Wording Errors in Abstract, Numeric Errors in Results, and Labeling Errors in Figure 2
JAMA Dermatology · 2019
In Reply We welcome the opportunity to discuss the comments by Chen et al regarding our systematic review and metaanalysis of suicidality in patients with atopic dermatitis (AD).In our study, we found a significantly higher risk of suicidal ideation and suicide attempts in those with AD compared with those without AD. 1 The letter by Chen and colleagues contains 2 errors.First, the authors stated that "15 of the reviewed studies included control groups of patients with different conditions."However, as stated in our article, 15 was the total number of studies in the systematic review.Of these, only 3 contained control groups of patients with other conditions 2-4 ; the vast majority of the studies had control groups that consisted of healthy people or those from the general population.Second, Chen et al state that the association was incorrectly estimated by merging crude odds ratios (ORs) with adjusted ORs.However, we did not merge crude ORs with adjusted ORs.As stated in our article, we used all crude ORs in our metaanalyses, in accordance with established methods.5 Of the studies that only reported adjusted ORs, we calculated crude ORs from the raw data provided.Chen and colleagues express concerns regarding combining heterogeneous control groups.In our meta-analysis, most studies contained controls that were not heterogeneous populations with skin diseases but rather healthy individuals or those from the general population.In comparison, it is rather commonplace in dermatology where controls are those with other skin diseases.2-4 We discussed these limitations in the original article. 1 Most importantly, we performed the meta-analyses again using only studies with healthy controls or control groups of the general population to compare these results to our original findings.For suicide attempts, we found that the strength of association was nearly identical to that of the original metaanalysis (OR, 1.39; 95% CI, 1.12-1.71vs OR, 1.36; 95% CI, 1.09-1.70,respectively).For suicidal ideation, we found that the strength of association was only slightly higher than what was reported in our original meta-analysis (OR, 1.59; 95% CI, 1.38-1.83vs OR, 1.44; 95% CI, 1.25-1.65,respectively).Thus, excluding patients with other diseases from the control group did not change the original conclusions.Chen and colleagues also expressed a concern regarding the use of crude ORs instead of adjusted ORs.However, using crude ORs is an established and commonplace method 6 that allows investigators to combine studies to increase the power.Regarding adjusted ORs, the studies in our meta-analysis adjusted for different confounders.These confounders are not interchangeable or equivalent.In fact, it has been recommended not to combine adjusted ORs in a meta-analysis where the adjusted factors were different among the studies.5 Thus, the alternative analytical methods proposed by Chen et al are not applicable for this study.In conclusion, the original findings were robust.The additional meta-analyses including only studies with controls who were healthy or from the general population confirmed the original findings and did not change the conclusions of the article.