Real-World Outcomes in Metastatic Renal Cell Carcinoma: Insights From a Joint Community-Academic Registry

Clinical Research Practices Original Contribution Real-World Outcomes in Metastatic Renal Cell Carcinoma: Insights From a Joint Community-Academic Registry By Michael R. Harrison, MD, Bradford R. Hirsch, MD, Daniel J. George, MD, Mark S. Walker, PhD, Connie Chen, PharmD, Beata Korytowsky, MA, Edward Stepanski, PhD, and Amy P. Abernethy, MD, PhD Duke University Medical Center; Center for Learning Health Care, Duke Clinical Research Institute, Durham, NC; ACORN Research; Memphis, TN; and Pfizer, New York, NY Abstract Introduction: As new therapeutics for metastatic renal cell carcinoma (mrcc) are quickly introduced to market, comparative randomized trial evidence guiding treatment decisions is lacking, especially in the second treatment exposure and beyond. As a demonstration case, we studied mrcc in real-world clinical settings by creating a joint community-academic retrospective mrcc registry to assess outcomes. Materials and Methods: For this overall survival (OS) analysis, the analytic cohort included all patients in the registry diagnosed between January 1, 2007, to May 31, 2011 (N 384). Patients were grouped by up to three treatment exposures according to each drug s mechanism of action: vascular endothelial growth factor tyrosine kinase inhibitor (VEGFR TKI), mammalian target of rapamicin inhibitor (mtor), or no systemic treatment (NSTx, which could include radiation or surgery). OS by exposure sequence was evaluated using Kaplan-Meier, pairwise comparison, and Cox regression analyses. Results: Median OS was 17.2 months. OS (months) for one exposure was: mtor 5.4, TKI 18.2, NSTx 18.4; for two exposures: mtor/tki 9.3, TKI/mTOR 13.9, TKI/TKI 35.2; and for three exposures: TKI/mTOR/TKI 20.9, TKI/TKI/mTOR 33.1. By pairwise comparison, OS for TKI, mtor/tki, TKI/mTOR, TKI/ TKI, TKI/mTOR/TKI and TKI/TKI/mTOR sequences was greater than mtor (all P.04); demographics confirmed that individuals treated with early mtor inhibition more commonly had adverse prognostic features. In Cox regression analysis, compared with the referent (TKI), TKI/TKI (hazard ratio 0.53; P.03) had a lower risk of death, and mtor (hazard ratio 2.16; P.002) had a higher risk of death. Conclusions: mrcc survival outcomes are different by pattern, with general findings consistent with trial-based expectations in similar patient populations. Real-world data can provide context around patterns of care and impact when experimental trial data are lacking. Introduction Comparative effectiveness research (CER) is an area of intense interest. CER is defined by the Agency for Healthcare Research and Quality as being designed to inform health care decisions by providing evidence on the effectiveness, benefits, and harms of different treatment options. 1 Learning health care is an extension of the CER continuum whereby the research results are intentionally fed back into an overall system of coordinated data so that clinical care and research are more tightly coupled and inform each other. When powered by real-time data, analytics, and clinical decision support, rapid learning health care leads to new insights into the effectiveness of different treatments and into the patient experience. CER and learning health care are pressing issues in oncology, especially in light of the variety of evolving treatments for different cancer types. However, there are limited data sources to support analyses and to generate new evidence about the role of new treatments in routine practice. Metastatic renal cell carcinoma (mrcc) is a clear illustration of this issue. In 2012, kidney and renal pelvis tumors accounted for approximately 4% of all cancer diagnoses in the United States, with 64,770 new cases and 13,570 deaths. 2 Of patients who present with localized disease, 20% to 30% develop recurrence (ie, advanced or metastatic disease), 3 and up to one third of all patients present with metastatic disease at initial diagnosis. 4 Until the mid-2000s, cytokine therapy (high-dose interleukin-2 or interferon-alfa) was the treatment of choice for mrcc, but poor response rate, marginal survival benefit, and significant toxicity limited applicability. 5 The evolution of molecular technologies has directed therapy toward mrcc-specific targets, most notably the vascular endothelial growth factor (VEGF) pathway and the mammalian target of rapamycin (mtor) pathway, both of which are related to the pathogenesis of clear-cell mrcc. 6 As of July 2012, seven distinct targeted therapies have been approved by regulatory agencies as treatments that improve overall survival (OS; temsirolimus) 7 and/or progression-free survival (sorafenib, sunitinib, everolimus, bevacizumab/interferon-alfa, pazopanib, and axitinib). 8-14 The rapid availability of these new treatments has given the clinician unprecedented flexibility in the treatment of patients with mrcc. And although practice guidelines have been published, 15 there are a multitude of conflicting approaches to the sequencing and prioritization of treatments influenced by personal clinical experience, published clinical trial evidence, and treatment-emergent adverse events. The clinician must prescribe sequential therapies from the available treatment options with little practical or trial-based evidence about which treat- Copyright 2013 by American Society of Clinical Oncology jop.ascopubs.org 1

