Bias Reduction through Analysis of Competing Events (BRACE) Correction to Address Cancer Treatment Selection Bias in Observational Data

Abstract

Purpose: Cancer treatments can paradoxically appear to reduce the risk of noncancer mortality in observational studies, due to residual confounding. Here we introduce a method, Bias Reduction through Analysis of Competing Events (BRACE), to reduce bias in the presence of residual confounding. Experimental Design: BRACE is a novel method for adjusting for bias from residual confounding in proportional hazards models. Using standard simulation methods, we compared BRACE with Cox proportional hazards regression in the presence of an unmeasured confounder. We examined estimator distributions, bias, mean squared error (MSE), and coverage probability. We then estimated treatment effects of high versus low intensity treatments in 36,630 prostate cancer, 4,069 lung cancer, and 7,117 head/neck cancer patients, using the Veterans Affairs database. We analyzed treatment effects on cancer-specific mortality (CSM), noncancer mortality (NCM), and overall survival (OS), using conventional multivariable Cox and propensity score (adjusted using inverse probability weighting) models, versus BRACE-adjusted estimates. Results: In simulations with residual confounding, BRACE uniformly reduced both bias and MSE. In the absence of bias, BRACE introduced bias toward the null, albeit with lower MSE. BRACE markedly improved coverage probability, but with a tendency toward overcorrection for effective but nontoxic treatments. For each clinical cohort, more intensive treatments were associated with significantly reduced hazards for CSM, NCM, and OS. BRACE attenuated OS estimates, yielding results more consistent with findings from randomized trials and meta-analyses. Conclusions: BRACE reduces bias and MSE when residual confounding is present and represents a novel approach to improve treatment effect estimation in nonrandomized studies.

Adjuvant Radiotherapy in Surgically Treated HPV-Positive Oropharyngeal Carcinoma with Adverse Pathological Features

“…Limitations in our study include its retrospective design and lack of standardized follow-up protocols. Furthermore, the NCDB lacks cause-specific outcomes data which may confound analyses of treatment effects in observational studies [ 40 ]. Importantly, the NCDB lacks data on a well described adverse feature, perineural invasion, as well as other pathologic risk factors such as close margins and other factors associated with survival, including tobacco use status and cumulative lifetime tobacco use, and therefore our analyses were unable to incorporate these potentially important prognostic indicators.…”

Treatment Considerations for Patients with Locoregionally Advanced Head and Neck Cancer with a Contraindication to Cisplatin

“…This study was limited by missing data on HPV status and additional key clinical characteristics, such as smoking and performance status, A separate analysis of the SEER-Medicare database similarly suggested that carboplatinbased regimens were not only superior to cetuximab, but were also similar in efficacy to cisplatin [56]. Unfortunately, selection bias is a common problem with retrospective studies in competing risks populations, even after controlling for confounding variables with propensity score models [57]. Collectively, these studies highlight the need to compare alternative treatment regimens in randomized trials.…”

Problems With Interpreting Treatment Effects in Observational Competing Risks Data

“…The authors did not describe the technique followed for hemostasis, and we have already demonstrated that the effect of hemostasis depends on the energy of the coagulation devices used compared with conventional hemostasis with knots. 2,3 In addition, an incidence of 8.3% and 2.5%, respectively, in the 2 groups regarding postthyroidectomy hemorrhage could be considered quite high, especially when the procedure is performed by an experienced endocrine surgeon. 4 Furthermore, in the prospective group, the authors decided to include only patients having risk factors for bleeding and they did not report or compare the incidence of postoperative bleeding and the variation of the neck circumference in patients without risk factors.…”

“…An observational study comparing cetuximab and cisplatin undoubtedly would have been marred by treatment selection bias and would require bias mitigation techniques such as BRACE. 2 Cetuximab and carboplatin-based treatment, on the other hand, are both considered alternative radiosensitizing approaches in cisplatin-ineligible patients (National Comprehensive Cancer Network category 2B); thus, physicians may regard them as somewhat interchangeable in both efficacy and toxic effects. As seen in the RTOG 1016 clinical trial, 3 however, cetuximab with radiotherapy was associated with some unique toxic effects and is not necessarily "less intensive" compared with carboplatin.…”