Introduction

As the average age of the global population increases, neurodegenerative diseases, like Parkinson’s disease (PD), are a rising cause of concern because of the consequential global economic and societal burdens. With a drastically increased prevalence in the past two decades, PD is one of the fastest-growing neurological diseases worldwide, affecting ~1–3% of the population over 60 years old1,2.

PD research has made substantial progress in deciphering what factors influence the disease’s risk, progression, and severity. Genetics, environmental factors, and clinical history all contribute to the etiology of PD. Therefore, each person with PD may be the result of a unique combination of a different set of factors, hence the diverse spectrum of symptoms observed3. This might be one of the reasons why treatments have not been successful across the entire PD population.

To make this even more complicated, there is one important factor in PD etiology, biological sex, which has been understudied4. Estrogen is believed to be a protective factor against PD and perhaps one of the reasons why men are about 1.5 times more likely to have PD than women5,6. Aside from risk, PD varies in presentation, severity, and treatment success between sexes7,8. For example, women with PD have increased mortality compared to men with PD9. Furthermore, one study found that men and women with PD have different rates of progression and severity when adjusted for demographic variables10. Regarding PD severity in motor symptoms, two studies showed that there was poorer postural stability in women and greater rigidity in men11,12. For the severity of non-motor symptoms, women reported a higher severity of fatigue and mood-related symptoms13,14. A few studies also relayed that women with PD tended to have greater severity of depression and impairment in activities of daily living11,12. Thus, while previous studies have described potential sex differences in the epidemiology and clinical presentation of PD between men and women, the role that sex-specific factors, and particularly Women-specific health factors (WSHFs), may play in the differences observed is still poorly understood. These WSHFs encompass the sexual and reproductive health experiences that revolve around changes in hormonal and immunological changes (i.e., menses, hormone-related medications, family planning, surgeries, and menopause) and have been overlooked in both research and clinical discussions15,16.

In fact, only a few studies have examined the association between the disease and certain individual WSHFs, such as menses, contraceptives, surgeries, hormonal disorders, pregnancies, hormone-replacement therapy, and menopause. Moreover, the results have been highly contradictory5,17,18,19, in part due to the lack of attention toward these sex-specific factors, as well as the neglect of gathering the necessary data.

Beyond the lack of representation of women’s health in research, WSHFs are often ignored or excluded in clinical trials in and outside the PD field. One study found that ~42% of clinical trial protocols required contraception or sterility for women without providing any explanation, making women have a lower chance of being included in a clinical trial and resulting in a decreased representation of WSHFs in clinical research. Without incorporating these WSHFs, the scientific community ignores the basic biology of women and continues to inaccurately represent the clinical manifestations and treatment outcomes for 50% of the population20.

This exploratory study seeks to shed light upon the role WSHFs play in the severity of PD in women by deploying a national questionnaire and analyzing the responses together with the participants’ genetic and clinical histories. Our study not only includes WSHFs as equally weighted variables amongst known variables in PD severity but also allows for the data to be collected in a comprehensive way through the questionnaire.

We hypothesized that certain WSHFs would be associated with greater PD severity even after adjusting for non-specific-sex factors. This study sets the stage for acknowledging sex-specific experiences in an age-related neurological disorder.

Results

PD GENEration population

The questionnaire was sent to 966 women with PD enrolled with the Parkinson’s Foundation PD GENEration. Of the returned surveys, 304 women gave complete responses, providing a response rate of 31.5%. Overall, respondents averaged 64.7 years in age (±9.1), with a mean age at diagnosis of 58.5 years (±10.4). Women completing the survey had an average of 6.2 years of disease duration (±5.3) since they were diagnosed with PD (Table 1). Since PD GENEration tests participants for seven known PD genes, we are able to see the genetic status of each respondent. Within our cohort, we have women with GBA (10%), LRRK2 (8%), PINK1 (1%), PRKN (4%), and VPS35 (1.3%) mutations21,22. The entirety of demographic information collected for every PD GENEration respondent by severity group can be found in Table 1.

