Abstract
Cookey et. al.’s paper 1 highlights the complex presentations of patients seen in early first-episode psychosis clinics with comorbid substance abuse. Previous studies, as the article suggests, tended to focus on cannabis use only or primarily on treatment outcomes. By approaching these patients from a broader perspective, the authors offer insight on the clinical characteristics of patients with comorbid alcohol-use disorder as well. The authors present several significant findings for their subgroups of patients. However, the way that the Alcohol, Smoking, and Substance Involvement Screening Test (ASSIST) was used to assign patient to groups as well as some of the results presented raise concerns about the clinical utility and validity of the findings.
The paper cited 2 to reduce threshold values from >11 to >4 for alcohol and >4 to >2 for cannabis reports at best a marginally satisfactory sensitivity/specificity of 0.79/0.64 for cannabis and 0.73/0.62 for alcohol, respectively. 2 At these lowered thresholds, consuming any amount of alcohol weekly or consuming cannabis once or twice in a 3-month period would satisfy the criteria but would not represent pathological or clinically significant conditions. 3 We acknowledge that using the original ASSIST values would have resulted in smaller group sizes and loss of between-group differences. However, we would like clarification on the pathological criteria used to differentiate substance use from dependence and caution the conclusions arising from data with the lower ASSIST threshold.
Furthermore, using multiple statistical tests, the authors increase the risk of false positive discovery. The Tukey–Kramer adjustments are too liberal and typically require a large sample and are therefore inappropriate. We suggest using Benjamini–Hochberg 4 adjustments to control for alpha inflation due to multiple tests, which is more rigorous and appropriate in this instance. Among the significant findings, that is, Positive and Negative Syndrome Scale-positive and Social and Occupational Functioning Assessment Scale (SOFAS) scores, one notes that they amount to a Cohen’s d of 0.68 and 0.7, respectively, which suggests that the study was only powered to find large effect sizes. Therefore, the study is prone to both type I and II errors. The negative findings may only be nonsignificant due to low statistical power, as such they should never be discussed as bona fide findings.
Notably, a meta-analysis by Koskinen et al. has already reported similar findings with younger schizophrenia patients with cannabis use disorder and older with alcohol use disorder. 5 The authors discuss the association of anxiety and better social and occupational functioning among alcohol users in the context of previous studies that have already explained possible reasons for these novel associations. However, with the possibility of both type I and II errors, we question what novel findings the study is introducing following corrections to the data analysis.
Highlighting the role of alcohol and cannabis in the early phase psychosis population is extremely important to further understand the complex impact of substance use on development and treatment of early phase psychosis. However, the methods of sample subdivision, and the questionable data analytic strategy, cast some doubts on the value of the reported findings.
