Response- and Progression-Based End Points in Trial and Observational Cohorts of Patients With NSCLC.

Importance: Response Evaluation Criteria in Solid Tumors (RECIST) are commonly used to assess therapeutic response in clinical trials but not in routine care; thus, RECIST-based end points are difficult to include in observational studies. Clinician-anchored approaches for measuring clinical response have been validated but not widely compared with clinical trial data, limiting their use as evidence for clinical decision-making. Objective: To compare response- and progression-based end points in clinical trial and observational cohorts of patients with non-small cell lung cancer (NSCLC). Design, Setting, and Participants: This retrospective cohort study used patient-level data from the IMpower132 trial (conducted April 7, 2016, to May 31, 2017) and a nationwide electronic health record (EHR)-derived deidentified database (data collected January 1, 2011, to March 31, 2022). Patients in the observational cohort were selected according to the inclusion and exclusion criteria of the IMpower132 trial. All patients in the observational cohort had stage IV NSCLC. Exposure: All patients were randomized to or received first-line carboplatin or cisplatin plus pemetrexed. Main Outcomes and Measures: End points included response rates, duration of response, and progression-free survival, compared between the trial and observational cohorts before and after weighting. Response rates for the observational cohort were derived from the EHR. Results: A total of 769 patients met inclusion criteria, 494 in the observational cohort (median [IQR] age, 67 [60-74] years; 228 [46.2%] female; 45 [9.1%] Black or African American; 352 [71.3%] White; 53 [10.7%] American Indian or Alaska Native, Asian, Hawaiian or Pacific Islander, or multiracial) and 275 in the trial cohort (median [IQR] age, 63 [56-68] years; 90 [32.7%] female; 4 [1.5%] Black or African American; 194 [70.5%] White; 65 [23.6%] American Indian or Alaska Native, Asian, Hawaiian or Pacific Islander, or multiracial). All 3 end points were comparable between the study cohorts. Trial patients had a higher number of response assessments compared with patients in the weighted observational cohort. The EHR-derived response rate was numerically higher than the objective response rate after weighting (100.3 of 249.3 [40.2%] vs 105 of 275 [38.2%]) due to higher rates of observed partial response than RECIST-based partial response. Among patients with at least 1 response assessment, the EHR-derived response rate remained higher than the objective response rate (100.3 of 193.4 [51.9%] vs 105 of 256 [41.0%]) due to a higher proportion of patients in the observational cohort with no response assessment. Conclusions and Relevance: In this study, response- and progression-based end points were similar between clinical trial and weighted observational cohorts, which increases confidence in the reliability of observational end points and can inform their interpretation in relation to trial end points. Additionally, the difference observed in response rates (including vs excluding patients with no response assessment) highlights the importance of future research adopting this 2-way approach when evaluating the relationship of EHR-derived and objective response rates.

Considerations for pooling real-world data as a comparator cohort to a single arm trial: a simulation study on assessment of heterogeneity.

BACKGROUND: Novel precision medicine therapeutics target increasingly granular, genomically-defined populations. Rare sub-groups make it challenging to study within a clinical trial or single real-world data (RWD) source; therefore, pooling from disparate sources of RWD may be required for feasibility. Heterogeneity assessment for pooled data is particularly complex when contrasting a pooled real-world comparator cohort (rwCC) with a single-arm clinical trial (SAT), because the individual comparisons are not independent as all compare a rwCC to the same SAT. Our objective was to develop a methodological framework for pooling RWD focused on the rwCC use case, and simulate novel approaches of heterogeneity assessment, especially for small datasets. METHODS: We present a framework with the following steps: pre-specification, assessment of dataset eligibility, and outcome analyses (including assessment of outcome heterogeneity). We then simulated heterogeneity assessments for a binary response outcome in a SAT compared to two rwCCs, using standard methods for meta-analysis, and an Adjusted Cochran's Q test, and directly comparing the individual participant data (IPD) from the rwCCs. RESULTS: We found identical power to detect a true difference for the adjusted Cochran's Q test and the IPD method, with both approaches superior to a standard Cochran's Q test. When assessing the impact of heterogeneity in the null scenario of no difference between the SAT and rwCCs, a lack of statistical power led to Type 1 error inflation. Similarly, in the alternative scenario of a true difference between SAT and rwCCs, we found substantial Type 2 error, with underpowered heterogeneity testing leading to underestimation of the treatment effect. CONCLUSIONS: We developed a methodological framework for pooling RWD sources in the context of designing a rwCC for a SAT. When testing for heterogeneity during this process, the adjusted Cochran's Q test matches the statistical power of IPD heterogeneity testing. Limitations of quantitative heterogeneity testing in protecting against Type 1 or Type 2 error indicate these tests are best used descriptively, and after careful selection of datasets based on clinical/data considerations. We hope these findings will facilitate the rigorous pooling of RWD to unlock insights to benefit oncology patients.

