Post-Marketing Safety Concerns with Esketamine: A Disproportionality Analysis of Spontaneous Reports Submitted to the FDA Adverse Event Reporting System

Gastaldon and colleagues introduce esketamine nasal spray as an adjunctive treatment for adults with treatment-resistant depression (TRD), noting its US FDA approval in March 2019 and subsequent approval by the European Medicines Agency. Regulatory authorities required a Risk Evaluation and Mitigation Strategy (REMS) because of concerns about misuse, dissociation and sedation. The authors point out that safety evidence available at approval came mainly from pre-marketing trials and pooled analyses, which showed higher rates of dissociation and lower acceptability versus placebo, but that rare or delayed adverse events (AEs) may not be captured by randomized controlled trials. Given these gaps, the study aims to examine post-marketing safety signals for esketamine using the FDA Adverse Event Reporting System (FAERS). The intent was to detect and characterise potential real-world AEs that might be rare, unexpected, or differ from those reported in regulatory trials, thereby informing further research and regulatory monitoring.

The investigators used a case/non-case disproportionality design applied to FAERS data, which mirrors a case-control approach: 'cases' were reports listing esketamine as the suspect drug, while 'non-cases' comprised reports for all other drugs in the database during the study period. Disproportionality compares observed versus expected counts of specific AEs among esketamine reports to identify signals suggestive of an association. Two statistical disproportionality measures were computed for all AEs with at least four reports: the Reporting Odds Ratio (ROR) and the Bayesian information component (IC), each with 95% confidence intervals. Conventional thresholds were applied (lower 95% CI > 1 for ROR and > 0 for IC), and a signal was considered present when both measures met significance. Analyses were performed using R packages PhViD and BCPNN. Main analyses compared esketamine to all other drugs in FAERS; sensitivity analyses restricted comparisons to venlafaxine for disease-related AEs to partially control for confounding by indication. The study also compared reports classified as serious versus non-serious across several variables, including age, sex, weight, esketamine dose, time-to-onset, antidepressant polypharmacy, and co-prescription of mood stabilisers, antipsychotics, benzodiazepines or somatic medications. AEs were grouped into four categories for interpretation: (a) expected AEs with a detected signal, (b) expected AEs without a detected signal, (c) disease-related AEs for which TRD itself is a risk factor, and (d) unexpected AEs. Non-parametric tests and chi-squared statistics were used for group comparisons where reported. The extracted text does not clearly state the exact study time window here, though other sections refer to the period around March 2019–March 2020.

The extracted text reports multiple safety signals and descriptive findings from FAERS. A time-to-onset listing given in the Methods-text shows counts and mean onset times for several commonly reported events: vomiting (n = 74; mean onset 14.3 days), depression (n = 65; 3.6 days), suicidal ideation (n = 64; 8.3 days), anxiety (n = 63; 11.5 days), increased blood pressure (n = 62; 10.6 days), dizziness (n = 56; 15.7 days), product dose omission (n = 45; 32.6 days), and feeling abnormal (n = 43; 8.4 days). These represent examples from a longer list provided in the supplementary material. Disproportionality analyses identified signals across the predefined AE categories, including somatic complications such as increased blood pressure, vomiting, nausea, ataxia, sensory disturbance and hypertension; full ROR and IC values were reported in supplementary tables but are not reproduced in the extracted text. Sex-stratified analyses found broadly similar safety signals in males and females, with a few exceptions: decreased therapeutic effect remained significant in both sexes, whereas drug ineffectiveness and dizziness lost significance in females and hospitalisation lost significance in males in the stratified tables referenced. Comparing serious versus non-serious reports revealed statistically significant differences. Mean esketamine dose was higher in serious versus non-serious reports (68.84 ± 15.69 mg vs 63.97 ± 17.25 mg; p = 0.001), and receiving 84 mg was associated with a higher frequency of serious AEs than 56 mg (χ2 = 5.79, p = 0.016). Females were more likely than males to feature in serious reports (χ2 = 125.29, p < 0.001), and higher weight was also associated with serious AEs (p = 0.043); age and time-to-onset did not differ between groups. Specific events more often reported as serious included completed suicide, dissociation, increased blood pressure, falls, hospitalisation, hypertension, loss of consciousness, sedation, and suicidal ideation; events more often non-serious included anxiety, device malfunction, feeling abnormal, drug inefficacy, dysgeusia, nausea, and product dose omission (χ2 = 817.8, p < 0.001). Reports classed as serious were more frequently associated with antidepressant polypharmacy (χ2 = 9.08, p = 0.003) and co-prescription of mood stabilisers (χ2 = 8.14, p = 0.004), benzodiazepines (χ2 = 19.48, p < 0.001), antipsychotics (χ2 = 25.62, p < 0.001), or somatic medications (χ2 = 16.79, p < 0.001). Sensitivity analyses using venlafaxine as comparator for disease-related AEs retained signals for self-injurious ideation, suicidal ideation, emotional disorder, depression, crying and depressed mood, but not for mental disorder, completed suicide, or suicide attempt. In the Discussion the authors note that the relative reporting ratio for suicidal ideation was approximately 24-fold higher versus other drugs and about 5–9 times higher versus venlafaxine; these figures are presented in the paper as indicative of the magnitude of the signal rather than as causal risk estimates.

