Sargent, Daniel J.

Purpose Biomarkers can add substantial value to current medical practice by providing an integrated approach to prediction using the genetic makeup of the tumor and the genotype of the patient to guide patient-specific treatment selection. We discuss and evaluate various clinical trial designs for the validation of biomarker-guided therapy. Methods Designs for predictive marker validation are broadly classified as retrospective (ie, using data from previously well-conducted randomized controlled trials [RCTs]) versus prospective (enrichment, unselected, hybrid, or adaptive analysis). We discuss the salient features of each design in the context of real trials. Results Well-designed retrospective analysis from well-conducted prospective RCTs can bring forward effective treatments to marker-defined subgroups of patients in a timely manner (eg, KRAS and colorectal cancer). Enrichment designs are appropriate when preliminary evidence suggest that patients with or without that marker profile do not benefit from the treatments in question; however, this may sometimes leave questions unanswered (eg, trastuzumab and breast cancer). An unselected design is optimal where preliminary evidence regarding treatment benefit and assay reproducibility is uncertain (eg, epidermal growth factor receptor and lung cancer). Hybrid designs are appropriate when preliminary evidence demonstrate the efficacy of certain treatments for a marker-defined subgroup, making it unethical to randomly assign patients with that marker status to other treatments (eg, multigene assay and breast cancer). Adaptive analysis designs allow for prespecified marker-defined subgroup analyses of data from an RCT. Conclusion The implementation of these design strategies will lead to a more rapid clinical validation of biomarker-guided therapy.

Purpose Intensive postoperative surveillance is associated with improved survival and recommended for patients with late stage (stage IIB and III) colon cancer. We hypothesized that stage I and IIA colon cancer patients would experience similar benefits. Patients and Methods Secondary analysis of data from the Clinical Outcomes of Surgical Therapy trial was performed by analyzing results according to TNM stage; early (stage I and IIA, 537 patients) and late (stage IIB and III, 254 patients) stage disease. Five-year recurrence rates were higher in patients with late (35.7%) versus early stage disease (9.5%). Early and late stage salvage rates, recurrence patterns and methods of first detection were compared by χ2 test. Results Salvage rates for early- and late-stage disease patients with recurrence were the same (35.9% v 37%; P = .9, respectively). Median survival after second surgery after recurrence was 51.2 and 35.8 months for early- and late-stage patients, respectively. Single sites of first recurrence did not significantly differ between early and late stage, but multiple sites of recurrence occurred less often in early-stage patients (3.6% v 28.6%, for early v late, respectively; P < .001). Methods of first detection of recurrence were not significantly different: carcinoembryonic antigen (29.1% v 37.4%), computed tomography scan (23.6% v 26.4%), chest x-ray (7.3% v 12.1%), and colonoscopy (12.7% v 8.8%), for early versus late stage disease, respectively. Conclusion Patients with early-stage colon cancer have similar sites of recurrence, and receive similar benefit from postrecurrence therapy as late-stage patients; implementation of surveillance guidelines for early-stage patients is appropriate.

Purpose Fluorouracil/leucovorin as the sole therapy for metastatic colorectal cancer (CRC) provides an overall survival of 8 to 12 months. With an increase in surgical resections of metastatic disease and development of new chemotherapies, indirect evidence suggests that outcomes for patients are improving in the general population, although the incremental gain has not yet been quantified. Methods We performed a retrospective review of patients newly diagnosed with metastatic CRC treated at two academic centers from 1990 through 2006. Landmark analysis evaluated the association of diagnosis year and liver resection with overall survival. Additional survival analysis of the Surveillance Epidemiology and End Results (SEER) database evaluated a similar population from 1990 through 2005. Results Two thousand four hundred seventy patients with metastatic CRC at diagnosis received their primary treatment at the two institutions during this time period. Median overall survival for those patients diagnosed from 1990 to 1997 was 14.2 months, which increased to 18.0, 18.6, and 29.3 months for patients diagnosed in 1998 to 2000, 2001 to 2003, and 2004 to 2006, respectively. Likewise, 5-year overall survival increased from 9.1% in the earliest time period to 19.2% in 2001 to 2003. Improved outcomes from 1998 to 2004 were a result of an increase in hepatic resection, which was performed in 20% of the patients. Improvements from 2004 to 2006 were temporally associated with increased utilization of new chemotherapeutics. In the SEER registry, overall survival for the 49,459 identified patients also increased in the most recent time period. Conclusion Profound improvements in outcome in metastatic CRC seem to be associated with the sequential increase in the use of hepatic resection in selected patients (1998 to 2006) and advancements in medical therapy (2004 to 2006).

