Mandrekar, Sumithra J.

2577 Background: The emergence of MTA in oncology has revolutionized the phase I design paradigm. The usual dose-limiting toxicity (DLT) endpoint is an oversimplification of the complex toxicity profile resulting in loss of information. Moreover, MTAs can induce less DLTs but more multiple moderate toxicities that may lead to dose reduction in long lasting treatment. We thus sought to develop a new toxicity endpoint, the Total Toxicity Profile (TTP), to incorporate multiple toxicities. Methods: We reanalysed the phase I trial that evaluated erlotinib (75, 100, 125, 150 mg/m2) in children with brain stem glioma (NCT00418327). Only 2/21 patients experienced protocol defined DLTs, but multiple moderate toxicities were observed. The TTP endpoint is a weighted combination of all toxicities experienced by the patient (Euclidean norm of the weights), where the weights (ranging from 0 to 10) reflect the clinical importance of each grade and type of toxicity. The weights and the acceptable toxicity target (target TTP) were elicited from 3 clinicians using hypothetical patient cohorts with various toxicity outcomes. They were asked to make a decision of where the next cohort of patients would be treated: higher, lower or same dose level. The target TTP was defined as the mean of TTPs of the cohorts associated with a decision to repeat the dose. We used the TTP-driven Quasi-Likelihood Continual Reassessment Method (QL-CRM) to re-estimate the recommended dose. Results: Clinicians identified 5 main dermatological toxicity types possibly attributable to erlotinib. The consensus weights (table below) as well as the target TTP (set at 4.66) were easily obtained, confirming the feasibility of the process. The weights differed from the CTCAE grades, confirming the interest of the approach. The recommended dose based on the retrospective analysis of all grades toxicity was lower than that identified by the DLT-driven CRM (100 instead of 125mg/m2). Conclusions: By neglecting multiple moderate toxicities, DLT-driven designs may over-estimate the recommended dose when investigating MTA. Our proposal is an appealing alternative design for MTA. [Table: see text]

7605 Background: Overall survival (OS) can be partitioned into progression-free survival (PFS) and PPS. PPS helps to understand trial results, especially when PFS and OS data on trial treatment effects are discordant. At ASCO 2011, we reported that the magnitude of difference in the PFS estimates using different DP date methods was large enough to alter trial conclusions. Here, we investigate the impact of the DP date method on 1) PPS estimates, and 2) predictive utility of DP metrics on subsequent OS (SOS). Methods: Individual patient (pt) data from 14 trials were pooled. DP date was determined using: reported progression date (RPD) (method 1, M1), one day after last progression-free (PF) scan (M2), and midpoint between last PF scan and RPD (M3). PPS was estimated using the method of Kaplan-Meier for the 3 DP date methods. A flexible landmark analysis at 2, 4, and 6 months (mos) using Cox proportional hazards model was used to assess the impact of DP status (progression versus no-progression) on SOS (using M1, M2, or M3). Results: Among NSCLC (SCLC), 87% (91%) of pts reported DP. As expected, the PPS estimates were the lowest for RPD, highest for M2, and in-between for M3 (a direct consequence of the DP date method); with no difference by arm for the randomized trials. Regardless of the DP date method, patients who were progression-free had improved SOS (NSCLC: Hazard ratio, HR<=0.33; p < 0.0001; SCLC: HR<=0.48; p < 0.002) at each landmark time point, with comparable concordance index (SCLC: 0.57-0.65; NSCLC: 0.63-0.67), i.e., ability to discriminate patients with different SOS outcomes. Conclusions: While the DP date methods do not impact the predictive utility of the DP metrics, they significantly impact PPS estimates. The translation of a significant treatment effect on PFS to an effect on OS is influenced by PPS (longer PPS dilutes effect on OS). Standards for declaring DP date are thus critical to trial design and for trial go/no-go decisions. [Table: see text]

6133 Background: Patient (pt) trial experience may give insight to quality of life (QOL) factors affecting accrual, retention and outcome. While pts are vital to clinical trial success, we know little about their post-trial opinions. This trial examined pt opinion of their trial and treatment decision making (TDM). Methods: Pts enrolled on designated North Central Cancer Treatment Group (NCCTG) phase II or III treatment trials completed the TDM Control Preferences Scale at baseline and the Was It Worth It (WIWI) satisfaction assessment at the ends of cycle 1 and active protocol treatment. The primary endpoint was the proportion of pts reporting worthwhile participation. Fisher Exact tests compared pt opinion across subgroups. Results: As of 01/15/2012, 264 pts were enrolled on 25 protocols and treated at 79 sites. 86% of pts were in phase II studies and 89% had stage IV disease. At the end of cycle 1, pts felt the trial was worthwhile (74%), would do it again (85%), and would recommend it to others (82%). 85% of pts reported undiminished QOL and only 7% rated the trial worse than expected. End of treatment responses were similar. Satisfaction rates varied by tumor site (p=0.04). 11% of pts having tumor response rated the trial as not worthwhile, and 66% of pts with progressive disease rated it worthwhile (see table). Pts with concordant preferred and actual TDM roles rated participation higher than pts with discordant TDM roles (see table). Conclusions: Most pts endorsed their clinical trial experience. Contrary to popular beliefs, treatment outcome did not have an overwhelming impact on pt satisfaction, and was just one of many factors such as TDM role discordance. Assessing pt satisfaction will inform future study design that can potentially improve pt accrual and retention. [Table: see text]

