Grothey, Axel

PURPOSE Several lines of treatment can be used sequentially in patients with metastatic colorectal cancer. We investigated the evolution of patient/tumor characteristics and their prognostic impact across treatment lines to develop an overall prognostic score (OPS). PATIENTS AND METHODS Individual patient data from 48 randomized trials were analyzed. The end point was overall survival (from random assignment to death). Missing data were imputed. The complete data set was then separated into construction (80%) and validation sets (20%). The Cox's model was used to define risk groups for survival using the OPS. The discrimination capability was assessed in each treatment-line via bootstrapping to obtain optimism-corrected calibration and discrimination C-indices. Internal validation was done in the validation set. RESULTS A total of 37,560 patients (26,974 in first-line [1L], 7,693 in second-line [2L], and 2,893 in third-line [3L]) were analyzed. Some clinical, biological, and molecular characteristics of patients/tumors included in therapeutic trials evolve over the lines. Seven independent prognostic variables were retained in the final multivariate model common to all lines: Eastern Cooperative Oncology Group performance status, hemoglobin, platelet count, WBC/absolute neutrophil count ratio, lactate dehydrogenase, alkaline phosphatase, and the number of metastatic sites. The OPS was used to define four patient subgroups with significantly different prognoses in 1L, 2L, and 3L, separately, with adequate C-indices: 0.65, 0.66, and 0.69 in the construction set and 0.65, 0.66, and 0.68 in the validation set, respectively. The OPS was not predictive, with 3L drugs ( v placebo) or subsequent line (2L/1L or 3L/2L) extending survival in all prognostic groups. CONCLUSION The same prognostic model using practical variables can be used before all treatment lines. The OPS could better stratify patients in future clinical trials and help to therapeutic decision in routine practice.

3517 Background: Incidence of early-onset colorectal cancer (eoCRC, age < 50) is steadily increasing. Decisions on adjuvant treatment (adjTx) regimen and duration should consider tx adherence, toxicity (tox) and expected outcomes in a population with life-expectancy longer than late onset CRC (loCRC, age ≥ 50). Methods: Individual patient data from stage II/III patients (pts) from 6 randomised trials in the IDEA database were used to compare characteristics, tx adherence, and adverse events of eoCRC to loCRC. To reduce the confounder of non-cancer-related deaths due to age/co-morbidities, time-to-recurrence (TTR) and cancer-specific survival (CSS) were compared by stratified Gay k-sample test. 5-year cancer-specific mortality (CSM) rate were estimated by adjusted cumulative incidence function. 3-year relapse-free survival (RFS) rate were compared by stratified and adjusted COX models. Results: Out of 16,349 pts included, 1564 (9.6%) were eoCRC. Compared to loCRC, eoCRC had lower percent of male pts (51% vs 57%, p < 0.01) better performance status (PS0 86% vs 80%, p < 0.01), similar T stage distribution (% T1-3/T4: 76/24 vs 77/23, p = 0.97), higher rate of N2 disease (24% vs 22%, p < 0.01), more likely to complete pre-planned duration of adjTx (83.2% vs 78.2%, p < 0.01) and received a higher tx intensity especially with 6 month tx (mean oxaliplatin dose intensity 75% vs 72%, p < 0.01; capecitabine 85% vs 78%, p < 0.01; 5FU 85% vs 82% p < 0.01). Gastrointestinal tox was more common in eoCRC (any grade nausea 58% vs 45%, p < 0.01; any grade vomiting 22% vs 16%, p < 0.01); haematological tox was more frequent in loCRC (62% vs, 69%, p = < 0.01); any grade neuropathy rate was similar (75%). Significant interaction was found between age and T stage for TTR (p = 0.04). Clinical outcome estimates and comparisons are shown in Table. Notably, high risk stage III (T4/N2) eoCRC had significantly lower 3-y RFS rate (54% vs 64%, HRadj 0.74, p < 0.01). Conclusions: eoCRC have better tx adherence than loCRC, as expected. While in high risk stage II and low risk stage III, cancer-specific outcomes are not different, in high risk stage III young age is negatively prognostic and associated with significantly higher relapse rate and risk of CRC death; this is despite a higher administered adjTx-intensity, suggesting a more aggressive disease biology. Clinical trial information: NCT00749450 (SCOT); NCT00646607 (TOSCA); NCT01150045 (CALGB/SWOG 80702); NCT00958737 (IDEA France) [Table: see text]