Purpose In February 2000, the criteria for measuring tumor shrinkage as an indicator of antitumor activity were redefined by the Response Evaluation Criteria in Solid Tumors (RECIST). This resulted in simplifying bidimensional to unidimensional measurement of lesions. Under RECIST, all lesions, up to 10, must be measured. Scanning and measuring multiple lesions is costly, time-consuming, and a disincentive to participation in clinical trials. We investigated whether fewer than 10 lesions can be measured without compromising the accuracy of assessing a regimen's activity. Patients and Methods Thirty-two North Central Cancer Treatment Group trials including 2,374 patients were analyzed. Twelve studies were conducted before RECIST; 20 were conducted post-RECIST. Agreement between objective status by cycle, confirmed response, overall response rate, and time to progression (TTP) was evaluated based on all 10 versus the largest one through five lesions. Results The median number of lesions reported on RECIST trials did not differ from pre-RECIST trials (median = 2.0). One lesion at baseline was reported in 49% of patients, two lesions in 28% of patients, three lesions in 12% of patients, four lesions in 6% of patients, and five lesions in 5% of patients in post-RECIST trials. Utilizing the largest two lesions produced excellent concordance with that using all lesions for all end points. In no trial did the overall response rate differ by more than 3% when two versus all lesions were considered. Evaluating more than two lesions did not significantly improve agreement. Conclusion Based on these trials, the assessment of more than two lesions did not alter the conclusions regarding a treatment's efficacy as judged by response rate or TTP.

Purpose The primary goal of this multicenter phase III trial was to determine whether overall survival (OS) of fluorouracil (FU) -refractory patients was noninferior when treated with second-line infusional fluorouracil, leucovorin, and oxaliplatin (FOLFOX4; arm B) versus irinotecan (arm A). Cross-over to the other treatment on disease progression was mandated. Patients and Methods Patients who experienced treatment failure with one prior FU-based therapy and had not received prior irinotecan or oxaliplatin, either for metastatic disease or within 6 months of adjuvant FU therapy, were randomly assigned to arm A (irinotecan 350 or 300 mg/m2 every 3 weeks) or arm B (FOLFOX4). Results A total of 491 patients were randomly assigned (arm A, n = 245; arm B, n = 246); 288 (59%) had experienced treatment failure with FU for metastatic colorectal cancer. Two hundred twenty-seven patients (46%) received protocol-mandated third-line therapy (arm A, 43%; arm B, 57%). Median survival was 13.8 months (95% CI, 12.2 to 15.0 months) for initial treatment with FOLFOX4 and 14.3 months (95% CI, 12.0 to 15.9 months) for irinotecan (P = .38; hazard ratio = 0.92; 95% CI, 0.8 to 1.1). Response rates (RR; 28% v 15.5%; P = .0009) and time to progression (TTP; 6.2 v 4.4 months; P = .0009) were significantly superior with FOLFOX4. In the nonrandom subset of patients who crossed over, RR and TTP improvements with FOLFOX4 continued into third-line treatment. Irinotecan therapy was associated with more grade 3 nausea, vomiting, diarrhea, and febrile neutropenia; FOLFOX4 was associated with more neutropenia and paresthesias. Conclusion In patients who experienced treatment failure with front-line FU therapy, OS does not significantly differ whether second-line therapy begins with irinotecan or FOLFOX4. FOLFOX4 produces higher RR and longer TTP. Both arms had notable OS in patients who experienced treatment failure with first-line FU therapy.

Abstract Phase II trials are screening trials that seek to identify agents with sufficient activity to continue development and those for which further evaluation should be halted. Although definitive phase III trials use progression-free or overall survival to confirm clinical benefit, earlier endpoints are preferable for phase II trials. Traditionally, tumor shrinkage of a predetermined degree (response) has been used as a surrogate of eventual survival benefit based on the observation that high response rates (RR), and particularly complete responses, in the phase II setting resulted in survival benefit in subsequent phase III trials. Recently, some molecularly targeted agents have shown survival and clinical benefit despite very modest RRs in early clinical trials. These observations provide a major conundrum, with concerns of inappropriate termination of development for active agents with low RRs being balanced by concerns of inactive agents being taken to late-phase development with resultant increases in the failure rate of phase III trials. Numerous alternate or complementary endpoints have been explored, incorporating multinomial endpoints (including progression and response), progression-free survival, biomarkers, and, more recently, evaluation of tumor size as a continuous variable. In this review, we discuss the current status of phase II endpoints and present retrospective analyses of two international gastrointestinal cancer studies showing the potential utility of one novel approach. Alternate endpoints, although promising, require additional evaluation and prospective validation before their use as a primary endpoint for phase II trials.

Abstract As the use of molecularly targeted agents, which are anticipated to increase overall survival (OS)and progression-free survival (PFS) but not necessarily tumor response, has increased in oncology, there has been a corresponding increase in the recommendation and use of randomized phase II designs. Such designs reduce the potential for bias, existent in comparisons with historical controls, but also substantially increase the sample size requirements. We review the principal statistical designs for historically controlled and randomized phase II trials, along with their advantages, disadvantages, and statistical design considerations. We review the arguments for and against the use of randomization in phase II studies, the situations in which the use of historical controls is preferred, and the situations in which the use of randomized designs is preferred. We review methods used to calculate predicted OS or PFS values from historical controls, adjusted so as to be appropriate for an experimental sample with particular prognostic characteristics. We show how adjustment of the type I and type II error bounds for randomized studies can facilitate the detection of appropriate target increases in median PFS or OS with sample sizes appropriate for phase II studies. Although there continue to be differences among investigators concerning the use of randomization versus historical controls in phase II trials, there is agreement that each approach will continue to be appropriate, and the optimal approach will depend upon the circumstances of the individual trial.