• 2019-10
  • 2019-11
  • 2020-03
  • 2020-07
  • 2020-08
  • br propensity score are potential


    propensity score are potential confounders between treatment and outcome. Once they are part of the pro-pensity score model, they cannot be part of the outcome model as it ABT-263 (Navitoclax) would mean that we did not adjust for them in the propensity score. The classic Cox model for abdominal cancer is presented in Supplementary Table I (online only).
    The outcomes of interest were time to admission to the hospital with abdominal cancer, time to death related to abdominal cancer, time to admission to the hospital with any cancer, and time to death because of any cancer.
    Adjusted survival curves were also plotted (Figs 1 and 2) using the previously calculated IPWs.29
    To adjust for the propensity of patients to receive open repair or EVAR, the G-formula technique28 (a causal infer-ence analysis) was employed for estimation of the hazard ratio (HR) and 95% confidence interval (CI) for the associ-ation between type of surgery (open repair or EVAR) and long-term survival. The model used for the G-computa-tion formula included the following covariates: age, year, sex, comorbidities as binary variables, deprivation quintile, strategic health authority, type of admission (elective or emergency), and type of operation (EVAR or open; Supplementary Table II, online only). A Cox propor-tional hazards model was fitted for each outcome of in-terest and used to calculate the potential outcomes for 
    both EVAR and open AAA repair to obtain the HR based on the average effects. The 95% CIs for the G-formula were obtained using bootstrapping with 500 repetitions.
    An additional analysis was performed within the EVAR group to compare primary and secondary end points of cancer-free survival in patients treated at centers using CT surveillance with those in centers using non-CT surveillance.
    All analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC) and R 3.3.2 software (http://www.
    Patients’ demographics. Of the 39,390 patients who underwent AAA repair during the study period and were included in this study, 14,419 patients underwent EVAR and 24,971 patients underwent open AAA repair. For the analysis of all cancer types, 14,150 and 24,465 pa-tients were included in the EVAR and open AAA repair groups, respectively. Moreover, for the purposes of anal-ysis, our logistic model was noted to have a good fit (C statistic of 0.812).
    Patients who underwent EVAR were older and pre-sented with a greater burden of medical comorbidities than those who underwent open surgery (Table I). The percentage of deaths caused by any cancer was higher
    Journal of Vascular Surgery Markar et al 1779
    Fig 2. Comparison of patients receiving endovascular aneurysm repair (EVAR) or open abdominal aortic aneu-rysm (AAA) repair, showing an increase in the proportion of patients admitted after AAA surgery with any type of cancer or cancer as cause of death in the EVAR group. Univariate Cox model: hazard ratio (HR), 1.09; 95% con-fidence interval (CI), 1.02-1.16; P ¼ .0163.
    Table I. Characteristics of the study participants
    All cancers
    AAA, Abdominal aortic aneurysm; EVAR, endovascular aneurysm repair.
    Categorical data are presented as percentage frequencies. a P values correspond to exact c2 tests and Wilcoxon-Mann-Whitney test for categorical and continuous variables, respectively.
    in the patients who underwent EVAR (EVAR, 18.0%; open, 14.7%; Tables II and III). The most common cause of death after open surgery was attributed to a ruptured aneurysm on presentation (EVAR, 5.4%; open surgery, 23.1%).
    common cause of death for emergency cases was aneu-rysm rupture on presentation in both procedures (open surgery, 40.5%; EVAR, 16.7%). The most common causes of death for elective cases were AAAs without mention of rupture for open surgery (10.7%) and chronic ischemic heart disease for patients undergoing EVAR (9.0%). The