Real-world effectiveness of ketamine in treatment-resistant depression: A systematic review & meta-analysis

Martinotti and colleagues frame treatment-resistant depression (TRD) as a common, severe condition with large health and societal costs and a poor response to conventional monoaminergic antidepressants. They note growing evidence implicating glutamatergic dysfunction in TRD and outline how ketamine and its S-enantiomer, esketamine, act as NMDA receptor antagonists and engage downstream mTOR/BDNF signalling and neuroplasticity. While intravenous ketamine has demonstrated efficacy in TRD and esketamine has shown benefit in randomized trials, the authors highlight a gap in knowledge about esketamine's safety and effectiveness in routine, unselected clinical settings because many RCTs exclude patients with common comorbidities and complex medical histories. This study therefore aims to evaluate the real-world effectiveness and safety of intranasal esketamine administered alongside standard antidepressants in a multicentre Italian sample of adults with TRD. The primary objective was to document change in depressive symptoms over one- and three-month follow-ups; a secondary objective was to characterise tolerability and adverse events in routine care.

The REAL-ESK study is an observational, retrospective, multicentre study that analysed 116 adults (61 female, 55 male; mean age 50 ± 12 years) treated with intranasal esketamine under an early-access programme across numerous Italian mental health centres. Eligibility required age ≥18, a current major depressive episode meeting TRD criteria (failure to achieve ≥50% symptom reduction after at least two adequate monoaminergic antidepressant trials, as judged by a psychiatrist), and concurrent treatment with an SSRI or SNRI consistent with national regulatory indications. Patients with absolute contraindications to esketamine per the Italian regulator (for example untreated arterial hypertension or prior cerebrovascular disorders) were excluded. Clinical and anamnestic data were retrospectively extracted from medical records at baseline (T0), one month (T1) and three months (T2). Clinician-rated scales used were the Montgomery–Åsberg Depression Rating Scale (MADRS), the 21-item Hamilton Depression Rating Scale (HAM-D-21) and the 21-item Hamilton Anxiety Rating Scale (HAM-A). The investigators also recorded treatment history, augmentation strategies (mood stabilisers, antipsychotics, benzodiazepines), comorbidities and clinically relevant events including discontinuation, relapse or remission. The study received local ethics approval and adhered to confidentiality and Helsinki Declaration principles. Sample size was estimated with G*Power for a repeated-measures ANOVA assuming an expected response of 40%, α = 0.05 and power = 95%, and allowing for a 20% rate of premature dropout. Statistical analyses used SPSS 20.0. Continuous outcomes are reported as mean ± SD; categorical data as counts and percentages. Changes from baseline were tested with paired Student t-tests and categorical comparisons used Pearson χ2 tests. Analyses were two-tailed with significance set at p < 0.05. The extracted text does not provide a detailed, prespecified statistical modelling plan (for example, mixed-effects models or intention-to-treat handling) beyond the described tests.