7510 Background: We previously demonstrated that PFS may be a candidate surrogate endpoint for OS in ES-SCLC using data from 3 randomized trials (Foster, Cancer 2011). Here, we sought to formally assess the patient- and trial-level surrogacy of PFS using data from 9 additional randomized phase II and III trials conducted by the NCI-funded cancer cooperative groups since 1986. Methods: Individual patient data from all 12 trials (3178 patients: 9 phase III and 3 phase II) were pooled. OS was the primary endpoint in all phase III trials; 3 phase III and 1 phase II trial were positive per protocol. Patient-level surrogacy (Kendall’s tau) was assessed using the Clayton copula bivariate survival model. Trial-level surrogacy was assessed via association of the log hazard ratios on OS and PFS across trials, including: weighted (by trial size) least squares regression of Cox model effects (R² WLS) and weighted (by trial size) correlation of the copula effects (R² Copula). One trial had 4 treatment arms thus 14 total two-arm comparisons were made. Results: With a median follow-up of 41.8 months in the 106 patients still alive, the median OS and PFS across trials were 9.7 months (95% CI: 9.5, 9.9) and 5.7 months (95% CI: 5.5, 5.8), respectively. There were 3120 PFS events in total (2564 disease progressions and 556 deaths without progression). The median time from progression to death was 4.1 months (95% CI: 3.9, 4.3). PFS showed modest association with OS at the patient-level (tau= 0.56) and at the trial-level (R² WLS = 0.58; R² Copula (standard error) = 0.55 (0.29)). The 95% CIs for the predicted HR for OS given observed HR on PFS under a weighted leave-one-out prediction always included the observed HR for OS; however such intervals were wide, suggesting uncertainty on the practical use of PFS as a surrogate for OS in this setting. Conclusions: PFS failed to demonstrate surrogacy for OS in ES-SCLC based on this large pooled analysis. Given that the difference in the median PFS and OS is less than 6 months, we recommend using OS as the primary endpoint in phase III trials of previously untreated ES-SCLC.

7502 Background: Prior studies suggest that adjuvant brachytherapy reduces local recurrence (LR) after sublobar resection (SR) for NSCLC. A multicenter randomized study was undertaken in patients (pts) with stage I NSCLC ≤3cm comparing SR to SR with brachytherapy (SRB). Methods: High-risk operable patients with NSCLC were randomized to SR or SRB. Brachytherapy involved placement of I125 seeds incorporated into Vicryl sutures or into Vicryl mesh placed over the staple line. Wedge or segmental resection was allowed. The primary endpoint was time to local recurrence (LR) defined as recurrence within the primary tumor lobe at the staple line (local progression), away from the staple line or within hilar nodes. The trial was designed to have 90% power to detect a hazard ratio (HR) of 0.315 in favor of the SRB arm using a one-sided α of 0.05 with a sample size of 100 eligible pts per arm. Follow-up CT scans were obtained serially for 36 months. Results: 224 pts were randomized; 213 (109 SR,104 SRB) were eligible. Median (range) age was 71 (49-87) yrs; 94 (44%) were male. No differences were found in baseline characteristics. Adverse events, previously reported, were not different between arms. Median (range) follow-up was 4.06 (0.04, 5.0) yrs. There was no difference between arms in time to LR (HR = 0.87; 5% CI: 0.41, 1.86, p=0.72) or to LR or death (LRD) (HR = 0.81, 95% CI: 0.50, 1.32, p=0.40). There was no difference between arms in pattern of LR (table). In subgroups of pts with potentially compromised surgical margin (margin < 1cm; margin:tumor ratio <1; positive staple line cytology) SRB did not reduce LR or LRD at 3-yrs. Overall 3-yr survival was similar for SR (71%) and SRB (72%) (p=0.81). Conclusions: LR remains a concern after sublobar resection. However, local progression at the staple line was low. This trial demonstrated that adjuvant brachytherapy does not impact oncologic outcomes. Clinical trial information: NCT00107172. [Table: see text]

