The cost‐effectiveness of radial access percutaneous coronary intervention: A propensity‐score matched analysis of Victorian data

Abstract Background Despite evidence of the comparative benefits of transradial access percutaneous coronary intervention (PCI) over transfemoral access, its uptake remains highly varied across Australia. Few studies have explored the implications of the choice of access site during PCI from the perspective of the Australian healthcare setting. We, therefore, performed a cost‐effectiveness analysis of radial versus femoral access PCI. Methods Data from the Victorian Cardiac Outcomes Registry (VCOR) were used to inform our economic analyses. Patients treated through either radial or femoral access PCI were propensity score‐matched using the inverse probability weighted (IPW) method, and the incidence of major bleeding and all‐cause mortality in the cohort was used to inform an economic model comprising a hypothetical sample of 1000 patients. Costs and utility data were drawn from published sources. The economic evaluation adopted the perspective of the Australian healthcare system. Results Among a cohort of 1000 patients over 1 year, there were 19 fewer deaths, and six fewer episodes of nonfatal major bleeding in the radial group compared to the femoral group. Total cost savings attributed to radial access was AUD $1 214 688. Hence, from a health economic point of view, radial access PCI was dominant over femoral access PCI. Sensitivity analyses supported the robustness of these findings. Conclusions Radial access is associated with improved patient outcomes and considerably lower costs relative to femoral access PCI. Our findings support radial access being the preferred approach for PCI across a variety of indications in Australia.


INTRODUCTION
International guidelines support percutaneous coronary intervention (PCI) as the preferred means for coronary revascularization in the setting of acute coronary syndromes (ACS). 1 In contemporary practice, access to the coronary arteries in PCI is achieved via the femoral or radial arteries. Although femoral access was traditionally favored for the ease of cannulation and direct access to the coronary arteries, there is considerable evidence of the benefits of radial access PCI. A recent Cochrane review found that relative to femoral access, radial access is associated with significant reductions in patient mortality, bleeding, and access-site complications. 1 This is supported by data from cardiac registries in the United States, United Kingdom, and Australia. [2][3][4][5] In addition to the greater safety profile of radial access, hospital length of stay (LOS) is considerably shorter relative to transfemoral PCI. 6,7 Despite an accumulating body of evidence pertaining to the comparative benefits of radial access, the uptake of transradial access PCI has been variable. 7,8 This may be attributed to patient factors, as well as operator preference and experience. 1,2,7 Studies have highlighted the potential cost savings and cost-effectiveness attributed to radial access PCI, including a recent Australian analysis of published data by our group. 3,6,7,9 Evidence from the Victorian Cardiac Outcomes Registry (VCOR) demonstrated that radial access is associated with improved patient outcomes and shorter LOS in Victoria, Australia, in line with international findings. 4,8 Furthermore, there has been considerable uptake in the number of radial access procedures over time, with the proportion of radial access PCIs overtaking femoral access in 2016. 4,8 Importantly, at present, a substantial proportion of PCIs are performed via transfemoral access and significant discrepancies persist in the uptake of radial access PCI across Victorian hospitals. 4 In this context, we performed a cost-effectiveness analysis of radial access PCI using data from VCOR to explore the health and economic benefits of radial access PCI.

