Abstract
Background:
Depressive disorders are recognized as a common neuropsychiatric disorder of Parkinson’s disease (PD). Reported frequencies vary widely among studies and depend on the diagnostic criteria, the methods of ascertainment used, and the population sampled.
Objective:
We aimed to evaluate the frequency of depressive disorders in PD and to investigate the relationship with PD clinical variables.
Methods:
A systematic review and meta-analysis of observational studies (community-based, prospective and retrospective cohort, case-control, and cross-sectional studies) reporting the frequency of depressive disorders in PD patients.
Results:
Electronic database search wielded 3,536 articles; an additional 91 were identified through citation chaining. 163 full-text articles were assessed for eligibility. Of these, 49 met the inclusion criteria for our analysis. The pooled frequency of depressive disorders was 30.7% (95% confidence interval [CI] 25.6 to 36.2; I2 = 95%; 49 studies; combined n = 10,039). The pooled frequency of major depressive disorder was 14.0% (95% CI 10.5 to 18.5; I2 = 88%; 23 studies; combined n = 5,218). Subgroup/meta-regression analyses were conducted to investigate the relationship between frequency and study inclusion criteria, methodology used for diagnosis, and study design. We found a statistically significant correlation between study design and depressive disorders frequency (ranging from 8% in the community-based study to 44% in the retrospective studies) and a statistically significant positive correlation between mean baseline PD duration and major depressive disorder frequency.
Conclusion:
The current meta-analysis found a global frequency of depressive disorders of 30.7% and major depressive disorder of 14.0%. Study design influenced the frequency of depressive disorders in PD. Mean baseline PD duration and major depressive disorder frequency were positively correlated.
Keywords
INTRODUCTION
Parkinson’s disease (PD) is mainly diagnosed based on typical motor symptoms of tremor, rigidity, bradykinesia, and motor impairment, but it has become increasingly accepted that non-motor symptoms are a prominent component of PD [1, 2].
Depressive disorders are widely recognized as one of the most common neuropsychiatric disorders in PD patients [3, 4]. They can develop at any phase during PD and frequently predate the onset of motor symptoms [5].
A neurobiological basis of depression in PD, interacting with psychological factors, is recognized, and depression could be, in part, the result of damage to serotoninergic neurotransmission as well as noradrenergic and dopaminergic mechanisms [6]. Untreated depression has a negative impact on PD prognosis and is associated with earlier initiation of dopaminergic therapy, greater functional disability, physical and cognitive deterioration, increased mortality, poorer quality of life, and increased caregiver distress [7–9].
Despite their high frequency in PD, depressive disorders remain underrecognized and undertreated [10, 11]. Reported frequencies of depressive disorders in PD depend on which definitions of depressive disorders and depressive subtypes, and diagnostic criteria are used, as well as the population sampled, and the frequency and duration of depressive symptoms [12].
The main subtypes of depressive disorders in PD include major depression, minor depression, and persistent depressive disorder or dysthymia, and are similar to those of depressive subtypes in non-PD populations [12, 13].
Non-major forms of depression are distinguished by the presence of milder and fewer symptoms compared to major depression [12]. Besides sustained mood changes and anhedonia, depressive disorders are characterized by cognitive and vegetative symptoms with sleep and appetite changes that often overlap with features of PD itself [12]. Furthermore, some PD-specific features, like dopaminergic medication effects on motor and mood states and increased psychiatric comorbidities like impulse control disorders or psychosis frequently, make the diagnosis of depressive disorders in PD harder [12].
To avoid underdiagnosis of PD depression, a National Institutes of Health work group recommended an inclusive approach in which overlapping somatic features are regarded as features of the depressive disorder [14]. This work group also recommended provisional diagnostic and assessment strategies to diagnose depressive disturbances in PD by adapting DSM criteria [13] for diagnosing depression in PD [14].
Estimates of the frequency of depressive disorders in PD vary widely among studies, with frequencies ranging from 2.7% to more than 90% [15–17]. There are few systematic reviews on the frequency of depressive disorders in PD [18–21], with important limitations or lacking information regarding the relationship of the frequency of depressive disorders with PD clinical variables.
Updated investigations on depressive disorders in PD are mandatory to promote adequate assessment and management of depression and to prevent negative impact on PD prognosis.
The current study aimed to evaluate the frequency of depressive disorders, major depressive disorder, and non-major depressive disorders (including minor depressive disorder and persistent depressive disorders) in PD patients.
MATERIALS AND METHODS
We conducted a systematic review evaluating the frequency of depressive disorders in PD.
