Exploratory subgroup identification in the heterogeneous Cox model: A relatively simple procedure.

For survival analysis applications we propose a novel procedure for identifying subgroups with large treatment effects, with focus on subgroups where treatment is potentially detrimental. The approach, termed forest search, is relatively simple and flexible. All-possible subgroups are screened and selected based on hazard ratio thresholds indicative of harm with assessment according to the standard Cox model. By reversing the role of treatment one can seek to identify substantial benefit. We apply a splitting consistency criteria to identify a subgroup considered "maximally consistent with harm." The type-1 error and power for subgroup identification can be quickly approximated by numerical integration. To aid inference we describe a bootstrap bias-corrected Cox model estimator with variance estimated by a Jacknife approximation. We provide a detailed evaluation of operating characteristics in simulations and compare to virtual twins and generalized random forests where we find the proposal to have favorable performance. In particular, in our simulation setting, we find the proposed approach favorably controls the type-1 error for falsely identifying heterogeneity with higher power and classification accuracy for substantial heterogeneous effects. Two real data applications are provided for publicly available datasets from a clinical trial in oncology, and HIV.

FIR: Efficacy, Safety, and Biomarker Analysis of a Phase II Open-Label Study of Atezolizumab in PD-L1-Selected Patients With NSCLC.

INTRODUCTION: The FIR phase II study (NCT01846416) evaluated the efficacy and safety of anti-programmed death-ligand 1 (PD-L1) atezolizumab in advanced NSCLC selected by tumor cell (TC) or tumor-infiltrating immune cell (IC) PD-L1 expression. METHODS: Patients with PD-L1 TC2/3 (PD-L1 staining on ≥5% of TC) or IC2/3 tumors (PD-L1 staining on ≥5% of IC; determined by SP142 PD-L1 immunohistochemistry assay) with paired fresh and archival histology samples were recruited into cohort 1 (chemotherapy-naive/>6 months between adjuvant chemotherapy and recurrence), cohort 2 (≥ second-line without brain metastases), or cohort 3 (≥ second-line with treated brain metastases). Patients received 1200 mg atezolizumab on day 1 (21-day cycles). Primary endpoint was investigator-assessed modified Response Evaluation Criteria in Solid Tumors, objective response rate (Response Evaluation Criteria in Solid Tumors v1.1). Secondary endpoints were overall survival, progression-free survival, duration of response, and safety. RESULTS: Patients (N = 138) were enrolled (137 evaluable for response: cohort 1, n = 31; cohort 2, n = 93; and cohort 3, n = 13). Investigator-assessed objective response rate was 32%, 21%, and 23% for cohorts 1, 2, and 3, respectively. Treatment-related adverse events were reported in 81%, 67%, and 69% of patients, respectively, including grade 3-4 treatment-related adverse events in 16%, 19%, and 15%, respectively. Moreover, 88.6% (86 of 97) paired baseline tumor samples had <5% change in TC/IC PD-L1 expression over time. CONCLUSIONS: Atezolizumab monotherapy showed clinical activity in patients with NSCLC, including those with brain metastases; safety was consistent with previous trials. Atezolizumab has completed phase III monotherapy studies in second-line. Front-line trials are ongoing, confirming these favorable results.

Estimation of treatment effects in weighted log-rank tests.

Non-proportional hazards have been observed in clinical trials. The log-rank test loses power and the standard Cox model generally produces biased estimates under such conditions. Weighted log-rank tests have been utilized to increase the test power; however, it is not intuitive how to interpret the test result in terms of the clinical effect. We propose a Cox-model based time-varying treatment effect estimate to complement the weighted log-rank test. The score test from the proposed model is equivalent to the weighted log-rank test, and a time-profile of the treatment effect can be obtained by fitting a time-varying covariate Cox model. Simulation results show that the proposed model preserves type-I error and achieve higher power than log-rank tests under non-proportional hazards scenarios. Whereas the standard Cox model produces biased effect estimates, the proposed model produces unbiased estimates if the weight function is correctly specified. It also achieves a better model fit and an enhanced flexibility to accommodate non-proportional hazards compared to the standard Cox model. The proposed approach makes the assumptions of the weighted log-rank test explicit and the validity of assumptions can be assessed based on prior knowledge or model goodness-of-fit. It also helps to translate the weighted log-rank test results into quantitative estimates of the treatment effect with intuitive interpretation. The proposed method can be routinely conducted to complement weighted log-rank tests, especially in the setting where non-proportional hazards are expected.

