Innocenti, Federico

Approximately 20 drugs have been shown to be effective for the treatment of colorectal cancer (CRC). These drugs are from several classes of agents and include cytotoxic drugs, therapeutics that target cell signaling pathways at the extracellular and/or intracellular levels, and combination therapies that contain multiple targeted agents and/or cytotoxic compounds. Targeted therapeutics can include monoclonal antibodies, fusion proteins, and small molecule drugs. The first introduced into clinical use was 5‐fluorouracil in the early 1960s and remains the foundation for most CRC treatments in both adjuvant therapy and in advanced (metastatic) treatment regimens. As with other cancers, the consideration of biomarkers has the potential to improve CRC therapy through patient stratification. The biomarkers can include germline genetic markers, tumor‐specific genetic markers, immune markers, and other biomarkers that can predict antitumor efficacy or the likelihood of toxicity prior to administration of a specific drug. Consistent with the benefit of considering biomarkers in treatment, many newer targeted therapies are developed and approved simultaneously with a companion diagnostic test to determine efficacy. This review will focus on biomarkers that have demonstrated clinical utility in CRC treatment; however, it is noted that many additional biomarkers have been theorized to contribute to drug response and/or toxicity based on known biological pathways but thus far have not attained widespread use in the clinic. The importance of pretreatment biomarker testing is expected to increase as future drug development will likely continue to focus on the concurrent development of companion diagnostics.

3588 Background: Growing data suggest dietary factors are associated with survival among pts with non-metastatic CRC. However, data on diet and survival among pts with advanced or metastatic disease are very limited. Methods: We prospectively examined dietary fat intake assessed at initiation of treatment for advanced or metastatic CRC in relation to OS and PFS. This analysis was conducted among 1,149 pts in the CALGB 80405 randomized controlled trial who completed a validated food frequency questionnaire. We examined intakes of saturated, monounsaturated, and polyunsaturated (total n-3, long-chain n-3, and total n-6) fats as well as animal and vegetable fats. Based on data from non-metastatic CRC and other cancers, we hypothesized that higher intakes of long-chain n-3 fatty acids and vegetable fats would be associated with longer OS and PFS and higher intakes of saturated fat and animal fat would be associated with shorter OS and PFS. We used Cox proportional hazards regression to estimate adjusted hazard ratios (HR) and 95% confidence intervals (CI). Results: Over a median follow-up of 6.1 years [y; interquartile range (IQR): 5.3, 7.2 y], we observed 974 deaths and 103 events of progression without death during follow-up. Participants in this analysis had a median age of 59 y (IQR: 51 to 67 y); 41% were female and 86% identified as white. We observed no statistically significant associations between any type of dietary fat and OS. However, vegetable fat was non-linearly associated with longer PFS (HR comparing 4th to 1st quartile: 0.78; 95% CI: 0.64, 0.96; p-trend: 0.10). We also observed a linear association between continuous saturated fat and PFS (HR per 5% kcal/d: 1.21; 95% CI: 1.03, 1.42; p-value: 0.02), perhaps driven by pts with high saturated fat intake. Conclusions: We observed no statistically significant associations between types of dietary fat and OS among pts with advanced or metastatic CRC. However, a healthy diet that includes vegetable fat and is modest in saturated fat may be associated with longer PFS. Future studies to replicate these findings and examine diet in relation to cancer survival in racially/ethnically diverse populations are needed. Support: K07CA197077, U10CA180821, U10CA180882, https://acknowledgments.alliancefound.org . Clinical trial information: NCT00265850.

