Incidence,Prevalence,and Mortality of Psychosis Associated with Parkinson’s Disease (1991–2010)

Study population

Statistical analysis

Continuous variables were summarized with medians and interquartile ranges (IQRs); categorical variables were summarized with frequency counts and percentages. Incidence rates of psychosis per 100 person-years were calculated using the number of incident cases of psychosis divided by the number of years at risk for the cohort during the specified period. Patients with partial follow-up within a period were counted proportionally. Sex-specific incidences were also calculated.

The risk of mortality for PD patients was assessed using a Cox proportional-hazards model [19]. The outcome of interest was follow-up time in years from the onset of PD to date of last follow-up, with an indicator as to whether the patient was deceased (1; event) or alive (0; censor) at last follow-up. The model included time-dependent covariates (TDC) as predictors for psychosis and unresolved psychosis. The TDC of psychosis was created using time from the onset of PD to first instance of psychosis (1; event) or last follow-up if no psychosis occurred (0; censor). Similarly, the TDC for unresolved psychosis were created using the time from the onset of PD to the presence of psychosis lasting longer than 6 months (1; event) or last follow-up, if psychosis resolved before 6 months or no psychosis occurred (0; censor). This model was adjusted for patient sex and age at onset of PD, as well as the presence of each of tremors and bradykinesia at PD onset. Age at onset was centered and scaled to 5-year increments.

In a secondary analysis, the relationship between mortality and the treatment of psychosis with antipsychotic medications was assessed using a Cox proportional hazards model. Only patients in the PDP cohort were included in this analysis. The outcome of interest was follow-up time in years from the onset of psychosis to date of last follow-up, with indication as to whether the patient was deceased (1; event) or alive (0; censor). In this secondary model we again included the same adjustments of sex, age, and symptoms as in the primary model.

All statistical analyses were conducted using R software version 3.6.1 (The R Foundation for Statistical Computing, Vienna, Austria) [20], and a p value of <0.05 was considered statistically significant.

Standard protocol approvals, registrations, and patient consents

The Mayo Clinic and Olmsted Medical Center Institutional Review Boards approved this study (ID reference number 07-008807), and participating patients (or their legally authorized representatives) provided informed written consent for the use of their medical information for research [2]. More detailed information about the Olmsted County population is reported elsewhere [14, 22]. Information about the different types of parkinsonism and the clinical and pathological characteristics of this cohort is also reported elsewhere [16, 23].

Data availability statement

All the relevant data have been shared and published in this article; data regarding case ascertainment of parkinsonism and methodology on case identification have been previously published [2].

Demographics, incidence, and prevalence

Among the 669 incident cases of parkinsonism in Olmsted County, MN, from 1991 to 2010, there were 311 cases of PD. Eleven cases were excluded for not having record of the cardinal signs of PD. Three additional cases were diagnosed with psychosis prior to the onset of PD and were excluded, resulting in a dataset of 297 incident cases of PD (178 (59.9%) male, 119 (40.1%) female). The median age at PD onset was 74 years of age (74 years for males, 75 years for females, IQR = 66–81). There was pathological confirmation in 24 (3.4%) patients within our cohort. The clinicopathologic correspondence positive predictive value was 83.3% (20/24) and the sensitivity for the patients with clinical diagnosis of PD was 100%. Of the clinicopathologic discrepancies two were pathologically consistent with Alzheimer’s disease, one with progressive supranuclear palsy, and one with vascular parkinsonism.

Of the 297 cases of PD, 114 (38.4%) patients developed psychosis. The median age of psychosis onset was 79 years of age (80 for males, 79 for females, IQR = 74–86). Among the 114 individuals with PDP, 93 (80%) had died at the time of abstraction with a median age at death of 84 years (84 for males, 85 for females, IQR = 79–90). The median age at death in the 183 individuals without psychosis was 86 years (86 for males, 88 for females, IQR = 80–91). Further details of patient demographics and characteristics can be found in Table 1.

Hallucinations were the most common symptom of psychosis, occurring in 99 of the 114 (87%) PDP patients. The median time to the onset of psychosis was 8.1 years (8.3 for females, 8.1 for males, IQR 5.7–10.6) after the onset of PD. The median time of follow up for the entire cohort was 18 years (95% CI: 17–20). 59 out of 114 (52%) PDP patients were treated with antipsychotics (25/49 [51% ] females, 34/65 [52% ] males). Table 2 provides details regarding the characteristics of psychosis for the PDP cohort.

