Prospective association of psychological pain and hopelessness with suicidal thoughts

Suicidal thoughts and behaviours remain a major public health problem, and earlier research has shown limited improvement in predicting suicidal behaviour despite repeated evaluation of similar risk factors. Contemporary suicide theories adopt an "ideation-to-action" framework that treats suicide ideation (SI) as distinct from suicidal behaviour and emphasise the need to identify short-term antecedents that precipitate transitions into states of heightened suicidal thinking. Identifying reliably predictive short-term markers is challenging because SI fluctuates and it is difficult to prospectively observe who will experience re-emergent SI over brief intervals. Ballard and colleagues used clinical trials of the glutamatergic agent ketamine as a model system to study short-term antecedents of SI, because ketamine often produces rapid but transient reductions in SI that then reappear within days. The study tested whether the interaction of two psychological constructs strongly linked to suicide risk—hopelessness (negative expectations about the future) and psychological pain (also termed psychache)—would prospectively predict re-emergence of SI after ketamine. As a secondary aim, the investigators examined whether other depressive symptom clusters (anhedonia, depressed mood, impaired sleep) showed similar prospective relationships with SI.

Data were pooled from the active treatment arms of four inpatient, randomized controlled trials conducted at the NIH Clinical Center between 2006 and 2018 (overall n = 108). The trials included participants with treatment-resistant major depressive disorder (MDD) or bipolar disorder (BP); three trials were crossover ketamine versus saline and one was a parallel trial of ketamine plus riluzole versus ketamine plus placebo (ketamine open-label). Only participants aged 18–65 in good physical health who received a ketamine infusion, completed relevant rating scales (MADRS, HAM-D, BDI), and had non-zero ratings on at least one primary outcome at two or more timepoints were included. Most participants were medication-free before randomisation (minimum washout intervals specified); those in the BP study could continue therapeutic lithium or valproate. Psychiatric diagnoses were established with the SCID for DSM-IV. Exclusion criteria included pregnancy, breastfeeding and medical contraindication to ketamine. All participants had provided informed consent and the protocol was IRB-approved. Suicidal ideation was measured using the MADRS suicidal thoughts item (0–6). Hopelessness was indexed by BDI item 2 (discouraged about future; 0–3). Psychological pain was represented by an empirically derived Negative Cognitions score composed of selected items from the MADRS, HAM-D and BDI that conceptually overlap with psychache (examples listed in the paper include pessimistic thoughts, guilt, thoughts of failure, self-criticism, worthlessness). The interaction of hopelessness and psychological pain was modelled as their product. Secondary, empirically derived scores captured Anhedonia, Depressed Mood and Impaired Sleep. All target variables were rescaled by their maximum possible scores to aid interpretation. Assessment timepoints included 40, 80, 120 and 230 minutes post-infusion and Days 1, 2, 3, 7 and 11 depending on the substudy. Two analysis sets were pre-specified: a short-term analysis (n = 105) using 230 min, Day 1, Day 2 and Day 3, and a long-term analysis (n = 45, RIL study only) using 230 min, Day 3, Day 7 and Day 11. Only active-treatment data (no saline/placebo arms) were included. Because the research question concerned within-subject longitudinal relationships among correlated constructs, the investigators used a random intercept cross-lagged panel model (RI-CLPM). The RI-CLPM separates stable between-subject differences from within-subject fluctuations and uses autoregressive paths to model temporal stability; a positive autoregressive parameter indicates that a value above an individual's mean at one timepoint tends to predict a value above that mean at the next timepoint. Evidence for the primary hypothesis would be a strong cross-lagged path from hopelessness/psychological pain to later SI combined with a weak reverse path. Multiple-group modelling across substudies was intended but did not converge, so the primary analyses pooled studies and constrained lagged and autoregressive parameters to equality over time to reduce model complexity. Missing data were assumed missing at random and were not imputed. The investigators avoided dichotomous notions of "statistical significance," reporting raw p-values and 95% confidence intervals in line with current guidance. Structural equation modelling was performed in Mplus Version 8.4 with processing via MplusAutomation in R version 4.0.2.