Search and inclusion criteria
A literature search was performed by one of the authors (GSD); the screening and decisions on inclusion were performed independently by three of the authors (IC, LP, and CS); data extraction and quality control were performed independently by four of the authors (IC, CS, AQ, and JV) and disagreements on screening or data extraction were resolved by discussion with two of the authors (GSD and IC). This systematic review followed the reporting principles of PRISMA. The protocol was registered in PROSPERO as Neuropsychiatric events in Parkinson’s Disease: a systematic review and meta-analysis (CDR42018118863).
The search strategy, developed for all databases, combined the terms Parkinson* with depression and was adapted from previous Cochrane systematic reviews of interventions for this condition, and The Cochrane Highly Sensitive Search Strategy for identifying randomized clinical trials (RCTs) in MEDLINE: sensitivity- and precision-maximizing version (2008 revision) was used. The search strategy was restricted to human participants only. All terms were searched as free-text and controlled vocabulary terms (i.e., Medical Subjects Headings (MeSH)).
The bibliographic databases MEDLINE, EMBASE, and the Cochrane Central Register of Controlled Trials (CENTRAL), and the clinical trials registries the WHO International Clinical Trials Registry Platform, and ClinicalTrials.gov, were searched for relevant material for inclusion.
In addition, the reference lists from the identified articles were cross-checked to identify any further potentially eligible studies.
Studies were included in this review if they: 1) were community-based, prospective and retrospective cohort studies, case-control studies, or cross-sectional studies; 2) were peer-reviewed papers written in English, Spanish, or French; 3) reported the frequency of depressive disorders, major depressive disorder, non-major depressive disorder, minor depressive disorder, or persistent depressive disorder; 4) included diagnosis made by clinical, structured or semi-structured interview, or through the administration of validated scales administered by experienced clinicians and not only by self -report questionnaire; 5) included 25 or more participants. Non-major depressive disorders included minor depressive disorder or “other specified depressive disorder, depressive episode with insufficient symptoms” [13], and persistent depressive disorder or dysthymia. Minor depressive disorder differs from persistent depressive disorder in its duration (less than two years) although the symptom profile is similar [13].
The last search was performed in December 2020.
Data extraction and outcomes
Three reviewers independently screened the titles and abstracts of the retrieved articles and read the full-text reports to determine whether they met the inclusion criteria. Disagreements were resolved through discussion, with a fourth reviewer serving as the final arbitrator. When necessary, reviewers sought additional information from the study authors to resolve questions regarding eligibility.
Four reviewers then extracted data from the individual studies identified for inclusion into a pre-piloted form (Microsoft Excel spreadsheet). This information included: authors; journal; year of publication; study design; setting; population demographics; sample size; intervention and control conditions; primary outcome; secondary outcomes; methodology for diagnosis (clinical or/and (semi) structured interview).
Data analysis
The analysis was performed in R version 4.0.3 using the meta-package [22]. All outcomes were binary. Therefore, the pooled proportion of events was calculated with 95% confidence intervals, using a Sidik-Jonkman random effect model [23]. Statistical heterogeneity was reported using I2 [24]. To assess the robustness of the results and to identify potential methodological biases and subpopulations in which outcomes differed, we also conducted a series of a priori defined meta-regressions (percentage of men, mean disease duration, mean age onset and, mean disease stage Hoehn & Yahr [25] and subgroup analyses (cognitive status, the methodology used for psychosis diagnosis and the study design).
To investigate the possibility of publication bias we created a funnel plot and conducted Egger’s test [26] for the primary outcome.
RESULTS
The electronic database search yielded 3,536 citations, with an additional 91 were identified through citation chaining. Of these, 163 were fully inspected, 57 removed and 106 included as relevant for neuropsychiatric events prevalence. Of these, 49 met the inclusion criteria for our analysis, all of which were published between 1989 and 2017. Forty-nine studies (1 community-based, 13 case-control, 18 cross-sectional, 2 retrospective, and 15 prospective cohort studies) with a total of 10,039 patients were included for the global frequency calculation of depressive disorders, 23 studies with a total of 5,218 patients were included for the calculation of the frequency of major depressive disorder, 7 studies for the calculation of the frequency of non-major depressive disorders, 6 studies for the calculation of frequency of persistent depressive disorder and 10 studies for the calculation of the frequency of minor depressive disorder.
Studies that used vague diagnostic criteria, were only published as an abstract, poster, or conference paper, had incomplete data, were duplicated publications, or were methodologically weak were excluded.
The flowchart of search results is displayed in Fig. 1. An overview of included studies is provided in the Supplementary Material.
