Percutaneous Revascularization for Ischemic Left Ventricular Dysfunction: Cost-Effectiveness Analysis of the REVIVED-BCIS2 Trial

BACKGROUND: Percutaneous coronary intervention (PCI) is frequently undertaken in patients with ischemic left ventricular systolic dysfunction. The REVIVED (Revascularization for Ischemic Ventricular Dysfunction)-BCIS2 (British Cardiovascular Society-2) trial concluded that PCI did not reduce the incidence of all-cause death or heart failure hospitalization; however, patients assigned to PCI reported better initial health-related quality of life than those assigned to optimal medical therapy (OMT) alone. The aim of this study was to assess the cost-effectiveness of PCI+OMT compared with OMT alone. METHODS: REVIVED-BCIS2 was a prospective, multicenter UK trial, which randomized patients with severe ischemic left ventricular systolic dysfunction to either PCI+OMT or OMT alone. Health care resource use (including planned and unplanned revascularizations, medication, device implantation, and heart failure hospitalizations) and health outcomes data (EuroQol 5-dimension 5-level questionnaire) on each patient were collected at baseline and up to 8 years post-randomization. Resource use was costed using publicly available national unit costs. Within the trial, mean total costs and quality-adjusted life-years (QALYs) were estimated from the perspective of the UK health system. Cost-effectiveness was evaluated using estimated mean costs and QALYs in both groups. Regression analysis was used to adjust for clinically relevant predictors. RESULTS: Between 2013 and 2020, 700 patients were recruited (mean age: PCI+OMT=70 years, OMT=68 years; male (%): PCI+OMT=87, OMT=88); median follow-up was 3.4 years. Over all follow-ups, patients undergoing PCI yielded similar health benefits at higher costs compared with OMT alone (PCI+OMT: 4.14 QALYs, £22 352; OMT alone: 4.16 QALYs, £15 569; difference: −0.015, £6782). For both groups, most health resource consumption occurred in the first 2 years post-randomization. Probabilistic results showed that the probability of PCI being cost-effective was 0. CONCLUSIONS: A minimal difference in total QALYs was identified between arms, and PCI+OMT was not cost-effective compared with OMT, given its additional cost. A strategy of routine PCI to treat ischemic left ventricular systolic dysfunction does not seem to be a justifiable use of health care resources in the United Kingdom. REGISTRATION: URL: https://www.clinicaltrials.gov; Unique identifier: NCT01920048.

By generating multiple imputed datasets, MICE allow for capturing the variability introduced by imputing missing values. 22This approach recognizes that there is inherent uncertainty in imputing missing data and provides a robust framework for conducting statistical analyses.After imputation, the M imputed datasets can be analysed separately using standard statistical methods, and the results can be combined using Rubin's rule to obtain valid and reliable estimates that account for the uncertainty introduced by the imputation process.
All the MICE models were estimated in STATA18 using the command ice.

Implementation in STATA:
The command ice is used to initiate the imputation process.The variable of interest with missing data, EQ5D score, is specified, along with several auxiliary variables including Hypertension, BMI, Ethnicity, and NYHA risk classification.These auxiliary variables are utilized in the imputation model to predict the missing values of EQ5D score (Table S3).It is important to note that the auxiliary variables should be complete without any missing values.The m option is set to 50, indicating that 50 imputed datasets will be created.The match option implies that the imputed datasets will be created using a matching algorithm.
In summary, the ice command performs multiple imputation using chained equations in Stata, incorporating auxiliary variables to predict missing values of the variable EQ5D score.The resulting imputed datasets with 50 imputations generated.

Assumptions:
The MICE (Multiple Imputation using Chained Equations) approach relies on several key assumptions when imputing missing data. 23These assumptions play a crucial role in ensuring the validity of the imputation process and the subsequent analyses conducted using the imputed data.There are some assumptions in MICE such as, Missingness Mechanism and Linearity and Correct Specification of Imputation Models.However, the main assumption of MICE is: Missing at Random (MAR): This assumption states that the missingness of data can be explained by observed variables and is not dependent on unobserved variables after accounting for the observed ones.In other words, the probability of data being missing can be predicted by the available information.

Number of Imputed Datasets:
To capture the uncertainty associated with imputed values, a total of 50 imputed datasets were generated using the imputation model.Multiple imputed datasets allow for comprehensive analysis by accounting for the variability resulting from the imputation process.

Merging Imputed Results -Rubin's Rule:
After generating the imputed datasets, the imputed results were merged using Rubin's rule.Rubin's rule combines the estimates and variances from each imputed dataset, yielding valid and reliable inferences that account for the uncertainty introduced by the imputation model. 12

Results of Imputation:
The imputed results provide a complete dataset, eliminating any missing values.This allows for a comprehensive analysis without the need to exclude cases due to missing data.

Note:
The BCIS jeopardy score is a simple method for estimating the amount of myocardium at risk on the basis of the particular location of coronary artery stenoses.The maximum score is 12 and a score ≥6 was required to be eligible for REVIVED.A score of 12 indicates proximal disease in all three major epicardial coronary arteries.Coef = coefficient, Std err.= standard error, BMI= body mass index, NYHA= New York heart association, BCIS=British cardiovascular intervention society, MI= myocardial infarction, PCI = percutaneous coronary intervention, CABG= coronary artery bypass grafting, EQ-5D-5L = EuroQol 5-Dimension 5-Level.

Table S2 . Unit cost of resources used Category Input Unit cost NHS main code
based on the NHS Reference Costs 2011/12 updated to 2021 prices, and estimated to be £17,707, £9,617 and £20,728 for ICD, CRT-P and CRT-D, respectively.24 * Health system advice suggested that the ranking of the unit costs for implantable devices in the NHS Reference Costs 2021/22, with CRT (£9,617) being more expensive than CRT-D (£7,108), was not as expected. 7Thus, devices were costed according to the methodology used to inform NICE's 2014 Technology Appraisal,

Table S4 . Average number of days of hospitalisation Percutaneous coronary intervention + optimal medical therapy (n = 108)
M= months, Y= years, N= number, CI= confidence interval * T-test was performed.Difference is not statistically significant, p-value = 0.2932.

Table S5 . Number (%) of patients with new implantable devices by visit Percutaneous coronary intervention + optimal medical therapy (n=347) Optimal Medical Therapy (n=353)
* Before baseline data on devices shown for completeness, however not included in the economic analysis.† Chi squared test was performed.Difference is not statistically significant, p-value =0.205.Note: 121 devices were implanted after baseline in the OMT group; one patient was missing the date of implantation, hence we present data on n=120 in the OMT group.

Table S6 . Distribution of medications by visit and treatment group
Note : PCI = percutaneous coronary intervention, OMT = optimal medical therapy, N = number, prop = proportion

Table S9 . Generalised Linear Models results
The BCIS jeopardy score is a simple method for estimating the amount of myocardium at risk on the basis of the particular location of coronary artery stenoses.The maximum score is 12 and a score ≥6 was required to be eligible for REVIVED.A score of 12 indicates proximal disease in all three major epicardial coronary arteries. Note: