The Effect of Repeated Ketamine Infusion Over Facial Emotion Recognition in Treatment-Resistant Depression: A Preliminary Report

Open-label study of six ketamine infusions administered over 12 days followed by a 4-week follow-up period was conducted at the Minneapolis VA Medical Center.

Subjects aged 18 to 70 years had recurrent Major Depressive Disorder by a Structured Clinical Interview for DSM-IV and moderate to severe depressive symptoms [Hamilton Depression Rating Scale-17-item (HDRS-17) ≥14]. TRD was defined by the lack of improvement from current depressive episode after at least two adequate antidepressant trials of different pharmacological classes. The Antidepressant Treatment History Form⁵ (rating antidepressant trial ≥3; confidence rating ≥3) was used to assess antidepressant trials.

Patients were excluded due to cognitive impairment [Mini-Mental State Examination (MMSE) ≤26]; unstable medical illness, pregnancy, use of barbiturates or narcotics, current/lifetime diagnosis of post-traumatic stress disorder, psychosis, mania/hypomania; CNS related disorders (e.g., Parkinson's disease, severe head trauma); substance use disorder within 6 months or positive urine toxicology at baseline; active suicidal ideation. The dosages of medications used for depression including augmenting agents remained stable for at least 2 months prior to the beginning of and during the study period. The local Institutional Review Board approved the study, and written informed consent was obtained from all subjects.

Six IV infusions of ketamine hydrochloride solution (Mylan Inc, St. Paul, MN) at 0.5 mg/Kg were administered over 40 minutes on a Monday-Wednesday-Friday schedule over a 12-day period. Vital signs and pulse oximetry were recorded every 10 minutes for 1 hour beginning 10 minutes before infusion. Severity of depressive symptoms [Montgomery-Åsberg Depression Rating Scale (MADRS)],⁶ psychotomimetic symptoms (Brief Psychiatric Rating Scale), and dissociative effects (Clinician-Administered Dissociative States Scale) were measured 30 minutes before infusion, at the end of infusion (t₀+40 min), and again at t₀+100 minutes and t₀+160 minutes. Patients who responded after the last dose of ketamine were followed weekly for 4 consecutive weeks or until relapse was observed. Response was defined as ≥50% improvement from MADRS baseline score. Remission was established by a MADRS ≤9. Relapse was defined as <50% of baseline MADRS score at that follow-up visit.

Facial emotion recognition task (FERT) was administered at baseline and at each follow-up visit after completion of the six ketamine infusions. Facial images were taken from the Facial Expressions of Emotion: Stimuli and Tests⁷ and presented by Superlab Pro on a laptop computer. The task aims to identify facial emotions of happiness, fear, and sadness randomly presented in three blocks corresponding to each target emotion. Each block was composed by 36 neutral, and 42 emotional expressions (equally divided in 25% and 50% intensity) presented at 500 ms of exposure. Following completion of FERT, participants had a gender recognition task following the same procedures as FERT. This task served as an active experimental control for aspects unrelated to facial emotion recognition such as visual and attentional processing.

Linear mixed models using maximum likelihood estimation and an unstructured covariance structure were used to analyze change in MADRS scores over time. Mixed models examined the change in FERT performance over time with intercept, time, and other covariates as fixed effects. Logistic regression models analyzed the change in FERT performance as predictors of depression relapse following treatment.

The sample was composed by 15 men; average age of 52 years; onset of first major depressive episode at age 32; had at least three lifetime episodes of major depression; and had a HDRS-17 mean score=20.3 (SD 4.3) and at least three adequate antidepressant trials during current episode of 17 months; three had a past history of substance disorder; 11 had at least one psychiatric hospitalization; four had prior suicidal attempts; and two had received electroconvulsive therapy for depression. Out of 15 subjects, two dropped out after the first infusion because of dissatisfaction of expected ketamine’s therapeutic effects.

For the sample of 13 subjects, there was a statistically significant decreased in MADRS score (F (5, 65) = 37.8, p <0.0001) from 29.9 (SD=2.3) at baseline to 7.0 (SD=2.3) after the last infusion. Out of 13 subjects, 12 (92.3%) achieved response criterion and nine (69.2%) remitted. Six out of 12 subjects (50%) remain in response status throughout the 4 weeks of follow-up. The mean time for those six subjects who relapsed after the last infusion was 16 days (range 7–28 days). Four patients relapsed within 1 week after the last infusion.

There was a statistically significant difference in the perception of target emotions at 25% intensity [F(2, 1498)=16.05,p<0.0001] and 50% intensity [F(2,1498)=38.3,p<0.0001]. Overall, happiness was the easiest emotion to identify at both intensities (Table 1). We also found a statistically significant higher accuracy in congruent target-stimuli trials (e.g., target the recognition of fear with fearful faces as stimuli) at 25% [F(4,1498)=171.88, p <0.0001] and 50% intensity [F(4,1498)=14.33, p <0.0001]. Under congruent conditions, happiness again had the highest accuracy (Table 1). For neutral facial stimuli, there was a significant difference in the recognition of emotions [F(2,1498)=38.3,p <0.0001] with perception of sadness showing the lowest accuracy.

Reaction times were omitted as a measure of FERT performance because of a large variation to complete the task; however, we found no speed-accuracy trade-off (r=−0.11,p <0.001) during FERT performance. We also found a no significant difference in the accuracy of gender recognition by emotions at neutral, 25% or 50% intensity (data not shown).

Based on FERT baseline performance, accuracy to identify some emotional faces may have resulted in a ceiling effect minimizing the probability to detect improvement, if any, during treatment infusion. We chose accuracy in the congruent target-stimuli conditions of happiness, sadness and fear at 25% intensity as predictor variable for further analysis. FERT performance was not found to be a significant predictor of change in MADRS measure through six infusions [(fear:F(1,13)= 0.138,p=0.72; happy: F(1,13)= 0.982,p=0.34; sadness:F(1,13)= 0.047,p=0.83)].

We found no significant changes in repeated measure of emotion recognition over time after completion of infusions [sadness(F(4,539)=0.70,p=0.59; fear(F(4,539)=1.87,p=0.11;happiness (F(4,539)=0.22,p=0.93)]. We also found no significant interaction of accuracy recognition of emotions and depression relapse (MADRS >50% of baseline score) [sadness (Wald χ²(1)=0.39, p=0.53; fear (Wald χ²(1)=0.95, p=0.33; happiness (Wald χ²(1)=1.39, p=0.24)].