Table 1 Characteristics of the study sample stratified by UPDRS parts I-IV (A-D) severity groups.
Full size table

As our previous work stated, the women’s questionnaire is composed of questions that are specific to a woman’s reproductive lifecycle. Participants in this cohort have a mean MoCA score of 27.4 (±2.3) (data not shown), indicating the women who completed this questionnaire had high cognitive aptitude and were able to complete the questionnaire without issues with impairment. Most respondents experienced their onset of PD during menopause (i.e., at least 1 year after their last menstrual cycle) (73.7%), compared to those that either experienced PD symptoms during perimenopause (8.8%) or during regular menses (17.5%). When asked about pregnancy-related experiences, the majority of respondents have had children (74.3%), although we noticed a higher incidence of infertility experiences (32.2%) or perinatal complications (31.2%) compared to non-PD women. It is important to note that within the United States, the national prevalence rate of history of primary and secondary infertility among women ranges from 15.5 to 19%23,24. A summary of all responses from the women’s questionnaire can be found in Table 2.

Table 2 Women’s health questionnaire responses.
Full size table

Univariate modeling

Variables were extracted from the questionnaire (WSHFs) and PD GENEration database. Thus, we were able to include other non-sex-specific factors for PD and comorbidity history in our modeling. From our univariate models, we found that certain factors, such as PD medications, depression, loss of smell, PTSD, B12 deficiency, peripheral neuropathy, and flair-up migraines, were associated with a moderate/severe UPDRS score within at least one of the UPDRS Parts in women. Interestingly, for women-specific health factors, we saw that medications feeling less effective during menses, tubal ligation, mastectomy, total hysterectomy, vaginal birth, gestational diabetes, peri- and postnatal depression, and perinatal bacterial infection were associated with a moderate/severe PD phenotype. A full summary of the univariate logistic regression with the sample size, OR (95% CI), and p values can be seen in Supplementary Table 1.

Multivariable modeling

To further assess the role of WSHFs and PD severity, we constructed multivariable logistics regression models by using variables that were significant in the univariate models and adjusting for age, disease duration, and medication. From Part I, having depression as a current diagnosis was associated with a moderate/severe PD severity phenotype (OR = 9.22 (3.31, 25.71), p < 0.001). In terms of pregnancy-related experiences, we found that having a natural (vaginal) birth was significantly associated with the moderate/severe cohort (OR = 4.48 (1.10, 18.20), p = 0.036), and postpartum depression demonstrated an association that did not reach conventional statistical significance (OR = 4.17 (0.91, 19.20), p = 0.067). Through our modeling, we noticed that “other comorbidities” was significant as well (OR = 3.91 (1.51, 10.14), p = 0.005). This was gathered as part of PD GENEration’s clinical history, where participants can include specific disorders they were diagnosed with; these comorbidities are listed in Supplementary Table 2. For Part II, we saw that years since diagnosis was significant (OR = 1.17 (1.09, 1.25), p < 0.001), which is expected as Part II assesses motor experiences in daily living. Depression and perinatal depression were significantly associated with the moderate/severe PD phenotype (OR = 2.33 (1.13, 4.80), p = 0.022 and OR = 6.72 (1.35, 33.43), p = 0.020, respectively). In Part III, the motor examination, having a LRRK2 mutation (OR = 6.74 (1.62, 28.00), p = 0.009), total hysterectomy (OR = 5.46 (1.74, 17.18), p = 0.004), and being diagnosed with B12 deficiency (OR = 6.22 (1.69, 22.95), p = 0.006) were significantly associated with higher PD severity. For Part IV, only years since diagnosis was significant (OR = 1.17 (1.06, 1.29), p = 0.002). Overall, the concordance statistic (c-statistic) for these four regression models demonstrated high accuracy for discerning PD severity for women within this cohort (Part I: 0.874, Part II: 0.805, Part III 0.791, and Part IV: 0.782). The univariate variables in the multivariable regressions with the c-statistic, OR (95% CI), and p values are summarized in Table 3.