3568 Background: Diets inducing an elevated insulin response have been associated with increased recurrence and mortality in patients with non-metastatic colorectal cancer, but it remains unknown if postprandial hyperinsulinemia also affects progression and mortality in mCRC patients. The goal of this study was to assess the influence of dietary insulin load (DIL) and dietary insulin index (DII) on survival of mCRC patients. Methods: This was a prospective cohort study of 1,177 patients with previously untreated mCRC enrolled in a phase III trial of systemic chemotherapy plus biologics who reported dietary intake within one month after chemotherapy initiation. DIL was calculated as a function of food insulin index and the energy content of individual foods reported on a food frequency questionnaire. DII was calculated by dividing DIL by total energy intake. The primary endpoint was overall survival (OS). Secondary endpoints were progression-free survival (PFS) and treatment-related adverse events (TRAEs). The primary statistical test was a test for trend, which was performed using the median value for each quintile of dietary insulin score as a continuous variable. Cox proportional hazards regression was used to adjust for potential confounders including assigned treatment arm, known prognostic factors, comorbidities, body mass index, and physical activity. Results: Higher DIL was significantly associated with worse OS (ptrend = 0.04); patients in the highest quintile survived 34.1 months, compared to 27.7 months in the lowest quintile (Cox hazard ratio [HR] 1.22, 95% confidence interval [CI] 0.99 - 1.51). Higher DII was non-significantly associated with worse OS (HR 1.18, 95% CI 0.94 - 1.48, ptrend = 0.09). There was no significant association between dietary insulin scores and PFS. The influence of dietary insulin scores on survival did not differ significantly by various molecular markers involved in the insulin signaling pathway, including C-peptide, adiponectin, IGF-1, IGFBP-3, and IGFBP-7. Higher dietary insulin scores were significantly associated with greater risk of any TRAE. Those with a DIL greater than the median had a 75.4% rate of any TRAE, compared to 70.8% in those with a DIL less than or equal to the median (HR 1.19, 95% CI 1.03 - 1.38, p=0.02); the most significant associations were with neutropenia (HR 1.30, 95% CI 1.05 - 1.61, p=0.01) and diarrhea (HR 1.43, 95% CI 1.00 - 2.06, p=0.05). Conclusions: Higher DIL was significantly associated with worse OS, and both higher DIL and DII were significantly associated with increased TRAEs, in patients with previously untreated mCRC. These findings may inform future dietary recommendations for patients with mCRC. Further investigation into the molecular mechanisms underlying these associations is warranted. Clinical trial information: NCT00265850.

3515 Background: CALGB/SWOG 80405 was a randomized phase III trial in first-line mCRC patients treated with Bev, Cet, or both, plus chemotherapy. No difference in OS was found between Bev and Cet. We tested the effect of immune signatures on OS in all the three arms of the study and analyzed differences in OS between the Cet and Bev arms. Methods: 578 primary tumors were profiled by RNAseq. Immune signatures of TGF-β, cytotoxic T cells, wound healing, macrophages, lymphocytes, and INF-γ, as well as relative frequencies of CD8+ T-cells, memory resting CD4+ T cells, memory activated CD4+ T cells, macrophages M1 and M2, and activated mast cells were measured. Multivariate Cox proportional hazard models were applied using elastic-net penalization with covariates (age, race, gender, all RAS and BRAF V600E mutations). For relevant signatures, optimal cut-offs for OS were calculated. Results: In all the three arms of the study, high expression of macrophages M2 (HR 6.81, 95% CI 3.56-30.16) and TGF-β (HR 1.37, 95% CI 1.03-2.10) conferred reduced OS compared to low expression; high expression of plasma cells (HR 0.52, 95% CI 0.27-0.83) and memory-activated CD4+ T cells (HR 0.34, 95% CI 0.10-0.65) conferred increased OS compared to low expression. Using optimal cut-offs from these 4 signatures, pts have been categorized as to whether they had either 4, 3, 2, 1, or 0 beneficial signatures associated with increased OS. In all arms of the study (N = 469, after accounting for covariates), the median (95% CI) OS decreased from 42.5 (35.8-47.8; N = 79), to 31.0 (28.8-34.4; N = 177), 25.2 (20.6-27.9; N = 144), and 17.0 (13.5-20.4; N = 69) months when the number of beneficial signatures decreased from 4, to 3, 2, and 0-1 (combined due to a low number of pts), respectively (p = 3.48e-11). In the Bev arm (N = 205), high expression of macrophages M2 conferred reduced OS compared to low expression (HR 6.6, 95% CI 2.7-67.1). In the Cet arm (N = 165), high expression of macrophages M2 conferred reduced OS compared to low expression (HR 4.3, 95% CI 2.1-79.8); high expression of plasma cells (HR 0.36, 95% CI 0.06-0.55) and memory activated CD4+ T cells (HR 0.37, 95% CI 0.03-0.98) conferred increased OS compared to low expression of either signatures. The plasma cell signature interacted with Bev and Cet on the OS of pts (interaction p = 0.009). Conclusions: Tumor immune signatures in mCRC pts are determinants of survival. In pts treated with Bev- and Cet-combination therapies that are standard of care, immune signatures affect response to therapy. These results, provide new markers for treatment selection and for the development of novel active combinations including immune checkpoint inhibitors. Support: U10CA180821, U10CA180882, U24CA196171; https://acknowledgments.alliancefound.org