Harrison et al ment selection, or sequence of selections, leads to the longest durable disease response or survival. In the absence of comparative randomized trials, it is informative to study what is occurring in clinical practice (ie, the real world ). To date there has only been one marketed product (axitinib) with published head-to-head clinical trial data against another targeted agent. We therefore developed a novel registry of patients with mrcc treated in community and academic settings. The mrcc registry includes several core features: (1) longitudinal data capture in order to follow the natural history of the disease in routine practice, (2) adequate capture of variables relevant to the patient and disease to allow robust analyses and minimize confounding, (3) use of mobile health technology to augment traditional collection mechanisms, and (4) robust capture of the traditional core outcomes such as survival and tumor response for all patients. The registry represents a new evidence development paradigm, utilizing data warehouses and information collected directly from patients in routine clinical care, with special attention to treatment selection and sequence to provide new insights into the relative effectiveness of the agents. Here we report the observed survival of patients with mrcc, categorized by treatment and mechanism of action sequences, organized to reflect how oncologists think when choosing each successive mrcc treatment. 15,16 Although the manuscript focuses on a relatively specific questionos among patients with mrcc, it also reflects a practical demonstration of what is possible with the capture of data from routine clinical practice. Materials and Methods We created a joint community and academic registry with one set of data elements, a standardized case report form (CRF), data dictionary, and statistical analysis plan. This collaboration resulted in a multicenter registry consisting of an academic research network (Duke Oncology Network; Durham, NC) and a community-based oncology network (ACORN Research; Memphis, TN). The institutional review boards affiliated with Duke University Health System and ACORN approved the registry. A waiver of informed consent was granted. All patients identified as meeting the following criteria were included in the registry: adults ( 18 years) diagnosed with mrcc between Janaury 1, 2007, and May 31, 2011, who received some portion of their mrcc care in the Duke Oncology Network or in a community oncology site affiliated with ACORN. Patients who were treated as part of clinical research trials were excluded because of protocol-influenced practice patterns. Because data collection was retrospective, we captured those elements that were being collected consistently in routine care at participating clinics. Data Sources The Duke University data warehouse was accessed to gather laboratory, clinical, financial, sociodemographic, procedural, treatment, and patient-reported data on Duke patients with mrcc that met the inclusion criteria. Review of paper medical records supplemented electronic data, as necessary. ACORN maintains a similar database of demographic, medical, treatment, and patient-reported outcome variables drawn from the electronic medical record systems, billing systems, and patientreported outcome repositories of 11 community-based oncology practices. Databases were queried for all potentially eligible cases. All cases identified as meeting the eligibility criteria were included until the target accrual goal ( 200 each by Duke and ACORN) at each site was reached. Cases were screened approximately in the order of their presentation within the database, which generally corresponded to the date of diagnosis. Each institution maintained its own independent screening list and clinical database. Each site s research coordinator confirmed eligibility and abstracted data into the common registry. Completed CRFs underwent quality review before entry into a secure repository and then were merged into a single registry. Data Analysis The initial analytic data set contained patients with data entered by October 31, 2011 (N 455; community 255, academic 200). Two patients were excluded as a result of missing treatment information. In order to avoid survival bias and provide adequate time for survival follow-up, we focused on cases diagnosed between January 1, 2007, and June 31, 2011, which excluded 68 patients, for a total of 385 patients (community 254, academic 131). One patient s tumor type was changed during pathology re-review; this patient was excluded. The final analytic cohort included 384 patients (community 254, academic 130; Appendix Figure A1, online only). One patient did not have adequate date information to contribute to the survival analyses; therefore, the OS analyses reflect 383 patients and summarize survival time from diagnosis of mrcc to death. The resulting data was analyzed in aggregate, by line of therapy, therapeutic setting, and exposure sequence. Analyses concentrated on 2007 and after, when targeted therapies were widely available for clinical use (approximately 1 year after approvals of sorafenib and sunitinib). Sensitivity analyses demonstrated that the 68 patients diagnosed before 2007 would have been predominantly assigned to the other or no systemic therapy groups. Because these patients would