|
Abstract
|
We read with interest the publication by Kridin and Ludwig1 investigating psychiatricsequalae from isotretinoin use in acne vulgaris. Using ICD codes and the TriNetX platform,authors extracted a cohort of patients prescribed isotretinoin and a propensity scorematched group prescribed oral antibiotics. They found no increase in the hazard ratio (HR) of depression, major depressive disorder, and suicidal attempts in the isotretinoin group, therefore concluding that isotretinoin does not increase the risk of suicide and depression.
We would like to commend and appreciate the authors for furthering knowledge in this controversial field. However, we herein highlight some concerns with the results and analysis.
First, baseline psychiatric co-morbidities were not considered. As physicians may be less willing to prescribe isotretinoin in patients with existing depression,2 the isotretinoin group may have a baseline lower prevalence and thus lower future risk of psychiatric events compared to the oral antibiotic group. This selection bias may be addressed through one of the following: limiting the inclusion criteria to those without existing psychiatric related ICD codes, performing subgroup or multivariable analysis to adjust for pre-existing psychiatric conditions.
Second, we are unclear about the propensity score matching process. Less relevant comorbidities such as heart disease and renal disease were included, while important covariates such as liver dysfunction, alcoholism, drug use and pregnancy were not. Additionally, an analysis of propensity matched cohorts should consider its 'paired' nature, such as through declaring the individual pairs as a random effect. Furthermore, given that some variables remained unbalanced after matching (e.g. smoking, unemployment), they should also be adjusted for in a multivariable cox regression.
A third concern is the long observation period. While the duration of observation is not explicitly mentioned, the survival curve shows time periods extending to almost 20 years of follow up. Isotretinoin is typically prescribed for months to short years,3 and psychiatric outcomes that occur many years after completion of treatment is unlikely medication related. An inappropriately lengthy follow up can 'dilute' the HRs/risk of any medication-related adverse effect. It is thus important to provide the 'number at risk' tables and risk35
ratios of each outcome at pre-defined earlier time points (e.g. 3 months, 6 months) after initiation of isotretinoin...
Language: en |