Substance use and suicidal thoughts, self-harm, and suicide attempts are both common among people identified as being at ultra-high risk (UHR) of psychosis, a group of mostly young people showing early warning signs that precede a small but meaningful proportion of first psychotic episodes. Whether those two facts are actually connected in this specific population, however, has not been clearly established.

This PRISMA-guided, PROSPERO-registered systematic review (CRD42024614852) searched Embase, MEDLINE, PsycINFO, and Web of Science for studies examining the relationship between substance use and suicidality in UHR populations, then had two reviewers independently extract data and assess study quality using the Mixed Methods Appraisal Tool (MMAT).

Twelve studies, reported across 13 papers and covering 2,198 participants, met the review’s inclusion criteria. Because the available studies varied too much in design and outcome measures to pool statistically, the authors used a narrative synthesis rather than a meta-analysis.

Across those 12 studies, researchers reported 40 separate analyses linking some measure of substance use to some measure of suicidality. Only 13 of those 40 analyses reached statistical significance, and the review notes that most of the significant findings were derived from a single study rather than replicated across multiple studies.

Of the substances examined, tobacco use showed the most consistent preliminary associations with suicidal ideation and self-harm. The review frames this as a preliminary signal rather than an established causal relationship.

Alcohol and cannabis were each linked to suicidal ideation and self-harm in some analyses, but not the majority, meaning the association appeared in certain studies or certain outcome measures without being a consistent finding across the evidence base.

Cannabis use appeared alongside alcohol as linked to ideation and self-harm in some analyses, without the review characterizing that link as strong or consistent.

Cocaine use was associated with suicidal ideation specifically, not with self-harm, suicide attempts, or suicide deaths, and that association appeared more clearly when researchers used continuous measures of cocaine exposure rather than simple yes/no measures.

When studies used composite substance-use variables, combining use of multiple substances into a single measure rather than analyzing each substance separately, those composite measures were largely unrelated to suicide outcomes.

This pattern suggests that if any real relationship exists between substance use and suicidality in this population, it may be substance-specific rather than a general effect of using drugs or alcohol broadly, though the review’s authors caution that the current evidence is not strong enough to confirm this pattern reliably.

The review’s authors identified major methodological limitations across the included studies, most notably a heavy reliance on cross-sectional data, which captures a single point in time and cannot establish which came first, the substance use or the suicidality, and the common use of dichotomous (yes/no) substance-use measures that cannot capture dose, frequency, or pattern of use.

Given these limitations and the low rate of replicated significant findings, 13 of 40 analyses, the authors concluded that current evidence is insufficient to establish a reliable association between substance use and suicidality in the UHR population, and called for rigorous longitudinal studies using validated, substance-specific measures.

Suicide risk assessment in early psychosis and UHR services already accounts for many established risk factors, including prior self-harm, depressive symptoms, and social isolation. Whether substance use belongs on that list, and how heavily it should be weighted, has been an open clinical question given how commonly substance use and early psychotic symptoms co-occur in this age group.

This review adds a rigorous but sobering data point to that conversation: despite substance use and suicidality both being prevalent in UHR populations, the current published evidence connecting the two directly remains sparse, inconsistent, and largely unreplicated. That does not mean no relationship exists, only that it has not yet been demonstrated with the kind of evidence needed to guide clinical practice confidently.

I read reviews like this one for what they quietly correct as much as for what they find. It would be easy to assume that substance use, including cannabis, meaningfully raises suicide risk in a population already known to be vulnerable, and to build clinical messaging around that assumption. This review is a reminder that assumption is not yet supported by strong evidence. Only 13 of 40 analyses reached significance, most from a single study, and the studies behind those findings were mostly cross-sectional, meaning they cannot tell us which came first.

For clinicians working with young people at ultra-high risk of psychosis, and for cannabis clinicians who may see this population as well, the honest clinical position right now is that we do not have reliable evidence connecting substance use, cannabis included, to suicidality in this specific group. That is not the same as saying there is no risk to monitor. Suicide risk in UHR populations is real and needs careful, individualized assessment regardless of substance use status. What this review tells us is that we should not lean on unproven claims about substance use as a suicide-risk marker until better, longitudinal, substance-specific research exists. Anyone working with a patient who is expressing suicidal thoughts should take that seriously and connect them with appropriate crisis and mental health support, independent of what this or any other single review does or does not show about substance use.

A Four-Step Reading Frame