The relationship between dissociation and antidepressant effects of esketamine nasal spray in patients with treatment-resistant depression

Chen and colleagues situate this post-hoc analysis within an ongoing debate about whether the transient dissociative effects produced by N-methyl-D-aspartate receptor (NMDAR) antagonists such as ketamine are causally linked to their rapid antidepressant effects. Earlier studies of intravenous ketamine in treatment-resistant depression (TRD) have reported mixed findings: some found correlations between dissociation and symptom improvement, while others did not. Esketamine nasal spray (ESK) has demonstrated antidepressant efficacy in several Phase III trials, but comparable analyses examining the relationship between dissociation and antidepressant response for ESK were lacking. The present study therefore aimed to evaluate whether clinically meaningful dissociative symptoms after ESK dosing are associated with, or mediate, antidepressant outcomes. Using data from large Phase III trials of ESK in TRD, the investigators carried out responder, correlation and causal mediation analyses to test relationships between peak Clinician-Administered Dissociative States Scale (CADSS) scores after dosing and changes in Montgomery-Åsberg Depression Rating Scale (MADRS) scores at defined post-dose timepoints. The analysis emphasises results from the short-term flexible-dose trial (TRANSFORM-2) and the randomized withdrawal maintenance trial (SUSTAIN-1), with complementary TRANSFORM-1 data presented in supplemental material, because those trials produced statistically significant primary efficacy results.

This is a post-hoc analysis of three randomized, double-blind, active-controlled, multicentre Phase III trials of esketamine nasal spray: TRANSFORM-1 (fixed-dose, n=346), TRANSFORM-2 (flexible-dose, n=227) and SUSTAIN-1 (maintenance, n=297 for randomized maintenance phase). In TRANSFORM-1 and TRANSFORM-2 patients received adjunctive oral antidepressant (AD) therapy plus either ESK (56 mg or 84 mg, fixed or flexible) or placebo nasal spray, administered twice weekly during a 4-week double-blind induction phase. SUSTAIN-1 enrolled patients who had achieved stable response or remission on ESK + AD during an optimization phase and randomised them to continue ESK + AD or switch to placebo nasal spray in a variable-duration, relapse-prevention maintenance phase; dosing frequency in SUSTAIN-1 was individualised (weekly or every other week) based on MADRS scores. Depressive symptoms were measured using the MADRS total score; dissociative symptoms were measured with the CADSS total and component scores recorded pre-dose, 40 minutes post-dose (Tmax), and 1.5 hours post-dose on treatment days. A CADSS total score >4 was predefined as indicating clinically meaningful dissociation (based on prior healthy participant ranges); additional sensitivity analyses used cutoffs of ≤2 and ≤8. Primary analyses reported here focus on TRANSFORM-2 and SUSTAIN-1, with TRANSFORM-1 outcomes in supplementary material because TRANSFORM-1's primary efficacy comparison was not statistically significant for one dosage arm. Statistical approaches included responder analyses (≥50% reduction in MADRS), Spearman correlation coefficients (with 95% CIs) between peak post-dose CADSS and MADRS change at specified timepoints, and causal mediation analysis to partition direct and indirect (mediated via CADSS) treatment effects. The mediation framework used ANCOVA models adjusting for region/country, class of oral AD (SNRI vs SSRI) and baseline MADRS; indirect and direct effects were estimated using parametric bootstrapping via the R "mediation" package. For SUSTAIN-1 an approach following Lange et al. was used (R package "timereg") to assess whether CADSS change on the first maintenance day mediated time to relapse. Temporal profiles of CADSS and MADRS over repeated dosing were estimated using mixed models for repeated measures (MMRM). The investigators note these are post-hoc analyses and specify covariates included in models.