3030 Background: Diarrhea, HFS, and hypertension are common toxicities of sorafenib. No markers are validated to predict patients at risk of these toxicities. This study aimed to identify genetic predictors of sorafenib-induced toxicities. Methods: A two-step, discovery-validation approach was used. The discovery set included 140 renal cell carcinoma patients from the TARGET study treated with sorafenib (400 mg twice daily) and genotyped for 1040 single-nucleotide polymorphisms (SNPs) in 56 genes. The three most statistically significant SNPs associated with grade ≥2 composite toxicity (either hypertension, diarrhea, HFS, or other skin toxicities, CTCAE v.3.0) were tested for association with grade 3 composite toxicity (either hypertension, diarrhea, or HFS, CTCAE v.4.0) in a validation set of 240 hepatocellular carcinoma patients from Alliance/CALGB 80802 treated with sorafenib (400 mg twice daily) alone or with doxorubicin. Associations between SNPs and composite toxicity was performed by logistic regression, with adjusting covariates (age, gender, race, and treatment arm, the latter two covariates for the validation set only). A meta-analysis odds ratio (OR) of each SNP-grade 3 toxicity association between the discovery and validation sets was obtained by inverse variance to point toward effects specific to a type of toxicity. Results: In the discovery set, the top three SNPs associated with grade ≥2 composite toxicity were rs12366035 (C>T, minor allele frequency, MAF 0.34) in VEGFB (p 0.0007), rs4035887 (G>A, MAF 0.49) in EPAS1 (p 0.0021), and rs4864950 (T>A, MAF 0.23) in KDR (p 0.0058). These SNPs were genotyped in the validation set and only rs4864950 in KDR was replicated. No grade 4 toxicities were reported. Similar to the discovery set (OR 2.41, 95% CI 1.29-4.51), the A allele of rs4864950 increased the risk of grade 3 composite toxicity (p 0.032, OR 2.12, 95% CI 1.70-4.27) in the validation set. Grade 3 toxicity prevalence in the discovery and validation sets were 3.6% and 7.4% diarrhea, 8.6% and 12.3% HFS, 3.6% and 8.8% hypertension, respectively. The meta-analysis of the two datasets showed that the A allele of rs4864950 increased the risk of grade 3 diarrhea (p 0.045, OR 3.09, 95% CI 1.03-9.29), grade 3 HFS (p 0.012, OR 2.57, 95% CI 1.24-5.37), but not grade 3 hypertension (p 0.207, OR 0.51, 95% CI 0.18-1.45). Conclusions: We provide the first evidence of clinical validity of a marker of sorafenib-induced diarrhea and HFS. Sorafenib inhibits VEGFR2 (coded by KDR), leading to epithelial hypoxia and causing diarrhea and HFS. Variant rs4864950 might affect the function VEGFR2, which, during VEGFR2 inhibition, increases the risk of diarrhea and HFS. This SNP is common and can be genotyped in patients before receiving sorafenib for a better risk assessment. Support: U10CA180821, U10CA180882, U24CA196171; https://acknowledgments.alliancefound.org ClinicalTrials.gov Id: NCT01015833.

343 Background: Alliance/CALGB 80802, a randomized phase III trial, evaluated sorafenib plus doxorubicin vs. doxorubicin in pts with HCC and showed no improvement in median overall survival (OS) (HR[95CI] 1.05[0.83-1.31]) or PFS (HR[95CI] 0.93[0.75-1.16]). In HCC surrogacy of tumor response with OS remains controversial, in part due to varying criteria used for response evaluation (e.g., RECIST1.1 and mRECIST). We evaluated the performance of several models to predict OS using pretreatment clinical and radiomic variables. Methods: In CALBG 80802, we segmented all measurable tumor lesions on sequential CT scans. A lesion’s imaging phenotype was deciphered with 23 uncorrelated quantitative imaging features measured at baseline and week (wk) 10 (first follow-up). An OS landmark survival analysis was conducted at wk 10. Patients were randomly assigned (3:1) to training (n = 92) and validation (n = 37) sets. In a training set, 6 random forest predictive models (6 signatures) used features that best predicted OS using 3 sets of variables: radiomics only (n = 23), clinical only (n = 9), radiomics and clinical (n = 32). Two time points (baseline only or baseline + wk 10) were assessed. Each signature's output was an individualized prediction and a continuous value ranging from 0 to 1 (from most to least favorable predicted OS). The primary endpoint was to compare these models' performance to predict OS using error rate (Harrell's concordance-index) in the validation set. Results: Of the 6 training signatures evaluated, the one achieving the highest performance in the validation set was an 8-feature signature combining radiomics and clinical variables measured at two time points (baseline + wk 10) with an error rate of 35.6%. The variables [rank of importance] (table) selected by the signature included baseline clinical features (albumin[1], AFP[2], Child-Pugh[4]), baseline radiomics features (component 17[3], component 1[5], component 9[7], tumor volume[8]) and wk 10 radiomics features (delta tumor volume[6]). Variable delta tumor volume [6] used a more enhanced estimation of tumor burden at baseline and a delta tumor volumetric measurement; compared to RECIST1.1 measurement of percentage change in unidimensional measurement of a subset of target lesions. The four quartiles of the signature were significantly associated with OS (Log-Rank, P < 0.0001). Conclusions: The selected combined radiomic and clinical composite signature provided the best prediction for OS in the 80802 study patients’ population. It is a suggested way forward to go beyond single anatomic measurement techniques such as RECIST or mRECIST. [Table: see text]