The incidence of psychosis within the first five years of PD was 1.97 cases per 100 person-years (2.22 males, 1.68 females). The overall incidence of PDP was 4.28 cases per 100 person-years (4.47 per 100 person years for females and 4.14 per 100 person years for males). Figure 1 shows the incidence of psychosis by disease year in males and females.

OPEN IN VIEWER

Fig. 1

Shows the incidence of PDP comparing male and female sex as well as both sexes combined. Incidence is calculated per 100 patient years. The number at risk table represents the number of PD patients who are still at risk of developing psychosis comparing the overall cohort.

The prevalence of psychosis at disease year 5, 10, and 15 years was 10.1%, 23.2%, and 39.6% respectively. The median duration of the first instance of psychosis was 2.45 years, with 63.2% of initial instances of psychoses lasting longer than 6 months.

Mortality and survival of PDP

Using the Cox proportional-regression model, we observed that PDP patients have an increased risk of death compared to PD patients (HR = 1.71, p = 0.027, 95% CI = [1.06, 2.76]). Figure 2 shows the expected survival of PDP patients compared with PD patients. Individuals with unresolved psychosis had no significant difference in mortality compared to patients with resolved psychosis (HR = 1.41, p = 0.19).

OPEN IN VIEWER

Fig. 2

Shows the estimated survival comparing PD patients to PDP patients.

Among the patients who have PD without psychosis, males had a 73.4% increase in the risk of death compared to females (HR = 1.73, p < 0.001, 95% CI = [1.29, 2.33]). However, within the PDP cohort there was no sex difference in risk of death (HR = 1.11, p = 0.70, 95% CI = [0.71, 1.73]). Model estimates regarding the risk of mortality in PD and PDP patients can be found in Tables 34 respectively.

Antipsychotic treatment in PDP

Out of the 114 PDP patients, 59 (52%) were treated with antipsychotics. There was no significant difference in the risk of death following the onset of psychosis among treated and untreated PDP patients after accounting for age, sex, and the presence of tremors or bradykinesia at PD onset (HR = 1.11, p = 0.60). Similarly, there was no difference in the risk of death following the treatment with antipsychotics after the onset of psychosis between males and females (HR = 1.11, p = 0.70).

Incidence and prevalence of PDP

Few epidemiological studies have explored the frequency of PDP due to the rarity and the lack of a universally accepted diagnostic criterion. Using the REP medical records-linkage system, we report an overall incidence of 4.3 cases per 100 person-years. Prevalence of psychotic symptoms was 38.4% in our population of PD patients. In a previous study, which used a prospective longitudinal design in Rogaland County, Western Norway, the observed incidence was 7.97 per 100 person-years after a 12-year follow-up [24]. Despite supporting similar results, the Norwegian results differ from ours as their study explored only hallucinations or delusions, whereas our study used the NINDS/NIMH Work Group clinical criteria [17]. Furthermore, the longer follow up of about 25 years of clinical observation, in-depth review of medical records, and the approach of our study may have provided a more accurate estimation of PDP. However, the higher incidence of PDP found in the prospective study may indicate an underreporting of psychotic symptoms within patient populations. This may be of importance to bring awareness to the significancy of educating PD patients about psychosis. This may also encourage healthcare providers to examine for psychotic symptoms when working with PD patients.

The prevalence of psychotic symptoms in PD has been reported varying from 16–75% [6, 25]. However, there is limited knowledge of the epidemiology of PDP using the most recent diagnostic criteria, which includes multiple psychotic symptoms [18]. The population-based study from Norway also found a prevalence of 41.9% [26]. A study from the UK also observed a prevalence of 26% of PDP, while 90.8% of those patients met the NINDS/NIMH Work Group clinical criteria for PDP [27]. However, a different cross-sectional study using this criterion reported a prevalence of 60% [25]. On the other hand, methodological differences and differences in the clinical definition should be considered while comparing this study to others. Importantly, the NINDS/NIMH Workgroup has recently proposed new criteria that we adopted but could not be used in previous studies, thus the discrepancies in the clinical definitions. In addition, using hospital-based or community-based cohorts, can lead to an overestimation in the frequency of PDP.