Effectiveness of bevacizumab exposure beyond disease progression in patients with non-small-cell lung cancer: analyses of the ARIES observational cohort study.

PURPOSE: Bevacizumab used in combination with first-line chemotherapy confers an overall survival (OS) benefit for patients with non-squamous non-small-cell lung cancer (NSCLC). This analysis from the ARIES observational cohort study (OCS) was initiated to evaluate the effect of bevacizumab use beyond disease progression (BBP) on clinical outcomes in patients with NSCLC receiving first-line treatment with bevacizumab and chemotherapy. METHODS: The ARIES OCS prospectively enrolled patients from 2006 to 2009 in the United States who had advanced non-squamous NSCLC, received bevacizumab with chemotherapy in the first-line setting, and survived progressive disease (PD). A dichotomous landmark analysis examined post-PD OS (ppOS) in patients who received BBP versus no BBP within 30 days post PD. A time-dependent Cox model assessed the effect of cumulative BBP exposure on ppOS. RESULTS: The ARIES OCS enrolled 1967 patients with first-line NSCLC; 1358 patients had first PD and were alive at the 30-day landmark (351 patients with BBP and 1007 patients with no BBP). The landmark analysis showed that BBP was associated with a lower risk of death (BBP versus No-BBP); hazard ratio [HR], 0.75; 95% confidence interval 0.65-0.86. In the cumulative exposure analysis of 1461 patients who had PD, HRs for ppOS decreased by approximately 4% for each additional 21-day interval of bevacizumab received. Protocol-specified bevacizumab-select adverse events occurred in 14% of BBP patients. CONCLUSIONS: BBP was associated with a lower risk of death in patients with NSCLC treated with first-line bevacizumab and chemotherapy. Copyright © 2016 John Wiley & Sons, Ltd.

Isolating the Role of Bevacizumab in Elderly Patients With Previously Untreated Nonsquamous Non-Small Cell Lung Cancer: Secondary Analyses of the ECOG 4599 and PointBreak Trials.

BACKGROUND: Patient-level data from 2 phase III studies in patients with previously untreated, advanced-stage, nonsquamous non-small cell lung cancer (NSCLC) were pooled to examine outcomes with bevacizumab and chemotherapy based on age. METHODS: Data from patients randomized to paclitaxel-carboplatin (PC)+bevacizumab in the Eastern Cooperative Oncology Group 4599 (E4599) and PointBreak studies were pooled and compared with E4599 patients randomized to PC alone. Patients were grouped by age: below 65, 65 to 74, 70 to 74, below 75, and 75 years or above. A multivariable model was used to calculate hazard ratios (HRs) and corresponding 95% confidence intervals (CIs) using time-to-event outcomes. Adverse events (AEs) were assessed by age group in each study. RESULTS: The PC+bevacizumab and PC arms comprised 901 and 444 patients, respectively. PC+bevacizumab was associated with significant increases in overall survival relative to PC in patients below 65 years (hazards ratio [HR], 0.75; 95% confidence interval [CI], 0.62-0.89), 65 to 74 years (HR, 0.80; 95% CI, 0.64-1.00), 70 to 74 years (HR, 0.68; 95% CI, 0.48-0.96), and below 75 years (HR, 0.78; 95% CI, 0.68-0.89) but not in those aged 75 years or above (HR, 1.05; 95% CI, 0.70-1.57). Increased incidence of grade ≥3 AEs was reported with PC+bevacizumab versus PC in patients below 75 years (63% vs. 48%; P<0.05) and 75 years or above (81% vs. 56%; P <0.05) in E4599. CONCLUSIONS: This analysis suggests that the survival benefits associated with PC+bevacizumab extend to patient subgroups below 75 years with advanced-stage NSCLC; no benefit, however, was observed for bevacizumab-eligible patients who were 75 years or above.

Erlotinib therapy after initial platinum doublet therapy in patients with EGFR wild type non-small cell lung cancer: results of a combined patient-level analysis of the NCIC CTG BR.21 and SATURN trials.