A systematic approach towards missing lab data in electronic health records: A case study in non-small cell lung cancer and multiple myeloma.

Real-world data derived from electronic health records often exhibit high levels of missingness in variables, such as laboratory results, presenting a challenge for statistical analyses. We developed a systematic workflow for gathering evidence of different missingness mechanisms and performing subsequent statistical analyses. We quantify evidence for missing completely at random (MCAR) or missing at random (MAR), mechanisms using Hotelling's multivariate t-test, and random forest classifiers, respectively. We further illustrate how to apply sensitivity analyses using the not at random fully conditional specification procedure to examine changes in parameter estimates under missing not at random (MNAR) mechanisms. In simulation studies, we validated these diagnostics and compared analytic bias under different mechanisms. To demonstrate the application of this workflow, we applied it to two exemplary case studies with an advanced non-small cell lung cancer and a multiple myeloma cohort derived from a real-world oncology database. Here, we found strong evidence against MCAR, and some evidence of MAR, implying that imputation approaches that attempt to predict missing values by fitting a model to observed data may be suitable for use. Sensitivity analyses did not suggest meaningful departures of our analytic results under potential MNAR mechanisms; these results were also in line with results reported in clinical trials.

Emulating Randomized Controlled Trials with Hybrid Control Arms in Oncology: A Case Study.

This proof-of-concept study retrospectively assessed the feasibility of applying a hybrid control arm design to a completed phase III randomized controlled trial (RCT; CheckMate-057) in advanced non-small cell lung cancer using a real-world data (RWD) source. The emulated trial consists of an experimental arm (patients from the RCT experimental cohort) and a hybrid control arm (patients from the RCT and RWD control cohorts). For the RWD control cohort, this study used a nationwide electronic health record-derived de-identified database. Three frequentist statistical borrowing methods were evaluated: a two-step Cox model, a fixed Cox model, and propensity score-integrated composite likelihood ("Methods 1-3"). The experimental treatment effect for hybrid control designs were evaluated using hazard ratios (HRs) with 95% confidence interval (CI) estimated from the Cox models accounting for covariate differences. The reduction in study duration compared to the RCT was also evaluated. All three statistical borrowing methods achieved comparable experimental treatment effects to that observed in the CheckMate-057 clinical trial, with HRs of 0.73 (95% CI: 0.59, 0.92), 0.74 (95% CI: 0.61, 0.91), 0.72 (95% CI: 0.59, 0.88) for Methods 1-3, respectively. Reduction in study duration time was 99-115 days when borrowing 30-38 events for Methods 1-3, respectively. This study demonstrated that it is feasible to emulate an RCT using a hybrid control arm design using three frequentist propensity-score based statistical borrowing methods. Selection of an appropriate, fit-for-use RWD cohort is critical to minimizing bias in experimental treatment effect.

Augmenting control arms with real-world data for cancer trials: Hybrid control arm methods and considerations.

