The High-Dose Aldesleukin “Select” Trial: A Trial to Prospectively Validate Predictive Models of Response to Treatment in Patients with Metastatic Renal Cell Carcinoma

High-dose aldesleukin (HD IL2) received FDA approval in 1992 due to its ability to produce durable responses in a small percentage of patients with mRCC (1). Data presented to the FDA showed an objective response rate (ORR) of 14% in 255 patients with mRCC treated in seven phase II trials (1). Although the clinical efficacy has improved incrementally in subsequent trials largely due to clinical selection factors, investigators have attempted to develop predictive biomarkers of response that might further narrow its application to patients most likely to benefit (2–4).

Retrospective analyses have suggested that clinical characteristics and pathologic features could predict for response (or resistance) to IL2 (1, 3, 5–8). Leibovich and colleagues conducted a multivariate analysis of patients who received IL2 after nephrectomy. This study showed that survival was inversely associated with lymph node involvement, constitutional symptoms, sarcomatoid histology, metastases involving sites other than bone and lung, multiple metastatic sites, and a TSH level > 2.0 mIU/L, and led to the creation of the University of California Los Angeles (UCLA; Los Angeles, CA) SANI (survival after nephrectomy and immunotherapy) score (9).

Several studies have shown that responses to immunotherapy are most frequently seen in patients with clear cell (cc) RCC (10–12). In a retrospective analysis of pathology specimens obtained from 163 patients who had received IL2 therapy, the response rate to IL2 was 21% for patients with clear cell tumor histology compared with 6% for patients with non–clear cell tumor histology (12). Among the patients with ccRCC, histologic subclassification based on the presence of “good” predictive features (e.g., more than 50% alveolar and no granular or papillary features) and the absence of “poor” predictive features (e.g., more than 50% granular or any papillary features) was associated with response to IL2. Application of this model was reported to produce an impressive ORR (52%) in a small, prospective study (13).

Carbonic anhydrase 9 (CA-9) has been identified as an IHC marker that might predict the outcomes of patients with renal carcinoma. In an analysis by Bui and colleagues, CA-9 expression in more than 85% of tumor cells (high CA-9 expression by IHC) from ccRCC was associated with improved survival and a higher ORR in IL2-treated patients (14). Building on this work, Atkins and colleagues developed a two-component model that combined histologic subclassification with IHC staining for CA-9 (15). In a retrospective analysis, this “integrated selection” model (ISM) was able to identify a “good” predictive features group that contained 26 (96%) of 27 responders to IL2. On the basis of these findings, other investigators began to incorporate CA-9 analysis into their IL2-based clinical trials and treatment selection decisions (16).

In an attempt to improve the therapeutic index of IL2, the Cytokine Working Group (CWG) designed and conducted the HD IL2 “Select” Trial. The primary objective of this study was to evaluate prospectively whether the ISM could identify a group of patients with advanced RCC and “good” predictive features who were significantly more likely to respond to HD IL2-based therapy than a historical, unselected patient population (1). During the course of this trial, retrospective analyses identified potential predictors of increased [e.g., programmed death-ligand 1 (PD-L1) expression, CA-9 gene SNP] and decreased (e.g., elevated pretreatment levels of fibronectin and VEGF) response to immunotherapy in patients with RCC (17–19). Additional tissue was collected and secondary study objectives were amended to determine whether other clinical and pathologic features could help to further refine the optimal population for HD IL2-based therapy.

Patients with mRCC of any histologic type and no prior systemic therapy were enrolled. Major eligibility criteria included an Eastern Cooperative Oncology Group (ECOG) performance status of 0 or 1; bidimensionally measurable and clearly progressive disease; adequate organ function, with serum creatinine ≤1.5 mg/dL or calculated creatinine clearance >60 mL/minute; forced expiratory volume in 1 second >2.0 liter/second or 75% of predicted value; and no evidence of ischemia on a cardiac stress test. Patients who had received prior systemic treatment and those with brain metastases, seizure disorders, organ allografts, history of another malignancy, or concurrent corticosteroid therapy were ineligible. The protocol was approved by the human investigational review board at each participating site and voluntary written informed consent was obtained from each patient.

The study was conducted by the CWG. Patients received IL2, 600,000 IU/kg/dose (Prometheus Laboratories Inc.) i.v. every 8 hours for 5 days (maximum of 14 doses) beginning on day 1 and again on day 15. One course generally consisted of 5 days of treatment, 9 days of rest, 5 more days of treatment, and 9 weeks of rest, followed by up to two additional courses of HD IL2 for patients who benefited and tolerated most of the planned IL2 doses. A treatment delay of up to 4 weeks was allowed for resolution of side effects between courses. Patients were eligible to receive a maximum of three courses of treatment.