Table 3 Summary of Multivariable Logistic Regression (MLR) Results for UPDRS parts I-IV (A-D).
Full size table

Discussion

This study is the first of its nature to look at WSHFs concerning PD severity in women within the U.S. through the deployment of a women-specific health questionnaire. We leveraged PD GENEration’s unique dataset of clinical and genetic health data from PD participants, coupled with the questionnaire responses, to develop multivariable logistic models to assess if WSHFs contribute to PD severity after adjusting for other non-sex-specific factors. Using Goetz et al.’s thresholds, we created mild vs. moderate/severe groups for PD phenotype for each subpart of the UPDRS assessment. Our analysis found that WSHFs, such as perinatal depression, delivery of children through vaginal birth, and history of a total hysterectomy, were associated with a severe PD phenotype, suggesting that WSHFs are important variables to be considered for PD severity in women. Aside from WSHFs, non-sex-specific factors, like major depression disorder, LRRK2 mutation carrier status, and B12 deficiency, were seen amongst women with a more severe phenotype compared to those with a mild PD phenotype. This study sets the stage for acknowledging the role of WSHFs in the severity of PD. While there have been studies looking at PD risk and some WSHFs, they overlook the growing population of women who have already been diagnosed with PD and thus have a declining health-related quality of life5,17,18,19,25,26,27,28,29,30,31,32,33.

Through our multivariable logistic regression models, we found that certain non-sex-specific and sex-specific factors were associated with a moderate/severe PD phenotype within each subpart of the UPDRS assessment. One such non-sex-specific factor was depression, for which there has been a great emphasis on the associations between it and PD. Depression (major depressive disorder) appeared in Parts I and II. While our study cannot definitively say if it was a pre-motor symptom of PD or if it was due to a more severe PD phenotype, depression appeared as a comorbidity of PD. Literature has shown that depression has been a prodromal symptom seen more commonly in women than in men, and as a result, women are more often misdiagnosed compared to men15,34,35. Moreover, in our study, perinatal depression was associated with a moderate/severe UPDRS Part II phenotype. Due to PD being more prevalent among older women, pregnancy-related factors are often neglected in PD studies, despite pregnancy having short-term and long-term effects on maternal health36,37. The few studies that have looked at pregnancy and PD have focused on women with young-onset PD, who usually present with a different profile38,39. One of the long-term effects of pregnancy is that it has been shown to cause neurological changes in women for them to transition to a more maternal mindset40,41,42,43. Thus, one might hypothesize that complications that can arise during this phase, such as perinatal depression, could alter the physiological adaptions required for the transition to motherhood, resulting in a woman being predisposed to having PD and, even more so, a more severe PD phenotype once diagnosed. More studies are needed to parse out the exact mechanism perinatal depression has in PD risk, progression, and severity. Furthermore, within this cohort, having a natural delivery of a child was significantly associated with having a moderate/severe phenotype of PD once diagnosed. In 2020, about 69% of American women experienced a natural delivery; however, the long-term implications of natural and cesarean deliveries remain unclear for maternal health and aging44,45. Additionally, due to a low percentage of women having children while diagnosed with PD, little is known about if cesarean or vaginal births play a role in the risk, progression, and severity of PD, but there is one study that looked at if having PD while pregnant resulted in women going with a particular delivery method. After surveying clinicians, it was observed that being diagnosed with PD would not impact a woman’s ability to have a vaginal delivery, and cesarean sections should be reserved as an alternative method of delivery, as is standard in care46. Other studies align with this notion and found little evidence for women with PD to have cesarean over vaginal deliveries, regardless of if a woman was currently on PD treatment and medication or not47,48. Overall, our study demonstrating that having a vaginal delivery was associated with a more severe PD phenotype highlights how pregnancy and postpartum choices can potentially impact women’s PD risk and severity when they are diagnosed years after giving birth. Thus, subsequent studies must further examine if and how the method of child delivery, particularly when the women are not diagnosed or affected by PD, impacts PD risk, progression, and severity.