9599 Background: We have previously identified overall a single-item measure for baseline quality of life (QOL) as a strong prognostic factor for survival (Tan, ASCO 2008), and that fatigue was an important component of patient QOL (Sloan, 2007). To explore whether patient-reported fatigue was supplemental or redundant to the prognostic information of overall QOL, we performed a patient-level pooled analysis of 43 NCCTG and MCCC oncology clinical trials of the effect of baseline fatigue on OS. Methods: 3,915 patients participating in 43 trials provided data at baseline for fatigue on a single-item 0–100 point scale. OS was tested for association with clinically deficient fatigue (CDF, score 0–50, n=1,497) vs not clinically deficient fatigue (nCDF, score 51–100, n=2,418). Cox proportional hazards models adjusted for the effects of overall QOL, performance score, race, disease site, age and gender. Results: Baseline fatigue was a strong predictor of OS for the entire patient cohort (CDF vs. nCDF: 31.5 mos vs >83.9 mos, p<0.0001). The effect sizes were consistent across different disease sites (GI, esophageal, head and neck, prostate, lung, breast and others). After controlling for covariates, including performance status and overall QOL, baseline fatigue remained a strong prognostic factor in multivariate models (CDF vs. nCDF: HR=1.23, p=0.02). Conclusions: Fatigue is a strong prognostic factor for OS independent of overall QOL and PS in a wide variety of oncology patient populations. Single-item measures of overall QOL and fatigue can help to identify vulnerable subpopulations among cancer patients. No significant financial relationships to disclose.

6512 Background: The use of randomized phase II designs with an experimental arm and a standard-treatment control arm (R2PII) instead of a conventional single-arm design is clearly increasing in oncology. In practice, sample size, related cost issues, the belief that historical controls are adequate, and the use of a standard-treatment control arm in a phase II setting are frequently raised objections to R2PII trials. As the expense and complexity of definitive phase III trials increase, the ability of phase II trials to provide reliable and accurate results is critical. Methods: We investigated the ability of single arm vs R2PII trials to provide accurate conclusions by modeling variability in historical controls, patient outcome drifts independent of the tested therapy, and patient selection effects. Simulations compared R2PII and single-arm designs with binary endpoints under realistic parameters (e.g. alpha = beta = 0.10, historical control success rate = 20%, target success rate = 40%). Results: In the absence of variability in historical controls, estimated false positive and negative rates in both designs mirror the designated specifications. However, even in the presence of a modest drift effect in the population (mean 5% absolute shift in true control success rate), the false positive rate in single-arm designs is inflated two to three fold (to 20%-30%), while the R2PII retains the desired error rates. Greater confidence in historical controls in the single-arm design corrects only a small portion of the deviations. Increasing the sample size in each trial inflates the false positive error rate further to as much as 50%. Varying several sets of parameters gave similar results. Conclusions: In the presence of variability in historical controls, patient drift and/or selection effects, the false positive error rate of a single arm design is unacceptably high. In contrast, the R2PII design is reliable and robust despite the complexities in patient outcome drift and selection effects, and variability in historical control success rates. Given the rapid improvements in outcomes of many tumor types, the R2PII design should be the preferred method to evaluate novel agents in oncology in spite of the associated costs and the use of a reference control arm. No significant financial relationships to disclose.