be included only in the registry if they survived to receive treatment after 2007, this would bias the OS estimate for these two treatment categories (survival bias). Median survival follow-up was 11.7 months. A total of 170 unique treatment regimens were identified in the initial analytic data set. Hence, patients were grouped by sequencing of treatments ( exposure sequence ) that reflects clinical decision making 16 : VEGF receptor tyrosine kinase inhibitors (TKI), mtor, or other (eg, bevacizumab, cytokines, chemotherapy, or nonapproved combinations). Patients who received no systemic treatment as of the data cutoff were classified as no systemic therapy (NSTx). Patients in the NSTx group could have had radiation or surgery. Receipt of a drug within one of these classes (TKI, mtor, NSTx, or other) was considered a systemic treatment exposure. Receipt of the same drug again was not considered an additional 2 JOURNAL OF ONCOLOGY PRACTICE Copyright 2013 by American Society of Clinical Oncology

Overall Survival in Patients With mrcc exposure, but later receipt of another drug in the same class was considered an additional exposure (eg, sunitinib then sorafenib would be classified as TKI/TKI). Patients who received a combination of therapies or a non-tki, non-mtor monotherapy within their first three exposures were classified as other. Patients were classified only by their first three exposures (or fewer if three exposures) into comparison groups for survival analysis. Uncommon exposure sequences with N 2 (TKI/ mtor/mtor, TKI/TKI/TKI, and mtor/tki/tki) were grouped as other. Variables were summarized using routine descriptive statistics. 2 goodness-of-fit tests, independent t tests, and analysis of variance were used to examine group differences. Kaplan-Meier survival analysis was used to estimate time to event. Log-rank tests were used to compare OS across groups. Cox regression (method of Hosmer and Lemeshow 17 ) was used to examine the effect of exposure sequence controlling for demographic, disease, and treatment characteristics; covariate rather than propensity score adjustment was used because of the number and sample size of regimen groups. Sidak adjustment was used for multiple comparisons. Results The analytic cohort (N 384) averaged 64.0 years of age (standard deviation [SD] 10.7), was 66% male, 72% white, had predominantly clear-cell mrcc (64%), and averaged 1.6 (SD 0.9) metastatic sites; 69% had undergone a prior nephrectomy. Overall, 29% of patients had favorable Memorial Sloan-Kettering Cancer Center (MSKCC) risk, and 9% had poor risk (Appendix Table A1, online only). Exposure Sequences Derived exposure sequences are presented in Figure 1, and patient characteristics by exposure sequence are detailed in Appendix Table A1. The number of patients per sequence ranged from a low of 11 (mtor/tki and TKI/TKI/mTOR) to a high of 109 (TKI). Twenty-six percent of patients received no systemic therapy and 10% received other, including bevacizumab (n 7), chemotherapy (n 3), combinations (n 7), and immunotherapy (n 16). The largest single group was TKI (28%). An additional 9% received monotherapy with an mtor, resulting in 37% of the cohort receiving only one line of therapy. Twenty-seven percent received two or more treatment exposures, with 19% receiving two treatment exposures and 8% receiving three treatment exposures (Figure 1). Mean age differed among groups (P.007). Patients in the other group were significantly younger, consistent with the administration of immunotherapy (predominantly interleukin-2) in one quarter of this group. There was a significant sex differene by exposure sequence distribution (P.032). The highest fraction of patients with favorable MSKCC risk was the TKI/TKI/mTOR exposure sequence, whereas the highest fraction of patients with a poor risk was the mtor exposure sequence (Appendix Table A1). The sequence with the lowest fraction of patients with a favorable MSKCC risk was mtor/ TKI. Patients in the mtor sequence trended toward worse Three exposures disease with more metastatic sites (median 2.0; P.067) and impaired performance status (18%; P not significant). OS No systemic treatment 26% Other 10% TKI/mTOR/TKI 5% TKI/TKI/mTOR 3% mtor/tki 3% TKI/mTOR 8% Two exposures TKI/TKI 8% TKI 28% mtor 9% One exposure Figure 1. Derived exposure sequences. mtor, mammalian target of rapamycin; TKI, tyrosine kinase inhibitor. A total of 216 deaths were recorded for 383 patients. Median OS, reflecting time from diagnosis of mrcc to death, was 17.2 months (95% CI, 13.5 to 23.4). Median OS significantly differed by exposure mechanism of action and sequence (P.001; Appendix Figure A2, online only, and Table 1), ranging from a low of 5.4 months (mtor) to a high of 35.2 months (TKI/TKI). Median OS for the NSTx group was 18.4 months. The upper bounds of the CI around OS for the TKI/ TKI, TKI/TKI/mTOR, and NSTx exposure sequences were not reached. Pairwise comparisons were performed between each of the seven exposure sequences plus the other and NSTx groups. Before adjustment for multiple comparisons, there were significant differences between 11 of the 36 possible pairwise combinations; after Sidak adjustment, five remained significant (P.05), all of which involved mtor as the first line of treatment in one or both of the pairs (Appendix Figure A3, online only). A Cox regression model of OS (referent TKI) controlled for significant covariates from among patient demographic and disease characteristics (Table 2). Patients who received mtor only (hazard ratio [HR] 2.2; P.002) demonstrated the highest risk of death, and patients who received TKI/TKI (HR 0.53; P.031), TKI/mTOR/TKI (HR 0.51; P.068), and other (HR 0.63; P.071) demonstrated the lowest risk. Patients treated with mtor alone retained this high risk even after adjustment for covariates; this suggests unidentified prognostic factors not included in the Cox model. Patients with more severe disease fared poorly: liver metastasis (HR 1.67; P.004), higher Fuhrman grade (III/IV v II; HR 1.75; P.014), or undocumented Fuhrman grade (v II; HR 1.58; P.038) demonstrated higher risk (Fuhrman grade was undocumented in 48%). Copyright 2013 by American Society of Clinical Oncology jop.ascopubs.org 3

Harrison et al Table 1. Median Overall Survival by Exposure Sequence* Variable TKI (n 109) TKI/TKI (n 31) TKI/TKI/mTOR (n 11) TKI/mTOR (n 34) TKI/mTOR/TKI (n 18) mtor (n 33) mtor/tki (n 11) Other (n 39) NSTx (n 98) Patients, No. 109 31 11 34 18 33 11 39 97 Events, No. 56 16 5 24 9 29 9 21 47 Median, months 18.20 35.19 33.08 13.93 20.93 5.39 9.26 29.63 18.43 95% CI 8.7 to 33.1 13.7 to NR 13.8 to NR 7.6 to 23.8 16.5 to 30.1 3.4 to 7.9 6.9 to 12.1 14.3 to 41.3 9.3 to NR Log-rank P*.001 Abbreviations: mtor, mammalian target of rapamycin; NR, not reached; NSTx, no systemic treatment; TKI, tyrosine kinase inhibitor. * Adjusted pairwise differences (P.05) for median survival: mtor mtor/tki, TKI/mTOR/TKI, TKI/TKI/mTOR, TKI/TKI, and Other. Table 2. Summary of Cox Regression Analysis of Overall Survival (final model) Parameter HR P 95% CI Reference group TKI alone TKI/TKI 0.53.031 0.30 to 0.95 TKI/TKI/mTOR 0.44.084 0.17 to 1.12 TKI/mTOR 1.042.868 0.64 to 1.69 TKI/mTOR/TKI 0.51.068 0.25 to 1.05 mtor 2.16.002 1.34 to 3.48 mtor/tki 1.31.464 0.63 to 2.71 Other 0.63.071 0.38 to 1.04 NSTx 0.95.80 0.64 to 1.41 Overall parameter.001 Stage IV v stage IV 1.08.746 0.69 to 1.68 Unknown v stage IV 0.63.041 0.40 to 0.98 Overall stage at diagnosis.003 Metastatic to liver (yes/no) 1.67.0043 1.18 to 2.38 Fuhrman III/IV v I/II 1.75.014 1.12 to 2.73 Unknown v I/II 1.58.038 1.03 to 2.44 Overall Fuhrman grade.046 MSKCC risk Intermediate v poor 0.52.003 0.33 to 0.80 Favorable v poor 0.21.001 0.12 to 0.36 Overall.001 Abbreviations: HR, hazard ratio; MSKCC, Memorial Sloan-Kettering Cancer Center; mtor, mammalian target of rapamycin; NSTx, no systemic treatment; TKI, tyrosine kinase inhibitor. Discussion In the absence of randomized clinical trial data, how does a clinician make a decision about which drug is right for a given patient? With the approval of seven targeted therapies for mrcc over the past six years, questions arise as to how these drugs are being used and the outcomes of sequential use in the real-world setting. Restrictive inclusion criteria in randomized clinical trials limit the treatment population and the generalizability to the broader population of patients with mrcc. Expanded access trials, such as that of sunitinib for first-line treatment of patients with mrcc, 18 have provided insights into real-world outcomes, but detailed data on practice patterns and outcomes across sequential lines of therapy are limited. Further, prognostic classification systems focus on pooled clinical trial data and may not be fully generalizable, 19-21 and most retrospective registries have been treatment based, focused on only one or two lines of therapy. 22 Our registry approach differs in that it represents a unique all-in look at mrcc across academic and community settings, treatment exposures, mechanisms of action, and treatment sequencing. Registries such as this will provide the backbone for CER and rapid learning health care in the future. The key output of this analysis is an overview of real-world practice patterns in mrcc and the associated outcomes. After a median follow-up of 11.7 months, 26% of patients received no systemic therapy, 28% received monotherapy (TKI was the most common treatment exposure), and 27% of patients received two treatment exposures (Appendix Figure A1). Patients treated with two exposures could be divided into those who received a TKI in both of the first two exposures (n 42, 11%) versus those who received mtor in one of the first two exposures (n 63; 16.4%) with a 40/60 split. An advantage of a registry is that, as more data accumulate, findings will be iteratively updated. Consistent with clinical trials experience, 7,8 patients treated with two TKIs in the first three treatment exposures (independent of order or number of exposures) had significantly superior survival compared with patients treated with an mtor alone (Table 1, Appendix Figure A3). After adjustment of covariates and compared with TKI alone, a Cox model showed that patients treated with two TKI exposures in the first three treatment exposures also had a significantly lower risk of death (although TKI/TKI/mTOR and TKI/mTOR/TKI demonstrated a trend only), whereas mtor demonstrated a significantly higher risk of death (Table 2). This study of a real-world community/academic registry offers broad insights into the care of patients with mrcc, with overarching results consistent with expectations. Despite wide variability in the treatment and management of patients with mrcc, real-world outcomes generally follow expectations based on prospective phase III clinical trials (Appendix Table A2, online only). Unlike clinical trial analyses, we were able to report on patients from diagnosis to the ultimate outcome of therapy as opposed to focusing only on a treatment window. We found that one quarter of our cohort received no systemic treatment at the data cutoff, a previously unidentified cohort needing further study. Although a direct comparison between clinical trials and our observational data is not possible, the trial-based data do provide some context with which to view our real-world results. Patients exposed only to an mtor inhibitor were at the highest risk for death compared with those who only received a 4 JOURNAL OF ONCOLOGY PRACTICE Copyright 2013 by American Society of Clinical Oncology