Search results flowchart.
Risk of bias
Due to the different study designs included in our review, we assessed the risk of bias using two methods, namely the Newcastle-Ottawa Scale (NOS) in case-control studies and cohort studies and the Quality Assessment Checklist for Prevalence Studies (adapted from Hoy et al. [27]) for prevalence studies. Studies were considered high quality on the NOS if they obtained more than four points. Full data are provided in the Supplementary Material.
Primary outcomes: Global frequency of depressive disorders and major depressive disorder
The pooled global frequency of depressive disorders was 30.7% (95% CI 25.6 to 36.2; I2 = 95%; 49 studies; combined n = 10,039; Fig. 2). Additionally, none of the regressors were able to meaningfully explain the statistical heterogeneity in the analysis. Regarding the subgroup analyses conducted, we found that none of the subgroups were statistically significant (Table 1).
Pooled global frequency of depressive disorders.
Meta-regressions and subgroup analyses for frequency of depressive disorders
We found no evidence of publication bias (p-value for linear test = 0.20). The funnel plot for this outcome can be found in the Supplementary Material.
The pooled global frequency of major depressive disorder was 14.0% (95% CI 10.5 to 18.5; I2 = 88%; 23 studies; combined n = 5,218; Fig. 3). Additionally, the only two regressors that were found to meaningfully reduce statistical heterogeneity for this outcome, namely mean baseline Hoehn & Yahr stage (R2 = 13.16%) and mean baseline PD duration (R2 = 20.42%), were positively correlated. Regarding the subgroup analyses conducted, we found that none of the subgroups were statistically significant (Table 2).
Pooled global frequency of major depressive disorder.
Meta-regressions and subgroup analyses for frequency of major depressive disorder
We found no evidence of publication bias (p-value for linear test = 0.20). The funnel plot for this outcome can be found in the Supplementary Material.
Secondary outcomes
The pooled estimated frequency for non-major depression was 20.6%, 20.0% for minor depression, and 11.3% for persistent depressive disorder. The full details about the secondary outcomes can be seen in Table 3. Full data and forest plots are provided in the Supplementary Material.
Secondary outcomes
DISCUSSION
This systematic review of the frequency of depressive disorders in PD includes recent evidence not previously included in other published systematic reviews. This is also the first systematic review that considers important PD clinical variables such as PD duration and mean Hoehn & Yahr stage to calculate the frequency of depressive disorders. In our study, the pooled global frequency of depressive disorders in PD patients was 30.7% (95% CI 25.6 to 36.2), major depressive disorder was 14.0% (95% CI 10.5 to 18.5), non-major depressive disorder was 20.6% (95% CI 13.4 to 30.4), minor depressive disorder was 22.0% (95% CI 14.8 to 26.4), and persistent depressive disorder was 11.3% (95% CI 6.6 to 18.5). Our global frequencies are similar to those reported in a previous systematic review published in 2008 by Reijnders [18] that included 36 studies. They found that clinically significant depressive symptoms were present in 35%, major depression in 17% of patients, minor depression in 22%, and persistent depression in 13% [18]. This review took the different settings (general population, general practice, outpatient and inpatient setting, and nursing home) and diagnostic approaches into account. They found a lower prevalence of depression in the general population compared to hospital outpatient and inpatient settings and higher prevalence rates with the use of structured interviews compared to unstructured clinical interviews [18]. A more recent review, published in 2016 by Goodarzi [19], whose primary objective was to review the diagnostic accuracy of screening tools for the detection of depression also checked for the pooled prevalence of depression. This latter study included 21 studies (2,028 patients), from 2000 to 2014, and found a prevalence of major depression of 22.9% and a prevalence of major and minor depression of 31.3%. This review only included adult PD outpatients and included self-rated tools [19].
Two earlier reviews on the prevalence of depression in PD have been published [20, 21] although they have important limitations regarding quality non-assessment of the included studies. Both were part of a more extensive review also including the etiology and/or treatment of depression. The first, published in 2001 by Slaugher [20], included 45 studies conducted between 1922 and 1998 and found a prevalence of depressed patients of 31% using different diagnostic approaches and 42% when using DSM-III and DSM-III-R criteria. Few studies included had evaluated the depression subtypes based on DSM-III and DSM-III-R criteria and found a prevalence of 22.5% of dysthymia (4 studies; n = 213 patients), minor depression in 36.6% (3 studies; n = 423 patients), and major depression in 24.8% (1,119 patients) [20].
