenoconversion, or a PPMI research diagnosis of PDRD. At each site visit, the Prodromal Diagnostic Questionnaire and Diagnostic Features assessment as completed by the site investigator are utilized by PPMI as research-based criteria for phenoconversion. The most likely diagnosis is indicated in the Prodromal Diagnostic Questionnaire, while the Diagnostic Features assessment describes exam features that support the diagnosis. The investigator determines the diagnosis based on examination and unstructured interview of the participant, but without access to results of DAT-SPECT or other biomarker data. For our study, phenoconversion was defined as a primary research diagnosis of an α-synucleinopathy: PD, MSA, or DLB. Other neurodegenerative diagnoses including progressive supranuclear palsy (PSP) were identified but not included as a positive outcome. Furthermore, the PPMI Consensus Committee reviews any prodromal or at-risk cohort cases where the investigator has specified a diagnosis of a parkinsonian disorder, i.e., phenoconversion (as described at ppmi-info.org). Using available data on motor and non-motor measures, as well as DAT-SPECT results, the committee specifies if the research data support the presence of a neurodegenerative parkinsonian disorder. Thus, for each participant who meets the phenoconversion outcome as specified above by the investigator, there is a designation as to whether the consensus committee found research data supporting phenoconversion. Because a research diagnosis may be delayed and/or inaccurate in the research setting, we also selected functional dependence in daily activities as a clinically meaningful outcome. Functional dependence was identified using the SE-ADL, which assesses functional status associated with mobility and is scored from 0 to 100 (fully dependent to fully independent) [28]. A score ≤ 70 indicates dependence with activities of daily living; it was the cutoff used to generate a binary variable for functional dependence. Both outcomes were evaluated every 6 months up to 60.

Covariates:

Other variables that could be confounders for a relationship between dysautonomia and outcomes of interest included age, self-identified gender, MDS-UPDRS-III, and MoCA, all determined from the baseline visit. Levodopa-equivalent daily dosage (LEDD), which is the net dosage of dopaminergic medication converted to the equivalent levodopa dosage [29], was determined at the baseline and every 6-month follow-up visit; this was identified as a covariate since levodopa may reduce motor signs associated with a PD diagnosis but may also induce autonomic signs/symptoms. In order to account for other medications that may induce autonomic signs or symptoms and confound the associations under study, for each SCOPA-AUT subscale, two binary variables were generated for relevant medication use at baseline: one for whether a participant was on a medication indicated for a relevant autonomic symptom, and one for whether a participant was on a medication that can induce a relevant autonomic symptom. Similarly, for cardiovascular dysfunction variables, eight binary variables were generated from the baseline visit: four for whether a participant was on a medication indicated for tachycardia, bradycardia, hypertension, hypotension, and four for whether a participant was on a medication that can induce tachycardia, bradycardia, hypertension, or hypotension. UpToDate^® was used to determine medication indications and side effects [30]; a medication was defined as an inducer if product labeling indicated a > 10% risk of that side effect. Lastly, for each SCOPA-AUT scale and objective measure of cardiovascular dysfunction, a binary variable was generated for the presence of an alternative diagnosis associated with a relevant autonomic symptom or feature of dysautonomia in order to account for other conditions aside from PD that may confound the associations under study.

Statistical analysis

Baseline demographic/clinical data were compared between the three cohorts (H/RBD, LRRK2, GBA) using Kruskal–Wallis, Pearson’s chi-squared, or Fisher’s exact test, as appropriate. Associations between dysautonomia measures and baseline demographic/clinical measures were evaluated using linear/logistic regression analysis, as appropriate, with dysautonomia measures serving as independent variables.

Cox proportional-hazards regression analysis was used to evaluate time to phenoconversion and SE-ADL ≤ 70 as separate outcomes in the total population for each dysautonomia measure (there were 0 deaths prior to reaching the outcome measures in this population, which is a competing risk). A prior-observation-carried-forward method was employed to account for any datapoints for outcome missing prior to the last visit. In multivariable analysis, baseline age, gender, MoCA, relevant dysautonomia-indicated/inducing medication use, relevant alternative diagnosis were incorporated as covariates, as well as every-6-month LEDD as a time-varying covariate. If a significant association was identified between a dysautonomia measure and outcome, a likelihood-ratio test was employed to evaluate whether an interaction existed between the measure and cohort. If the p value associated with the likelihood-ratio test statistic was ≤ 0.10, regression analysis was repeated stratified by cohort. Since functional dependence as measured on the SE-ADL is not PD-specific, we also evaluated whether phenoconversion was associated with functional dependence in this population by performing Cox proportional-hazards regression analysis with time to SE-ADL ≤ 70 as the outcome and phenoconversion at every 6-month visit as a time-varying variable, adjusting for baseline age, gender, MoCA, and time-varying LEDD in the multivariable model. For all survival analyses, a post-estimation analysis of Schoenfeld residuals was performed to evaluate the proportional hazards assumption.