Overall Survival in Patients With mrcc TKI, a risk that persisted even after controlling for covariates. Patients treated with an mtor inhibitor had a median OS of just 5.4 months, with a relatively tight confidence interval (95% CI, 3.4 to 7.9) compared with other exposure sequences. The median survival of our patients initially treated with an mtor sequence was shorter than the 10.9 months reported by Hudes et al for patients in the temsirolimus (an mtor) arm of the phase III Global Advanced Renal Cell Carcinoma (ARCC) trial. 7 Whether the Global ARCC patients received a second treatment exposure was not reported, but with a progressionfree survival (PFS) of 3.8 months on the temsirolimus arm, some patients may have received second-line therapies. Interestingly, of the 44 patients who received an mtor first line, only 11 went on to receive a TKI. Median OS of the patients in the mtor/tki group was 9.2 months, approaching that reported by Hudes et al. In general, patients treated with mtor inhibitors alone or in the first-line setting had a poor MSKCC risk, consistent with the current clinical guidelines for mtor inhibitors. 23 The overwhelming majority of mtor inhibitor use in our cohort, especially in the first exposure, was with temsirolimus. These results might reflect factors not identified in the Cox regression and do not imply that an mtor is an inferior treatment. Rather, the results most likely are a reflection of a constellation of factors that result in a patient being selected to receive an mtor. Most patients who received two treatment exposures beginning with a TKI had relatively long median OS, approaching 3 years (Table 1). The exception to this was the TKI/mTOR group, whose median OS of only 14 months was still significantly greater (P.003) than that in the mtor group. Retrospective analyses have been conflicting with regard to TKI/TKI versus TKI/mTOR sequencing comparisons. 24,25 However, prospective data regarding TKI/ TKI versus TKI/mTOR are available from the INTORSECT study, 26 which randomly assigned patients with progression of disease on sunitinib (TKI) to either temsirolimus (mtor) or sorafenib (TKI). While there was no significant difference in PFS (temsirolimus 4.28 months, sorafenib 3.92 months; HR 0.87; 95% CI, 0.71 to 1.07; P.1933), sorafenib was statistically superior for the secondary end point of OS (temsirolimus 12.27 months, sorafenib 16.64 months; HR 1.31; P.014). The survival difference between these sequences was more pronounced in our cohorts, but OS is not directly comparable to INTORSECT as a result of differences in the starting point from which OS was measured. Differences in patient selection, number of patients, and variations in efficacy of mtor inhibitors (ie, everolimus v temsirolimus), and VEGF TKIs could also play a role. Finally, TKI monotherapy revealed a different picture of median OS as compared with mtor inhibitors. After controlling for covariates, only the TKI/TKI (HR 0.53; 95% CI, 0.30 to 0.95; P.031) sequences had a significantly decreased risk for death versus TKI monotherapy, although TKI/mTOR/TKI, other, and TKI/TKI/mTOR (Table 2) sequences each demonstrated a strong trend. Data from the prospective phase III AVOREN trial of interferon-alfa with or without bevacizumab as first-line treatment of mrcc showed that patients treated on the bevacizumabcontaining arm who then received a second-line TKI had a median OS of 38.6 months, 10 in line with the median OS for our TKI/TKI (35.2 months) and TKI/TKI/mTOR (33.1 months) groups. In addition, patients treated with a TKI/TKI exposure sequence (either sorafenib 3 sunitinib or sunitinib 3 sorafenib) in a retrospective Czech database had median OS of 30.0 and 35.4 months, respectively, 27 similar to our TKI/TKI group s survival of 35.2 months. A key outcome of this analysis was the identification of new areas in need of study. As an example, the substantial number of patients who did not undergo systemic treatment for mrcc (n 98; 26%) needs to be better understood. This group has not been reflected in clinical trials, expanded access programs, treatment-based registries, or observational reports. Heng et al 10 recently reported on a cohort who did not meet eligibility for clinical trials and concluded that these individuals had inferior outcomes and needed clinical trials designed specifically for this population. The no-treatment group in our cohort had a median OS of 18.4 months with a wide CI (95% CI, 9.3 to not reached). This suggests great variability within the outcomes of this group and warrants further exploration. Limitations This work has several important