Huang, Erich P.

PURPOSE The mode of action of targeted cancer agents (TCAs) differs from classic chemotherapy, which leads to concerns about the role of RECIST in evaluating tumor response in trials with TCAs. We investigated the performance of RECIST using a pooled database from 50 clinical trials with at least one TCA. METHODS We examined the impact of the number of target lesions (TLs) on within-patient variability of tumor response. The prognostic effect of TL response (at 12 weeks or on study on the basis of a maximum five TLs) on survival was studied through landmark and time-dependent Cox models adjusted for baseline tumor load, occurrence of new lesions, or unequivocal progression of nontarget disease. RESULTS Data were obtained from 23,259 patients with cancer (36% lung, 28% colorectal, 11% breast, and 25% other); 15,620 received TCAs, predominantly transduction or angiogenesis inhibitors, as a single agent (37%), combined with other TCAs (7%), or as chemotherapy (56%); 28% received chemotherapy only; and 5% received best supportive care or placebo. A total of 17,222 patients contributed to the analyses. Within-patient variability decreased with increasing number of TLs, similarly for TCAs (with/without chemotherapy) and chemotherapy only. Mixed responses occurred proportionally in all treatment classes. Landmark analyses showed an ordinal relationship between percentage change from baseline to 12 weeks and overall survival, and demonstrated a clear distinction between tumor shrinkage and progressive disease according to RECIST. Time-dependent analysis showed no marked improvement in the ability to predict survival on the basis of TL tumor growth compared with nontarget progression or new lesion occurrence, regardless of treatment. Similar results were seen for major tumor types and different classes of TCAs. CONCLUSION This work reinforces that RECIST version 1.1 perform well for response assessment of TCAs.

2534 Background: The Response Evaluation Criteria in Solid Tumors (RECIST) v1.1 were derived from an international collaborative effort supported by data from clinical trials (16 studies, 9147 patients) on cytotoxic chemotherapy (CT), providing a standard tool for response assessment. RECIST’s role has been questioned for TCA. Using a pooled individual patient database (IPD) from clinical trials performed by industry and cooperative groups, we assessed whether modifications to RECIST are required to evaluate antitumor activity of TCA. Methods: Data were collected from phase 2 and 3 clinical trials testing TCA in solid tumors. To study the occurrence of mixed responses, the variability of response of lesions within patients was studied. Furthermore, response was correlated with survival through landmark analyses and time dependent Cox models. Results: Clinical data were obtained from 23,259 patients, mainly with lung (36%), colorectal (28%) or breast cancer (11%). 15,620 patients (67%) received a TCA, mainly transduction or angiogenesis inhibitors, either as single agent (37%) or combined with other TCAs (7%) or CT (56%); 28% received CT only and 5% best supportive care or placebo. Within-patient variability reduced as the number of lesions used for response assessment increased, and did so similarly for TCAs (+/- CT) and CT. Mixed responses seemed to occur similarly across these treatment categories as well. Landmark analyses showed improving overall survival by % tumor shrinkage and a clear distinction between the effect of tumor shrinkage and progressive disease (PD) according to RECIST 1.1. This was confirmed by time dependent analysis. In addition target lesion growth showed no marked improvement in overall survival prediction over and above the other components of RECIST 1.1 PD (new lesions, non-target PD), regardless of treatment (TCA, CT or both) received. Similar results were seen focusing on major tumor types and classes of TCA. Conclusions: Using a large IPD dataset we demonstrated that RECIST 1.1 performs equally well for response assessment of TCA as for CT. No modifications are required.

BACKGROUNDThe objective of this study was to demonstrate that computer‐extracted image phenotypes (CEIPs) of biopsy‐proven breast cancer on magnetic resonance imaging (MRI) can accurately predict pathologic stage.METHODSThe authors used a data set of deidentified breast MRIs organized by the National Cancer Institute in The Cancer Imaging Archive. In total, 91 biopsy‐proven breast cancers were analyzed from patients who had information available on pathologic stage (stage I, n = 22; stage II, n = 58; stage III, n = 11) and surgically verified lymph node status (negative lymph nodes, n = 46; ≥ 1 positive lymph node, n = 44; no lymph nodes examined, n = 1). Tumors were characterized according to 1) radiologist‐measured size and 2) CEIP. Then, models were built that combined 2 CEIPs to predict tumor pathologic stage and lymph node involvement, and the models were evaluated in a leave‐1‐out, cross‐validation analysis with the area under the receiver operating characteristic curve (AUC) as the value of interest.RESULTSTumor size was the most powerful predictor of pathologic stage, but CEIPs that captured biologic behavior also emerged as predictive (eg, stage I and II vs stage III demonstrated an AUC of 0.83). No size measure was successful in the prediction of positive lymph nodes, but adding a CEIP that described tumor “homogeneity” significantly improved discrimination (AUC = 0.62; P = .003) compared with chance.CONCLUSIONSThe current results indicate that MRI phenotypes have promise for predicting breast cancer pathologic stage and lymph node status. Cancer 2016;122:748–757. © 2015 American Cancer Society.