Background: Hybrid controlled trials with real-world data (RWD), where the control arm is composed of both trial and real-world patients, could facilitate research when the feasibility of randomized controlled trials (RCTs) is challenging and single-arm trials would provide insufficient information. Methods: We propose a frequentist two-step borrowing method to construct hybrid control arms. We use parameters informed by a completed randomized trial in metastatic triple-negative breast cancer to simulate the operating characteristics of dynamic and static borrowing methods, highlighting key trade-offs and analytic decisions in the design of hybrid studies. Results: Simulated data were generated under varying residual-bias assumptions (no bias: HRRWD = 1) and experimental treatment effects (target trial scenario: HRExp = 0.78). Under the target scenario with no residual bias, all borrowing methods achieved the desired 88% power, an improvement over the reference model (74% power) that does not borrow information externally. The effective number of external events tended to decrease with higher bias between RWD and RCT (i.e. HRRWD away from 1), and with weaker experimental treatment effects (i.e. HRExp closer to 1). All dynamic borrowing methods illustrated (but not the static power prior) cap the maximum Type 1 error over the residual-bias range considered. Our two-step model achieved comparable results for power, type 1 error, and effective number of external events borrowed compared to other borrowing methodologies. Conclusion: By pairing high-quality external data with rigorous simulations, researchers have the potential to design hybrid controlled trials that better meet the needs of patients and drug development.

Concordance of Clinician-Documented and Imaging Response in Patients With Stage IV Non-Small Cell Lung Cancer Treated With First-Line Therapy.

Importance: In observational oncology studies of solid tumors, response to treatment can be evaluated based on electronic health record (EHR) documentation (clinician-assessed response [CAR]), an approach different from standardized radiologist-measured response (Response Evaluation Criteria in Solid Tumours [RECIST] 1.1). Objective: To evaluate the feasibility of an imaging response based on RECIST (IRb-RECIST) and the concordance between CAR and imaging response based on RECIST assessments, and investigate discordance causes. Design, Setting, and Participants: This cohort study used an EHR-derived, deidentified database that included patients with stage IV non-small cell lung cancer (NSCLC) diagnosed between January 1, 2011, to June 30, 2019, selected from 3 study sites. Data analysis was conducted in August, 2020. Exposures: Undergoing first-line therapy and imaging assessments of response to treatment. Main Outcomes and Measures: In this study, CAR assessments (referred to in prior publications as "real-world response" [rwR]) were defined as clinician-documented changes in disease burden at radiologic evaluation time points; they were abstracted manually and assigned to response categories. The RECIST-based assessments accommodated routine practice patterns by using a modified version of RECIST 1.1 (IRb-RECIST), with independent radiology reads. Concordance was calculated as the percent agreement across all response categories and across a dichotomous stratification (response [complete or partial] vs no response), unconfirmed or confirmed. Results: This study found that, in 100 patients evaluated for concordance, agreement between CAR and IRb-RECIST was 71% (95% CI, 61%-80%), and 74% (95% CI, 64%-82%) for confirmed and unconfirmed response, respectively. There were more responders using CAR than IRb-RECIST (40 vs 29 with confirmation; 64 vs 43 without confirmation). The main sources of discordance were the different use of thresholds for tumor size changes by RECIST vs routine care, and unavailable baseline or follow-up scans resulting in inconsistent anatomic coverage over time. Conclusions and Relevance: In this cohort study of patients with stage IV NSCLC, we collected routine-care imaging, showing the feasibility of response evaluation using IRb-RECIST criteria with independent centralized review. Concordance between CAR and centralized IRb-RECIST was moderate. Future work is needed to evaluate the generalizability of these results to broader populations, and investigate concordance in other clinical settings.

Characterization of a Real-World Response Variable and Comparison with RECIST-Based Response Rates from Clinical Trials in Advanced NSCLC.

