Statistics from Altmetric.com
We read with interest the study by Villumsen et al on the risk of Parkinson’s disease (PD) in patients with IBD.1 This study took advantage of the long series of nation-wide data from 1977 to 2014 in the Danish Population and Patient Registers. For each person ever living in Denmark during this period, the register data include dates of birth, immigration, emigration, death, diagnosis with IBD and diagnosis with PD. This allows for a cohort design with accumulation of person-years and disease outcomes for the entire population divided by exposure status. In the present case, the exposed cohort would include person-years and PD cases after an IBD diagnosis, and the non-exposed cohort would include the remaining person-years and PD cases in the dataset. Poisson regression could be used to estimate the rate ratio for exposed versus non-exposed observations.
Given these possibilities, it is surprising that many studies choose to sample the person-years in a case–control design instead of using the entire dataset. The sampling will inevitably imply less statistical power, and sampling of controls is complicated and might easily go wrong. The study by Villumsen et al is an example of where the sampling in our opinion went wrong. The authors started by dividing the entire dataset from 1977 to 2014 into those ever diagnosed with IBD and those never diagnosed with this disease. Then, they used the first group as their exposed cohort, and sampled persons from the second group to constitute their non-exposed cohort, and calculated the HR of PD between the exposed and the non-exposed cohorts.
In the Villumsen et al study, the date of diagnosis of each IBD case was used as the matching date in selection of controls, and PD diagnoses prior to these matching dates were excluded from the analysis. This procedure implies that all PD cases after IBD are included in the analysis, but part of the PD diagnoses made in non-IBD-exposed person-years are excluded from the analysis. Villumsen et al have therefore slightly overestimated the risk of PD following an IBD diagnosis. Pitfalls like this can be avoided by using the full dataset.
Contributors EL wrote the first draft and MSP and SNL edited. All authors approved the final letter.
Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
Competing interests None declared.
Provenance and peer review Not commissioned; internally peer reviewed.
Patient consent for publication Not required.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.