Responder analyses in TRANSFORM-2 showed no significant differences in antidepressant response rates between patients who did versus did not exhibit clinically meaningful dissociation (CADSS >4) after the first or last ESK doses. At day 2, 12% of patients with CADSS ≤4 and 21% with CADSS >4 showed a ≥50% MADRS reduction (p=0.182). By day 28, responder rates were 70% (CADSS ≤4) versus 69% (CADSS >4) after the initial dose (p=0.854). Considering dissociation after the final ESK dose (day 25), 68% (CADSS ≤4) versus 71% (CADSS >4) were responders at day 28 (p=0.721). Similar distributions were reported for TRANSFORM-1 in supplementary analyses. Correlation analyses produced Spearman coefficients close to zero and statistically nonsignificant. In TRANSFORM-2, correlations between CADSS after the first ESK dose and MADRS change were r = -0.05 (95% CI: -0.23, 0.14) at day 2 and r = -0.08 (95% CI: -0.27, 0.12) at day 28. The correlation between CADSS after the last dose (day 25) and MADRS at day 28 was r = -0.16 (95% CI: -0.36, 0.04). Analyses of CADSS components, including depersonalisation, likewise showed no significant correlations. Causal mediation analyses found insufficient evidence that dissociation mediated antidepressant effects. In TRANSFORM-2 the indirect (mediated) effects had 95% CIs that included zero, whereas the direct effect of ESK on MADRS change—independent of CADSS—was statistically significant: LS mean differences were -3.62 (95% CI: -6.62, -0.74) after the first dose and -4.83 (95% CI: -8.37, -1.28) after the last dose. SUSTAIN-1 mediation analysis similarly showed no mediation of time to relapse by CADSS change (indirect effect estimate 0.12, CI including zero), while the direct effect of continued ESK on delaying relapse was significant (LS mean difference -2.44; 95% CI: -4.04, -0.84). Temporal analyses demonstrated that peak CADSS scores declined across consecutive ESK doses (main effect of study day and day × treatment interaction p<0.0001), whereas the antidepressant advantage of ESK + AD over AD + placebo on MADRS persisted through the 4-week induction phase. The treatment difference in MADRS did not change significantly between day 2 and day 28 (interaction p=0.5341). Average CADSS scores were generally <4; in TRANSFORM-2 only 3.8% of patients had CADSS >4 at day 1 and 4.2% at day 25. The authors additionally report that CADSS magnitudes after ESK nasal spray were substantially lower than those typically reported after a single IV ketamine infusion (median CADSS ~6–8 vs ~25), and sensitivity analyses using alternative CADSS thresholds (≤2 and ≤8) produced the same null association.

Chen and colleagues interpret the convergent findings from responder, correlation, mediation and temporal analyses as evidence that dissociative symptoms elicited by esketamine nasal spray are neither necessary for, nor predictive of, antidepressant benefit in patients with TRD within the dose range studied. They contrast these results with prior mixed findings from IV ketamine studies, noting that some earlier reports identified correlations between CADSS subscales and later antidepressant response whereas others did not. The authors highlight several factors that may explain differences between studies of IV ketamine and ESK nasal spray: stereochemical differences, route and formulation (IV versus nasal), dosing frequency (single infusion versus twice-weekly dosing), concomitant oral antidepressants in the ESK trials, timing of clinical assessments, and differences in the magnitude and variance of CADSS scores that affect statistical sensitivity. They note that CADSS peak magnitudes after intranasal ESK were much lower on average than those reported after standard IV ketamine, which could reduce the power to detect an association. The discussion also summarises mechanistic considerations: while NMDAR antagonism is implicated in both dissociative and antidepressant effects, distinct downstream mechanisms may separate the two phenomena; a recent preclinical hypothesis implicating HCN1 in dissociation is mentioned but not resolved. Limitations acknowledged by the authors include the post-hoc nature of the analyses, timing of assessments that precluded examining some temporal relationships (for example an isolated association between dissociation after a single dose and antidepressant response several days later without intervening doses), potential noise from placebo/nocebo effects and small treatment effect sizes, and limited sensitivity to detect weak correlations in individual study samples. They note that combined analyses across studies were conducted to mitigate sample size limitations and that sensitivity analyses using different CADSS cutoffs and subcomponents corroborated the null findings. Clinically, the investigators suggest dose selection for ESK should prioritise antidepressant efficacy and tolerability rather than attempting to induce dissociative symptoms, since dissociation was neither necessary nor sufficient for response. Finally, they call for additional research to clarify the potentially distinct mechanisms by which NMDAR antagonism produces antidepressant and dissociative effects.