A meta-analytic investigation of problem gambling and self-harm: A causal inference perspective.

Problem gambling may be an underappreciated treatment target for reducing self-harm. Multivariate studies examining the relationship between problem gambling and self-harm have returned inconsistent results, perhaps due to insufficient power or differences in study quality.We conducted a series of meta-analyses examining the effect of problem gambling on self-harm outcomes of varying severity. We assessed the sensitivity of results to study characteristics, publication bias, and influential cases. To highlight threats to causal inference, we evaluated each study for residual confounding bias.In total, two types of meta-analyses were adequately powered: those assessing (a) the effect of problem gambling (binary) on suicide ideation and (b) problem gambling (binary) on suicide attempt (k = 14 unique studies between the two meta-analyses). For both meta-analyses, we found a positive, statistically significant average effect of problem gambling on the respective self-harm outcome. We did not detect any influential studies among these two meta-analyses, nor residual confounding. However, these findings are limited by the overall quality of included literature. PET-PEESE analysis detected substantial potential publication bias in the main results, with both meta-analytic effects being reduced to statistical nonsignificance. All included studies also used a cross-sectional design for their analysis of interest, thus simultaneity bias cannot be ruled out.While these findings suggest there is an appreciable relationship between problem gambling and at least two forms of self-harm (suicide ideation and attempts), well-designed longitudinal research is needed to make more firm conclusions. (PsycInfo Database Record (c) 2022 APA, all rights reserved).