3533 Background: PFS has previously been established as a surrogate for OS based on IPD from first-line mCRC trials conducted before 1999. As mCRC treatment (trt) has advanced in the last decade and OS has increased from 1 to 2 years, this surrogacy required re-examination. Methods: IPD from 16,762 pts, median age 62, 62% male, 53% ECOG PS 0 were available from 22 1st line mCRC studies conducted from 1997-2006; 12 tested targeted (anti-angiogenic and/or anti-EGFR) regimens. The relationship between PFS (first event of progression or death) and OS was evaluated at patient-, trt-arm-, and trial-levels using correlation (corr.) coefficients and R2 (closer to 1 the better) from weighted least square (WLS) regression of arm-specific survival rates and trial-specific hazard ratios (HRs), estimated using Cox and Copula bivariate models. The concordance of significance (CoS) of the treatment effects (TEs) on both endpoints was calculated. Results: 44% pts received a targeted regimen. Median PFS was 8 and OS was18 months. The corr. between PFS and OS was modest at all three levels with low CoS in TE comparisons (see Table). Analyses limited to trials testing targeted trts, non-strategy trials, or superiority trials did not improve surrogacy. Conclusions: In modern mCRC trials, where survival post-first progression exceeds time to first progression, PFS TEs do not reliably predict TEs on OS. Nonetheless, until alternative endpoints of clinical benefit can be validated, PFS remains a relevant primary endpoint for 1st line mCRC trials, as our data demonstrate that the ability for any agent to produce an OS benefit from a single line of therapy is challenging. [Table: see text]

3514 Background: In the CORRECT phase III trial, the multikinase inhibitor regorafenib (REG) demonstrated significant improvement in overall survival (OS) and progression-free survival (PFS) vs placebo (Pla) in patients with metastatic colorectal cancer (mCRC) whose disease had progressed on other standard therapies. An exploratory biomarker subanalysis was conducted to identify protein biomarkers with potential predictive or prognostic value. Methods: Fifteen proteins of interest, many of which are involved in angiogenesis, were quantified by multiplex immunoassay or ELISA in baseline plasma samples collected at study entry from 80% (611/760) of patients. Potential predictive and prognostic effects were evaluated. Results: The biomarker subpopulation was representative of the overall study population in terms of OS and PFS. Using OS as the clinical endpoint, Tie-1 was the only protein whose level demonstrated significant correlation with efficacy (low protein group: REG/Pla, HR 0.87; high protein group, HR 0.56; interaction, p=0.035). Using PFS as the clinical endpoint, von Willebrand factor (VWF) was the only protein whose level demonstrated significant correlation with efficacy (low protein group: REG/Pla, HR 0.39; high protein group, HR 0.60; interaction, p=0.02). Following correction for multiple testing, neither Tie-1 nor VWF data retained statistical significance. Baseline levels of IL-8 and placental growth factor (PlGF) were found to have prognostic value for OS (IL-8: high/low protein levels, HR 3.48, p<0.001; PIGF: HR 1.81, p=0.002). IL-8 was also prognostic for PFS (high/low protein levels: HR 1.63, p<0.001). Conclusions: None of the plasma proteins examined showed significant predictive value for REG efficacy after multiple testing correction. The association between baseline levels of Tie-1/VWF and REG efficacy may be a hypothesis to be tested in further trials. Clinical trial information: NCT01103323.

e14564 Background: Recent long-term clinical trial follow-up suggests that adding Ox to a fluoropyrimidine (FP) may provide less benefit in adjuvant CC than originally believed (Tournigand JCO 2012). In addition, emerging data on tissue biomarkers such as microsatellite mismatch repair (MMR) and the 12-gene Recurrence Score (RS) suggest these assays may be helpful in better defining treatment benefit, and a new report provides evidence that RS may specifically predict the value of Ox (O'Connell ASCO 2012). In view of these developments we sought to document current use of adjuvant chemotherapy and physician perceptions about these decisions. Methods: U.S.-based oncologists were recruited to complete a survey and contribute treatment data for their last three patients (pts) with stage II disease and last pt with a stage III tumor. Results: 25 gastrointestinal cancer clinical investigators and 77 practicing oncologists participated, providing data from a total of 408 cases (92% treated since 2011). Findings from the two groups are similar and are presented in aggregate. A median of 9 and 15 pts annually were evaluated with Stage II and III tumors, respectively. Clinicians stated that they require a median of a 5% absolute reduction in relapse rate to use a FP alone and an additional 5% reduction to add Ox. Ox/FP was administered to 95 of 102 pts with stage III tumors (93%) and 85 of 306 pts with stage II disease (28%), particularly to those under age 70 or with T4 tumors (Table). MMR was evaluated in 127 pts (42%) with stage II and 20 pts (20%) with stage III disease. RS was obtained for 46 pts (15%) with stage II tumors, but only 2 pts (2%) with stage III. Conclusions: Adjuvant Ox is employed widely in spite of the fact that it may provide less benefit than many oncologists report requiring to initiate its use. However, several recent research developments may be laying the groundwork for a change in this treatment practice. [Table: see text]