As HD IL2 is an FDA-approved treatment regimen, only serious adverse events (SAE) according to CTCAE version 3.0 were reported during the conduct of the trial. Response and progression were assessed according to standard World Health Organization (WHO) criteria and were initially determined by investigator assessment of radiographs (20). Patients were evaluated for response during weeks 8 and 12 of each course. To be eligible for more than one course of treatment, patients must have had at least stable disease (SD) with evidence of minor tumor regression or objective response and had to meet baseline eligibility criteria for organ function. Progression-free survival (PFS) was calculated from the date of IL2 initiation to the date of disease progression, or death on treatment as per the evaluating physician, or censored at the last documented tumor assessment for patients whose disease had not progressed. All patients who achieved a complete response (CR), partial response (PR), and SD for more than 6 months had their CT scans audited by independent radiologists to confirm their response and response duration. Overall survival (OS) was calculated from the date the first dose of IL2 was administered to the date of death or censored at the last documented contact with the patient. Data were updated through October 31, 2013.

Laboratory investigations were done in conjunction with the Tissue Acquisition Pathology and Clinical Data (TAPCD) Core of the Dana-Farber/Harvard Cancer Center (Boston, MA) Kidney Cancer SPORE. Examination of tumor tissue obtained before exposure to IL2 was performed using a number of modalities to identify potential predictors of response to treatment. As a requirement of study enrollment, patients consented to allow access by the investigators to the original hematoxylin and eosin-stained slides used to confirm the diagnosis of RCC for use in correlative studies. Paraffin-fixed tissue, obtained from representative, satisfactory tissue blocks of primary or metastatic tumors from participating subjects, was reviewed by the TAPCD for histologic features and stained for CA-9 using the M75 antibody according to a previously published protocol (15). Patients were then classified into the “good” or “poor” pathologic predictive factor group as proposed by the ISM (15). Additional, exploratory laboratory investigations were performed in collaboration with coinvestigators according to prepublished protocols [e.g., VEGF and fibronectin levels, CA-9 SNP and PD-L1 IHC (≥5% tumor membrane staining was considered “positive”)] based on preliminary data suggesting that they might predict for response to immunotherapy in RCC (17–19, 21).

The primary objective was to prospectively determine whether the response rate to high-dose IL2 for patients with mRCC and “good” pathologic predictive features by the ISM was significantly higher than a historical, unselected patient population. The primary endpoint, ORR was defined by WHO criteria as used in previous IL2 trials. The accrual goal was 110 patients, to enroll 66 patients in the ISM “good” pathologic predictive factor group, assuming 60% of patients would be classified as having “good” factors and a response rate between 30% and 40% (12, 15). The sample size provided >80% power for a one-sample exact test with two-sided α = 0.05 to detect response proportion of 0.30 versus 0.14 and assures confidence interval (CI) precision (half-width) less than ±0.14 (1). Secondary objectives included prospectively estimating the response rate to high-dose IL2 for patients with mRCC and “poor” pathologic predictive features by the ISM and comparing this response rate to the response rate of patients with “good” pathologic predictive features. Hypothesis-generating, univariate analyses were planned to explore whether other clinical [e.g., Memorial Sloan Kettering Cancer Center (MSKCC) and UCLA SANI score)] plasma (e.g., VEGF and fibronectin levels) and tumor [(e.g., IHC markers (PD-L1, B7-H3) or CA-9 gene SNP] features might be associated with HD IL2 responsiveness to further refine the optimal population for HD IL2-based therapy (9, 11, 22). Comparisons on ORR and durable remission rate (PFS > 3 years, all evaluable patients either had progression within 3 years or were followed for more than 3 years) were assessed by the Fisher exact test. Distributions of PFS and OS were summarized with the Kaplan–Meier estimates.

Between November 2006 and July 2009, 123 patients were enrolled at 13 participating institutions. All patients met the eligibility criteria. Three patients withdrew consent before receiving therapy and were not included in the analysis. Ninety-nine percent of patients had undergone prior nephrectomy, 70% were intermediate risk by MSKCC prognostic criteria, and 85% intermediate risk by UCLA SANI score (Table 1). Tumor (98%) and blood (94%) samples were collected on most patients. During the first course of therapy, patients received an average of 17 doses of the planned 28 doses (61%) of HD IL2. Doses were held due to treatment-related AEs that were typical of the prior published experience with this regimen (1, 3). Common SAEs included were hypotension, renal failure, and hyperbilirubinemia. Treatment-related AEs were reversible, except in two cases of treatment-related death. A 47-year-old male patient died during course 1, week 1 as a result of complications from a myocardial infraction. A 70-year-old male patient died following course 1, week 1 as a result of a cardiac arrhythmia.

The ORR of all 120 patients to HD IL2 confirmed by independent review was 25% (95% CI, 17.5–33.7%; Table 2). This response was substantially greater than a historical response rate of 14% (P = 0.0014; ref. 1). Three patients had CRs (2.5%) and 27 had PRs (22.5%). Nine patients (7.5%) achieved SD that lasted more than 6 months. The overall response rate to HD IL2 confirmed by investigator assessment was 28.3% (95% CI, 20.5–37.3%; Table 2). Nine patients were classified as having CRs (7.5%) and 25 had PRs (20.8%). By investigator assessment, some degree of tumor regression was seen in 49 patients (42%; Fig. 1). The median response duration for the cohort was 20.6 months (95% CI, 6.9–42.7; Fig. 2A). The median PFS was 4.2 months (95% CI, 2.5–4.7; Fig. 2B) and 13 patients (11%) achieved a durable remission by remaining progression free for at least 3 years. Seven patients developed disease progression following IL2 and were treated with local therapy (e.g., surgery or radiation) but have yet to require systemic therapy. Eighty patients received at least one cycle of VEGF-targeted therapy following IL2, which contributed to a median OS from time of IL2 initiation of 42.8 months (95% CI, 35.6–51.9; Fig. 2B).