INTRODUCTION: Effectiveness metrics for real-word research, analogous to clinical trial ones, are needed. This study aimed to develop a real-world response (rwR) variable applicable to solid tumors and to evaluate its clinical relevance and meaningfulness. METHODS: This retrospective study used patient cohorts with advanced non-small cell lung cancer from a nationwide, de-identified electronic health record (EHR)-derived database. Disease burden information abstracted manually was classified into response categories anchored to discrete therapy lines (per patient-line). In part 1, we quantified the feasibility and reliability of data capture, and estimated the association between rwR status and real-world progression-free survival (rwPFS) and real-world overall survival (rwOS). In part 2, we investigated the correlation between published clinical trial overall response rates (ORRs) and real-world response rates (rwRRs) from corresponding real-world patient cohorts. RESULTS: In part 1, 85.4% of patients (N = 3248) had at least one radiographic assessment documented. Median abstraction time per patient-line was 15.0 min (IQR 7.8-28.1). Inter-abstractor agreement on presence/absence of at least one assessment was 0.94 (95% CI 0.92-0.96; n = 503 patient-lines abstracted in duplicate); inter-abstractor agreement on best confirmed response category was 0.82 (95% CI 0.78-0.86; n = 384 with at least one captured assessment). Confirmed responders at a 3-month landmark showed significantly lower risk of death and progression in rwOS and rwPFS analyses across all line settings. In part 2, rwRRs (from 12 rw cohorts) showed a high correlation with trial ORRs (Spearman's ρ = 0.99). CONCLUSIONS: We developed a rwR variable generated from clinician assessments documented in EHRs following radiographic evaluations. This variable provides clinically meaningful information and may provide a real-world measure of treatment effectiveness.

Exposure-response analyses of trastuzumab emtansine in patients with HER2-positive advanced breast cancer previously treated with trastuzumab and a taxane.

PURPOSE: In the phase III EMILIA study, trastuzumab emtansine (T-DM1) significantly improved progression-free survival (PFS) and overall survival (OS) versus capecitabine plus lapatinib (control) in previously treated human epidermal growth factor receptor 2-positive advanced breast cancer. Using EMILIA data, we evaluated the T-DM1 exposure-response relationship. METHODS: Exposure-response relationships were examined with four exposure metrics [model-predicted and observed minimum concentration (C min) and area under the concentration-time curve from time zero to day 21 of T-DM1 at cycle 1] and multiple efficacy (OS, PFS, objective response rate) and safety (grade ≥ 3 adverse events, grade ≥ 3 thrombocytopenia, grade ≥ 3 hepatotoxicity) endpoints. RESULTS: An apparent exposure-response trend was observed between model-predicted exposure metrics and efficacy; trends for observed exposure metrics were shallower and often not significant. Although median OS and PFS were numerically longer in patients with higher versus lower model-predicted cycle 1 C min, OS and PFS hazard ratios for T-DM1-treated patients in the lowest exposure quartile (Q1) versus control were < 1 after adjusting for baseline risk factors (e.g., ECOG status, tumor burden, measurable disease, and number of disease sites). No meaningful exposure-response relationship was observed for any safety endpoints. CONCLUSION: Exposure-response relationships for efficacy were inconsistent across exposure metrics; model-predicted cycle 1 C min showed the strongest exposure-response trend. The Q1 subgroup based on model-predicted cycle 1 C min had numerically similar or better OS and PFS versus control following covariate adjustment. The approved T-DM1 dose (3.6 mg/kg every 3 weeks) has a positive benefit-risk ratio versus control, even for the T-DM1 Q1 subgroup.

Relationship between Tumor Biomarkers and Efficacy in EMILIA, a Phase III Study of Trastuzumab Emtansine in HER2-Positive Metastatic Breast Cancer.