A total of 116 patients entered the study; by T1 ten patients (8.62%) had discontinued (seven for inefficacy, two for excessive side effects, one for severe psychomotor agitation), leaving 106 patients for the one-month analysis. By T2 a further five patients (4.31%) had discontinued and ten patients had not yet reached T2 at data cut-off, yielding 91 patients in the three-month analysis. At baseline the cohort was clinically severe: mean MADRS 35 ± 8.53 and long illness duration (mean 19 ± 11.05 years). Lifetime antidepressant trials averaged 3.28 ± 1.89. Personality disorders were present in 15% and substance use disorders in 6%; 64% had no other psychiatric comorbidity. Many patients were receiving augmentation: 56.9% on a mood stabiliser and 57.7% on an antipsychotic in addition to antidepressant therapy. Depressive symptoms declined significantly after treatment. Mean MADRS fell from 35 ± 8.53 at T0 to 22.27 ± 9.81 at T1 (t = 15.79, df = 95, p < 0.0001) and to 14.69 ± 9.88 at T2 (t = 18.07, df = 81, p < 0.0001). Suicidal ideation (MADRS item 10) also decreased significantly: mean 2.13 ± 1.58 at T0, 1.00 ± 0.55 at T1 (t = 9.12, df = 95, p < 0.0001) and 0.94 ± 0.10 at T2 (t = 8.64, df = 81, p < 0.0001). Clinical response (≥50% reduction on MADRS or HAM-D-21) occurred in 33 patients (28.4%) at T1 and in 68 patients (64.2%) at T2 (χ2 = 12.69, df = 1, p < 0.0001). Remission rates (MADRS <10 or HAM-D-21 <7) were 11.2% (13 patients) at T1 and 40.6% (43 patients) at T2 (χ2 = 12.43, df = 1, p < 0.0001). Notably, only 29% of T2 remitters had been remitters at T1; many T2 remitters had not responded at one month. Adverse events were consistent with prior trials. Dissociative symptoms were recorded in 39.7% of patients, sedation in 28.4% and transient hypertension in 10.3%. Manic symptoms occurred in 2.6%, psychomotor agitation in 1.7%, anxiety in 2.6% and headache in 2.6%; 27.6% reported no side effects. Severe adverse events prompting discontinuation occurred in three patients at T1 (2.58%), including one case of severe psychomotor agitation. The authors report no observed signs of abuse, withdrawal or long-term cognitive, urogenital or hepatic toxicity within the study period. Comparisons by comorbidity status showed no significant differences in response or remission between patients with and without psychiatric comorbidities (T1 response p = 0.121; T1 remission p = 0.339; T2 response p = 0.741; T2 remission p = 0.257). However, patients receiving augmentation with mood stabilisers or antipsychotics had lower response rates (T1 p = 0.023; T2 p = 0.010). The extracted text notes an increase in the proportion receiving the 84 mg dose from 26.5% at T1 to 38.5% at T2, but a full dosing schedule and timing are not clearly reported in the extraction.

Martinotti and colleagues interpret their findings as supporting the effectiveness and tolerability of intranasal esketamine in a heterogeneous, clinically severe, treatment-seeking TRD population treated in routine care. They emphasise that, compared with RCT samples, the study cohort had higher rates of psychiatric comorbidity, longer illness duration, higher unemployment and more frequent augmentation therapies, yet achieved three-month response (64.2%) and remission (40.6%) rates comparable to randomized trials. A key observation discussed is the marked increase in response and remission between one and three months: many patients who remitted at three months had not responded at one month. The authors argue this suggests the induction phase should not be the sole time window for evaluating esketamine benefit in routine practice. They propose several, non-mutually exclusive explanations: greater baseline illness severity and chronicity in their sample may require longer exposure for neuroplastic changes (mTOR/BDNF-mediated synaptic potentiation) to translate into clinical improvement; conservative dose escalation in real-world practice may delay therapeutic effect; and lower adherence outside trial settings could blunt early responses. Regarding safety, the investigators report no novel adverse signals and rates comparable to RCTs, with infrequent manic switches and no observed addictive behaviours during follow-up. They note lower effectiveness among patients on mood stabilisers or antipsychotics and discuss two possible causes: greater baseline severity in those patients or pharmacodynamic interactions that diminish esketamine's antidepressant effect — a topic they recommend for further study. The authors acknowledge important limitations: the open-label, retrospective design without placebo or active comparator, absence of inter-rater reliability assessments across centres, and the unstructured method for recording side effects (data extracted from clinical records rather than standardised instruments). They contend the study's strengths include its multicentre nature and use of an unselected clinical sample, which they present as complementary to RCT data by illustrating outcomes in real-world practice and populations typically excluded from trials. The investigators recommend cautious interpretation and call for further research to clarify dosing strategies, the impact of concomitant psychotropic medications and longer-term safety and effectiveness.