This study also found that carrying a LRRK2 mutation was associated with a moderate/severe phenotype in Part III of UPDRS. LRRK2 (leucine-rich repeat kinase 2) gene encodes for a ROCO family protein and is a major contributor to PD risk, as mutations in LRRK2 have been linked to dopaminergic nerve cell death and impaired dopamine neurotransmission49. LRRK2 mutations can cause autosomal dominant PD, which has been present in up to 40% of familial PD cases in certain ethnic groups50. Differences in PD risk and presentation between men and women with LRRK2 p.G2019S mutations have been well characterized in the field51,52,53. While a LRRK2 mutation carrier status is a known risk factor for PD for both men and women, overall, women have a relatively higher incidence of having a LRRK2 p.G2019S mutation compared to men. Studies also observed differences in presentation and medication dosages between men and women with LRRK2 mutation status51,54. Our study showed that the LRRK2 mutation was associated with a more severe PD phenotype in our women cohort, opposite to previous reports combining both males and females55,56. Our results should be considered with caution as our sample size was small (n = 11), and therefore further studies are needed to determine if genetic mutations in women cause a more severe motor severity. Interestingly, only one study has briefly mentioned no differences in severity between women with positive LRRK2 carrier status and those without, however, the majority of studies do not present sex-stratified analyses51.

B12 deficiency is another non-sex-specific factor that was significantly associated with a higher UPDRS Part III subscore. Overall, B12 has been characterized as an important vitamin in maintaining PD nutrition. Studies have shown that low B12 is linked to cognitive impairment, higher Hoehn and Yahr scores, and dementia-related symptoms57,58. B12 deficiency is commonly seen in pregnant women, vegetarians, the elderly, and those who have neurodegenerative disorders59,60. The DATATOP study, a large two-year study of patients with early-onset PD, showed that a more rapid progression of PD symptoms occurred in those with lower B12 levels compared to those with higher levels, suggesting that increasing B12 levels may slow PD progression61,62. Therefore, it is reasonable to observe that those with a moderate/severe phenotype in our study had a B12 deficiency compared to those with mild PD severity. However, this factor needs to be fully explored in future studies to determine if B12 is a validated risk factor or modifies PD severity, as low B12 levels may be due to PD itself. While this association appeared in Part III (motor examination), interestingly, there was also a trend of significance in Part II (motor experiences of daily living) of this subcohort (OR = 2.27 (1.13, 4.80) p = 0.075). Since literature has suggested that B12 deficiency has been associated with peripheral neuropathy63,64, this observation of statistical significance in our cohort suggests a possible connection between B12 deficiency and the trend of peripheral neuropathy in women with a more severe PD phenotype.

In our study, we defined total hysterectomies as the removal of the entire uterus, cervix, and ovaries/fallopian tubes. Although the removal of the ovaries and fallopian tubes is not included in the definition of total hysterectomies, they are often included in the procedure based on the patient’s medical history. Total hysterectomies were significantly associated with a more severe PD phenotype in the PD GENEration questionnaire respondents. Hysterectomies are one of the most commonly performed procedures done in women (second after cesarean sections), and they treat gynecologic malignancies and benign gynecologic diseases65,66. About 1 in 3 women in the United States have a hysterectomy by the age of 6067, and one study approximates that about 600,000 of these procedures are performed annually in the United States68. Removal of the ovaries leads to patients taking exogenous hormones, which have been studied in relation to PD risk. However, across these studies, there is no conclusive answer to if exogenous hormones contribute to PD risk. One study found that hysterectomies, with or without unilateral oophorectomy (removal of one ovary), were significantly associated with PD risk25. Other studies found similar results with women being at a higher risk for PD if they either had unilateral or bilateral oophorectomies before experiencing menopause or if they used estrogen alone after a hysterectomy18,19,30. On the other hand, one study found that anemia, a condition that is more common in women and can lead to hysterectomies, was associated with PD risk in women, even after adjusting for hysterectomies, which indicates that the procedure is not a confounding variable that impacts PD risk69,70. In our study, the association between PD severity and hysterectomies with the removal of ovaries indicates that this procedure is not only a risk factor but also a modifier for the severity of the disease itself. It is critical for further studies to elucidate a definitive answer on the implications partial and total hysterectomies have in increasing the risk of PD and modulating PD severity, especially since they are prevalent procedures in gynecological care.

