Influence of Concomitant Percutaneous Transluminal Angioplasty with Percutaneous Coronary Intervention on Clinical Outcomes of Stable Lower Extremity Arterial Diseases
The data source
This study is a retrospective observational study conducted using the Korea National Health Insurance Data Sharing Service (KNHISS) Custom Search Database. KNHISS is a data sharing service operated by KNHIS that provides a cohort sample database as well as custom databases to facilitate the use of national health information for research purposes. Detailed information about the KNHISS custom database has been described earlierseven and is shown in Supplementary Data 1.
All procedures in this study protocol adhere to the ethical principles of the Declaration of Helsinki. The Institutional Review Board of Hanyang Guri University Hospital (GURI-2019-03-032) and the KNIHS Ethics Committee (NHIS-2019-1-561) reviewed and approved the protocol for this study. Informed consent of the subjects was exempted because the data was anonymized.
From the KNHISS database, we included patients with stable LEAD who underwent PCI between January 2014 and December 2015. ATP was reimbursed in this period when a patient was (1) diagnosed with symptomatic LEAD and (2) had at least one lesion with a stenosis ≥ 70% in diameter. Patients who had undergone prior PCI or ATP between January 2012 and December 2014 or those with diagnosis or reimbursement codes suggesting acute limb ischemia, CLI, or limb amputations were excluded. Due to the research ethics policy declared by KNHISS, only 30% of the total population were randomly selected to produce the custom database for this study.
Definitions of Clinical Features and Outcomes
Estimated glomerular filtration rate (eGFR) was calculated using the chronic kidney disease–epidemiology collaboration equation. Chronic renal failure (CKD) was defined as an eGFR ≤ 60 mL/min/1.73 m2. The duration of follow-up was defined from the day of the first outpatient visit after index PCI until the day the first clinical event occurred, when claims ceased to emerge in the database, or when 5 years elapsed after ICP index. . All patients were followed for at least 3 years. Comorbidities and clinical events were identified using KCD-7 codes. The Charlson Comorbidity Index (CCI) was used to represent the severity of comorbidities as previously described8. The detailed KCD-7 codes used to define comorbidities and clinical findings are described in Supplementary Table 1. Myocardial infarction (MI) was defined as a new diagnosis of MI codes (I21-I24) at the time of admission, stroke was defined as new diagnosis of stroke codes (I60-I64) at admission, and CV death was defined as death from CV disease (I20-I25, I42 -I43, I50), cerebrovascular events (I60-I69) or peripheral vascular diseases (I70-I74).
Coronary revascularization was defined as a composite of PCI and coronary bypass. Repeat coronary revascularization was defined as coronary revascularization after an index hospitalization. A major adverse CV event (MACE) was defined as a composite of CV death, MI, stroke, and repeated coronary revascularization. Bleeding events were defined as new diagnostic codes for bleeding events requiring admission or transfusion (Supplementary Table 1) based on the Global Utilization of Streptokinase and Tissue Plasminogen Activator for Occluded Coronary Artery (GUSTO) criteria9. End-stage renal disease (ESRD) has been defined as a new diagnostic code for ESRD (N18.5). CLI was defined as a new diagnostic code for diabetic foot ulcer/gangrene or lower extremity amputation (Supplementary Table 1).
PCI and PTA were defined as insurance claim codes representing procedures (Supplementary Table 2) with those representing designated devices for their respective procedures, including plain angioplasty balloons, drug-coated balloons, stents bare metal (BMS) or drug-eluting stents (DES) at index hospitalization (Supplementary Table 3). In contrast, coronary artery bypass graft surgery was defined using procedure codes only (Supplementary Table 2). Concomitant ATP at the time of PCI was defined as ATP performed during the same index hospitalization period as the PCI was performed. ATP after discharge was defined as ATP at first readmission after an index hospitalization.
Since patients undergoing concomitant PTA at the time of PCI were expected to have more severe atherosclerotic CV disease with more comorbidities than those undergoing PCI alone, we performed propensity score matching (PSM) to balance the covariates between the two groups. Propensity scores were created using a multivariate logistic regression model including age, sex, BMI, waist circumference (WC), eGFR, MI at admission baseline, number of coronary stents used, number of vessels involved, ≥ median income, prior bleeding events, prior MI, prior LEAD, prior cerebrovascular events, prior hemiplegia, diabetes, hypertension, chronic kidney disease, peptic ulcer, ESRD and ICC as covariates. The matching procedure was performed using the nearest neighbor method in a ratio of 1:5. The quality of PSM was assessed using absolute standardized mean differences (SMD).
Categorical variables were compared using a chi-square test, while continuous variables were compared using Student’s t-test or Mann-Whitney’s U-test. Kaplan-Meier survival analysis with a log-rank test was used to compare cumulative incidences of clinical outcomes, including all-cause death, CV death, MACE, ESRD, amputation and the ATP after the exit between the ATP + PCI group and the PCI group only. Cox proportional hazard models were used to assess the association of concurrent ATP with clinical outcomes after the PCI index. Multivariate Cox proportional hazard models were used to adjust for confounding influences and strengthen associations between concurrent ATP and clinical outcomes. Multivariate models included age, sex, BMI, waist circumference, MI at initial admission, number of coronary artery stents used, ≥ median income, prior bleeding events, prior MI , prior LEAD, prior stroke, diabetes, hypertension, peptic ulcer, DAPT duration, and statin use as covariates. Full multivariate models were reduced by a backward variable selection process with a threshold of p> 0.05 to minimize overfitting bias and multicollinearity between variables.
To estimate the influence of unmeasured confounding factors, E-values were estimated for the hazard ratio (RR) and the 95% confidence interval (CI). The E-value is a simple and powerful sensitivity analysis tool to assess the strength of a risk-outcome associationten. The E-the value of an HR indicates the minimum HR that an unmeasured confounder should have with both the causative factor and the outcome to fully explain the apparent risk-outcome association, while the E-the value for an upper or lower limit of the CI indicates the minimum HR that the confounder should have with the causal factor and the result for the CI to include the null value. A top E-value indicates a strong causal relationship that would survive in the presence of a confounder with an HR E -value with causal factor and result.
Subgroup analyzes were performed to detect the presence of differential impacts of concurrent ATP on the risk of MACE, ESRD, and limb events (composite of amputation and ATP after discharge) in various subsets of patients. Mediation analyzes were performed in the PSM cohort to identify the causal mediating effects of limb and kidney outcomes on the association between concurrent ATP and all-cause death using a resampling technique. bootstrap11. Parametric survival regression models with a Gaussian distribution, including limb and kidney outcomes as time-varying covariates, were used as model objects for mediators and outcome in mediation analyses. A mediation analysis for each mediator was performed in a set of 1000 bootstrap samples, and the mean total, direct, and indirect causal mediation effects on the coefficients of the survival regression parametric models were reported.
All statistical analyzes other than mediation analyzes were performed using commercially available statistical software SAS 7.1 (SAS Institute, Cary, NC, USA). Mediation analyzes were performed using statistical software R-4.3 (R Core Team, R Foundation for Statistical Computing, Vienna, Austria) and its “survival” and “mediation” packages in RStudio-1.3 (RStudio Team , RStudio Inc., Boston, MA, USA). A p-valueof