Three sensitivity analyses were performed to assess the robustness of the survival analysis results. In the first, binary variables were derived from the SCOPA-AUT total scale and subscales and regression analysis was repeated in the total population; a SCOPA-AUT score was considered positive if at least one subscore was positive, while a SCOPA-AUT subscore was considered positive if the response for at least one question in the subscale was ≥ 2 (i.e., symptom occurs regularly or often). In the second, since there were several participants missing a phenoconversion measure at the visit 6 months prior to the visit of the event, regression analysis for time to phenoconversion was repeated in the total population assuming the event occurred at the prior visit in those participants. In the third, PPMI Consensus Committee designations for phenoconversion were used instead of the original research diagnoses determined by site investigators. Another sub-analysis was performed to evaluate whether exclusion of participants from a given Cox model due to an absent dysautonomia variable at baseline may have contributed to selection bias by repeating regression analysis for both time to phenoconversion and functional dependence using absence/presence of the dysautonomia variable as the independent variable for each measure.

For all analyses, a p value associated with the test statistic of ≤ 0.05 was considered significant. Stata version 16.0 was used [31].

Results

Baseline demographic/clinical data

Table 1 describes baseline demographic/clinical data in the total population, as well as among the three cohorts. There were 443 total participants with at least one dysautonomia measure who met inclusion criteria, 440 who completed at least part of the SCOPA-AUT, 442 who had supine vital sign testing, and 441 who had both supine/standing vital sign testing. Compared to those in the LRRK2 and GBA cohorts, participants in the H/RBD cohort were older and less likely to be female, had a higher MDS-UPDRS-III and SCOPA-AUT, were more likely to have supine hypertension and neurogenic orthostatic hypotension, and were more likely to be on an autonomic symptom-inducing medication and cardiovascular-indicated medication, while H/RBD and GBA participants were more likely to have an alternative diagnosis associated with dysautonomia than LRRK2 participants. Table 2 lists common indicated and inducing medications for each dysautonomia variable as well as alternative diagnoses associated with autonomic signs present in the population.

Table 1.

Baseline demographic/clinical data

Variable	Total	H/RBD	LRRK2	GBA	p a
Number of participants	443b	62c	199d	182e	–
Demographic/clinical variables
Age median, years (IQR)	62.34 (57.33–67.50)	67.25 (65.33–72.50)	61.33 (56.00–66.17)	61.83 (57.33–66.75)	< 0.01f
Female, %	54.63	22.58	58.29	54.63	< 0.01g
Total MDS-UPDRS median (IQR)	7 (4–13)	10.5 (7–17)	6 (3–11)	7 (3–13)	< 0.01f
MDS-UPDRS-III median (IQR)	2 (0–4)	3 (1–6)	2 (0–4)	1 (0–3)	< 0.01f
MoCA (IQR)	27 (25–29)	27 (25–28)	27 (25–29)	27 (26–29)	0.52f
SCOPA-AUT scales
SCOPA-AUT median (IQR)	7 (4–11)	10 (6–16)	7 (4–11)	7 (4–11)	< 0.01f
SCOPA-GI median (IQR)	1 (0–2)	2.5 (1–5)	1 (0–2)	1 (0–2)	< 0.01f
SCOPA-U median (IQR)	3 (2–5)	4 (2–6)	3 (2–5)	3 (2–5)	0.10f
SCOPA-CV median (IQR)	0 (0–1)	0 (0–1)	0 (0–.5)	0 (0–0)	0.20f
SCOPA-TR median (IQR	1 (0–2)	1 (0–2)	1 (0–2)	1 (0–2)	0.60f
SCOPA-PM median (IQR)	0 (0–1)	0 (0–1)	0 (0–.5)	0 (0–1)	0.70f
SCOPA-FS median (IQR)	2 (0–3)	2 (1–4)	1 (0–2)	2 (0–3)	0.03f
SCOPA-ms median (IQR)	0 (0–2)	2 (0–4)	0 (0–1)	1 (0–2)	< 0.01f
Cardiovascular measures
Hypertension while lying, %	27.15	37.70	20.60	30.77	0.01g
Low pulse pressure, %	27.89	22.95	30.65	26.52	0.44g
Neurogenic orthostatic hypotension, %	8.16	24.59	5.03	6.08	< 0.01g
Medications/diagnoses associated with dysautonomia
Any dopaminergic medication, %	1.58	1.61	0.50	2.75	0.19h
Any symptom-indicated medication, %	36.12	41.94	31.66	39.01	0.19g
Any symptom-inducing medication, %	60.27	75.81	53.27	62.64	< 0.01g
Any cardiovascular-indicated medication, %	32.05	46.77	27.64	31.87	0.02g
Any cardiovascular-inducing medication, %	23.93	27.42	19.60	27.47	0.16g
Any alternative diagnosis, %	9.03	11.29	4.02	13.74	< 0.01h