This is an exploratory study that addresses sex-specific experiences from menses to menopause; however, there are some limitations. First, since this study primarily gathers WSHF history from a questionnaire, like other survey studies, it relies heavily on the response rate. Our response rate is low, at 31.5%, while 60–80% is generally expected71,72. Although we are below average, a further limitation of our questionnaire response rate is selection bias as our questionnaire was deployed via email to PD GENEration, leaving those who are neither able to join PD GENEration nor have regular email access unable to participate. As PD GENEration is rapidly growing, we hope to continue deploying this questionnaire to more women with PD. As this is an exploratory study for the purpose of hypothesis generation, we presented our results without adjusting for multiple comparisons. Though given the large number of associations tested in this study, there is the possibility of Type I errors. This study used a questionnaire based on previous literature looking at women’s health and PD and WSHFs that are of interest to researchers and clinicians looking at sex differences. Because of the study design and variable selection approach, a causal relationship between sex-specific factors and PD cannot be established. It is possible that some of our findings may have been affected by confounding or reverse causality. Further studies are needed to expand the sex-specific factors to analyze, ideally based on Directed Acrylic Graphs (DAGs). Furthermore, as we implemented this questionnaire retrospectively without a medical history for each participant, the study cannot define the causality of PD severity. This limits the definitive conclusiveness of the role the discussed factors have in PD severity, leaving future studies to validate these findings. We have also begun distributing this questionnaire across North America Cleveland Clinic campuses. Currently, this questionnaire has been circulated nationally in the United States, and we are expanding our efforts to Latin America through the Latin American Research Consortium (LARGE-PD)73. This international effort will help better understand WSHF differences between regions and sociocultural norms, overcoming the lack of generalizability this study currently faces regarding women with PD in various environments. These greater efforts will hopefully overcome the low response rate (due to selection bias) currently present in this study.

Another limitation is that this study uses PD severity. PD severity varies across literature and clinical standpoints, as many factors contribute to categorizing PD patients’ current health-related quality of life. We implemented Martinez-Martin et al.’s thresholds for severity, but because our study uses only one UPDRS time-point for analysis, this could be a misrepresentation of the respondents’ PD severity phenotype and result in misclassification bias. The UPDRS assessment can be influenced by many factors, such as the time medication was taken before the assessment, stress, “on” and “off” states, and whether the participant or clinician answered the questions74. Thus, there is a lack of generalizability as we are using a one-time assessment as the basis of severity alongside participant-reported outcomes. A larger population of women with PD is needed to validate the findings presented in this study. Further studies will be needed to determine whether these thresholds represent the moderate/severe PD phenotype and use multiple UPDRS assessments. Progression of PD is another critical factor in the health-related quality of life, and it varies from person to person. Thus, studying the effects of WSHFs in PD progression will significantly benefit women with PD and the scientific community.

Lastly, the sample size is another limitation of this study. Due to the overall n = 304, we have small sample sizes between our phenotype groups when looking at each WSHF. A larger sample size would allow us, with greater statistical power, to look at smaller effect sizes each WSHF has with PD severity. With a higher response rate, we hope to overcome this limitation by validating our findings in a larger sample size of women with PD.

Despite these limitations, our study has considerable strengths. We created a scientific questionnaire that addresses multiple WSHFs, some of which have never been addressed before18. Our questionnaire was a collaboration between neurologists, PD experts, and women’s health specialists and was validated for comprehension at Cleveland Clinic. Unlike other previous questionnaires, this one is publicly available for the sole purpose of incorporating WSHF questions into routine PD assessments. As many questionnaires used in the clinic are focused on gender-neutral factors, we hope this study sheds light on the purpose of acknowledging sex-specific aspects in the pursuit of bettering the treatment and care for those suffering from PD. Furthermore, this study demonstrates the importance that sex-specific factors have and that the field must further explore the role of these factors in PD risk and severity.

