Increased risk of Parkinson’s disease after depression

Methods.

Results.

A total of 1,358 depressed and 67,570 nondepressed subjects were included in the analyses. The descriptive characteristics of this study population are displayed in table 1. PD was diagnosed in 278 subjects. Among depressed subjects, 19 (1.4%) developed PD, and among nondepressed subjects, 259 (0.4%) developed PD. Follow-up ranged from 1.0 to 25.3 years among depressed subjects and from 0.2 to 25.4 years among nondepressed subjects. Mean age (SD) at the start of the study was 43.8 (13.5) years among depressed subjects and 37.8 (14.7) years among nondepressed subjects. Compared with nondepressed subjects, there were more females in the depressed group and also more subjects with a lower level of education. Type of health insurance was about equally distributed among subjects with and without depression.

In table 2, the associations between depression and PD are shown. The hazard ratio (HR) with its corresponding 95% CI after correction for age, sex, and socioeconomic status was 3.13 (1.95 to 5.01) for depressed vs nondepressed subjects. In subgroups of males and females, the estimated HR (95% CI) adjusted for age and socioeconomic status were 3.60 (1.67 to 7.80) for males and 2.85 (1.57 to 5.17) for females. In subgroups of age at diagnosis of depression, the HR (95% CI) were 2.38 (0.74 to 7.67) for subjects under 50 and 3.27 (1.96 to 5.47) for subjects 50 years or older. After inclusion of subjects with depressive complaints in our exposed group, the HR (95% CI) was 2.84 (1.83 to 4.41) with adjustment for age, sex, and socioeconomic status.

Discussion.

In this study, we observed a positive association between depression and subsequent risk for PD, which was significant after adjustment for age, sex, and socioeconomic status. The results in subgroups based on sex and age at diagnosis of depression were consistent with this overall result. The estimated HR were in the order of magnitude of 3, which, in epidemiologic terms, indicates a quite large effect. Although we are not aware of other prospective studies on depression preceding PD, our results are supported by a number of case studies that have previously been published.8-11⇓⇓⇓

The subjects included in our study were a largely unselected group of persons because almost all people in the Netherlands are registered in a general practice. The population in the RNH, therefore, may be considered representative for the regional general population. To ensure the quality of the RNH database, instruction and training sessions are held for participating general practice as well as regional consensus groups and quality control interventions. In addition, an automated thesaurus and automated checking for erroneous or missing entries further increase the validity of the data.15 Reliability and completeness have been shown by comparisons between the RNH’s cancer data and the data of the regional cancer registry17 and between the RNH’s epilepsy data and the Maastricht Epilepsy Case Register.15

Misclassification of disease is not likely to have occurred in our study. Because PD does have serious implications for the patient’s quality of life, it is not expected that many patients with the disease remain undiagnosed. Furthermore, because a diagnosis of PD was made on clinical grounds, using the criteria of the ICHPPC-2 and in most cases by the neurologist, it is also unlikely that a patient was coded as having PD when this was, in fact, not the case. The ICHPPC-2 criteria for PD14 are comparable with the criteria of the UK Parkinson’s Disease Society Brain Bank18 for diagnosis of the parkinsonian syndrome. They include tremor, bradykinesia, and muscular rigidity but not postural instability. They also include most exclusion criteria (Step 2) as specific ICPC codes are used for patients with diseases that are used as exclusion criteria mentioned in the UK Parkinson’s Disease Society Brain Bank and only one code per disease can be used. The RNH procedure does not include the supportive criteria (Step 3), however.

We do not expect that detection bias might have influenced our results. If PD had been diagnosed more often in depressed subjects because of a potential higher consultancy frequency compared with nondepressed subjects, we would also have found positive results for other diseases. However, for investigated endpoints such as diabetes, epilepsy, MS, and stroke, we observed no associations. Furthermore, because the HR for PD was highly significant, the chance of a type 1 error is negligible. Also, detection bias is unlikely because data collection in the RNH is not specifically focused on either PD or depression.

Some readers may criticize the criteria used to diagnose depression. Depressive disorder was diagnosed according to the criteria of the ICHPPC-2,14 which are used in the RNH. The six criteria of this classification are largely comparable with the nine symptoms of a depressive episode as described in the Diagnostic and Statistical Manual of Mental Disorders, Fourth Revised Edition (DSM-IV) of the American Psychiatric Association.19 In the ICHPPC-2 criteria list, two symptoms of the DSM-IV classification are not included (reduced appetite and loss of energy/tiredness), while two other symptoms are combined to one item (slow mentation and decreased interest and activities). It has frequently been reported that not all depressions are recognized or diagnosed as such by the general practitioner.20-22⇓⇓ In an attempt to evaluate possible exposure misclassification, we performed a sensitivity analysis, adding subjects with depressive complaints (ICPC code P03) to our group of depressed subjects while they were consequently removed from our reference group. Risk estimates were essentially unchanged, indicating that our results are robust for such changes.

Assuming that the observed association is real, how can we explain this finding? As patients had no diagnosis of PD at the time their depression started, psychological mechanisms in the sense of dealing with illness seem unlikely to explain these findings. Therefore, we sought for a biologic explanation. In our view, the “serotonin hypothesis” for depression in PD may offer a plausible explanation for the increased incidence of PD in patients with prior depressive disorder. This hypothesis considers the lowered brain serotonin activity in patients with PD as a risk factor for depression.23,24⇓ Many studies have demonstrated lower serotonin activity in patients with PD, either as decreased 5-hydroxyindolacetic acid concentration in CSF or as degeneration of serotonergic nerve cells in postmortem neurochemical or autoradiographic studies.25-28⇓⇓⇓ As serotonin has an inhibitory function on dopamine release in the striatum, the reduction of serotonin activity is seen as a compensatory mechanism for the reduction in dopamine activity in PD.29,30⇓ At the same time, this compensatory reduction of serotonin activity increases the risk for depression.31 Thus, patients with PD may have a biologic vulnerability to depressive disorder that is specific for PD. As the reduced activity of serotonin already exists before the onset of motor symptoms, the risk for depression is also increased long before parkinsonian symptoms become apparent. It is possible that the increased incidence of PD in patients with a prior history of depression is a reflection of the presence of this biologic risk factor for depression in still symptom-free, preclinical stages of PD. Our findings may thus be explained by the specific over-representation of patients with preclinical PD among the patients with depressive disorder. The “serotonin hypothesis” also raises the conceptual issue of whether depression in this context should be seen as a separate disease entity or rather, because of its possible pathophysiologic association with neurotransmitter changes of preclinical PD, as an intrinsic symptom or first clinical manifestation of this disease.