IQR interquartile range

^aAll values associated with comparison between H/RBD, LRRK2, GBA

^bSCOPA-AUT includes 352, SCOPA-GI/CV/TR/PM includes 440, SCOPA-U includes 439, SCOPA-FS includes 173, SCOPA-MS includes 180, hypertension while lying includes 442, low pulse pressure/neurogenic orthostasis includes 441

^cSCOPA-AUT includes 50, SCOPA-FS includes 13, SCOPA-MS includes 37, hypertension while lying/low pulse pressure/neurogenic orthostasis includes 61

^dSCOPA-AUT includes 159, SCOPA-GI/U/CV/TR/PM includes 196, SCOPA-FS includes 84, SCOPA-MS includes 76

^eSCOPA-AUT includes 143, SCOPA-U includes 181, SCOPA-FS includes 76, SCOPA-MS includes 67, low pulse pressure/neurogenic orthostasis includes 181

^fKruskal–Wallis test

^gPearson’s chi-squared test

^hFisher’s exact test

Table 2.

Common examples of medications indicated to treat, medications that may induce, and diagnoses associated with autonomic symptoms or dysautonomia

Dysautonomia measure	Indicated medications	Inducing medications	Side effect description examplesa	Diagnoses
SCOPA-GI	Antihistamines, calcium carbonate, docusate, domperidone, ondanseron, polyethylene glycol, proton pump inhibitors, senna	Anastrazole, duloxetine, escitalopram, fluoxetine, lamotrigine, metformin, oxycodone, paroxetine, valproic acid, venlafaxine	Constipation, diarrhea, dyspepsia, nausea, reduced appetite, vomiting	Celiac disease, diabetes mellitus, diverticulitis, hypothyroidism, irritable bowel syndrome, lactose intolerance, ulcerative colitis
SCOPA-U	5-alpha reductase inhibitors, alpha antagonists, saw palmetto	Alprazolam	Difficulty in micturition	Diabetes mellitus, hypothyroidism, prostatic hyperplasia
SCOPA-CV	-	Alprazolam, atenolol, carvedilol, clonazepam, gabapentin, oxycodone, paroxetine, pramipexole, pregabalin, sertraline, tamsulosin, trazodone, valproic acid, valsartan, venlafaxine	Dizziness	Celiac disease, diabetes mellitus, hypothyroidism, vertigo
SCOPA-TR	Estradiol, estrogen	Anastrazole, bupropion, citalopram, nifedipine, paroxetine, sumatriptan, sildenafil, venlafaxine	Diaphoresis, hot flashes	Diabetes mellitus, hypothyroidism
SCOPA-PM	-	-	-	Cataracts, diabetes mellitus
SCOPA-FS	Estradiol, estrogen	Escitalopram, fluoxetine, paroxetine	Decreased libido, sexual disorder	Diabetes mellitus, hypothyroidism
SCOPA-MS	Phosphodiesterase inhibitors, testosterone	Escitalopram, fluoxetine, paroxetine	Decreased libido, sexual disorder	Diabetes mellitus, hypothyroidism
Hypertension while lying, low pulse pressure, neurogenic orthostatic hypotension	Angiotensin-converting enzyme inhibitors, angiotensin-II receptor antagonists, beta antagonists, calcium channel antagonists, diuretics	Anastrazole, atenolol, bupropion, liraglutide, metoprolol, pramipexole, quetiapine, ramipril, tamsulosin, testosterone	Tachycardia, bradycardia, hypertension, hypotension	Diabetes mellitus, hypertension, hypothyroidism

^aSide effect descriptions from UpToDate^® for identifying inducers of a particular dysautonomia variable; a medication was defined as an inducer if product labeling indicated a> 10% risk of a relevant autonomic symptom or sign

Table 3 describes cross-sectional associations between dysautonomia measures and baseline demographic/clinical measures in the total population. Higher SCOPA-AUT, SCOPA-GI, SCOPA-U, SCOPA-FS, SCOPA-MS scores, as well as supine hypertension and neurogenic orthostatic hypotension, were associated with higher age, while low pulse pressure was associated with lower age. Higher SCOPA-U and supine hypertension were associated with lower odds of female gender, while higher SCOPA-TR, SCOPA-PM were associated with higher odds of female gender. Higher SCOPA-AUT, SCOPA-GI, SCOPA-U, SCOPA-TR, SCOPA-PM, and supine hypertension were associated with higher MDS-UPDRS-III. Higher SCOPA-AUT, SCOPA-GI, SCOPA-U, SCOPA-PM, and supine hypertension were associated with lower MoCA, while low pulse pressure was associated with higher MoCA. Higher SCOPA-AUT, SCOPA-GI, SCOPA-U, and supine hypertension were associated with higher likelihood of being on a medication indicated for that measure, while higher SCOPA-AUT, SCOPA-GI, SCOPA-CV, SCOPA-FS were associated with higher likelihood of being on a medication that could induce symptoms of that measure. Only supine hypertension was associated with likelihood of having an alternative diagnosis aside from PD that could be associated with that dysautonomia measure.

