The cohorts

The study population comprised four cohorts, three exposed and one unexposed. Event of interest (outcome) was purchase of APD. Because PD is extremely rare among young people, only persons 30 years or older were included. The three exposed groups included all Danish inhabitants who during the five year study period purchased antidepressants, lithium, or antidiabetics (index drugs). The unexposed group comprised a large random sample from the general population. Persons from the unexposed group would change to one of the exposed groups if they purchased any index drugs during the study period.

Methodological aspects

The Danish prescription database does not hold individual information on the indication for prescription. Our data therefore have to be interpreted as indirect evidence. The included drug classes all have well defined indications, for which they primarily, but not exclusively, are prescribed. An important question is to which degree drug consumption corresponds to specific disorders. Antidepressants are mainly prescribed for depressive and anxiety disorders,¹⁰ but also for obsessive‐compulsive disorder, social phobia, eating disorders, neuropathic pain conditions, etc. Lithium is only used in affective disorders and antidiabetic drugs only for diabetes mellitus. Most APDs are prescribed for treatment of PD but a minor part may be prescribed for a few other conditions, like Lewy body dementia, multiple system atrophy, and restless legs syndrome.¹¹ The whole group of anticholinergics (N04A‐) were excluded because their primary use presumably is other conditions than PD. We have no indications that these factors can explain our results, as there is no evidence showing that patients with anxiety/affective disorders more often develop Lewy body dementia, multiple system atrophy, or restless legs syndrome than do patients with diabetes or persons from the general population.

Choice of control groups

Two control groups were included. One comprising persons from the general population who had not purchased any index drugs and a second of persons who had purchased antidiabetics, which allowed for an estimate of the possible effect of Berkson's bias¹² (an increased risk for persons diagnosed with one disorder of receiving additional diagnoses). However, we did not find indications of Berkson's bias, as persons who were diagnosed with diabetes did not have increased risk of getting diagnosed with a second disorder (such as, PD) compared with the unexposed control group. Diabetes has been shown to cause both diffuse and focal changes in the brain, especially if inadequately regulated,¹³ but to the best of our knowledge, diabetes has not been associated with the development of PD.

Possible biases

A number of potential biases have been avoided because of the study design and access to the comprehensive Danish registries, but it should be noticed that only patients who received prescriptions were included.

The risk of selection bias and information bias, particularly recall bias, is minimised. Data are routinely collected by electronic registration of all prescriptions and sales of medicinal products. The registries are nationwide and inclusive of the entire Danish population. The rate of emigration from Denmark is low (table 1​1).). None of the included drugs can be obtained without a prescription from a medical doctor. The treatment setting therefore is of no significance in data collecting. Medical treatment is tax financed with easy access and with a 75% reimbursement from the state on all prescriptions. Thus, the risk of bias attributable to socioeconomic factors can be regarded low.

Possible effects of antidepressants or lithium on the risk of developing PD

The basic pathophysiological feature of PD is reduced availability of dopamine in substantia nigra but the aetiology of PD remains elusive. Neuroprotective as well as neurotoxic drug effects should be considered. Available data show that both antidepressants and lithium have neuroprotective effects^14,15,16,17 with the strongest evidence for lithium while no evidence points to neurotoxic effects.