Gastaldon and colleagues interpret the FAERS analysis as the first real-world pharmacovigilance assessment of esketamine and conclude that the drug shows clear potential for serious and some unexpected AEs. They highlight four main observations: a rapid increase in the number of esketamine reports over the study window, detection of rare AEs not seen in regulatory trials, very large disproportionality estimates for several events (notably dissociation, sedation and 'feeling drunk'), and a high reporting ratio for suicidal ideation. On reporting trends, the authors point out that reports per month rose from a mean of about 65 in April–December 2019 to about 123 in the first three months of 2020, effectively doubling. They consider several explanations: a true rise in events, a Weber effect (reporting peak after approval), or notoriety bias linked to regulatory attention and REMS, but argue that notoriety alone is unlikely to account for the magnitude and persistence of the signals. Regarding rare and unexpected events, the investigators list occurrences such as self-injurious ideation, logorrhoea, panic attacks, paranoia, ataxia, akathisia, autoscopy and mania, and recommend these be followed up in analytical studies. The discussion situates the disproportionality findings relative to pre-marketing evidence: for example, the strong signal for dissociation aligns with pooled trial data showing substantially elevated dissociation risk in esketamine arms. The authors stress that signals for suicidal and self-injurious ideation remained even when comparing to venlafaxine, although signals for completed suicide and attempts did not; they caution that TRD itself raises baseline suicide risk, complicating causal interpretation. Potential mechanisms posited for increased suicidal ideation include baseline illness severity, withdrawal effects, dissociation or intoxication-like reactions, delayed psychiatric reactions, or lack of efficacy, but these are presented as hypotheses rather than established causes. Limitations are acknowledged in detail: pharmacovigilance data cannot establish causality, suffer from under-reporting and variable report quality, lack denominators so incidence cannot be calculated, and are vulnerable to confounding by indication and channelling bias since esketamine is used in severe TRD. Notoriety bias related to REMS is also possible. The authors note that they attempted to mitigate confounding with sensitivity analyses against venlafaxine and used shrinkage methods to reduce false positives; nevertheless, they call for validation of signals through prospective studies, pragmatic trials, observational cohorts and case-control investigations. Finally, implications for practice are discussed: clinicians should prescribe esketamine cautiously, particularly for patients already at risk of suicide, monitor for misuse potential and rare psychiatric reactions, and be attentive to vulnerable subgroups such as females and patients on polypharmacy. The authors recommend extended monitoring after administration and suggest that, if validated, labelling should be updated to clarify timing and risks of the identified AEs.

The authors conclude that this FAERS pharmacovigilance analysis identified new and unexpected safety signals for esketamine, including serious AEs with rapid onset that were more frequently reported in females, at higher doses, and in patients receiving multiple concomitant medications. They state these findings reinforce existing safety concerns and call for urgent clarification of the increased reporting of suicidal ideation through further post-marketing surveillance and targeted research, so that a reasonable risk–benefit assessment can be established.