Table 3.

Associations between baseline dysautonomia and demographic/clinical variables in total population

Dysautonomia measure	Number	Δ mean agea (95% CI)	p	Female ORb (95% CI)	p	Δ mean MDS-UPDRS-IIIa (95% CI)	p	Δ mean MoCA (95% CI)	p	Relevant dysautonomia-indicated medication use ORc,d (95% CI)	p	Relevant dysautonomia-inducing medication use ORe (95% CI)	p	Relevant alternative diagnosis ORf (95% CI)	p
SCOPA-AUT	352	0.21 (0.09–0.33)	<0.01	1.02 (0.99–1.06)	0.24	0.08 (0.02–0.13)	<0.01	–0.06 (–0.10–0.03)	<0.01	1.07 (1.03–1.10)	<0.01	1.12(1.07–1.17)	<0.01	0.98 (0.90–1.06)	0.58
SCOPA-GI	440	0.64 (0.31–0.97)	<0.01	1.01 (0.93–1.11)	0.78	0.32 (0.15–0.49)	<0.01	–0.22 (–0.33–0.11)	<0.01	1.24 (1.12–1.36)	<0.01	1.15 (1.04–1.26)	<0.01	0.94 (0.72–1.22)	0.63
SCOPA-U	439	0.54 (0.30–0.79)	<0.01	0.93 (0.86–0.99)	0.03	0.32 (0.19–0.44)	<0.01	–0.14 (–0.23–0.06)	<0.01	1.24 (1.11–1.38)	<0.01	1.15 (0.97–1.36)	0.10	1.09 (0.92–1.30)	0.33
SCOPA-CV	440	0.57 (−0.33–1.47)	0.21	1.14 (0.89–1.47)	0.28	0.62 (0.16–1.08)	<0.01	–0.14 (–0.44–0.16)	0.35	-	-	1.30 (1.01–1.66)	0.04	0.71 (0.27–1.90)	0.50
SCOPA-TR	440	–0.26 (–0.64–0.13)	0.19	1.23 (1.10–1.39)	<0.01	0.28 (0.08–0.47)	<0.01	–0.04 (–0.17–0.09)	0.52	1.04 (0.88–1.25)	0.64	0.99 (0.86–1.13)	0.81	0.83 (0.52–1.32)	0.42
SCOPA-PM	440	0.44 (–0.50–1.37)	0.36	1.62 (1.22–2.17)	<0.01	0.90 (0.43–1.37)	<0.01	–0.36 (–0.67–0.05)	0.02	-	-	-	-	0.73 (0.22–2.43)	0.61
SCOPA-FS	173	0.74 (0.13–1.31)	0.01	-	-	0.05 (–0.21–0.32)	0.69	–0.16 (–0.38–0.06)	0.15	1.00 (0.78–1.28)	0.99	1.29 (1.03–1.63)	0.03	0.87 (0.41–1.87)	0.73
SCOPA-MS	180	2.09 (1.45–2.73)	<0.01	-	-	0.16 (–0.17–0.48)	0.35	–0.06 (−0.26–0.14)	0.57	1.13 (0.86–1.47)	0.38	1.21 (0.97–1.50)	0.10	0.73 (0.32–1.70)	0.47
Hypertension while lying	442	3.91 (2.39–5.43)	<0.01	0.58 (0.38–0.89)	0.01	1.13 (0.33–1.92)	<0.01	–0.97 (–1.49–0.45)	<0.01	2.00(1.29–3.10)	<0.01	1.29 (0.80–2.09)	0.29	2.87 (1.25–6.58)	0.01
Low pulse pressure	441	–3.03 (–4.56–1.52)	<0.01	1.43 (0.94–2.19)	0.10	–0.13 (–0.93–0.67)	0.75	0.67 (0.39–0.85)	0.01	0.64 (0.40–1.02)	0.06	0.91 (0.55–1.48)	0.70	0.50 (0.17–1.50)	0.22
Neurogenic orthostatic hypotension	441	4.58 (2.08–7.09)	<0.01	0.82 (0.41–1.61)	0.56	0.39 (–0.91–1.70)	0.55	–0.61 (–1.47–0.25)	0.17	1.07 (0.52–2.21)	0.86	0.90 (0.40–2.03)	0.79	-	-