Data source
VCOR is a state-wide cardiac clinical quality registry in Victoria, established in 2012 for the purposes of monitoring and benchmarking hospital performance and outcomes post-PCI. 10 Statewide coverage was achieved in 2017, and all public and private PCI-capable centers currently participate and contribute data to VCOR. 10,11 Hospital-appointed data managers collect data pertaining to patient characteristics at baseline, demographic characteristics, and procedural outcomes. Additional patient follow-up is performed at 30 days for data on key patient outcomes, including mortality and major adverse cardiac and cerebrovascular events (MACCE), a composite of death, myocardial infarction, stroke, and target vessel revascularization. Additional details on VCOR have been described elsewhere. 10,11 The primary VCOR data set was linked to the Victorian Admitted Episodes Data set (VAED) to estimate the costs of PCIs. 12,13 The VAED contains admissions, diagnostic and procedural data across all Victorian hospitals; variables in the VAED reflect hospital activity for funding purposes. 12,13 Linkage with the National Death Index (NDI) was also performed to allow the estimation of long-term mortality for patients undergoing PCI. 10 Pearson's χ 2 tests for categorical variables, and univariable linear regression modeling for continuous variables, were used to explore differences in patient and procedural characteristics between radial and femoral treatment arms. Generalized linear regression modeling (GLM) was used to overcome the high positive skew associated with patient LOS and door-to-balloon/device time parameters. 14 Propensity score analyses were undertaken to reduce confounding arising from differences in characteristics of patients undergoing radial versus femoral PCIs. Additionally, the patient risk profile for PCI in Victoria has evolved, with patients presenting with greater risk over time. 9 Inverse probability weighting (IPW) was used to construct a synthetic cohort in which the distribution of patient and procedural characteristics at baseline was independent of treatment assignment. 15 Any bias attributed to differences in patient characteristics between the radial or femoral groups on key outcomes was therefore minimized. 15 The following variables were used in predicting the use of radial access: age (<75 years and ≥75 years); sex; indigenous status; body mass index (BMI); in-hours hospital arrival (between 08:00 and 18:00 on a workday); ACS category (non-ACS, UA, NSTEMI, STEMI); cardiogenic shock or out-of-hospital cardiac arrest (OHCA) requiring intubation; medicated diabetes mellitus; peripheral vascular disease; cerebrovascular disease; chronic oral anticoagulation therapy; prior coronary artery bypass grafting; previous PCI; use of glycoprotein IIb/IIIa inhibitors; use of thienopyridine or ticagrelor; estimated glomerular filtration rate (eGFR); required mechanical ventricular support; lesion complexity (American College of Cardiology/American Heart Association Type A/B1 vs. Type B2/C lesions); unprotected left main PCI and in-stent restenosis PCI. Balance on baseline covariates was evaluated using absolute standardized differences, with a value <10% considered as acceptable standardized bias. 16 Propensity score weights were trimmed at the 5th and 95th percentiles to account for the effect of outlier weights in the model. 15 The propensity score distribution for the IPW model is presented in Supporting Information, Appendix A.
Univariable logistic regression analyses were performed following IPW-matching to explore differences in the incidence of patient clinical outcomes between radial and femoral groups. These are presented in Supporting Information, Appendix B.

ECONOMIC EVALUATION
An economic model was developed to simulate the clinical and cost outcomes of radial versus femoral access for a hypothetical sample of 1000 individuals profiled on the IPW-matched cohort.

EFFECTIVENESS
Key outcomes considered were all-cause mortality at 0-30 days and 31 days to 1 year, and major nonfatal bleeding at 0-30 days following index PCI. These outcomes were selected as the clinical benefit attributed to radial access are reductions in major bleeding and mortality events within the 12-month period following PCI. The incidence of Bleeding Academic Research Consortium (BARC) Type 3 bleeding in the propensity-matched population was used to inform the rate of nonfatal major bleeding events. Similarly, the incidence of all-cause mortality at 30 days, and from 31 days to 1 year was informed by the incidence of allcause mortality in the propensity-matched population for the relevant period. The incidence of key outcomes considered in the economic model is presented in Table 1.

UTILITY VALUES
The utility values considered in the economic model are presented in Table 1.
Patients undergoing PCI for non-ACS indications were assigned a utility of 0.91 (95% confidence interval [CI]: 0.90-0.91). 18 Patients who underwent PCI for ACS indications were assigned a utility of 0.80 (95% CI: 0.79-0.81) to reflect that they had an initial ACS event. This utility value was derived from the estimated utility of patients following an ACS event in the Valsartan in Acute Myocardial Infarction (VALIANT) trial. 19 A disutility of −0.03 was applied to patients who experienced a major bleeding event, in line with a recent study by Doble et al. 20 This disutility was only applied to the initial 30-day period, as major bleeding was considered an acute event occurring within the initial 30 days from the index procedure.

COSTS
Key cost inputs used in the economic model are presented in Table 1.

PROCEDURAL COSTS
Procedural costs were estimated using the Casemix funding method, in which a "weighted inlier-equivalent separation" (WIES) weight is assigned to each episode of care. 21,22 The WIES reflects the cost of an episode of care relative to the average cost across all episodes of care and is multiplied by the WIES price set for a given financial year to estimate the cost for an episode of care. Each WIES weight was converted to dollar payments through the application of public sector payment rates for a given financial year to estimate the procedural cost of PCI. 23 Univariable GLM with gamma distribution and log-link was performed using the IPW-matched population to estimate the average procedural costs associated with PCIs. Costs were stratified by sex and indication for PCI in the GLM. All costs were adjusted for inflation to 2021 AUD$ based upon the Health Price Index (HPI). 24

