Assessing measures of suicidal ideation in clinical trials with a rapid-acting antidepressant

Rapid-acting antidepressants such as the NMDA receptor antagonist ketamine can reduce depressive symptoms within hours, and emerging work has suggested they may also reduce suicidal thoughts rapidly. However, most suicide assessment instruments were developed for longer treatment timelines, and there are unresolved methodological questions about measuring suicidal ideation in studies of fast-acting interventions. Key concerns include whether single items drawn from depression scales suffice instead of longer suicide-specific instruments, whether commonly used measures are sensitive to change over hours to days, and whether repeated questioning over a short interval affects validity. Ballard and colleagues used data from two double-blind, randomized, placebo-controlled, crossover ketamine trials to address these issues empirically. By administering several clinician-rated and self-report suicide measures repeatedly over minutes to days after infusion, the study aimed to compare convergent validity across instruments and evaluate their sensitivity to rapid change, with the goal of informing measurement choices in trials of rapid-acting anti-suicidal treatments.

Data were pooled from two inpatient, randomized, crossover, placebo-controlled trials of ketamine for treatment-resistant depression (both unipolar major depressive disorder and bipolar I/II depression) conducted at the NIMH. Sixty participants were included (reported in Results). Diagnosis was confirmed with the SCID-I/P. Eligible participants were aged 18–65, in a current major depressive episode without psychotic features, with minimum severity thresholds (HAM-D 17-item score >= 18, or MADRS >= 20 or 22 at screening and pre-infusion). Patients were medication-free for at least two weeks prior (five weeks for fluoxetine), except bipolar participants who remained on therapeutic lithium or valproate. One of the trials excluded acutely suicidal patients at consent defined as MADRS item 10 > 4 or by clinical judgment; participants whose suicidality increased during taper or trial were not systematically excluded. Ketamine hydrochloride 0.5 mg/kg was administered intravenously over 40 minutes and saline was used as placebo. Four instruments assessed suicidal ideation: the clinician-rated HAM-D 17-item (single suicide item scored 0–4), clinician-rated MADRS (single suicide item scored 0–6), the self-reported Beck Depression Inventory (BDI) suicide item (0–3), and the clinician-administered Scale for Suicide Ideation (SSI). For the SSI the researchers analysed both the first five items (SSI5) and the full 19-item total score (SSITotal). Baseline assessment occurred 60 minutes pre-infusion (covering the last 24 hours). Post-infusion assessments were at 40, 80, 120, and 230 minutes, and Days 1, 2, and 3, with reporting windows covering symptoms since the last assessment. Analyses focused on both static convergent validity at 230 minutes and Day 3 and sensitivity to change from baseline to those time points. Cut-offs for 'any suicidal ideation' were predefined: HAM-D >= 1, BDI >= 1, MADRS >= 2, or SSI5/SSITotal >= 2. Pearson correlations assessed agreement at single time points and for change scores. Linear mixed models evaluated trajectories across seven post-infusion time points, including fixed effects for time, intervention, their interaction, and a fixed intercept; models were restricted to participants endorsing any ideation at baseline on the specific measure and controlled for baseline ideation. A compound symmetry covariance structure and restricted maximum likelihood estimation were used. Cohen's d was calculated for differences between ketamine and placebo at 230 minutes and Day 3. As a post-hoc analysis, the earliest time point showing a significant drug–placebo difference was determined. IBM SPSS v21 was used and significance was set at p < .05, two-tailed.

Sixty participants were analysed: 23 with MDD and 37 with BD. The sample was 62% female (n = 37) with a mean age of 41.6 years (SD 11.3). Lifetime suicide attempts were reported by 48% (n = 28) and 18% (n = 11) reported more than one attempt. Mean illness duration was 25.3 years. At the 230-minute and Day 3 time points, correlations between the suicide assessments were generally high. All pairwise correlations exceeded r = .60, and correlations among the HAM-D suicide item, MADRS suicide item, and SSI5 were particularly strong (r > .80). The BDI item and the SSITotal showed lower correlations overall. Correlations of change from baseline to 230 minutes and Day 3 were all above r = .40, with the strongest change correlation observed between SSI5 and the BDI; correlations between the BDI and other single-item clinician measures were lower. Linear mixed models examining trajectories across the seven post-infusion time points showed significant drug effects of ketamine on suicidal thoughts for all measures except SSITotal (p < .01 for the significant measures). Effect sizes across 230 minutes and Day 3 were largest for the MADRS suicide item, the BDI suicide item, and SSI5, indicating these measures were more sensitive to change. In a post-hoc analysis the first time point at which ketamine differed from placebo was 40 minutes for all assessments (p < .05). Given concerns about ketamine's psychotomimetic effects at 40 minutes, the 80-minute time point was also examined; at 80 minutes all assessments except SSITotal showed a significant ketamine–placebo difference. The SSITotal results diverged from other measures; one participant with an SSITotal score > 20 was identified as an extreme outlier. Excluding that participant did not render the SSITotal model significant (drug effect F(1,314) = 2.02, p = .14; interaction F(6,285) = 1.15, p = .33).

Ballard and colleagues interpret their findings to indicate that single-item suicide ratings embedded within depression scales can converge closely with a brief clinician-rated suicide measure, and that some short measures detect rapid reductions in suicidal thoughts after ketamine. In particular, the HAM-D and MADRS suicide items correlated strongly with the first five SSI items, and the MADRS item, the BDI suicide item, and the SSI5 showed moderate sensitivity to rapid change as reflected by effect sizes. The longer SSI total score was less sensitive to short-term change and showed reduced agreement with other measures, a discrepancy the authors attribute to administration variability: the SSI proceeds from five to 19 items only when initial responses are positive, potentially introducing heterogeneity and extra variance. Differences between self-report (BDI) and clinician-rated items were also noted as a possible source of measurement variance. The investigators highlight that trajectories of suicidal ideation over three days differed significantly between ketamine and placebo for most scales, and that no evidence emerged for an iatrogenic increase in suicidal thoughts due to repeated assessment. Several limitations are acknowledged. The sample consisted of treatment-resistant depressed inpatients not selected for acute suicide risk, limiting generalisability to actively suicidal populations. Assessment timeframes varied across the three days so the 230-minute and Day 3 assessments are not strictly comparable, and longer follow-up (for example seven days) was not consistently available in this dataset. All measures relied on patient-reported or clinician-rated items; no implicit measures or biological markers of suicide risk were collected. The authors note that suicidal thoughts are a proxy for suicide risk and that lack of validated biomarkers remains an obstacle. Based on these results, the authors suggest that brief measures such as the MADRS suicide item, the BDI suicide item, and the first five SSI items may be particularly useful for trials of rapid-acting interventions because they balance brevity with sensitivity to change. They recommend further studies that enrol participants for acute suicidality, incorporate additional suicide-specific instruments (for example the Suicide Status Form or Columbia Suicide Severity Rating Scale), and explore implicit or neurobiological measures alongside self-report and clinician-rated tools to better characterise rapid changes in suicide risk.

With the emergence of rapid-acting treatments for suicidal risk, assessment instruments that are quick to administer yet sensitive to change over hours to days are needed. The present analysis indicates that single items from depression scales correlate with longer suicide assessments, and that the MADRS suicide item, the BDI suicide item, and the first five items of the SSI may be particularly well suited to detecting rapid changes in suicidal ideation in clinical trials.