Benjamin Haaland

E20570 Background: More evidence is needed to select the best treatment strategy for patients harboring EGFR mutations and its subtypes Methods: We performed a systematic search and included studies reporting OS and/or PFS efficacy estimates by EGFR mutation status or subtypes (Del19/L858R). Hazard ratios of competing treatments were pooled using a Bayesian hierarchical model incorporating both within and between study heterogeneities. Treatment benefits were evaluated using posterior hazard ratios with 95% credible intervals (CrI) and ranked by surface under the cumulative ranking curve (SUCRA) for OS benefit Results: 4,177 records were screened and 20 trials were meta-analyzed. Statistically significant OS and PFS benefits were seen with (i) first-line intercalated chemotherapy+erlotinib in EGFR mutation positive, and (ii) first-line afatinib in Deletion 19. In L858R, no OS benefit was seen although treatments showed PFS benefits Conclusions: Among all available strategies, first-line intercalated chemotherapy+erlotinib in EGFR mutation positive and afatinib in Del 19 are the only strategies showing both OS and PFS benefits

Recent implementations of local approximate Gaussian process models have pushed computational boundaries for non-linear, non-parametric prediction problems, particularly when deployed as emulators for computer experiments. Their flavor of spatially independent computation accommodates massive parallelization, meaning that they can handle designs two or more orders of magnitude larger than previously. However, accomplishing that feat can still require massive supercomputing resources. Here we aim to ease that burden. We study how predictive variance is reduced as local designs are built up for prediction. We then observe how the exhaustive and discrete nature of an important search subroutine involved in building such local designs may be overly conservative. Rather, we suggest that searching the space radially, i.e., continuously along rays emanating from the predictive location of interest, is a far thriftier alternative. Our empirical work demonstrates that ray-based search yields predictors with accuracy comparable to exhaustive search, but in a fraction of the time - bringing a supercomputer implementation back onto the desktop.

217 Background: Recently Abiraterone (A) and Enzalutamide (E), new drugs with different novel mechanisms of action, have been approved for the treatment of metastatic castration-resistant prostate cancer (CRPC) based on the COU-AA-301 and AFFIRM phase III trials, after docetaxel failure. The lack of direct comparative trials has introduced a new treatment dilemma for the practicing oncologist: is one drug more effective than the other? Methods: We performed an indirect comparative effectiveness analysis between E and A using the results of the AFFIRM and COU-AA-301 pivotal trials. Due to the lack of study-to-study replication of comparisons, these indirect comparisons are subject to the assumption that the relative effects of each drug relative to placebo are the same across studies. Results: 1) Overall survival (OS)- the respective hazard ratios (95% confidence intervals {CI}) in the AFFIRM and COU-AA-301 trials for OS for E vs. placebo and A vs. placebo were 0.63 (0.53-0.75) and 0.66 (0.56-0.79). The indirect estimate of the hazard ratio (HR) (95% CI; p-value) for E vs. A was 0.96 (0.75-1.22; 0.72) 2) Time to PSA progression- the respective HR(95% CI) in the AFFIRM and COU-AA-301 trials for time to PSA progression for E vs. placebo and A vs. placebo were 0.25 (0.20-0.30) and 0.58 (0.46-0.73). The indirect estimate of the HR (95% CI; p-value) for E vs. A was 0.43 (0.32-0.58; 4.3 × 10^-8). 3) Radiographic progression Free Survival (rPFS)- The respective HR (95% CI) in the AFFIRM and COU-AA-301 trials for time to PSA progression for E vs. placebo and A vs. placebo were 0.40 (0.35-0.47) and 0.67 (0.59-0.78). The indirect estimate of the HR (95% CI; p-value) for E vs. A was 0.60 (0.49-0.74; 7.1 × 10^-7). 4) PSA response- the respective odds ratios (95% CI) in the AFFIRM and COU-AA-301 trials for time to PSA progression for E vs. placebo and A vs. placebo were 77.08 (34.02-174.70) and 5.49 (3.84-7.84). The indirect estimate of the odds ratio (95% CI; p-value) for E vs. A was 14.04 (5.75-34.28; 6.6 × 10^-9). Conclusions: In this indirect comparative effectiveness analysis, E appears to be more effective than A in terms of time to PSA progression, rPFS and PSA response. OS was not different.

576 Background: C225, a monoclonal antibody against the epidermal growth factor receptor, has been shown to inconsistently improve response rates (RR), progression-free survival (PFS) and overall survival (OS) in the first-line treatment of advanced CRC Pts with K-ras WT tumors. Methods: We performed a meta-analysis of four trials where K-ras WT Pts received a Fp (capecitabine (C) or bolus (b) or infusional (CI) 5-fluorouracil (5-FU)) and oxaliplatin (oxali) or irinotecan (CPT) ± C225 (CRYSTAL, OPUS, COIN and NORDIC VII trials) and one trial, where K-ras WT and mutant Pts received C225 with capecitabine (C) and oxali or CPT (AIO study). We sought to determine if the choice of Fp affects the response to C225. A mixed effects model similar to that of DerSimonian and Laird was fit by restricted maximum likelihood and used to obtain an overall estimate of the effect of C225 in the presence of CI 5-FU, an indirect estimate of the decrease in the effect of C225 in the presence of C/b5-FU relative to CI 5-FU, and an estimate of the study-to-study variability. Results: Only Pts treated with CI 5-FU based chemo derived benefit from C225. Relative to CI 5-FU, Pts treated with C or b5-FU based doublet chemo had a decrease in RR, PFS and OS. The choice of oxali or CPT did not affect responses to C225. Conclusions: The lack of benefit for C225 with C or b5-FU chemo is unexpected. A possible explanation is increased toxicity with C225, which led to dose reduction of C only in the C225-arm of the COIN study; however, increased toxicity was not seen in the NORDIC VII study. Pending further study, only CI 5-FU regimens should be used with C225. [Table: see text]

311 Background: Advanced pancreatic cancer (PC) is a virulent disease, where gemcitabine is considered the standard-of-care. A recent phase III evaluation demonstrated superiority of FOLFIRINOX (bolus and infusional 5-fluorouracil (FU), irinotecan, oxaliplatin) in patients with an Eastern Cooperative Oncology Group performance status (ECOG PS) of ≤1. Recently, three studies were presented in abstract form, comparing gemcitabine/S-1 (GS) to gemcitabine. S-1 is a mixture of tegafur (an oral 5-FU prodrug), a dihydropyrimidine dehydrogenase inhibitor and an agent designed to reduce the gastrointestinal toxicity of 5-FU. All three trials demonstrated significant improvements in progression-free survival (PFS) while one study also showed a statistically significant improvement in overall survival (OS). Methods: We performed a meta-analysis of the three trials. Results: Seven hundred and seventy patients were randomized to receive GS vs. gemcitabine; 75% had metastatic disease and 65% had an ECOG PS of 0. GS was associated with superior RR (hazard ratio (HR) 0.348, p=3.06×10^-7), PFS (HR 0.64, p=7.26×10^-9) and OS (HR 0.79, p=2.47×10^-2) compared to gemcitabine. Patients receiving GS were more likely to experience nausea, diarrhea, rash and stomatitis (mostly grade 1/2); neutropenic fever occurred in <2% of patients. One study demonstrated superior quality-of-life for GS. Conclusions: GS should be considered a first-line option for the treatment of advanced PC in Asian patients. This regimen should serve as the reference for future trials and comparison with FOLFIRINOX in patients with ECOG PS ≤1 is warranted.