Mortality and survival of PDP

Although several studies have shown that PDP increases caregiver burden and risk of admission to nursing home care [28], few studies have explored the survival of patients with PDP compared to PD. A population-based study from Norway reports that patients with PDP have a 98% increased risk of death compared to PD patients [26]. Our results show that PDP patients have a 71.4% increase in the risk of death; importantly, individuals with a psychosis lasting longer than 6 months (hence defined as “unresolved”) did not have a higher rate of mortality as compared to those with a resolved psychosis. Thus, duration of psychosis does not seem to be associated to an increased risk of mortality in our study. Furthermore, nearly 70% of PDP cases in our population had a psychosis that lasted longer than six months, showing that the past treatment within this population was not effective. Although a number of studies have reported ambiguous results about the efficacy of antipsychotics on PDP symptoms, few have explored their impact on mortality [29–34]. A recent study showed that the off-label use of antipsychotics to treat patients with PDP led to an increased rate of mortality [35]. Our results show that treatment with antipsychotics does not have a statistically significant effect on mortality compared to PDP patients who went untreated. Although we do not observe an association, our results may support that the use of antipsychotics may potentially improve symptoms without decreasing the survival of these patients [29–34]. Notably, we observed that none of the patients experiencing psychosis were prescribed clozapine within our cohort. While clozapine is a drug that can be effective on the treatment of psychosis, the need of a diligent compliance on the management of possible drug-induced neutropenia or agranulocytosis limits the use in the clinical practice, especially in non-neurological/psychiatric providers.

Our study also supports that the clinical phenotype upon presentation of PD symptoms impacts the survival of PD patients. We report that patients who present with tremor predominant (TD) phenotype of PD have a decreased risk of death compared to patients who present with postural instability/gait difficulty (PIGD) phenotype. It has been shown that individuals with PIGD phenotype experience greater subjective intellectual, motor, and occupational impairment than people with TD [36]. These individuals also have a greater incidence of neuropsychiatric disorders compared to individuals presenting with TD phenotype [37, 38]. However, when examining the effect of phenotype on mortality of PDP, we find that there is no significant difference between TD and PIGD. To our knowledge, the clinical phenotype of PD on the mortality related to PDP is relatively unexplored. These results may continue to support that these differing phenotypes may have pathological and molecular differences, which impact disease course and survival. Our findings further highlight the relevance of an individualized treatment of these differing phenotypes. Although these phenotypes have differing impacts on mortality in PD, they did not have a significant difference on mortality in PDP.

Another relevant observation was that the treatment with antipsychotic medications did not further impact the mortality of patients with TD or PIGD phenotypes in PDP. These results further suggest that the current standards of care for patients with PDP may not be effective at treating PDP or decreasing the mortality of PDP.

Recent studies have been exploring the effect of sex on the clinical features, disease course, response to treatment, mortality, and pathophysiology in patients with PD. It has been shown that male sex is associated with higher incidence, earlier disease onset, more severe motor symptoms, hastened disease progression, and more frequent cognitive symptoms [39]. Our study finds that both men and women have a similar incidence of PDP as well as a similar frequency of presenting psychotic symptom and no significant difference in mortality when comparing men and women with PDP.

Our study has several strengths, including its long follow-up period (25 years) and having access to all medical history of each patient, which allows us to minimize the effect of selection bias on our results. However, we also acknowledge several limitations; additionally, confounders and reverse causality were considered in the interpretation of the data, but our study design and statistical adjustments minimized this risk. First, using a retrospective study design may overlook under-reporting of psychotic symptoms. In fact, it is possible that the provider writing the clinical notes could have omitted psychosis symptoms; therefore, there is the risk that our percentage of PDP (38%) could be underestimation. However, to minimize this effect, a study abstractor reviewed all cases of PD using the most recent criteria to classify PDP. Secondly, due to PDP’s rarity, in 114 cases, we only had 59 cases treated with antipsychotics, limiting more granular analyses. Thus, the accuracy of the documented clinical parameters may be affected. Using a retrospective study design, we could not confirm that patients who were prescribed anti-psychotic medications used them; also, it was not possible to have detailed report of visual symptoms in many cases; also, the clinicians used the term “hallucinations” to include “illusions” (hence, the absence of illusions in Table 2); moreover, we could not distinguish between visual and auditory hallucinations, since they were both reported as “hallucinations”. Hence, we were not able to break down our analysis between these two symptoms. Additionally, the study period used may no longer reflect the current incidence or treatment efficacy. Importantly, our study focused only on incident cases of PD that developed psychosis; we excluded all the cases of PD that later developed dementia; indeed, exploring the epidemiological figures of psychosis in PD-AD, PDD and DLB was beyond the scope of the study. We could not analyze PD patients with dementia (PDD) because we did not have detailed, consistently collected data on cognitive decline and dementia. Therefore, we excluded PDD cases from our study. Finally, our cohort lacked detailed pathological confirmation in most cases and, thus, is a clinically defined PD cohort.