Another strength of this study is that we look at exogenous hormone therapy, such as estrogen, in relation to PD severity. Previous studies have considered estrogen as a protective factor against PD; although this has not been reproducible across the field, there are limited studies that have addressed estrogen in PD severity28,29,31,33,75. While only 28% of our cohort utilized exogenous hormone replacement, we did not see exogenous hormone replacement associated with the mild phenotype. Moreover, we did not see any birth control forms associated with PD severity, despite the majority of participants having a history of use. Further studies are needed to better understand the role of estrogen and hormones in PD severity.

Another strength of this study is that it includes genetic, demographic, sex-specific, and clinical variables when looking at PD severity. Previous literature that looked at WSHFs used electronic medical records or a questionnaire that didn’t incorporate genetics, leading them to be neglected in PD etiology. We had a moderate cohort of those with LRRK2 and GBA mutations. Our study showed that amongst our cohort, having a LRRK2 genotype was significantly associated with a moderate/severe phenotype. Thus, including genetic data in these models elucidates that genetic predisposition may be related to having a more severe PD phenotype in women.

Overall, this study used a questionnaire deployed across the United States to determine what WSHFs women with PD experienced. With this questionnaire, we were able to create a dataset of 304 women and their genetics, women-specific experiences, and clinical history. We found that by creating score thresholds for each UPDRS subpart, we could determine if WSHFs were associated with PD severity when adjusted for age, age at diagnosis, and PD medication. We found that depression and vaginal birth (Part I); depression and perinatal depression (Part II); LRRK2 mutation status, B12 deficiency, and total hysterectomy (Part III) were significantly associated with a moderate/severe PD phenotype compared to a mild phenotype. This is one of the few studies that emphasize the role of WSHFs in PD. Future studies with a larger population of diverse women are needed to validate these findings, preferably with multiple UPDRS scores and a more detailed clinical history, as this study is an exploratory effort to address if WSHFs affect PD severity in women with PD. In regards to the limitations stated previously, such as the variables analyzed, causality, and the generalizability of findings, the results of this study should be considered with caution. The results observed in this study can shape future studies with larger cohorts to incorporate sex-specific factors when building models and explore the casualty of these factors as seen in other PD studies76. Nevertheless, this study sets the groundwork for acknowledging the role WSHFs in conjunction with non-sex-specific factors may play in PD and the potential benefit the scientific community can gain for therapeutics and clinical guidance if we further investigate the role sex-specific factors have in PD etiology.

Methods

Questionnaire development and implementation

With the gap in information regarding WSHFs and PD, we developed and implemented a nationally distributed questionnaire, which is often used to generate large datasets of patient-reported outcomes within the epidemiological field77. Our previous publication explains how we developed this questionnaire in detail78. In summary, we compiled a list of WSHFs based on their possible association with the disease or their lack of inclusion in PD data thus far. Then, neurologists and movement disorder specialists at Cleveland Clinic along with PD and women’s health experts advised on further modifications of the questionnaire. It was then formatted using a combination of short-text (age and additional comments to be typed if needed), checkboxes, and Likert-type rating scales for symptom changes with the guidance of the Cleveland Clinic Quantitative Health Sciences patient-reported outcomes team. Our primary concern was to limit recall bias, as our subjects skewed toward being above the age of 50 and post-menopausal, so we implemented branching logic, allowing the participant to answer questions that are relevant based on their age of diagnosis (during menses, before or after menopause) and women-specific experiences (including birth control, pregnancies, surgeries, and hormone-replacement therapy) that can occur around that age range. The questionnaire, which can be found in its entirety in our previous work78, is on REDCap, a secure web application used for building online surveys and databases79. To validate that our questionnaire was comprehensible and could be completed in a reasonable time, movement disorder specialists at CCF administered it with a tablet in the waiting room to ten women with PD and verified that these subjects were able to answer the questions comprehensively in a timely manner. The participants took an average of 20 min to complete the questionnaire.