BACKGROUND: The clinical benefit of erlotinib in treating epidermal growth factor receptor (EGFR) wildtype non-small cell lung cancer (NSCLC) has been questioned. We examined the impact of erlotinib in confirmed EGFR wildtype patients in two placebo-controlled phase III trials: the National Cancer Institute of Canada Clinical Trials Group BR.21 (BR.21) and Sequential Tarceva in Unresectable Non-Small Cell Lung Cancer (SATURN) trials. METHODS: Combined re-analysis of progression-free survival (PFS) and overall survival (OS) in patients with known wildtype EGFR, estimated by Kaplan-Meier curves and compared by two-sided log-rank test. Cox proportional hazards model was used to estimate hazard ratios (HR) adjusted for potential confounders. Additional analyses assessed comparability of patients with known and unknown EGFR mutation status to determine generalizability of the two study populations. RESULTS: Mutation status was known in 25% (n=184 of 731) of the BR.21, and 49% (n=437 of 889) of the SATURN populations, of which 82% (n=150) and 89% (n=388) respectively had wildtype EGFR. HR for PFS was 0.71 (95% CI, 0.59-0.85; P<0.01) and for OS was 0.72 (95% CI, 0.59-0.88; P<0.01). Baseline characteristics and outcome (PFS and OS) distributions were similar for patients with known and unknown EGFR status, suggesting generalizability of the EGFR wildtype data. Erlotinib benefit was sustained in all clinical subsets. CONCLUSIONS: Erlotinib provided a consistent and significant improvement in survival for patients with EGFR wildtype NSCLC in both studies, individually and in combination. The benefit of erlotinib does not appear to be limited to patients with activating mutations of EGFR.

Effectiveness and safety of post-induction phase bevacizumab treatment for patients with non-small-cell lung cancer: results from the ARIES observational cohort study.

Data from randomized, controlled trials suggest that post-induction phase (IP) treatment with bevacizumab may benefit patients with advanced non-small-cell lung cancer (NSCLC). Real-world clinical practice, however, can involve variable use and patterns of treatment in broader patient populations. To assess the effect of bevacizumab on post-IP overall survival (OS) following IP chemotherapy + bevacizumab, analyses were conducted in patients enrolled in the Avastin(®) Registry--Investigation of Effectiveness and Safety (ARIES) observational cohort study (OCS) who received post-IP bevacizumab. ARIES was a large, prospective OCS of patients who received chemotherapy in combination with bevacizumab for the first-line treatment of NSCLC. This unplanned, post hoc analysis included patients who received chemotherapy and bevacizumab and who did not have progressive disease through the completion of IP treatment. A dichotomous analysis compared outcomes in patients who did and did not receive bevacizumab before a landmark date of day 30 post IP. A cumulative exposure analysis used a time-dependent Cox regression model to assess the effect of cumulative post-IP bevacizumab exposure on post-IP OS. In the dichotomous analysis, the duration of post-IP OS was significantly longer in patients who received post-IP bevacizumab; median post-IP OS was 15.6 vs. 11.3 months, respectively (hazard ratio [HR] = 0.80; 95 % confidence interval 0.71-0.91; P < 0.001). The cumulative exposure analysis observed that each additional cycle of cumulative bevacizumab exposure decreased the HR for post-IP OS by 2.7 %, on average. In conclusion, post-IP bevacizumab exposure was associated with improved post-IP OS in patients with advanced NSCLC who were enrolled in the ARIES OCS.

Safety and effectiveness of bevacizumab-containing treatment for non-small-cell lung cancer: final results of the ARIES observational cohort study.