325 Background: Alliance/CALGB 80802 randomized phase III trial evaluated DO+S vs. S in pts with aHCC, and showed no improvement in median OS. Multi-drug resistant pathway mitigation by the Ras/Raf/MEK/ERK pathway and bFGF-mediated activation of Raf-1 promotes the formation of antiapoptotic Raf-1 and ASK1 complex, induced by anthracyclines. S efficiently blocks NS5A-recruited c-Raf mediated HCV replication and viral gene expression. Once inhibited by S, VEGF expression of HepG2 may limit HCV cellular entry. Release of Raf-1-Ask-1 dimer and inhibition of Raf-1 via S putatively differ in the presence or absence of DO. We hypothesize treatment with S reduces HCV titer levels (TL) and influence pts’ outcome. Methods: In 80802 HCV pts, TL were evaluated in both arms at baseline and post-baseline at Day 1 of Cycles 2, 3, and every 2 cycles and at progression or discontinuation of therapy. HCV undetectable (HCV-UN) levels were defined as < 50 copies/mL. TL were evaluated in relation to OS and PFS. HCV RNA levels were measured by TaqMan PCR and by genotype. Results: Of 356 pts, 83 were HCV+ with more Black/African American (25/50 = 50%) vs. White (54/239 = 23%) or other race groups (4/67 = 6%) (p < 0.0001). HCV titer data were available on 54 pts (S: 28, DO+S: 26). At baseline, 12 pts (S: 7, DO+S: 5) were HCV-UN, and post-baseline HCV TL did not significantly differ between treatment arms; one patient in each arm went from detectable (HCV-D) to HCV-UN. Post-baseline, 40 pts were HCV-D vs. 14 who were HCV-UN (S+DO: 8, S: 6 pts). Except for the two pts who became HCV-UN, baseline HCV-D vs. HCV-UN titers was similar to that status post-baseline. PFS and OS between HCV-D and HCV-UN both at baseline and post-baseline are delineated in the table. Conclusions: We observed that S did not influence HCV TL. Pts treated with DO+S vs. S had worse PFS if they had HCV-UN, and further that higher levels of HCV titers at baseline were associated with significantly improved PFS. Given the small sample size, these findings warrant further prospective evaluation. Support: U10CA180821, U10CA180882, U24CA196171; Bayer, Bristol-Myers-Squibb, and Sanofi. https://acknowledgments.alliancefound.org . Clinical trial information: NCT01015833. [Table: see text]

Abstract Purpose: Irinotecan/5-fluorouracil (5-FU; FOLFIRI) or oxaliplatin/5-FU (FOLFOX), combined with bevacizumab or cetuximab, are approved, first-line treatments for metastatic colorectal cancer (mCRC). We aimed at identifying germline variants associated with survival in patients with mCRC treated with these regimens in Cancer and Leukemia Group B/SWOG 80405. Experimental Design: Patients with mCRC receiving either FOLFOX or FOLFIRI were randomized to either cetuximab or bevacizumab. DNA from peripheral blood was genotyped for approximately 700,000 SNPs. The association between SNPs and overall survival (OS) was tested in 613 patients of genetically estimated European ancestry using Cox proportional hazards models. Results: The four most significant SNPs associated with OS were three haplotypic SNPs between microsomal glutathione S-transferase 1 (MGST1) and LIM domain only 3 (LMO3, representative HR, 1.56; P = 1.30 × 10−6), and rs11644916 in AXIN1 (HR, 1.39, P = 4.26 × 10−6). AXIN1 is a well-established tumor suppressor gene in colorectal cancer, and rs11644916 (G&gt;A) conferred shorter OS. Median OS for patients with the AA, AG, or GG genotypes was 18.4, 25.6, or 36.4 months, respectively. In 90 patients with stage IV colorectal cancer from The Cancer Genome Atlas (TCGA), rs11649255 in AXIN1 [in almost complete linkage disequilibrium (LD) with rs11644916], was associated with shorter OS (HR, 2.24, P = 0.0096). Using rs11648673 in AXIN1 (in very high LD with rs11644916 and with functional evidence), luciferase activity in three colorectal cancer cell lines was reduced. Conclusions: This is the first large genome-wide association study ever conducted in patients with mCRC treated with first-line standard treatment in a randomized phase III trial. A common SNP in AXIN1 conferred worse OS and the effect was replicated in TCGA. Further studies in colorectal cancer experimental models are required.