Economic evaluation of subcutaneous ketamine injections for treatment resistant depression: A randomised, double-blind, active-controlled trial - The KADS study

Major depressive disorder carries substantial clinical and economic burden, and roughly 30% of patients do not respond adequately to standard antidepressant or psychosocial treatments, creating a need for effective interventions for treatment resistant depression (TRD). Previous clinical research has evaluated racemic ketamine and esketamine (S-ketamine), with esketamine receiving regulatory approval in several countries for TRD, but large Phase III trials of racemic ketamine are fewer. Economic evaluations to date have mainly modelled intranasal esketamine or compared esketamine to intravenous ketamine, often extrapolating benefits far beyond trial durations; none had specifically evaluated subcutaneous racemic ketamine within a trial setting. Chatterton and colleagues report the planned within-trial cost-utility analysis embedded in the KADS randomised, double-blind, active-controlled Phase III trial. The study set out to assess the cost-effectiveness of subcutaneous racemic ketamine versus midazolam (an active control intended to improve blinding) for adults with TRD from both an Australian health sector perspective and a societal perspective, using trial data on resource use, quality of life and productivity over a 4-week treatment period with a 4-week follow-up (total 8 weeks).

KADS was a multicentre, randomised, double-blind, parallel-group Phase III trial conducted at six Australian specialist mood disorder centres and one New Zealand centre. Adults aged over 18 with Major Depressive Disorder of at least 3 months' duration, insufficient response to at least two adequate antidepressant trials, stable concomitant antidepressant dosing for ≥4 weeks, and a baseline MADRS score ≥20 were eligible. Randomisation used permuted blocks and allocation was concealed from study personnel except trial statisticians, the trial pharmacist and the data safety monitoring board. Participants received subcutaneous injections twice weekly for four weeks of either racemic ketamine hydrochloride or midazolam hydrochloride. The trial initially used fixed doses (Cohort 1) but was amended to a response-guided, flexible dosing schedule (Cohort 2) permitting escalation for non-responders; the economic evaluation focused primarily on Cohort 2 because of its larger sample size and superior antidepressant efficacy of ketamine in that cohort. The economic evaluation was a pre-specified, within-trial cost-utility analysis undertaken from Australian health sector and societal perspectives and followed CHEERS guidelines. Intervention and monitoring costs were estimated using micro-costing and presented in 2018–19 Australian dollars. Base-case 1 included costing both ketamine and midazolam; base-case 2 excluded the cost of midazolam on the rationale that it was a protocol-driven control cost. Intervention components included an initial psychiatrist consultation, medication, dispensing fees, and staff monitoring time. Health service use outside the trial, transportation, carer time and productivity losses (absenteeism and presenteeism) were captured via a self-reported resource use questionnaire at baseline, week 4 (end of RCT) and week 8, with optional linkage to administrative MBS and PBS records for some participants. Unit costs were drawn from Australian sources; lost work hours were valued using an average wage rate plus on-costs. The primary outcome for the economic analysis was quality-adjusted life years (QALYs) over 4 and 8 weeks, derived from AQoL-8D utility scores using area under the curve. Statistical analyses were performed in Stata 17 on a modified intention-to-treat (mITT) basis including participants who received at least one treatment; the base-case focused on Cohort 2 (flexible-dose). Missing cost and utility data (30% missing RUQ, 8% missing AQoL-8D in Cohort 2) were addressed by a three-step approach: descriptive assessment of missingness, multiple imputation with chained equations (40 imputations), and sensitivity analyses including complete-case analyses. Cost differences were estimated using generalised linear models (GLM) with gamma family and log link (adjusted for baseline costs), and QALY differences using GLM with gaussian family and identity link (adjusted for sex, age and baseline utility). Incremental cost-effectiveness ratios (ICERs) were calculated as mean cost differences divided by mean QALY differences for each base-case, perspective and follow-up period, with 95% confidence intervals obtained by nesting multiple imputation within 1000-bootstrap iterations. Ketamine was considered cost-effective if ICERs fell below an Australian willingness-to-pay (WTP) threshold of $50,000 per QALY gained. A range of one-way sensitivity analyses examined the impact of additional covariates (study site), alternative valuation methods for absenteeism/presenteeism, use of MBS/PBS administrative data, complete-case analyses, varying medication and dispensing prices, additional psychiatrist consultations, and actual (recorded) monitoring times. A multivariate sensitivity analysis explored joint uncertainty in key intervention cost parameters, including initial consultation fees, wholesale medication prices and nurse-to-patient monitoring ratios.

