Anti-suicidal effects of IV ketamine in a real-world setting

This was a retrospective analysis of de-identified clinical data collected from a community IV ketamine clinic in Chicago between January 2018 and December 2020. After exclusion of patients not treated for depressive symptoms, those without a baseline Quick Inventory of Depressive Symptomatology-Self Report (QIDS-SR), and those who received only one infusion, the analytic sample comprised 295 outpatients diagnosed with a primary mood, anxiety or trauma‑related disorder per DSM-5 criteria. The institutional review board at Baylor College of Medicine provided a waiver for secondary analysis of the anonymised data. Patients received an acute induction course of IV ketamine administered three times weekly (Monday-Wednesday-Friday), with most courses comprising six to eight infusions determined by patient and physician. Infusions were given over 40–60 minutes by a board-certified physician; the initial dose was weight‑based (0.50 mg/kg) and adjusted to achieve a tolerated, mild dissociative effect. Ondansetron was provided as needed for nausea. Symptom measures included the QIDS-SR for depressive symptoms and the 14‑item Concise Health Risk Tracking Self‑Report (CHRT‑SR) for suicidality; both were administered prior to each infusion. The Childhood Trauma Questionnaire (CTQ) was completed at the initial clinic visit to index childhood maltreatment across five domains, and a maltreatment load score (0–5) was derived. The CHRT‑SR covers propensity (pessimism, helplessness, perceived social support, despair), impulsivity, and active suicidal ideation; item scores range 0–4 for a maximum total of 56. The primary analytic approach was growth mixture modelling (GMM) to identify latent trajectory groups in CHRT scores across the first six clinic visits. Both linear and quadratic growth patterns were tested with 1–4 class solutions (eight models in total). Time intervals between infusions were included as a covariate to account for individual variation. Model selection used entropy, Bayesian Information Criterion (BIC), and likelihood ratio testing (BLRT); the adjusted Lo–Mendell–Rubin test was reported but not used when non-significant. After trajectory identification, groups were compared on demographic and clinical variables using paired t‑tests for within-group change, analysis of variance (ANOVA) and chi-square tests for between-group comparisons, and Bonferroni corrections for pairwise tests. Analyses were performed using R, SAS and Mplus.

The final sample numbered 295 patients; over half were male (approximately 59.6%) and the most common clinical diagnosis was treatment‑resistant depression (TRD), present in 83.4% of patients. Prior suicide attempts were reported by 4.4% and alcohol use disorder by 1.36%. On the CTQ, clinically significant childhood physical neglect and emotional neglect were reported by about 29.5% and 26.1% of patients, respectively, with sizeable proportions also reporting physical (24.1%) or emotional (33.9%) abuse. Growth mixture modelling indicated that quadratic growth models fit the data better than linear models. Within quadratic specifications, the BLRT supported 2‑ and 3‑class solutions but not a 4‑class model, and practical interpretability led to selection of a three‑class quadratic model as optimal. The three identified trajectories of CHRT total scores across six visits were: (1) Moderate Baseline CHRT Total, Gradual Improvement (MC‑GI), n=170 (57.6%), baseline mean CHRT‑Total 20.08 (SD=8.37) with a statistically significant gradual decrease by the end of treatment (p < .001; Cohen's d=1.08); (2) Severe Baseline CHRT Total, No Improvement (SC‑NI), n=63 (21.4%), baseline mean 34.21 (SD=6.99) with little change through Visit 6 (Visit 6 mean 33.66, SD=8.32; t = -0.03, p = .511, d=0.10); and (3) Severe Baseline CHRT Total, Rapid Improvement (SC‑RI), n=62 (21.0%), baseline mean 31.10 (SD=6.79) with a large and rapid reduction to mean 9.05 (SD=7.36) by the end of treatment (t=26.40, p < .001; d=3.93). After three infusions, the SC‑RI group’s CHRT score was lower than that of the MC‑GI group. Similar three‑pattern trajectories emerged for the CHRT subscales of propensity, impulsivity and active suicidal ideation. Group comparisons at baseline showed no significant differences in primary diagnosis, age, gender, weight or ketamine dose (mg/kg) across the three trajectories. CTQ maltreatment experiences were broadly similar, although a smaller proportion of the SC‑NI group had low CTQ maltreatment (65.12%) compared with the MC‑GI group (84.11%; p=0.024). Baseline depression severity on the QIDS‑SR was lower in the MC‑GI group (14.87 ± 0.35) than in the SC‑RI (18.40 ± 0.50) and SC‑NI (19.37 ± 0.45) groups (corrected p < .001). The MC‑GI group also had the lowest baseline suicidal propensity (14.38 ± 0.49) and impulsivity (3.45 ± 0.16) versus SC‑RI (22.59 ± 0.63; 5.13 ± 0.24) and SC‑NI (24.76 ± 0.62; 4.29 ± 0.28) (both corrected p < .001). Baseline CHRT Risk scores were highest in the SC‑NI group (5.16 ± 0.31), followed by SC‑RI (3.38 ± 0.35) and MC‑GI (2.25 ± 0.19), with significant pairwise differences (corrected p-values reported).

O'brien and colleagues interpret their findings as evidence that IV ketamine administered in a community clinical setting is associated with at least some anti‑suicidal effects for a majority of patients: 79% of the sample fell into trajectories showing either gradual or rapid improvement in suicidality, while a distinct 21% subgroup with severe baseline suicidality showed no benefit across five infusions. The three trajectory patterns were consistent across overall CHRT scores and the subdomains of propensity, impulsivity and active ideation, suggesting coherent symptom change profiles. The authors highlight the study’s novelty in assessing anti‑suicidal effects of IV ketamine in a naturalistic, community‑based treatment setting using an empirically validated suicidality measure. They note a potentially important clinical signal: patients presenting with the highest baseline CHRT Risk scores (for example, risk scores ≥ 4) did not improve and may have worsened, so such individuals may warrant close monitoring during and after treatment. Several limitations are acknowledged. The analyses were retrospective and not pre‑specified at data collection, and the dataset was not originally designed to evaluate suicidality trajectories; this may limit statistical power and the stability of identified classes. The study relied on self‑report measures, lacked a placebo or controlled comparison to address expectancy effects, and was missing key sociodemographic variables (years of education, ethno‑racial status, primary language, socioeconomic status), which restricts generalisability. The authors therefore caution that larger, prospectively designed studies are needed. For future research, the investigators recommend replication and investigation of biological and behavioural markers that might distinguish rapid responders from non‑responders, with the clinical aim of improving treatment planning and reducing the risk associated with persistently elevated suicidal ideation after an acute course of IV ketamine.