6520 Background: We previously reported that alternative cutpoints and alternate categorical metrics to RECIST standards provided no meaningful improvement in OS prediction. Here, we sought to identify and validate continuous, longitudinal TM-based metrics for predicting OS. Methods: Data from 13 trials (1739 patients with sufficient imaging-based measurements from breast, lung or colon cancer) were randomly split (60:40) into training and validation datasets and landmarked at 12-weeks (w). Multiple candidate metrics that capture tumor curve trajectory were considered: first slope (or % change) from baseline to 6 w, last slope (or % change) from 6 w to 12 w, and indicator of increase (<5% (includes no increase or a decrease) vs. ≥5%) from the prior assessment (baseline to 6 w, 6 w to 12 w). Multivariable Cox models stratified by study and number of baseline target lesions (<3 vs. >3) and adjusted for baseline tumor burden (ratio of the sum of lesions/number of lesions) were fit for each tumor type. Predictive ability was assessed via concordance (c)-index (0.5: no association to 1.0: perfect association), and hazard ratios/p-values (HR/p). Results: Patients whose tumor size increased by at least 5% from the prior assessment, as well as whose tumors showed a positive slope (or % change) from 6 to 12 w (indicating tumor growth) had worse OS. The first slope (or % change) metric was less consistent, with an initial decrease in slope suggesting worse OS in some cases. Predictive ability of these metrics varied across tumor type with poorest prediction in Breast (Table). Conclusions: Continuous TM-based metrics performed no better than categorical RECIST metrics for OS prediction. [Table: see text]

7555 Background: In a retrospective exploratory analysis of E4599, patients (pts) > 70 yo had a higher frequency of and more severe toxicities without apparent survival benefit from the addition of B to C+paclitaxel. We hypothesized that in this pt population, B will have better safety and efficacy profile when used in combination with C+P. Methods: Pts >/= 70 yo with previously untreated stage IIIB/IV (TNM 6th ed) nonsquamous NSCLC, ECOG PS 0-1, measurable disease and adequate organ function were eligible. C at AUC 6, P at 500 mg/m2 and B at 15 mg/kg were administered on day 1 of each 21-day cycle for up to 6 cycles followed by maintenance P+B in patients with CR, PR or SD. The primary endpoint was 6-month progression-free survival (PFS) rate. The treatment would be considered promising based on a single arm one-stage binomial design if 34 or more successes out of 55 patients were observed. This design had an exact significance level of 0.05 at 93% power to detect a true success rate of at least 70%. Polymorphisms in VEGFA, FPGS, GGH, SLC19A1 and TYMS in germline DNA were correlated with treatment outcome. Results: 58 eligible pts were enrolled; 29 males/29 females. Median age was 75. Median treatment cycles received was 6. Grade 3 or higher adverse events (AE) were reported in 49 (85%) pts. There were no treatment-related deaths. The most common grade 3/4 AEs(regardless of attribution) were hypertension (10%), fatigue (28%), dehydration (9%), neutropenia (43%) and thrombocytopenia (21%). There were 3 (5%) grade 3/4 hemorrhagic events. 8 (14%) had grade 4 neutropenia and 3 (5%) had grade 4 thrombocytopenia. Grade 3/4 ischemic/thromboembolic events occurred in 6 pts (10%). Thirty-four out of the first 54 (63%, 95% CI: 48.7-75.7%) evaluable pts met the primary endpoint (4 pts were lost to follow-up prior to 6 months). The confirmed ORR was 37.9% (95% CI: 25.5-51.6%). Median time to treatment failure was 4.8 months (95% CI: 3.9-6.4). Median PFS was 7.1 months (95% CI: 5.9-11.7), median OS was 13.7 months (95% CI: 9.4-15.7). Results of SNP analysis will be presented. Conclusions: C+P+B followed by maintenance P+B is an active and tolerable first-line regimen for elderly patients with good PS.

6006 Background: In March 1998, the Common Toxicity Criteria version 2.0 was implemented and introduced the collection and reporting of attribution of adverse events to study treatment. Collection and reporting of attribution adds time and cost to the clinical trial process, and its use while reporting adverse events (AE) in the literature has varied. We investigate whether attribution adds value to the interpretation of AE data. Methods: Patients on the placebo arm of trial 97–24–51 (CAI vs. Placebo) were chosen since the true relationship of the event to the study treatment is known (“unrelated”). Attribution was collected per CTCv2.0 and categorized as “not related” (not related or unlikely) and “related” (possible, probable, or definite). All reported AEs and the maximum severity for a given AE were evaluated. Patterns were evaluated using generalized estimating equations adjusting for multiple events per patient. Changes in attribution over time for the same event were computed. Results: 84 patients on the placebo arm experienced a total of 1013 AEs. 47% of AEs were reported as “related” with 36% reported as possibly related. Known CAI AEs were more likely to be reported as “related” (p = 0.005). Pulmonary AEs were more likely to be reported as “not related” (p = 0.0006). No patterns were observed by gender, age, PS, severity or treatment cycle of AE. When the same event was reported on the same patient in multiple cycles, the attribution category changed at least once 36% of the time with 25% changing from “related” to “unrelated” or vice versa. Conclusion: Almost 50% of events on a placebo arm of a phase III trial were reported as being attributed to study treatment and 25% of the time the same event was not consistently attributed. These data suggest that attribution is difficult to determine, unreliable, and thus its value in large phase III trials is questionable. [Table: see text] No significant financial relationships to disclose.