The second study, published in 2005 by Veazey [21], only included 16 studies, from 1972 to 2001, and gave a range of prevalence of depression between 7% and 76%. The prevalence of major depressive disorder in studies that used clinical interviews to establish DSM criteria ranged from 7.3% to 32% vs. 27.3% to 76% when self-report questionnaires were used. In community-based prevalence studies, the reported rate of MD was 7.7%. When populations of outpatients were used, the rates of MD ranged from 7.3% to 32% [21].
Estimates of the frequency of depressive disorders in PD vary widely among studies [15–17].
The variability in reported frequency of depressive disorders may be attributed to the use of different inclusion criteria in PD, the different criteria used to define depressive disorders, inconsistencies in the type of depressive disorder included in the study, as well as the methodology used for psychiatric diagnosis. Inclusion criteria in PD samples differ in PD duration, disease stage, age of onset, and cognitive status all of which could influence the frequency of depressive disorders.
To investigate the relationship between the frequencies of both major depressive disorder and depressive disorders and study inclusion criteria (duration disease, age onset PD, disease stage [Hoehn & Yahr stage] [25], and cognitive status), we conducted a subgroup/meta-regression analysis. Regarding depressive disorder frequency, none of the regressors were able to meaningfully explain the statistical heterogeneity in the analysis. None of the subgroups were statistically significant according to subgroup analysis. Regarding major depressive disorder frequency, the only two regressors that were found to meaningfully reduce statistical heterogeneity for this outcome, mean baseline Hoehn & Yahr stage (R2 = 13.16%) and mean baseline PD duration (R2 = 20.42%), were positively correlated. None of the subgroups were statistically significant according to subgroup analysis.
The methodology used for the diagnosis of depressive disorders (methods of ascertainment, e.g., the use of self-report questionnaires only versus the administration of different scales by a clinic or a clinic, or structured/semi-structured interview) could influence the frequency of depressive disorders. Studies using a rating scale to measure depression may yield higher frequencies than those using a clinical or semi-structured interview [18, 28] as reported by Goodarzi 2016. [19]. We only included studies in which the diagnosis was made by clinical, structured/semi-structured interview, or through the administration of validated scales by experienced clinicians and excluded studies that diagnosed depressive disorders only by self-report questionnaire. We observed a higher overall frequency when using a clinical or/and structured/semi-structured interview methodology (15%) in major depressive disorder compared with solely applying a rating scale (9%) although the difference in frequency was not statistically significant. The same trend was found when comparing the calculation of the frequency of depressive disorders, resulting in a frequency of 32% when using a clinical or/and structured/semi-structured interview methodology vs. 22% when solely applying a rating scale. If we had included all studies the importance of methodological ascertainment would likely have been more evident.
The global frequency of depressive disorders differed according to the study design, ranging from 8% in the community-based study to 44% in the retrospective studies (interaction p-value < 0.001). Regarding the global frequency of major depressive disorder, ranging from 11% in the prospective cohort studies to 16% in the cross-sectional studies, we did not find statistically significant differences in the proportion of major depression between different study designs.
The frequency of depressive disorders seems to be inadequately documented because there are insufficient community-based studies on frequency as well as a need for large and well-designed prospective studies including different stages and disease duration. None of the included studies evaluated the global frequency of major depressive disorder in advanced-stage patients.
This study has several limitations. First, few studies were included to review the frequency of non-major depressive, minor depressive, and persistent depressive disorders. Second, several studies reporting the frequency of disorders in samples lower than 25 patients were excluded as well as those in which diagnosis was made based solely on self-reported scales which can also be interpreted as a strength. Third, the limitations of subgroup and sensitivity analyses are well established and the fact we do not have statistically significant results does not imply that the examined covariates are not of clinical importance. The main strength of this review is its inclusive approach, with a focus on different subtypes of depressive disorders and a focus on the relationship with PD variables. Another strength of this review was the assessment of both the risk of bias and publication bias.
CONFLICT OF INTEREST
Joaquim J. Ferreira: no conflict of interest to report. No specific funding was received for this work. Received grants from GlaxoSmithKline, Grunenthal, Fundação MSD (Portugal), TEVA, MSD, Allergan, Novartis, Medtronic. He received consultancy fees from GlaxoSmithKline, Novartis, TEVA, Lundbeck, Solvay, BIAL, Merck-Serono, Merz, Ipsen, Biogen, Acadia, Allergan, Abbvie, Sunovion Pharmaceuticals, Zambon, Affiris. He also participated in advisory boards for Bial and expert testimony to Novartis.
The other authors have no conflict of interest to report.