limitations. First, because it was an observational study, the clinicians actively selected treatments, with unmeasured factors likely affecting treatment decisions. Inferences about the superiority of one treatment sequence over another are limited. Second, we did not directly capture the factors that led to physician s rationale behind changes in treatment exposures. These should be included in prospective databases to provide important information about contemporary mrcc management decisions. For example, there could be wide variation regarding how physicians determine progression on the basis of clinical and radiographic data. Third, data were collected retrospectively, with some missing data regarding known prognostic variables. Finally, grouping treatment exposures by drug mechanism may obscure important drug-specific information. For example, there may be differences in outcomes for everolimus versus temsirolimus or sunitinib versus sorafenib. Sequencing may also be important and affected by specific agents. Future prospective studies will need to break this out by specific drug to elucidate key differences between drugs and sequences that have not been tested in randomized trials. Conclusions and Next Steps As data in the registry accumulate, understanding about treatment exposures and outcomes will grow. Overall, this realworld registry provides evidence that outcomes in routine clinical practice may be similar to those observed in more selective, controlled clinical trials. Interesting insights have been gathered through this analysis, including data about differences in outcomes between mtor and TKI treatment exposures. Patients with mrcc who did not receive systemic therapy for a substantial period of time after diagnosis were identified and warrant further analysis. Registries are intended to augment, Copyright 2013 by American Society of Clinical Oncology jop.ascopubs.org 5

Harrison et al but do not take the place of, prospective clinical trial data. However, new therapies are rapidly becoming available in diseases such as mrcc without corresponding large, prospective clinical comparative trials with which to inform practice. Realworld registries can provide valuable insights in these settings. Overall, this mrcc registry represents an exciting new paradigm, but it is not adequate or complete. Further opportunities abound. First is the need for more robust observational cohorts in which outcomes, adverse events, and toxicities can be monitored across community settings. Next, observations should be operationalized into actionable information through real-time analytics, novel data visualization, patient education, and decision support solutions. The impact of reinvestment of lessons from a learning health care cycle should be monitored through continuously updating quality metrics. Finally, datasets should be made available to basic scientists to support basic discovery. This practical demonstration model is a step along the pathway to a learning health system. Acknowledgment Supported by Pfizer. Medical writing support was provided by Donald T. Kirkendall of Duke University Medical Center with funding from Pfizer. Previously presented in part as the American Society of Clinical Oncology Genitourinary Symposium, February 2-4, 2012, San Francisco, CA, and the Kidney Cancer Symposium, October 14-15, Chicago, IL. We acknowledge the important contributions of Laura Roe, Ursula Rogers, and Donald T. Kirkendall, PhD, ELS (all Duke employees) in the preparation of this article. Authors Disclosures of Potential Conflicts of Interest Although all authors completed the disclosure declaration, the following author(s) and/or an author s immediate family member(s) indicated a financial or other interest that is relevant to the subject matter under consideration in this article. Certain relationships marked with a U are those for which no compensation was received; those relationships marked with a C were compensated. For a detailed description of the disclosure categories, or for more information about ASCO s conflict of interest policy, please refer to the Author Disclosure Declaration and the Disclosures of Potential Conflicts of Interest section in Information for Contributors. Employment or Leadership Position: Connie Chen, Pfizer (C); Beata Korytowsky, Pfizer Inc. (C); Amy P. Abernethy, Advoset (C), Orange Leaf Associates (C) Consultant or Advisory Role: Michael R. Harrison, Exelixis (C), AVEO (C), Novartis (C); Daniel J. George, Pfizer (C), Novartis (C), Molecular Insight/Progenix (C), Dendreon (C), Aveo (C), Astellas (C), Bayer (C), Genentech/Roche (C), Teva (C), Exelixis (C), BMS (C), Viamet (C); Amy P. Abernethy, Novartis (C), Pfizer (C) Stock Ownership: Connie Chen, Pfizer; Beata Korytowsky, Pfizer Honoraria: Michael R. Harrison, Prometheus; Daniel J. George, Pfizer, Novartis, Dendreon Research Funding: Michael R. Harrison, Bristol-Myers Squibb, Pfizer, Exelixis, Dendreon, Argos; Bradford R. Hirsch, Pfizer, Dendreon, Bristol Myers Squibb, GlaxoSmithKline; Daniel J. George, Pfizer, Inc, Novartis, Genentech/Roche, Millenium/Takeda, Janssen, Exelixis, GlaxoSmithKline, Genentech/Roche, Molecular Insight/Progenix; Mark S. Walker, Pfizer; Edward Stepanski, Pfizer; Amy P. Abernethy, Pfizer, Helsinn, Amgen, Kanglaite, Alexion, Biovex, DARA, MiCo Expert Testimony: None Patents, Licenses or Royalties: None Other Remuneration: None Author Contributions Conception and design: All authors Financial support: Connie Chen Administrative support: Connie Chen, Edward Stepanski, Amy P. Abernethy Collection and assembly of data: Mark S. Walker Data analysis and interpretation: All authors Manuscript writing: All authors Final approval of manuscript: All authors Corresponding author: Amy P. Abernethy, MD, PhD, DUMC Box 3436, Duke University Medical Center, Durham, NC 27710; e-mail: amy.abernethy@dm.duke.edu. DOI: 10.1200/JOP.2013.001180; published online ahead of print at jop.ascopubs.org on November 26, 2013. References 1. Agency for Healthcare Research and Quality: What is comparative effectiveness research? http://effectivehealthcare.ahrq.gov/index.cfm/what-iscomparative-effectiveness-research1/ 2. Siegel R, Naishadham D, Jemal A: Cancer statistics, 2012. CA Cancer J Clin 62:10-29, 2012 3. Rouvière O, Bouvier R, Négrier S, et al: Nonmetastatic renal-cell carcinoma: Is it really possible to define rational guidelines for post-treatment follow-up? Nat Clin Pract Oncol 3:200-213, 2006 4. Motzer RJ, Bander NH, Nanus DM: Renal-cell carcinoma. N Engl J Med 335:865-875, 1996 5. Hutson TE: Targeted therapies for the treatment of metastatic renal cell carcinoma: Clinical evidence. The Oncologist 16:14-22, 2011 (suppl 2) 6. Mulders P: Vascular endothelial growth factor and mtor pathways in renal cell carcinoma: Differences and synergies of two targeted mechanisms. BJU Int 104:1585-1589, 2009 7. Hudes G, Carducci M, Tomczak P, et al: Temsirolimus, interferon alfa, or both for advanced renal-cell carcinoma. N Engl J Med 356:2271-2281, 2007 8. Escudier B, Bellmunt J, Négrier S, et al: Phase III trial of bevacizumab plus interferon alfa-2a in patients with metastatic renal cell carcinoma (AVOREN): Final analysis of overall survival. J Clin Oncol 28:2144-2150, 2010 9. Escudier B, Eisen T, Stadler WM, et al: Sorafenib in advanced clear-cell renal-cell carcinoma. N Engl J Med 3556:125-134, 2007 10. Heng DY, Choueiri TK, Lee J-L, et al: An in-depth multicentered populationbased analysis of outcomes of patients with metastatic renal cell carcinoma (mrcc) who do not meet eligibility criteria for clinical trials. J Clin Oncol 30:286s, 2012 (suppl 5; abstr 4536) 11. Motzer RJ, Escudier B, Oudard S, et al: Efficacy of everolimus in advanced renal cell carcinoma: A double-blind, randomised, placebo-controlled phase III trial. Lancet 372:449-456, 2008 12. Motzer RJ, Hutson TE, Tomczak P, et al: Sunitinib versus interferon alfa in metastatic renal-cell carcinoma. N Engl J Med 356:115-124, 2007 13. Rini BI, Halabi S, Rosenberg JE, et al: Phase III trial of bevacizumab plus interferon alfa versus interferon alfa monotherapy in patients with metastatic renal cell carcinoma: Final results of CALGB 90206. J Clin Oncol 28:2137-2143, 2010 14. Sternberg CN, Davis ID, Mardiak J, et al: Pazopanib in locally advanced or metastatic renal cell carcinoma: Results of a randomized phase III trial. J Clin Oncol 28:1061-1068, 2010 15. Hudes GR, Carducci MA, Choueiri TK, et al: NCCN Task Force report: Optimizing treatment of advanced renal cell carcinoma with molecular targeted therapy. J Natl Compr Cancer Netw 9:S1-S29, 2011 (suppl 1) 16. Sonpavde G, Choueiri TK, Escudier B, et al: Sequencing of agents for metastatic renal cell carcinoma: Can we customize therapy? Eur Urol 61:307-316, 2012 17. Hosmer DW, Lemeshow S: Applied Survival Analysis: Regression Modeling of Time to Event Data. New York, NY: John Wiley & Sons, 2008 18. Gore ME, Szczylik C, Porta C, et al: Safety and efficacy of sunitinib for metastatic renal-cell carcinoma: An expanded-access trial. Lancet Oncol 10:757-763, 2009 6 JOURNAL OF ONCOLOGY PRACTICE Copyright 2013 by American Society of Clinical Oncology

Overall Survival in Patients With mrcc 19. Manola J, Royston P, Elson P, et al: Prognostic model for survival in patients with metastatic renal cell carcinoma: Results from the International Kidney Cancer Working Group. Clin Breast Cancer 17:5443-5450, 2011 20. Mekhail TM, Abou-Jawde RM, Boumerhi G, et al: Validation and extension of the Memorial Sloan-Kettering prognostic factors model for survival in patients with previously untreated metastatic renal cell carcinoma. J Clin Oncol 23(4):832-841, 2005 21. Motzer RJ, Mazumdar M, Bacik J, et al: Survival and prognostic stratification of 670 patients with advanced renal cell carcinoma. J Clin Oncol 17:2530-2540, 1999 22. Heng DY, Xie W, Regan MM, et al: Prognostic factors for overall survival in patients with metastatic renal cell carcinoma treated with vascular endothelial growth factor targeted agents: Results from a large, multicenter study. J Clin Oncol 27:5794-5799, 2009 23. National Comprehensive Cancer Network: Kidney cancer. National Comprehensive Cancer Network and American Cancer Society; 012 24. Busch J, Seidel C, Kempkensteffen C, et al: Sequence therapy in patients with metastatic renal cell carcinoma: Comparison of common targeted treatment options following failure of receptor tyrosine kinase inhibitors. Eur Urol 60:1163-1170, 2011 25. Heng DY, Lee J-L, Harshman LC, et al: A population-based overview of sequences of targeted therapy in metastatic renal cell carcinoma (mrcc). J Clin Oncol 30, 2012 (suppl 5; abstr 387) 26. Hutson TE, Escudier B, Esteban E, et al: Temsirolimus vs sorafenib as second line therapy in metastatic renal cell carcinoma: Results from the INTORSECT