OR odds ratio, 95% CI 95% confidence interval

^aValues represent β coefficients for linear regression models

^bOR missing if cohort with dysautonomia variable only includes one gender

^Clncludes relevant symptom-indicated medication use for SCOPA-AUT total score and subscores and any cardiovascular-indicated medication use for cardiovascular dysfunction variables

^dOR missing if no patients were on a relevant medication

^eIncludes relevant symptom-inducing medication use for SCOPA-AUT total score and subscores and any cardiovascular-inducing medication use for cardiovascular dysfunction variables

^fOR missing if no patients with positive variable did not have a relevant alternative diagnosis

Association of dysautonomia with time to phenoconversion/functional dependence

Median follow-up was 30 months (range 6–60); Supplementary Fig. 1 depicts Kaplan–Meier estimates for censorship at last follow-up for each cohort. Forty participants reached phenoconversion, 13 reached SE-ADL ≤ 70, and nine reached both. Of the 40 participants who phenoconverted, the investigator research diagnosis at first visit of the outcome was PD in 36, DLB in three, and MSA in one. One GBA participant who did not reach phenoconversion or functional dependence reached a diagnosis of PSP.

Table 4 describes the results of Cox proportional-hazards regression analysis in the total population for time to phenoconversion and SE-ADL ≤ 70. On univariable analysis, SCOPA-AUT, SCOPA-GI, SCOPA-U, SCOPA-PM, SCOPA-FS were associated with increased phenoconversion risk, but only SCOPA-GI and SCOPA-FS were associated with increased risk on multivariable analysis; respective HRs (95% CI) were 1.13 (1.01–1.27), 3.26 (1.39–7.61). On univariable analysis, SCOPA-AUT, SCOPA-GI, SCOPA-CV, neurogenic orthostatic hypotension were associated with increased risk of SE-ADL ≤ 70, but only SCOPA-CV and neurogenic orthostatic hypotension were associated with increased risk on multivariable analysis; respective HRs (95% CI) were 1.87 (1.16–2.99), 5.45 (1.40–21.25). The assumption of proportional hazards was only violated for multivariable analysis of the association between SCOPA-AUT and time to SE-ADL ≤ 70 (p value associated with test statistic ≤ 0.05).

Table 4.

Associations between dysautonomia measures and phenoconversion/functional-dependence hazard in total population

Dysautonomia measure	Phenoconversion		Functional dependence
	Statistics	p	Statistics	p
SCOPA-AUT
Events/total months of follow-up	27/11466	–	5/11856	–
Univariable HR (95% CI)	1.07 (1.02–1.12)	< 0.01	1.12 (1.02–1.24)	0.03
Multivariable HRa (95% CI)	1.05 (0.99–1.11)	0.11	1.09c (0.94–1.28)	0.25
SCOPA-GI
Events/total months of follow-up	40/14214	–	13/14736	–
Univariable HR (95% CI)	1.17 (1.05–1.31)	< 0.01	1.35 (1.16–1.57)	< 0.01
Multivariable HRa (95% CI)	1.13 (1.01–1.27)	0.04	1.15 (0.98–1.36)	0.1
SCOPA-U
Events/total months of follow-up	40/14160	–	13/14682	–
Univariable HR (95% CI)	1.10 (1.00–1.21)	0.04	1.06 (0.90–1.26)	0.5
Multivariable HRa (95% CI)	1.04 (0.93–1.15)	0.51	0.97 (0.79–1.19)	0.77
SCOPA-CV
Events/total months of follow-up	40/14214	–	13/14736	–
Univariable HR (95% CI)	1.21 (0.87–1.69)	0.26	1.95 (1.30–2.92)	<.01
Multivariable HRa (95% CI)	1.02 (0.76–1.40)	0.83	1.87 (1.16–2.99)	0.01
SCOPA-TR
Events/total months of follow-up	40/14214	–	13/14736	–
Univariable HR (95% CI)	1.05 (0.90–1.23)	0.54	1.08 (0.83–1.40)	0.57
Multivariable HRa (95% CI)	1.02 (0.87–1.20)	0.8	1.02 (0.72–1.43)	0.93
SCOPA-PM
Events/total months of follow-up	40/14214	–	13/14736	–
Univariable HR (95% CI)	1.37 (1.00–1.88)	0.05	1.09 (0.57–2.11)	0.79
Multivariable HRa (95% CI)	1.28 (0.90–1.82)	0.17	0.74 (0.30–1.79)	0.5
SCOPA-FS
Events/total months of follow-up	6/5532	–	6/5532	–
Univariable HR (95% CI)	2.02 (1.29–3.16)	< 0.01	1.90 (0.91–3.97)	0.09
Multivariable HRa (95% CI)	3.26 (1.39–7.61)	< 0.01	1.61 (0.63–4.12)	0.32
SCOPA-MS
Events/total months of follow-up	21/5994	–	3/6324	–
Univariable HR (95% CI)	1.04 (0.81–1.33)	0.78	1.18 (0.67–2.09)	0.57
Multivariable HRa (95% CI)	1.04 (0.77–1.41)	0.81	1.17 (0.53–2.56)	0.7
Hypertension while lying
Events/total months of follow-up	40/14292	–	13/14814	–
Univariable HR (95% CI)	1.13 (0.58–2.23)	0.72	1.60 (0.52–4.90)	0.41
Multivariable HRb (95% CI)	0.56 (0.26–1.21)	0.14	0.75 (0.19–2.96)	0.68
Low pulse pressure
Events/total months of follow-up	40/14244	–	13/14766	–
Univariable HR (95% CI)	0.84 (0.41–1.72)	0.63	0.75 (0.21–2.74)	0.67
Multivariable HRb (95% CI)	0.91 (0.43–1.93)	0.8	1.18 (0.28–5.09)	0.82
Neurogenic orthostatic hypotension
Events/total months of follow-up	40/14244	–	13/14766	–
Univariable HR (95% CI)	1.79 (0.75–4.28)	0.19	6.00 (1.96–18.35)	< 0.01
Multivariable HRb (95% CI)	1.53 (0.60–3.93)	0.37	5.45 (1.40–21.25)	0.01