Of 181 randomised participants, three withdrew consent and 174 adults (mean age 45, SD 15) received at least one allocated dose and were analysed in the mITT population. Cohort 1 (fixed-dose) comprised N = 73 and showed no difference in average QALYs at 4 or 8 weeks; subsequent reporting therefore concentrated on Cohort 2 (flexible-dose; N = 106 for the economic analyses reported). Using micro-costing, the average intervention cost per participant over 4 weeks in Cohort 2 was $975 (SE $27) for ketamine and $923 (SE $26) for midazolam. Additional health care use and lost productivity costs were collected and analysed; cost differences by category between ketamine and midazolam were not statistically significant in aggregate tests (p > 0.05). From the health sector perspective, base-case 1 (including midazolam costs) produced lower average health sector costs for ketamine at end of RCT and over 8 weeks (negative mean differences), though these differences were not statistically significant. In base-case 2 (excluding midazolam costs), average health sector costs were significantly higher for ketamine at end of RCT (adjusted mean difference $1005; 95% CI $350 to $1984) and over 8 weeks (adjusted mean difference $868; 95% CI $33 to $2089). The higher costs in base-case 2 principally reflect assigning zero intervention cost to the control arm. Considering societal costs (healthcare plus productivity, transport and carer time), base-case 1 showed higher total costs for ketamine at end of RCT (adjusted mean difference $978; 95% CI -$762 to $4469) and over 8 weeks (adjusted mean difference $1167; 95% CI -$1124 to $5324), though these were not statistically significant. In base-case 2, societal costs were significantly higher for ketamine at end of RCT (mean difference $2040; 95% CI $263 to $5489) and remained higher but non-significant over 8 weeks (mean difference $2233; 95% CI -$102 to $6412). Health-related quality of life (AQoL-8D) utilities were significantly higher in the ketamine group at end of RCT (adjusted mean difference 0.096; 95% CI 0.040 to 0.152) and numerically higher at 8 weeks (adjusted mean difference 0.025; 95% CI -0.015 to 0.065, not statistically significant). Adjusted mean QALY differences favoured ketamine at both time points: 0.004 QALYs at end of RCT (95% CI 0.002 to 0.006) and 0.008 QALYs over 8 weeks (95% CI 0.003 to 0.014). Cost-effectiveness results varied by base-case and perspective. Under the health sector perspective, ketamine was dominant (lower cost and improved health) in base-case 1 at both end of RCT and over 8 weeks. In contrast, base-case 2 produced ICERs that exceeded the $50,000 per QALY threshold: at 4 weeks $251,250 (95% CI $268,715 to $6,531,253) and over 8 weeks $108,500 (95% CI $49,306 to $3,609,179). From the societal perspective, ketamine increased QALYs but at higher costs in both base-cases, and ICERs exceeded the WTP threshold. Probabilities of ketamine being cost-effective at a $50,000/QALY threshold showed wide variation: under the health sector perspective and base-case 1 the probability was 89% at end of RCT and 91% over 8 weeks, whereas in base-case 2 these probabilities dropped to 0.2% and 2.2% respectively. Under the societal perspective, probabilities were 30% and 32% for base-case 1, and 0% and 4.6% for base-case 2. Sensitivity analyses generally did not materially change conclusions for base-case 1. Sub-analyses using MBS/PBS administrative data included 36 participants (34% of Cohort 2) and produced a slightly higher but non-significant health sector cost and no significant QALY difference. Complete case analyses included 81 participants for health sector and 74 for societal perspectives. In base-case 2, using wholesale hospital medication prices rendered health sector and end-of-RCT societal cost differences non-significant; alternative monitoring time measures and dispensing fee assumptions affected statistical significance of societal and health sector costs. Adding an additional weekly psychiatrist consultation increased costs significantly in favour of higher ketamine costs. A multivariate sensitivity analysis yielded a dominant health sector ICER over 8 weeks (95% CI: dominant to $1,127,748) with a 67% probability of being cost-effective under those joint parameter assumptions.

Chatterton and colleagues interpret their findings as indicating that subcutaneous racemic ketamine is likely to be cost-effective from an Australian health sector perspective when the control (midazolam) intervention and monitoring costs are included for both arms (base-case 1). They emphasise that when midazolam costs are excluded (base-case 2) — reflecting the view that midazolam is a protocol-driven blinding agent rather than a clinically indicated comparator — ketamine is not cost-effective because intervention-related costs are higher and ICERs exceed common WTP thresholds. The authors suggest several mechanisms for observed improvements in the midazolam group, including non-specific psychosocial support arising from frequent staff contact and monitoring or a Hawthorne effect of being observed within a trial, which complicates choice of comparator and interpretation of within-trial cost-effectiveness. From the societal perspective, ketamine did not represent value for money in this trial, driven in part by slightly higher measured productivity losses in the ketamine group and wide confidence intervals owing to the small number of participants reporting work productivity effects. These trial-based results are placed in context with prior modelled economic evaluations of esketamine, which generally found esketamine unlikely to be cost-effective under health sector or payer perspectives; the present study complements those findings while highlighting the importance of formulation, dosing and monitoring intensity. The authors note that racemic ketamine drug costs were lower than esketamine but that intensive one-on-one, protocol-driven monitoring required in a Phase III trial increased total intervention costs. They recommend that extrapolating trial results to longer horizons (for example 5 years) is an important next step but beyond the current analysis. Key limitations acknowledged by the investigators include the short follow-up and brief 4-week administration period, trial-specific intensive monitoring that may overestimate real-world costs, reliance on self-reported resource use and productivity with attendant missing data (mitigated by multiple imputation and sensitivity checks including MBS/PBS linkage), and limited statistical power for cost outcomes. Strengths they highlight are the pre-planned economic analysis embedded within the RCT, collection of AQoL-8D utility data to estimate QALYs, use of a societal perspective, and extensive sensitivity analyses to explore parameter uncertainty. The authors conclude that longer-term evaluations comparing racemic ketamine with current best practice for TRD are warranted.

Subcutaneous racemic ketamine is likely to represent value for money from a health sector perspective in treatment resistant depression when the costs of midazolam and trial monitoring are assigned to both arms. Further research evaluating longer-term costs and effects of racemic ketamine versus best current practice for TRD is recommended.