Trial. Presented at the European Society for Medical Oncology Congress, Vienna, Austria, September 28-October 2, 2012 27. Buchler T, Klapka R, Melichar B, et al: Sunitinib followed by sorafenib or vice versa for metastatic renal cell carcinoma data from the Czech registry. Ann Oncol 23:395-401, 2011 28. Motzer RJ, Hutson TE, Tomczak P, et al: Overall survival and updated results for sunitinib compared with interferon alfa in patients with metastatic renal cell carcinoma. J Clin Oncol 27:3584-3590, 2009 29. Escudier BJ, Belldegrun AS, Melichar B, et al: Final results of the phase III, randomized, double-blind AVOREN trial of first-line bevacizumab (BEV) interferon-2a (IFN) in metastatic renal cell carcinoma (mrcc). J Clin Oncol 27:239s, 2009 (abstr 5020) 30. Rini BI, Halabi S, Rosenberg JE, et al: Bevacizumab plus interferon-alpha versus interferon-alpha monotherapy in patients with metastatic renal cell carcinoma: Results of overall survival for CALGB 90206. J Clin Oncol 27:289s, 2009 (abstr 5019) Copyright 2013 by American Society of Clinical Oncology jop.ascopubs.org 7

Harrison et al Appendix Table A1. Mean and SD or No. and Percentage of Patient Characteristics by Exposure Sequence Characteristic Overall Cohort (N 384) TKI (n 109) TKI/TKI (n 31) TKI/TKI/mTOR (n 11) TKI/mTOR (n 34) TKI/mTOR/TKI (n 18) mtor (n 33) mtor/tki (n 11) Other (n 39) NSTx (n 98) Patient Age, years 2 Mean 64.0 64.8 64.2 63.1 65.7 58.7 65.8 64.7 58.3 65.1 SD 10.7 11.6 9.4 10.5 9.9 9.3 10.9 8.8 10.6 10.1 Male sex No. 252 63 21 6 29 9 24 9 30 61 % 65.8 57.8 67.7 54.5 87.9 50.0 72.7 81.8 76.9 62.2 Ethnicity White No. 276 81 20 9 26 9 20 6 33 72 % 71.9 74.3 64.5 81.8 76.5 50.0 60.6 54.5 84.6 73.5 Minority No. 108 28 11 2 8 9 13 5 6 26 % 28.1 25.7 35.5 18.2 23.5 50.0 39.4 45.5 15.4 26.5 Disease Metastatic sites No. 1.6 1.5 1.5 1.5 1.6 1.6 2.0 2.1 1.4 1.5 % 0.87 0.83 0.78 0.82 0.86 0.63 1.02 1.04 0.61 0.95 Clear cell No. 245 71 21 7 22 9 15 6 26 68 % 63.8 65.1 71.0 63.6 64.7 50.0 45.5 54.5 66.7 68.7 Nephrectomy No. 265 77 19 8 21 11 21 7 32 70 % 69.0 70.6 61.3 72.7 61.8 61.1 63.6 63.6 82.1 70.7 MSKCC risk Favorable No. 111 35 5 4 12 6 7 1 14 27 % 28.9 32.1 16.1 36.4 35.3 33.3 21.2 9.1 35.9 27.6 Intermediate No. 240 64 23 7 20 11 18 10 20 67 % 62.5 58.7 74.2 63.6 58.8 61.1 54.5 90.9 51.3 68.4 Poor No. 33 10 3 0 2 1 8 0 5 4 % 8.6 9.2 9.7 5.9 5.6 24.2 12.8 4.1 Abbreviations: MSKCC, Memorial Sloan-Kettering Cancer Center; mtor, mammalian target of rapamycin; NSTx, no systemic treatment; SD, standard deviation; TKI, tyrosine kinase inhibitor. Table A2. OS in Selected Published Prospective Clinical Trials of First-Line Treatments and a Large Retrospective Database of Anti-VEGF Therapy Naïve Patients Study Type Agent (exposure) Second-Line Exposures Final Median OS (months) v IFN- Motzer et al 28 RCT Sunitinib (TKI) NR 26.4 v 21.8 P.051 Escudier et al 29 (AVOREN) RCT Bevacizumab IFN- (other) Median OS for second-line 23.3 v 21.3 P.129 TKI 38.6 months Rini et al 30 (CALGB) RCT Bevacizumab IFN- (other) NR 18.3 v 17.4 P.07 Hudes et al 6 (Global ARCC) RCT Temsirolimus (mtor) NR 10.9 v 7.3 P.007 Heng et al 22 Observational (of anti-vegf therapy) Sunitinib, sorafenib, or bevacizumab (TKI or other) NR 22.0 Abbreviations: ARCC, Advanced Renal Cell Carcinoma; CALGB, Cancer and Leukemia Group B; IFN-, interferon alpha; mtor, mammalian target of rapamycin; NR, not reported; OS, overall survival; RCT, randomized clinical trial; TKI, tyrosine kinase inhibitor; VEGF, vascular endothelial growth factor. 8 JOURNAL OF ONCOLOGY PRACTICE Copyright 2013 by American Society of Clinical Oncology

Overall Survival in Patients With mrcc Community (n = 255) Academic (n = 200) Accrued to registry (n = 455) Missing treatment data (n = 2) (n = 453) mrcc diagnosis before 2007 (n = 68) (n = 385) Screen failure upon QA review (n = 1) Analytic cohort (n = 384) Figure A1. Derivation of the sample cohort. mrcc, metastatic renal cell carcinoma; QA, quality assurance. Overall Survival (proportion) 1.0 0.8 0.6 0.4 0.2 Log-rank P <.001 No Chemo TKI/TKI Other TKI/mTOR/TKI TKI/TKI/mTOR TKI/mTOR mtor/tki mtor TKI 0 10 20 30 40 50 Time (months) Figure A2. Kaplan-Meier survival curve by exposure sequence. See Table 1 for pairwise differences between exposure sequences. Chemo, chemotherapy; mtor, mammalian target of rapamycin; TKI, tyrosine kinase inhibitor. Copyright 2013 by American Society of Clinical Oncology jop.ascopubs.org 9

P Harrison et al 1.0 0.8 0.6 Adjusted P value 0.4 0.2 0.05 0 mtor v mtor/tki TKI/mTOR/TKI v mtor TKI/TKI/mTOR v/ mtor Raw P value TKI/TKI v mtor Other v mtor No Chemo v mtor TKI/mTOR v mtor TKI v mtor TKI/TKI/mTOR v mtor/tki Other v mtor/tki TKI/TKI v mtor/tki TKI/mTOR/TKI v mtor/tki Other v TKI/mTOR TKI/TKI v TKI/mTOR TKI/TKI/mTOR v TKI/mTOR TKI/mTOR v TKI/mTOR/TKI No Chemo v mtor/tki Other v TKI No Chemo v TKI/mTOR Comparison TKI v TKI/TKI Other v TKI/mTOR/TKI TKI v TKI/TKI/mTOR Other v TKI/TKI/mTOR No Chemo v Other TKI v TKI/mTOR/TKI TKI v mtor/tki TKI/TKI v TKI/mTOR/TKI No Chemo v TKI/TKI TKI/TKI v TKI/TKI/mTOR TKI v TKI/mTOR No Chemo v TKI Other v TKI/TKI No Chemo v TKI/TKI/mTOR TKI/TKI/mTOR v TKI/mTOR/TKI No Chemo v TKI/mTOR/TKI TKI/mTOR v mtor/tki Figure A3. Pairwise comparisons by exposure sequence. Chemo, chemotherapy; mtor, mammalian target of rapamycin; TKI, tyrosine kinase inhibitor. 10 JOURNAL OF ONCOLOGY PRACTICE Copyright 2013 by American Society of Clinical Oncology