INTRODUCTION: Bevacizumab, a recombinant humanized monoclonal antibody against vascular endothelial growth factor, was approved by the US Food and Drug Administration for the treatment of advanced non-small-cell lung cancer (NSCLC) in combination with carboplatin and paclitaxel. ARIES (Avastin Regimens: Investigation of Effectiveness and Safety), a prospective observational cohort study, evaluated outcomes in a large, community-based population of patients with first-line NSCLC. METHODS: From 2006 to 2009, ARIES enrolled patients with locally advanced or metastatic NSCLC who were eligible for bevacizumab, excluding those with predominantly squamous histology. Patients were required to provide informed consent and to have initiated bevacizumab with chemotherapy within 4 months before enrollment. There were no protocol-defined treatments or assessments. The dosing of bevacizumab and chemotherapy, and the choice of chemotherapy regimen, was at the discretion of the treating physician. RESULTS: ARIES enrolled 1967 patients with first-line NSCLC. At study closure, median follow-up was 12.5 months (range, 0.2-65.5). Median age was 65 years (range, 31-93), and 252 patients (12.8%) identified as never smokers. Median progression-free survival was 6.6 months (95% confidence interval, 6.3-6.9), and median overall survival was 13.0 months (95% confidence interval, 12.2-13.8) with first-line bevacizumab plus chemotherapy. Incidences of bevacizumab-associated adverse events (19.7% overall) were consistent with those in randomized controlled trials of bevacizumab in NSCLC. CONCLUSION: Results from ARIES demonstrate similar outcomes to randomized controlled trials of bevacizumab when added to standard chemotherapy in a real-world patient population with advanced NSCLC.

The association of rash severity with overall survival: findings from patients receiving erlotinib for pancreatic cancer in the community setting.

OBJECTIVES: This retrospective study examined pancreatic cancer patients who received combination gemcitabine and erlotinib to determine if the association between rash and outcomes observed in clinical trials would be observed in 'real-world' community oncology settings. METHODS: Medical records from 10 community oncology practices were used to identify eligible patients. Rash severity was classified as High (moderate/severe) versus Low (absent/mild) based on medical record review. Kaplan-Meier analysis assessed progression-free survival (PFS) and overall survival (OS) by rash status from a landmark of 42 days after treatment initiation. Cox regression with time-varying covariates tested whether high-severity rash predicted longer OS and PFS. RESULTS: The High Severity group (n = 34) had longer median OS from the landmark than the Low Severity group (n = 134; 7.58 months vs 5.03 months, P = 0.0339). Cox regression analysis (n = 174) confirmed a reduced risk of death with High Rash Severity (hazard ratio [HR] = 0.67, P = 0.0389). Progression-free survival results showed a similar pattern (median PFS 2.37 months from landmark vs 2.04 months for High vs Low Severity groups, P = 0.0485). CONCLUSIONS: Results from this community sample were consistent with findings from randomized clinical trials, showing that longer OS is predicted by high-severity rash in erlotinib-treated pancreatic cancer patients.

Bevacizumab maintenance in patients with advanced non-small-cell lung cancer, clinical patterns, and outcomes in the Eastern Cooperative Oncology Group 4599 Study: results of an exploratory analysis.

INTRODUCTION: The Eastern Cooperative Oncology Group (ECOG) 4599 study showed a significant survival benefit with the use of bevacizumab (BV) in combination with carboplatin and paclitaxel (CP) in comparison with CP chemotherapy alone in patients with previously untreated advanced, metastatic or recurrent non-small-cell lung cancer (NSCLC). Such results were achieved using BV as maintenance therapy until progressive disease. Because current data on single-agent BV maintenance in non-small-cell lung cancer are limited, we present a retrospective analysis of safety and efficacy outcomes for patients who received maintenance BV after induction treatment and the maintenance-eligible population of the control arm in ECOG 4599. METHODS: Landmark analyses were conducted in patients in both the CP and CP+BV groups who were alive and progression free through the completion of six cycles + 21 days. The BV maintenance population consisted of patients in the CP+BV arm, who were alive without progressive disease before the start of maintenance (maintenance-nonprogressor population). CP nonprogressors were those patients in the CP-alone arm without progressive disease after six cycles of CP + 21 days. RESULTS: Two hundred and seventeen patients (51%) were alive, progression free, and eligible for maintenance therapy six cycles + 21 days after induction CP+ BV compared with 134 patients (30%) in the CP-alone arm. Postinduction progression-free survival was significantly longer in the BV maintenance group relative to CP nonprogressors (4.4 versus 2.8 months; hazards ratio [HR] 0.64; p < 0.001). One-year overall survival rates were 75% for the BV maintenance group versus 69% in the CP nonprogressor group. Two-year overall survival rates were 34% for the BV maintenance group versus 25% in the CP nonprogressor group. Median postinduction overall survival (OS) was also significantly longer for the BV-maintenance group compared with CP nonprogressors (12.8 versus 11.4 months; HR 0.75; p = 0.030). Within the subgroup having complete response or partial response after induction, the progression-free survival and OS hazard ratio estimates were 0.59 (95% [confidence interval] CI: 0.41-0.84) and 0.78 (95% CI: 0.53-1.14), respectively. In the maintenance setting, BV was associated with a less-than 1% rate of grade 3 or 4 hematological toxicities, no grade 3 or 4 nausea, vomiting or diarrhea, and no grade 5 toxicities. CONCLUSIONS: In this retrospective analysis of patients in the ECOG 4599 study, who were alive, progression free, and on-study 21 days after six cycles of induction therapy, significant reductions in HRs for progression (0.64, p < 0.001) and survival (0.75, p = 0.03) were associated with BV treatment during induction and maintenance compared with CP induction therapy alone and suggestive of possible benefit because of bevacizumab maintenance.