PURPOSE: HER2-positive breast cancer is heterogeneous. Some tumors express mutations, like activating PIK3CA mutations or reduced PTEN expression, that negatively correlate with response to HER2-targeted therapies. In this exploratory analysis, we investigated whether the efficacy of trastuzumab emtansine (T-DM1), an antibody-drug conjugate comprised of the cytotoxic agent DM1 linked to the HER2-targeted antibody trastuzumab, was correlated with the expression of specific biomarkers in the phase III EMILIA study. EXPERIMENTAL DESIGN: Tumors were evaluated for HER2 (n = 866), EGFR (n = 832), and HER3 (n = 860) mRNA expression by quantitative reverse transcriptase PCR; for PTEN protein expression (n = 271) by IHC; and for PIK3CA mutations (n = 259) using a mutation detection kit. Survival outcomes were analyzed by biomarker subgroups. T-DM1 was also tested on cell lines and in breast cancer xenograft models containing PIK3CA mutations. RESULTS: Longer progression-free survival (PFS) and overall survival (OS) were observed with T-DM1 compared with capecitabine plus lapatinib in all biomarker subgroups. PIK3CA mutations were associated with shorter median PFS (mutant vs. wild type: 4.3 vs. 6.4 months) and OS (17.3 vs. 27.8 months) in capecitabine plus lapatinib-treated patients, but not in T-DM1-treated patients (PFS, 10.9 vs. 9.8 months; OS, not reached in mutant or wild type). T-DM1 showed potent activity in cell lines and xenograft models with PIK3CA mutations. CONCLUSIONS: Although other standard HER2-directed therapies are less effective in tumors with PI3KCA mutations, T-DM1 appears to be effective in both PI3KCA-mutated and wild-type tumors. Clin Cancer Res; 22(15); 3755-63. ©2016 AACR.

Trastuzumab emtansine in human epidermal growth factor receptor 2-positive metastatic breast cancer: an integrated safety analysis.

PURPOSE: The antibody-drug conjugate trastuzumab emtansine (T-DM1) combines the cytotoxic activity of DM1 with the human epidermal growth factor receptor 2 (HER2) -targeted, antitumor properties of trastuzumab. T-DM1 has shown activity in phase I and II single-arm studies in patients with pretreated HER2-positive metastatic breast cancer (MBC) and has demonstrated superior efficacy and improved tolerability versus standard MBC treatments in randomized phase II and III studies. This analysis, combining available data from all single-agent T-DM1 studies to date, was conducted to better define the T-DM1 safety profile. PATIENTS AND METHODS: Six studies in patients with HER2-positive MBC who received T-DM1 3.6 mg/kg every 3 weeks and follow-up data from patients in an extension study were analyzed. Analyses included adverse events (AEs) by grade; AEs leading to death, drug discontinuation, or dose reduction; and select AEs. RESULTS: Among 884 T-DM1-exposed patients, the most commonly reported all-grade AEs were fatigue (46.4%), nausea (43.0%), thrombocytopenia (32.2%), headache (29.4%), and constipation (26.5%). The most common grade 3 to 4 AEs were the laboratory abnormalities of thrombocytopenia (11.9%) and increased AST serum concentration (4.3%). These were manageable and not generally associated with clinical symptoms. There were 12 AE-related deaths. AEs resulted in dose reductions in 17.2% of patients and drug discontinuations in 7.0%. CONCLUSION: In this analysis of 884 T-DM1-exposed patients, grade 3 or greater AEs were infrequent and typically asymptomatic and manageable. This favorable safety profile makes T-DM1 treatment suitable for exploration in other breast cancer settings.

Population pharmacokinetics of trastuzumab emtansine (T-DM1), a HER2-targeted antibody-drug conjugate, in patients with HER2-positive metastatic breast cancer: clinical implications of the effect of covariates.