13013 Background: The glutathione metabolic pathway is directly involved in the inactivation of platinum compounds. Our earlier work indicated that genotypes corresponding to varied glutathione-related enzyme activities may predict survival in patients with advanced lung cancer. Herein, we evaluated the role of glutathione-related genotypes on clinical outcomes in stage IIIB/IV non-small cell lung cancer patients who were stable or responding from one prior platinum-based chemotherapy. Methods: DNA samples from patients who had tumor regression or stable disease after treatment with one platinum based chemotherapy regimen were analyzed using 6 polymorphic DNA markers that encode 5 important enzymes in the glutathione metabolic pathway. The contrasting genotypes used in the analysis were GCLC (homozygous repeat 77 vs. heterozygous 7*), GPX1 (CC vs. other), GSTP1-I105V (AA vs. other), GSTP1-A114V (CC vs. other), GSTM1 (null vs. present), and GSTT1 (null vs. present). Multivariable Cox proportional hazards models adjusted for age, gender, treatment arm, ECOG performance status, stage, and prior tumor response, were used to evaluate the prognostic significance of the genotypes at each locus. Results: Data from 112 patients with a median follow-up of 10.6 (range: 0.4–56.3) months was used. No significant difference in time to disease progression was observed for any of the genotypes. Among the 6 genomic DNA markers, the GCLC genotype was an important prognostic factor for overall survival after adjusting for other factors. In particular, patients carrying a 77 repeat genotype (homozygous) had a significantly worse survival (Hazard ratio = 1.67, 95% CI: 1.05–2.63). Conclusions: Genotypes of glutathione-related enzymes, especially GCLC, may be used as host factors in predicting patients’ prognosis after platinum-based chemotherapy. Further investigation to define and measure the effects of these genes in chemotherapeutic regimens is needed. (This work was partly supported by NIH research grants CA77118, CA80127, and CA84354.) No significant financial relationships to disclose.

18037 Background: Platinum compounds are major chemotherapeutic agents in lung cancer and are metabolized by the glutathione pathway enzymes. We reported that genotypes of glutathione-related enzymes, especially GCLC, may be host factors in predicting survival of patients with stage IIIB/IV non-small cell lung cancer who were stable or responding from prior platinum-based chemotherapy (Yang et al., ASCO 2006). NCCTG investigators previously correlated allelic variants in 5-FU metabolizing genes with QOL in colorectal cancer patients (Sloan et al., ASCO 2004). We sought to evaluate the role of glutathione pathway genotypes on QOL. Methods: DNA samples were analyzed using 6 polymorphic DNA markers in the glutathione metabolic pathway. The contrasting genotypes in the analysis were GSTM1 and GSTT1 (null vs. present), GSTP1-I105V (AA vs. GA/GG), GSTP1-A114V (CC vs. CT/TT), GPX1 (CC vs. TT/TC), and GCLC (homozygous repeat 77 vs. heterozygous 7*). Forty-six patients completed the FACT-L and the UNISCALE QOL questionnaires at baseline and week-8, and we have results for all 6 genotypes. A clinically significant decline (CSD) in QOL was defined as a 10% decrease from baseline to week-8. A multivariate logistic regression model was used to evaluate the association of all 6 genotypes with a CSD in QOL. Results: Patients carrying the GPX1-CC genotype had a CSD in the UNISCALE (Odds Ratio (OR): 10.0; p=0.02), total FACT-L score (OR: 7.9; p=0.03), the FACT-L physical well being construct (OR: 13.7; p=0.02), and the FACT-L functional well being construct (OR: 5.3; p=0.05). Other constructs, i.e., emotional, social and family, and additional concerns were not significantly associated with the genotypes. Conclusions: Genotypes of glutathione-related enzymes, especially GPX1 may be inherited factors in predicting patients’ QOL after platinum-based chemotherapy. Further investigation to define and measure the direct or indirect effects of these genes on QOL is critical, particularly, via drug responses, toxicities, and disease recurrence. (This work was partly supported by NIH grants CA77118, CA80127, and CA84354.) No significant financial relationships to disclose.