1557 Background: Peripheral neuropathy (PN), a common and intrusive side effect of chemotherapy in clinical trials, has been assessed via the clinician-rated common terminology criteria for adverse events (CTCAE) and/or patient-reported outcome (PRO) measures (Morton, ASCO, 2005). We assessed the relative sensitivity of CTCAE and PRO measures for evaluating interventions for preventing and ameliorating CIPN. Methods: Data from285 patients in two prevention trials (N05C3, N04C7) and 432 patients from three intervention trials (N00C3, N01C3, N06CA) were analyzed separately. CTCAE version 3.0 item for neuropathy was compared to 2 NCCTG numerical analogue CIPN items, 6 EORTC CIPN20 individual items, 2 McGill Pain Questionnaire items, and one Brief Pain Inventory item. The sample size provided over 80% power to detect a true Spearman rank correlation of 0.15 assessed the relationship between CTCAE and PRO measures. Results: Results from both the prevention and treatment trials were similar in terms of the relationship between the CTCAE and PRO measures. No correlation coefficients between the CTCAE and any of the PRO items at baseline were above 0.35, with the majority around 0.2. Higher scores for the numbness items were observed than for the CTCAE in 47% of the patients at baseline and 39% during treatment in the treatment trials. Many patients had severe numbness PRO scores but mild CTCAE scores (12%) at baseline and (8%) during treatment in the treatment trials. Results for the numbness and tingling PRO measures were redundant. The correlations of the CTCAE with pain were weak, all below 0.35 at baseline with differences as much as 29 points on average on a 0-100 transformed scale. Conclusions: The CTCAE and PRO measures of PN measure different constructs. The CTCAE includes both functional and ADL/motor aspects which may or may not be related to the singular constructs of pain, numbness, tingling, or adjectival characteristics of pain assessed by the PRO measures. The two measures cannot be used in place of each other.

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]

3502 Background: Regorafenib (REG) is an oral multi-kinase inhibitor. The CORRECT trial was conducted to evaluate REG in patients (pts) with mCRC who had progressed after all approved standard therapies. Methods: Enrollment criteria included documented mCRC and progression during or ≤3 months after last standard therapy. Pts were randomized 2:1 to receive best supportive care plus either REG (160 mg od po, 3 wks on/1 wk off) or placebo (PL). The primary endpoint was overall survival (OS). Secondary endpoints included progression-free survival (PFS), overall response rate, disease control rate, safety and quality of life (QoL). Efficacy analyses across prespecified subgroups were evaluated using univariate Cox regression. Results: 760 pts were randomized (REG: 505; PL: 255). The OS primary endpoint was met at a preplanned interim analysis. OS and PFS were significantly improved in REG arm compared to PL arm: hazard ratio (HR) for OS 0.77 (95% CI 0.64-0.94, 1-sided p=0.0052), median OS 6.4 vs 5.0 mos; HR for PFS 0.49 (95% CI 0.42-0.58, 1-sided p<0.000001), median PFS 1.9 vs 1.7 mos. Comparable OS and PFS benefits were observed in exploratory subgroup analyses by region, age, time from diagnosis of mCRC to randomization, prior lines of treatment, and KRAS status (shown in table). The most common grade 3+ AEs related to REG were hand-foot skin reaction (16.6%), fatigue (9.6%), hypertension (7.2%), diarrhea (7.2%) and rash/desquamation (5.8%). QoL data will be presented. Conclusions: REG demonstrated statistically significant improvement in OS and PFS over PL, as well as comparable efficacy benefits across pt subgroups analyzed. [Table: see text]