Using classification and regression tree modelling to investigate treatment response to a single low-dose ketamine infusion: Post hoc pooled analyses of randomized placebo-controlled and open-label trials

Abstract

Background: Evidence suggests that clinical markers, such as comorbid anxiety, body weight, and others can assist in predicting response to low-dose ketamine infusion in treatment resistant depression patients. However, whether a composite of clinical markers may improve the predicted probability of response is uncertain. Methods: The current study investigated the results of our previous randomized placebo-controlled and open-label trials in which 73 patients with treatment-resistant depression (TRD) received a single ketamine infusion of 0.5 mg/kg. Clinical characteristics at baseline, including depression severity, duration of the current episode, obesity, comorbidity of anxiety disorder, and current suicide risk, were assessed as potential predictors in a classification and regression tree model for treatment response to ketamine infusion. Results: The predicted probability of a composite of age at disease onset, depression severity, duration of current episode, and obesity/overweight was significantly greater (area under curve = .736, p = .001) than that of any one marker (all p > .05). The most powerful predictors of treatment response to ketamine infusion were younger age at disease onset and obesity/overweight. The strongest predictors of treatment nonresponse were longer duration of the current episode and greater depression severity at baseline. Discussion: Depression severity, duration of the current episode, obesity, and age at disease onset may predict treatment response versus nonresponse to low-dose ketamine infusion. However, whether our predicted probability for a single infusion may be applied to repeated infusions would require further investigation. Clinical trial registration: UMIN Clinical Trials Registry (UMIN000023581 and UMIN000016985).