A nationwide population study of severe head injury and Parkinson's disease

4.1. Strengths 

Most reported studies were based on retrospective analyses, which may have a systematic recall bias. Because most injuries occurred decades ago verification of the injuries was often difficult. Our approach was prospective which strengthens the study. This study comprised a national population, thus selection bias was avoided and a high statistical power was obtained, which also strengthens the study.

The latency between head injury and PD diagnosis has been discussed in other studies. Most of the retrospective studies that found an association between head injury and PD, observed a long latency between head injury and disease [4], [6]. The present study was designed with a 10-year latency period and a 12–24-year follow-up for PD. Some prior studies have had even longer follow-up periods but that might cause dilution of the effect. However, since Goldman et al. [4] reported 20–30 mean years between head injury and PD, the follow-up period in the present study can be considered relatively short in brief.

In prior studies the risk of PD has been associated with severity of head trauma [4]. In the present study, we only included injuries that were severe enough to require hospitalisation.

We found no association between head trauma and PD during the follow-up period, thus change in mobility due to PD does not seem to have increased the risk for head injury during the years preceding PD diagnosis (no effect–cause bias). Diagnosis of PD might occur in connection with the treatment of the head injury (detection bias). This was avoided due to the defined latency period.

4.2. Weaknesses 

Prospective cohort studies with exposure assessment before onset of PD might be subject to registration bias, when case finding is based on hospital contacts. Patients outside medical care, most likely with early or mild disease, are not registered, and about 30–40% of all PD patients are treated by medical specialists outside the hospital system. However, most new cases of PD including patients treated at medical specialists are brain scanned in the hospital system. We have therefore no reason to believe that underreporting has given rise to bias.

All people in Denmark who were at least 20 years old as on 1 January 1981 were followed for hospitalisation due to head trauma during 1981–1993. Exposure to head injury was considered during several years, and therefore misclassification with regard to lifetime history of head injuries was reduced. The critical time period during which patients are at risk of PD is unknown, and therefore whether early, late or average lifetime exposures should be studied is unknown [2]. We had no information about persons younger than 20 years. Therefore we can only make inference about head injuries in adulthood. Neither did we have any information on head injuries prior to 1981. Based on simulations we found that this loss of information is very unlikely to have biased our results significantly towards unity. It is, however, known that there is an inverse relationship between smoking and PD [10] but the reason for that is still open for discussion. It is possible that head injury subjects are more likely to be smokers for instance because they are more risk prone. If so our result may be biased slightly toward unity.

The hospital register did not contain any information on outpatients prior to 1995. Hence we could not exclude people who had PD prior to the start of the follow-up. This was, however, not a major problem. Since our aim was to investigate relative rates we did not have any interest in the absolute incidence rates. The important thing is that the follow-up of people at hypothetical risk due to head injury was performed in exactly the same way as it was in the reference population.

4.3. Previous research 

Head injury is an inconsistently reported risk factor for PD. Goldman et al. [4] carried out a review of 11 epidemiological studies of head injury and PD and found relative risks between 0.6 and 6.2. Four studies showed no association between head trauma and PD. Most of the reported studies were inconclusive.

4.4. Conclusion 

Our result suggests that there is no association between severe head injury among adults and subsequent PD.