HR hazard ratio, 95% CI 95% confidence interval

^aAdjusted for baseline age, gender, MDS-UPDRS-III, MoCA, relevant symptom-indicated medication use, relevant symptom-inducing medication use, alternative diagnosis, time-varying LEDD

^bAdjusted for baseline age, gender, MDS-UPDRS-III, MoCA, cardiovascular-indicated medication use, cardiovascular-inducing medication use, alternative diagnosis, time-varying LEDD

^cAssociation violates proportional hazards assumption

Only the association between SCOPA-CV and time to SE-ADL ≤ 70 was modified by cohort, with a p value associated with the likelihood-ratio test statistic of 0.08. The associations between SCOPA-GI, SCOPA-FS and time to phenoconversion were not modified by a particular cohort (data not shown), while effect modification could not be evaluated for the association between neurogenic orthostatic hypotension and SE-ADL ≤ 70 since only H/RBD participants with a positive score reached the outcome. Table 5 describes the results of Cox proportional-hazards regression analysis stratified by cohort for the association between SCOPA-CV and time to SE-ADL ≤ 70. The association was driven by the H/RBD cohort (HR 2.70, 95% CI 1.16–6.29). Given a low event rate, univariable and multivariable HRs could not be calculated for the GBA cohort.

Table 5.

Associations between SCOPA-CV and functional-dependence hazard stratified by cohort

	H/RBD		LRRK2		GBA
	Statistics	p	Statistics	p	Statisticsa	p
Events/total months of follow-up univariable HR (95% CI)	7/3132	–	4/6954	–	2/4650	–
Multivariable HRb (95% CI)	2.18 (1.35–3.52)	< 0.01	1.21 (0.40–3.65)	0.73	–	–
	2.70 (1.16–6.29)	0.02	1.38 (0.48–3.99)	0.56	–	–

HR hazard ratio, 95% CI 95% confidence interval

^aHR missing due to low event rate and inability to calculate 95% CI

^bAdjusted for baseline age, gender, MDS-UPDRS-III, MoCA, relevant symptom-indicated diagnosis, time-varying LEDD

In Cox proportional-hazards regression analysis of the association between phenoconversion as a time-varying variable and time to SE-ADL ≤ 70 as the outcome, HR (95% CI) was 13.96 (4.34–44.91) in the univariable model and 9.27 (2.54–33.82) in the multivariable model.

Supplementary Table 1 describes regression analysis results in the total population for time to phenoconversion or SE-ADL ≤ 70 using binary SCOPA-AUT variables. For phenoconversion, the multivariable association for SCOPA-GI did not remain significant, and HR could not be calculated for SCOPA-FS since all participants with a positive score and no participants with a negative score reached the outcome. For functional dependence, the multivariable association with SCOPA-CV remained significant. Supplementary Table 2 describes regression analysis results in the total population for time to phenoconversion assuming the event occurred at the visit 6 months prior to the recorded event for any missing datapoints. There were three participants who had missing datapoints, and results were similar to those of the primary analysis. Supplementary Table 3 describes regression analysis results in the total population for time to phenoconversion using PPMI Consensus Committee designation instead of the original site investigator research diagnoses. A total of 24 instead of 40 participants reached PDRD, and HRs for both SCOPA-GI and SCOPA-FS remained significant, in addition to neurogenic orthostatic hypotension. Supplementary Table 4 describes results of regression analysis using absence/presence of the dysautonomia variable in primary analysis as the independent variable for each measure. Male participants excluded in primary analysis due to an absent SCOPA-MS score were more likely to reach SE-ADL ≤ 70 than those who were included on multivariable analysis (HR 8.54, 95% CI 1.27–57.27). Otherwise, participants with absent variables were not more or less likely to reach the outcomes of interest compared to participants without absent variables on multivariable analysis.