Model Checking Techniques for Assessing Functional Form Specifications in Censored Linear Regression Models.

In this paper we develop model checking techniques for assessing functional form specifications of covariates in censored linear regression models. These procedures are based on a censored data analog to taking cumulative sums of "robust" residuals over the space of the covariate under investigation. These cumulative sums are formed by integrating certain Kaplan-Meier estimators and may be viewed as "robust" censored data analogs to the processes considered by Lin, Wei & Ying (2002). The null distributions of these stochastic processes can be approximated by the distributions of certain zero-mean Gaussian processes whose realizations can be generated by computer simulation. Each observed process can then be graphically compared with a few realizations from the Gaussian process. We also develop formal test statistics for numerical comparison. Such comparisons enable one to assess objectively whether an apparent trend seen in a residual plot reects model misspecification or natural variation. We illustrate the methods with a well known dataset. In addition, we examine the finite sample performance of the proposed test statistics in simulation experiments. In our simulation experiments, the proposed test statistics have good power of detecting misspecification while at the same time controlling the size of the test.

Phase II study of carboxyamidotriazole in patients with advanced renal cell carcinoma refractory to immunotherapy: E4896, an Eastern Cooperative Oncology Group Study.

BACKGROUND: The current study evaluated the response rate and 6-month time to disease progression of the antiangiogenesis agent carboxyamidotriazole (CAI) in patients with metastatic renal cell carcinoma (RCC). METHODS: Fifty-seven patients with histologically confirmed metastatic RCC that progressed after biologic therapy (interferon or interleukin-2) were enrolled. Four patients were ineligible. CAI was administered orally as a 28-day cycle. Response and time to disease progression were evaluated. RESULTS: Fifteen of 53 eligible patients received > 5 cycles, but 13 patients eventually discontinued treatment because of progressive disease. The majority of toxicities were Grade 1. However, Grade 3/4 toxicities did occur, the majority of which were gastrointestinal in nature. One of 47 patients evaluable achieved a partial response (1.9%) lasting 172 days. Six of 53 patients were alive and disease progression free at 6 months from the start of treatment (11.3%). The median overall survival was 12.5 months. The survival periods in the low-risk, intermediate-risk, and poor-risk groups were 16.2 months, 20.9 months, and 5.8 months, respectively. CONCLUSIONS: Patients in trials of second-line therapy appear to have a better prognosis than previously considered, in part because they are eligible for another clinical trial. CAI was found to have little to no effect on the natural history of progressive RCC.

Functional form diagnostics for Cox's proportional hazards model.

We propose a new type of residual and an easily computed functional form test for the Cox proportional hazards model. The proposed test is a modification of the omnibus test for testing the overall fit of a parametric regression model, developed by Stute, González Manteiga, and Presedo Quindimil (1998, Journal of the American Statistical Association93, 141-149), and is based on what we call censoring consistent residuals. In addition, we develop residual plots that can be used to identify the correct functional forms of covariates. We compare our test with the functional form test of Lin, Wei, and Ying (1993, Biometrika80, 557-572) in a simulation study. The practical application of the proposed residuals and functional form test is illustrated using both a simulated data set and a real data set.