PURPOSE: Trastuzumab emtansine (T-DM1) is an antibody-drug conjugate comprising the humanized monoclonal antibody trastuzumab linked to DM1, a highly potent cytotoxic agent. A population pharmacokinetic (PK) analysis was performed to estimate typical values and interindividual variability of T-DM1 PK parameters and the effects of clinically relevant covariates. METHODS: Serum samples were collected from 671 patients with human epidermal growth factor receptor 2-positive locally advanced or metastatic breast cancer (MBC) who received single-agent T-DM1 in five phase I to phase III studies. Nonlinear mixed-effects modeling with the first-order conditional estimation method was used. RESULTS: A linear two-compartment model with first-order elimination from the central compartment described T-DM1 PKs in the clinical dose range. T-DM1 elimination clearance was 0.676 L/day, volume of distribution in the central compartment (V c) was 3.127 L, and terminal elimination half-life was 3.94 days. Age, race, region, and renal function did not influence T-DM1 PK. Given the low-to-moderate effect of all statistically significant covariates on T-DM1 exposure, none of these covariates is expected to result in a clinically meaningful change in T-DM1 exposure. CONCLUSIONS: T-DM1 PK properties are consistent and predictable in patients. A further refinement of dose based on baseline covariates other than body weight for the current 3.6 mg/kg regimen would not yield clinically meaningful reductions in interindividual PK variability in patients with MBC.

Response as a predictor of survival in patients with recurrent glioblastoma treated with bevacizumab.

Development of effective therapies for recurrent glioblastoma multiforme (GBM) and reliable, timely evaluation of their benefit are needed. Understanding the relationship between objective response (OR) and survival is important for determining whether OR can provide an early signal of treatment activity in clinical trials. We performed a landmark analysis to evaluate the association between OR and survival at 9, 18, and 26 weeks for 167 patients with recurrent GBM who participated in BRAIN, a phase II trial that evaluated efficacy of bevacizumab alone or in combination with irinotecan, using the Cox regression models adjusted for age, baseline Karnofsky performance score, first vs second relapse, and treatment arm. Hazard ratios (HRs) and P-values for survival between responders and nonresponders were calculated. Additional analyses were performed to test robustness, validity, fit, and accuracy of the models. The relationships between progression-free survival (PFS) and survival and between OR and PFS were also explored. There were 55 responders and 112 nonresponders across the 2 treatment arms in BRAIN. OR status at 9, 18, and 26 weeks was a statistically significant predictor of survival (HR ≤ 0.52, P < .01). PFS was also a statistically significant predictor of survival at each landmark (HR ≤ 0.25, P < .0001). The association between OR and PFS was not statistically significant, likely due to inadequate statistical power for the analysis. Clarifying the relationship of OR and survival is important for determining whether OR can be a reliable predictor of the benefit of a therapeutic agent in patients with recurrent GBM.

Corticosteroid use in patients with glioblastoma at first or second relapse treated with bevacizumab in the BRAIN study.

BACKGROUND: Vascular endothelial growth factor inhibitors have corticosteroid-sparing effects in patients with high-grade gliomas. We assessed corticosteroid use in patients with recurrent glioblastoma treated with bevacizumab (BEV) in the BRAIN study (J Clin Oncol 2009;27:4733-4740). METHODS: BRAIN was a phase II, multicenter, randomized, noncomparative trial of BEV alone (n = 85) or in combination with irinotecan (CPT-11) (n = 82) in adults with recurrent glioblastoma. Median corticosteroid dose for patients who used corticosteroids at baseline was summarized by treatment arm; the percentage of patients who had sustained (≥50% corticosteroid dose reduction for ≥50% of time on study drug) or complete (discontinuation of corticosteroid for ≥25% of time on study drug) reduction in corticosteroid dose overall and by objective response and progression-free survival was calculated. The incidence of corticosteroid-related adverse events was summarized. RESULTS: In each treatment group, 50% of patients were using systemic corticosteroids at baseline. The majority of those experienced a reduction in dose while receiving BEV-based therapy. Thirteen (30.2%) BEV and 20 (46.5%) BEV + CPT-11 patients had a sustained reduction of corticosteroid dose; 7 (16.3%) BEV and 9 (20.9%) BEV + CPT-11 patients had a complete reduction of corticosteroid dose. The majority of patients who had an objective response or progression-free survival >6 months experienced corticosteroid dose reduction. Approximately 64% of patients who used corticosteroids while receiving BEV-based therapy experienced infection. CONCLUSION: BEV may have corticosteroid-sparing effects in patients with recurrent glioblastoma. Corticosteroid reduction may positively affect patient health-related quality of life. Given the exploratory nature of the analyses in a noncomparative study, these results should be interpreted cautiously.