Discussion

In this study, we have described cross-sectional associations between dysautonomia and clinical features in a population with different risk markers for PD, as well as longitudinal associations between those dysautonomia measures and risk of phenoconversion and functional dependence. Furthermore, we have evaluated whether particular phenotypic or genetic marker-defined cohorts drive such longitudinal associations to identify whether certain subpopulations with specific combinations of dysautonomia and risk markers are at particularly elevated risk of PD-associated outcomes. The results of the longitudinal analysis remained significant after adjustment for medications and alternative diagnoses associated with autonomic signs and symptoms, which are potential confounders in the associations under study.

At baseline, H/RBD participants were older and had greater motor abnormalities and dysautonomia than gene mutation carriers. This likely reflects the different inclusion criteria utilized by PPMI for the different cohorts, including different age cutoffs as well as the fact that the H/RBD cohort was DAT-SPECT deficit-enriched [19]. Furthermore, both hyposmia and RBD are not simply risk markers but also prodromal signs often associated with underlying neurodegenerative disease. Therefore, participants would be expected to have higher rates of other prodromal features including dysautonomia and mild parkinsonian signs compared to participants with gene mutations.

In general, higher dysautonomia severity was correlated with older age. This is consistent with a higher prevalence of primary and medication-induced dysautonomia with older age in the general population but may also be due to the fact that prevalence of PD and therefore PPD, which are associated with dysautonomia, rises later in life [2]. The inverse association with pulse pressure likely reflects the fact that pulse pressure increases with older age due to atherosclerosis [32]. More severe dysautonomia was also associated with higher motor scores on the MDS-UPDRS-III, which is consistent with studies that have described associations between SCOPA-AUT subscores and neurodegeneration as measured on DAT-SPECT [33–35]. Whether these associations reflect a direct role of nigral dopaminergic neurons in autonomic function or more severe PD-associated pathology in other nervous system regions in individuals with greater nigral degeneration is unclear. Similarly, more severe autonomic features were in general associated with lower scores on MoCA, which is consistent with literature describing associations between dysautonomia and cognitive impairment in PD populations [36, 37]. Likelihood of female gender was variably associated with dysautonomia measures, which may reflect sex-specific rates of particular autonomic symptoms generally as opposed to PD-specific phenomena. For example, the inverse association between SCOPA-U and female gender may be explained by higher rates of certain urinary symptoms associated with benign prostatic hyperplasia in men.

The most notable observations from our study relate to the survival analysis. In the total population, SCOPA-GI and SCOPA-FS were associated with phenoconversion hazard at follow-up on univariable and multivariable analysis. While constipation is known to be a risk marker for PD based on nonenriched population studies [12–14], these results corroborate those from a recent prospective study suggesting that constipation increases risk of phenoconversion in individuals who have RBD [18]. The fact that the association was not reproduced when treating SCOPA-GI as a binary variable suggests that assessing symptom frequency may be a more sensitive tool for determining risk than simply identifying symptom presence/absence. While increased prevalence of female sexual dysfunction has been reported in PD [38, 39], it has not been established as a risk marker for PD incidence, and therefore has not been incorporated into the MDS research criteria as has male sexual dysfunction [3]. However, our results suggest that self-reported symptoms including vaginal dryness and difficulty reaching orgasm may be independently associated with increased PD risk. We did not find a significant interaction between SCOPA-FS and cohort, suggesting the relationship was not modified by any particular risk marker.

While gastrointestinal and female sexual dysfunction were associated with hazard of phenoconversion, SCOPA-CV and neurogenic orthostatic hypotension were associated with hazard of functional dependence on univariable and multivariable analysis. While both symptomatic and neurogenic orthostatic hypotension have been incorporated into the MDS research criteria [2, 3], these data suggest that not only do these features predict phenoconversion, but may also be associated with disease severity and risk of functional impairment. They also suggest that certain items of the SCOPA-AUT may represent a useful metric for predicting disease course in PPD. Individuals with PD may have worsening impairments in activities of daily living 5–7 years before diagnosis, and these are correlated with worsening motor signs [22]. Therefore, simple clinical assessments that can predict worsening functional status may be just as valid prognostic tools as markers that predict overt phenoconversion, and may be able to identify at-risk populations sooner. Regardless, it should be noted that the SE-ADL does not distinguish between causes of impaired mobility so it cannot be determined definitively whether functional dependence was associated with parkinsonism in any participant. However, regression analysis suggested a strong association between phenoconversion and hazard of functional dependence in this cohort, since phenoconversion over time was associated with a > 9 times higher risk of reaching functional dependence in the multivariable Cox model. We also attempted to account for the possibility that individuals who reached functional impairment were simply more intolerant of dopaminergic medication due to autonomic side effects and therefore had inadequately treated parkinsonism by adjusting for LEDD longitudinally.

