Substance use is common among people with major depressive disorder, and clinicians have expressed concern that co-occurring use might reduce response to neuromodulation treatments. Clinical guidance has at times discouraged offering transcranial magnetic stimulation (TMS) to patients with substance use, but empirical data addressing that concern are limited. This study used electronic health record data to test whether frequency of substance use is associated with antidepressant response to TMS, operationalized as change in PHQ-9 scores following a standard course of TMS.
Investigators extracted retrospective data from an academic medical center’s electronic health records on patients who received a standard course of TMS for major depressive disorder. The analytic dataset comprised 219 TMS treatment courses.
Substance use frequency was recorded for multiple categories: alcohol, cannabis, nicotine, stimulants, benzodiazepines (prescription), opioids, inhalants, psychedelics, and other drugs. The primary clinical outcome was change in Patient Health Questionnaire-9 (PHQ-9) score from pre-TMS to post-course.
Covariates included pre-TMS PHQ-9 score, patient age, sex, and number of TMS sessions received. The study was conducted under institutional review board approval (Vanderbilt University Medical Center IRB #251256) with a waiver of informed consent due to the retrospective design.
The authors used analysis of covariance (ANCOVA) and multiple regression models to assess whether substance use frequency predicted change in PHQ-9 score. Models controlled for pre-TMS PHQ-9 score, age, sex, and number of TMS sessions. Additional regression comparisons examined individual levels of use (for example, daily use compared with no use) within substance categories.
Results report both statistical significance testing (p values) and effect-size estimates (Cohen’s d) to frame the magnitude of any observed associations.
Data were extracted from 219 TMS courses. The most frequently reported substances were:
Other substance categories (cannabis, nicotine, opioids, inhalants, psychedelics, and other drugs) were also captured in the record abstraction, though the report focuses on overall associations across categories rather than reporting prevalence for each remaining class in detail.
Across all substance categories analyzed, the presence or frequency of substance use was not associated with change in PHQ-9 score following a standard course of TMS. All comparisons returned p values greater than 0.05. Reported effect sizes (Cohen’s d) ranged from 0.00 to 0.30, indicating small or negligible effects where differences were observed.
When individual levels of use were compared in multiple regression models (for example, daily use versus no use), level of substance use was similarly not associated with change in PHQ-9 score; again, all reported p values exceeded 0.05. The authors examined the range of plausible effects and found that estimates of the impact of substance use frequency on PHQ-9 change were generally below the minimal clinically important difference for the PHQ-9.
Although statistical testing alone can indicate whether an association is unlikely to be due to chance, the authors additionally interpreted findings relative to the minimal clinically important difference for the PHQ-9. The magnitude of plausible effects of substance use frequency on PHQ-9 change was generally smaller than that minimal threshold, leading to the interpretation that low to moderate substance use is unlikely to have a meaningful clinical impact on antidepressant response to TMS.
In this retrospective, single-center analysis of 219 TMS treatment courses for major depressive disorder, low- to moderate-level substance use was not associated with reduced antidepressant response to TMS. The authors conclude that low-level substance use should not by itself be used as an exclusion criterion for TMS treatment in patients with depression.
The study protocol received IRB approval from Vanderbilt University Medical Center (IRB #251256) with a waiver of informed consent because of its retrospective design. The authors declared no competing interests. Data may be available on reasonable request, as indicated in the source report.