Safety and efficacy of combining sunitinib with bevacizumab + paclitaxel/carboplatin in non-small cell lung cancer.

INTRODUCTION: Bevacizumab (B) improves survival of patients with metastatic, nonsquamous non-small cell lung cancer. Based on encouraging results from preclinical studies combining B with sunitinib (S), a phase II, randomized, open-label study (Study Assessing the Blockade of both VEGF Receptor and ligand to enhance Efficacy in Lung) was initiated to assess clinical outcomes of adding S to paclitaxel (P)/carboplatin (C) + B (PCB) for first-line treatment of locally advanced, metastatic, or recurrent nonsquamous non-small cell lung cancer. METHODS: Study enrollment was to occur in three phases. In the first phase, patients received PC + B (15 mg/kg every 3 weeks), +/-S (25 mg daily, 2 weeks on, 1 week off). If tolerated, the second phase would include a third cohort receiving 37.5 mg S. The third phase would consist of PCB +/- highest tolerable dose S. RESULTS: Between March 2007 and January 2008, 26 patients were randomized to receive PCB and 30 to PCB + S 25 mg. Because of poor tolerability, none of the patients were escalated to 37.5 mg S. Median treatment duration was 10.3 weeks for PCB and 6.0 weeks for PCB + S. Thirty-five percent of patients on PCB + S required S dose reduction, 52% required S treatment interruption, and 59% discontinued S because of adverse events, most frequently hematologic events (neutropenia, thrombocytopenia, and leukopenia) and fatigue. Patients receiving PCB + S required more B interruptions (38% versus 19% for PCB) and discontinuation (52% versus 35%) because of adverse events. Survival data were limited by small sample sizes and limited treatment duration. Overall survival was not mature at time of analysis: median 6.6 months for PCB + S and not reached for PCB. Two out of 25 efficacy-evaluable patients randomized to the PCB + S cohort had confirmed partial responses, compared with 5 of 19 randomized to the PCB cohort. CONCLUSIONS: The addition of S to PCB was not well tolerated because of toxicities. This combination should not be studied further at these doses and schedules.

Retrospective evaluation of the clinical and radiographic risk factors associated with severe pulmonary hemorrhage in first-line advanced, unresectable non-small-cell lung cancer treated with Carboplatin and Paclitaxel plus bevacizumab.

PURPOSE: Severe (grade >or= 3) pulmonary hemorrhage (PH) in advanced non-small-cell lung cancer was observed in two prospective, randomized, phase II (N = 99) and phase III (N = 878) trials of bevacizumab plus carboplatin and paclitaxel. Retrospective case-control and cohort analyses were conducted to identify associated radiographic and clinical risk factors for PH. PATIENTS AND METHODS: Six patients with PH from the phase II trial, 15 potential PH patients with hemorrhage at any site from the phase III trial, and their matched controls were evaluated with review of baseline and on-treatment radiographs by an independent radiology facility, blinded to patient/control status. Patients with severe (grade >or= 3) PH from each trial were matched with up to three controls based on sex, age group, histology (phase II), or sex and age group (phase III). RESULTS: Seven PH patients in the phase III trial were identified as severe PH. Six of the patients were early onset (occurred < 150 days of initiating bevacizumab) and one was late onset. Baseline tumor cavitation, not tumor location, was identified as the only potential risk factor for patients with early onset. Combined analysis of severe PH patients from the phase II and phase III trials (n = 13), compared with their pooled matched controls (n = 42), did not identify any additional baseline radiographic or clinical variables associated with PH. CONCLUSION: PH was an uncommon event. Based on these analyses, baseline tumor cavitation may be a potential risk factor for PH. No other baseline clinical variables were predictive for PH although the number of events was small.