Based on stratified analysis, the association between SCOPA-CV and functional dependence was driven by the H/RBD cohort. However, given the fact that this cohort was DAT-SPECT deficit-enriched, it is unclear whether clinical markers or underlying nigral degeneration modified the association. While the significant association for neurogenic orthostatic hypotension corroborates the association for symptomatic orthostasis as identified by the SCOPA-CV, effect modification could not be identified since only participants in the H/RBD cohort with the autonomic variable reached the outcome. Regardless, these data are consistent with results of a prospective study in a cohort with overt PD suggesting that individuals with RBD and orthostatic hypotension (as well as mild cognitive impairment) at baseline have a more malignant phenotype [40]. They may also be consistent with literature suggesting that RBD significantly increases the risk of neurodegenerative parkinsonian disorders in patients with PAF [15], although it should be noted that a diagnosis of PAF cannot be definitively determined in this population since, although we adjusted for medication use and alternative etiologies of cardiovascular dysfunction, advanced autonomic testing such as tilt table and quantitative sudomotor axon testing were not performed in PPMI. Regardless, given the known association between PAF and development of other synuclein-mediated conditions [15], it is possible that associations seen in the H/RBD cohort were driven by participants with undiagnosed PAF. Lastly, while prior studies have identified prodromal features in PAF patients that may help predict conversion to PD/DLB vs. MSA [15–17], only 1 participant reached a diagnosis of MSA in our cohort and so similar associations could not be evaluated in participants with neurogenic orthostatic hypotension.

There are several key limitations in our study. Given that participants were recruited at PPMI-associated sites via convenience sampling and the different inclusion criteria for different cohorts [19], our study may have suffered from selection bias. Further selection bias may have been introduced for analyses involving SCOPA-AUT, SCOPA-U, SCOPA-FS, and SCOPA-MS, since individuals with catheters or who replied not applicable on sexual dysfunction items were not included. Indeed, in our sub-analysis treating absence/presence of a dysautonomia measure as an independent variable for each measure, individuals who had absent SCOPA-MS scores had a higher risk of reaching functional dependence than those who did not. One explanation for this result is that prodromal individuals with sexual dysfunction experience faster disease progression but are also more likely to respond not applicable on the SCOPA-AUT due to sexual inactivity. In this case, not applicable responses for sexual dysfunction items may lead to underestimation of true associations, which should be kept in mind in any future studies evaluating relationships between self-reported sexual dysfunction and PD-related outcomes. Longitudinal associations may also have been missed given limited follow-up for many participants as well as a limited number of participants who reached the outcomes of interest; larger population studies have already identified associations between dysautonomia measures such as urinary and cardiovascular dysfunction and later phenoconversion [14, 25, 41], which was not observed in our primary analysis. While the SE-ADL was designed to evaluate functional impairment associated with PD, we cannot be sure that functional dependence in these participants was a result of parkinsonism versus another etiology, and therefore it is uncertain whether the observed associations between dysautonomia and SE-ADL ≤ 70 are PD-specific. However, as mentioned previously, the strong association between phenoconversion and functional dependence in regression analysis suggests that functional dependence was related to development of PDRD in this population. We did not incorporate statistical methodology to account for multiple comparisons such as a Bonferroni correction given the exploratory nature of this study and the concern for committing a type II error. However, since multiple associations were evaluated, we cannot entirely exclude the possibility of spurious results. Lastly, given the large number of covariates incorporated into each model and the limited number of outcomes, overfitting is a potential limitation of this study and identified associations may not be replicable in alternative datasets; future studies that perform detailed autonomic assessments in prodromal populations will be needed to validate our results. While the Parkinson At Risk Syndrome (PARS) study collected data on constipation, lightheadedness, and urinary symptoms, which did not predict phenoconversion, the sample size was small and the relationship with functional dependence was not evaluated [42].

In conclusion, our data suggest that assessment of gastrointestinal symptoms on the SCOPA-AUT may be a valuable tool for further risk stratification for phenoconversion in individuals with other PD risk markers, and that female sexual dysfunction as measured by the SCOPA-AUT may be an important variable for inclusion in PPD diagnostic criteria. Furthermore, the SCOPA-AUT as well as orthostatic testing may be valuable assessments for identifying subpopulations at particularly high risk for disease progression and functional impairment. Lastly, the presence of specific risk marker combinations, particularly cardiovascular symptoms along with other prodromal signs, may be associated with increased risk of functional impairment. Future prospective studies with larger sample sizes will be valuable to validate our findings in order to accurately identify specific subpopulations at higher risk for PD-related outcomes who may be recruited into experimental studies.