The Parkinson’s Foundation launched the PD GENEration study (PD GENE; ClinicalTrials.gov identifier: NCT04057794), which provides genetic testing for seven PD-related genes by a CLIA-certified laboratory and genetic counseling at multiple locations in North America at no cost to the individuals diagnosed with PD. At the same time, they collect critical clinical and demographic information from each patient21. Each participant is seen and clinically assessed by a movement disorder specialist, who inputs their demographic and clinical data. We have partnered with the Parkinson’s Foundation to implement our questionnaire because, in the past, PD GENEration has successfully distributed various research questionnaires from other institutions to patients on their email list who consented to receive questionnaires designed for research80,81. Starting in mid-November 2021, PD GENEration released our questionnaire cross-sectionally to women with PD who had participated in PD GENEration in order to retrospectively inquire about their women-specific experiences and how these experiences may have impacted their PD health-related quality of life. An email reminder was sent in mid-December, and the questionnaire closed on December 2021. This study was performed with written informed consent under an IRB regulated protocol.

PD GENEration has IRB approval (IRB # 20-596) to share de-identified patient data (genetics, clinical information, and demographics) with us in addition to the responses to our questionnaire for research purposes. PD GENEration exported the participants’ responses alongside their genetic and clinical data. This study is approved by Cleveland Clinic (IRB # 21-1138).

Quality control

As this is a patient-reported study, we filtered the variables in the questionnaire and demographic data from PD GENEration, and we removed variables where at least 80% of responses were missing or “not available”.

MDS-UPDRS severity scale

The Unified Parkinson’s Disease Rating Scale (MDS-UPDRS) is the most widely used clinical scale for rating motor and non-motor symptom severity in PD. The MDS-UPDRS has four parts: I: Non-motor Experiences of Daily Living; II: Motor Experiences of Daily Living; III: Motor Examination; IV: Motor Complications82. At the time of recruitment, PD GENEration had either a clinician or study coordinator administer the MDS-UPDRS.

Missing data in clinical rating scales is problematic because assessments are usually time-locked to an office visit and cannot be retrospectively filled in with reliability. To overcome the missingness in the dataset, we followed the protocol set by Goetz et al. that established thresholds for the maximum number of scores that can be missing from each part. Goetz et al. found that only one missing item from Part I, two from Part II, seven from Part III, and zero from Part IV can be allowed for an individual to be included in subsequent analyses83. If the number of missing items is within the threshold, we followed a protocol implemented by Goetz et al. to create a prorated total score for UPDRS Parts I-IV83.

Following these adjustments, we divided our cohort into “mild” and “moderate or severe” UPDRS groups based on the maximum triangulation cut-off values stated in Martinez-Martin et al.84. We decided to group them by “mild” vs. “moderate/severe” due to the low number of women with a “severe” UPDRS score. Thresholds for moderate/severe were Part I: >21; Part II: >29; Part III: >58; and Part IV: >12. Mild vs. moderate/severe distributions for each UPDRS part are presented in Supplementary Table 3.

Statistical analysis

Demographics, clinical characteristics, genetics, questionnaire responses, and UPDRS scores were summarized using descriptive statistics. Comparisons were made using t-test or Satterhwaite t-test (if assumption of equal variance was not met) for continuos variables and chi-square or Fisher’s exact test for categorical variables. Univariate logistic regression models were constructed for each UPDRS subpart to evaluate the associations between PD severity and WSHFs, genetics, and clinical variables. For “check all that apply” questions, dummy variables were created for each response option. Variables with fewer than five positive responses were not included in the analysis. Furthermore, for Likert-type questions, “N/A” and “Don’t recall” were set to missing. Questions were excluded if more than 80% of responses were “N/A” or “Don’t recall.” As our dataset contains novel variables that have neither been explored nor been well understood in the PD field so far, we first sought to see which variables may be associated with predicting PD severity using a univariate logistic regression model. If a variable had a very low or high prevalence, the Firth method was used in our regression modeling85,86. For each UPDRS subpart, multivariable logistic regression models were constructed from the variables of interest that were gathered from the univariate regression models. These included age, disease duration, and medication, as determined a priori, along with WSHF variables that were significant in the univariate logistic regression models at p < 0.05. Model assumptions and multicollinearity were assessed. All statistical analyses were conducted using SAS version 9.4 (SAS Institute Inc, Cary, NC). Statistical significance was established throughout at p < 0.05.

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.