COST OF ACUTE EVENTS
Costs associated with procedural complications, including bleeding and mortality, were captured in the procedural costs. In estimating the cost of acute events occurring outside of the procedure and up to 1 year, Australian Refined-Diagnosis Related Groups (AR-DRG) data for hospitalizations were used. 17 In lieu of available data on the costs of major bleeding due to PCI, the cost of a major bleeding event was assumed to be equivalent to the cost of gastrointestinal hemorrhage, in line with other studies which have considered the costs of major bleeding in the setting of cardiovascular disease in Australia. [25][26][27] The cost of patient death from all causes occurring outside of the acute period (31 days to 1 year) was based on the weighted average of AR-DRG codes F60B (Circulatory Disorders, Admitted for AMI Without Invasive Cardiac Investigation Procedure, Transferred <5 Days) and B70D (Stroke and Other Cerebrovascular Disorders, Transferred <5 Days). This is as mortality in the initial year following PCI is attributed to cardiovascular causes. 28 It was conservatively assumed that hospitalization would only occur in only 50% of deaths.
Therefore, only 50% of all deaths occurring outside of the initial hospital stay would incur hospitalizations costs, in line with previous studies. 25,27

MODEL OUTCOMES
The main outcome of interest for the cost-effectiveness analysis was the incremental cost-effectiveness ratio (ICER) in terms of cost per QALY gained and cost per year of life saved (YoLS) for radial access compared with femoral access PCI.

SENSITIVITY ANALYSES
A series of deterministic and probabilistic sensitivity analyses (PSA) was performed to explore the degree of uncertainty around key model parameters. 29 Key input parameters were varied by the upper and lower limit of the 95% CI, or by ±15% or ±25% around the estimate (Table 1).
Additional sensitivity analyses involved stratifying the radial and femoral groups by year of procedure, and the exclusion of high-risk patient subgroups. 4,30 These include patients with high-risk conditions requiring mechanical ventricular support or presenting with OHCA requiring intubation, patients aged >80 years, patients with an eGFR <90 ml/min/ 1.73 m 2 , and patients with Type B2/C lesion complexity. 4,30 Univariable logistic regression analyses were performed following IPW-matching to explore differences in the incidence of patient clinical outcomes between radial and femoral groups for these high-risk patient subgroups (Supporting Information, Appendix C). For the PSA, a second-order Monte Carlo simulation with 10 000 iterations was performed using the ranges and probability distributions presented in Table 1.

Baseline characteristics
Baseline and procedural characteristics of the unadjusted VCOR population are presented in Table 2 Table 2.

IPW ANALYSIS
A total of 28 982 patients were included in the propensity score-

COST-EFFECTIVENESS ANALYSES
The incidence of nonfatal major bleeding and all-cause mortality in each of the radial and femoral groups in the IPW-matched population was used to inform the model (see Table 1 Bleeding (

SENSITIVITY ANALYSES
The results of the one-way, deterministic sensitivity analyses are presented in Table 4.
Based on the one-way sensitivity analyses, the model was most sensitive to procedural costs and the year for which PCI occurred.
However, the dominance of radial PCI over femoral PCI was main- that is, radial access was both health and cost-saving relative to femoral access PCI.  Table 4 and Supporting Information, Appendix D).
Second, in the total VCOR cohort, 37% of patients were treated in private hospitals, but cost data were not available from private hospitals contributing data to VCOR. Hence, only unit costs from the public sector were applied. Public and private costs are likely to be different, due to differences in clinical characteristics between patients undergoing PCI in public and private hospitals, differences in patient management and differences in the relative efficiency between public and private hospitals. 4,39,40 However, procedural costs were varied in sensitivity analyses, and the results remained consistent in terms of radial access being cost saving.
Third, the impact of access sites on the incidence of MACCE was not explicitly considered in our analyses. MACCE is less likely to occur with radial access, 7 and hence we likely underestimated the cost-savings attributed to radial access PCI. However, this would not have altered the conclusion of our study.
Finally, the time horizon of our evaluation was limited to the 12-month period following index PCI. Although the evidence for the benefits of radial access lies within the short-term period following PCI, additional longer-term studies are warranted to examine additional longer-term survival benefits and potential cost-savings attributed to radial access. 1 A key strength of our study lies in the large and contemporaneous cohort considered for IPW-weighting. A similar analysis of patient outcomes using data captured between January 1, 2013, to December 31, 2014, in VCOR had identified significant reductions in major bleeding for patients with radial access PCI, but no association between access site selection and patient mortality at 30 days. 8 In contrast, our analyses drew upon a larger data set spanning 4 years and better reflected the current literature in support of the benefits of radial access. Importantly, our analyses provide real-world evidence for increasing the uptake of radial access PCI across Australia.