Participant-level descriptives are provided in the paper's Table (extracted text does not reproduce the table). As an exploratory index of within-subject stability, the intraclass correlation coefficient (ICC) for SI was 0.71 (95% CI 0.66, 0.76) for the short-term dataset and 0.62 (95% CI 0.51, 0.72) for the long-term dataset, indicating moderate within-subject consistency across timepoints. Short-term follow-up (n = 105): The simplified RI-CLPM showed adequate fit. Autoregressive parameters for SI and for the hopelessness/psychological pain construct were moderate (SI 95% CI: 0.026, 0.47; hopelessness/psychological pain 95% CI: -0.001, 0.49), implying only modest carry-over of within-person elevations from one timepoint to the next. Crucially, the cross-lagged parameters of substantive interest—hopelessness/psychological pain predicting later SI and SI predicting later hopelessness/psychological pain—were weakly positive and did not differ from each other (t = -0.15, p = 0.89). Thus, in the short-term data there was no evidence that the interaction of hopelessness and psychological pain prospectively predicted next-day SI. Long-term follow-up (n = 45): Model fit was reported as good, although chi-square issues prevented estimation of some fit indices. The SI autoregressive parameter was similar to the short-term result (95% CI: -0.027, 0.507). By contrast, the autoregressive parameter for hopelessness/psychological pain indicated that within-person elevations tended to persist across time (95% CI: 0.189, 0.851). The cross-lagged parameters did not support the hypothesised direction: hopelessness/psychological pain was not associated with later SI (95% CI: -0.185, 0.293), whereas SI moderately predicted later hopelessness/psychological pain (95% CI: 0.078, 0.678). The difference between those paths trended toward conventional thresholds (t = -1.81, p = 0.07) but the authors emphasise confidence intervals and raw p-values rather than dichotomous significance claims. Secondary outcomes: Models for Anhedonia, Depressed Mood and Impaired Sleep fitted the data well. For each, autoregressive parameters were moderately strong, but neither the path from the secondary symptom to later SI nor the reverse path from SI to later secondary symptom differed from zero. In summary, none of the examined depressive symptom clusters—including the hopelessness-by-psychological-pain interaction—predicted next-day SI in these post-ketamine datasets, while in the longer-term dataset SI predicted modest subsequent increases in hopelessness/psychological pain.

Ballard and colleagues interpret these negative prospective findings to indicate that, in this dataset of participants who received ketamine, the co-occurrence of hopelessness and psychological pain did not serve as an early warning predictor of next-day re-emergence of SI. The authors note that other depressive symptom clusters (depressed mood, anhedonia, impaired sleep) likewise failed to predict short-term changes in SI, which limits the ability of standard rating-scale measures to identify imminent SI in this context. Several explanations are offered. First, the setting is a pharmacological trial in which ketamine produces rapid, often transient reductions in SI and depressive symptoms; therefore, the temporal dynamics of SI after ketamine may differ from naturalistic fluctuations and from the patterns on which psychological theories of SI were developed. Second, the apparent association in the long-term data whereby SI predicted later hopelessness and psychological pain could reflect the distress caused by the reappearance of SI itself—qualitative reports cited by the authors include participants who feel "bereft" when SI returns and one quoted saying, "…hopefully this is going to get better with longer treatment, but if it doesn't it, you know, it makes me think that, what's the point of living"—suggesting that the subjective reaction to SI re-emergence may elevate negative cognitions. The authors highlight parallels with ecological momentary assessment (EMA) studies, which also show large within-day variability of SI and limited success in predicting short-term changes using constructs such as hopelessness or depressed mood. They therefore suggest that SI variability itself may be an important construct to study and that future work could focus on digital phenotyping or more fine-grained sampling to disentangle treatment response, emotional reactions to symptom return, and everyday variability in SI. Limitations are emphasised: this was a secondary analysis using proxy measures rather than dedicated scales for hopelessness or psychological pain; rating schedules were not consistently daily across studies, reducing datapoints; the sample was a convenience clinical-trial sample with no a priori sample-size calculation; a relatively small sample necessitated simplifying constraints to the RI-CLPM, and an underpowered structural equation model may produce unstable results. Given these constraints, the authors advise caution in interpreting the modest finding that SI predicted later hopelessness/psychological pain and they recommend replication with more targeted measures and denser sampling. Clinically, they underscore the importance of close monitoring of individuals with a history of SI in the days following ketamine administration because SI may re-emerge quickly and with little warning.