Although MSKCC risk group was not associated with objective response or durable remission (PFS >3 years), patients with a high UCLA SANI score failed to respond to IL2 (0/8 objective responses; Table 3) and had a significantly lower PFS [median 1.4 months (CI, 0.2–2.5), P < 0.001 by log-rank test]. There were no responses seen in patients with non–clear cell histology (0/5 objective responses). Objective and durable responses to IL2 were not associated with any pathologic classification (e.g., “good” risk clear-cell histology group, high CA-9 staining or “ISM” “good” risk group; Table 3).

ORR was not associated with CA-9 SNP status or plasma VEGF or fibronectin levels (data not shown). Response was positively associated with tumor expression of PD-L1 (P = 0.01) and B7H-3 (P = 0.08; Table 3) by IHC staining (23). Durable remission (PFS > 3 years) was positively associated with tumor expression of PD-L1 (P < 0.01) but not B7-H3 (P = 0.73) by IHC staining.

In this prospective biomarker study, the clinical results revealed a response rate (25%) for the entire cohort that was substantially higher than the initial experience with high-dose IL2 in patients with mRCC (1). Toxicity was typical of our prior published experience with this regimen (1, 3, 4, 24). There were two treatment-related deaths. Previous research by our group and others that identified potential clinical and pathologic predictors of response to immunotherapy likely contributed to enhanced pretreatment screening (e.g., fewer patients enrolled with non–clear cell tumor histology, high UCLA SANI score, or without prior nephrectomy than on previous IL2 studies) and an improvement in antitumor activity for the entire cohort (3, 10, 12, 15, 22, 25–27). For example, a smaller percentage of patients received HD IL2 with their primary tumor in place on the current study than on an earlier CWG phase III randomized trial of HD IL2 (<1% vs. 30%; ref. 3). In addition, the availability of other treatment options (e.g., VEGF and mTOR pathway inhibitors) for patients with mRCC likely altered the referral pattern to academic medical centers toward those patients perceived, on clinical grounds, to be most likely to benefit from IL2. As has been reported in other studies, patients experienced durable remissions of their disease (3-year PFS rate 11%). The encouraging OS demonstrated in this cohort has been reported by other investigators (3, 28–30). It is likely the result improved patient selection (as discussed above) and of the sequential application of IL2-based immunotherapy followed, in most cases, by molecularly targeted therapy.

Although the application of previously identified clinical and pathologic selection criteria improved outcomes for the entire cohort when compared with the original experience with HD IL2, analysis of the proposed “ISM” through central histology review and IHC staining for CA-9 was unable to further improve the selection criteria. The clinical outcomes seen in this study were consistent with the more recent experience with HD IL2 (2, 3). There are several potential explanations for this result. They include: host factors (e.g., patient immune response) and the complex biology of IL2, which not only stimulates and expands CD8 and natural killer cells, but also stimulates regulatory T cells and activation-induced T-cell death which may play a larger role in determining response to IL2 than had previously been thought (31, 32); tumor factors are important but within a patient cohort enriched for ccRCC histology and low/intermediate UCLA SANI score, markers other than CA-9 are more predictive, or analyzed samples were not representative given the lack of standards for tumor processing at community centers and the existence of significant tumor heterogeneity (33, 34). Given that this latter issue likely impacts many kidney cancer translational research projects, efforts to standardize tissue collection should be considered in future trials.

The CR rate in this cohort was lower than that has been reported in several prior HD IL2 trials in RCC (2, 3). The decline in the CR rate could be secondary to the application of independent radiology review and the enhanced resolution provided by modern CT scans. In this study, the CR rate by investigator assessment (7.5%) was greater than that reported by independent review (2.5%). Despite this discrepancy, the durable remission rate (PFS > 3 years) of 11% was consistent with the rate reported in earlier trials (3). Responses occurred in patients in all MSKCC risk classifications, but not in the limited number of patients enrolled with non–clear cell histology or high UCLA SANI score. On the basis of the results of this study, although MSKCC risk classification does not identify responders, other clinical and pathologic features (e.g., UCLA SANI high score and non–clear cell histology) may identify patients who are unlikely to respond to HD IL2 and probably should not receive it.

Although our data did not validate the hypothesis that the ISM could predict response to HD IL2, this trial provides further evidence for the need to confirm the value of proposed biomarkers in well-designed, prospective studies. Given that the activity of HD IL2 was similar in both the high and low CA-9 IHC staining risk groups, this assay should not be routinely performed as a tool to identify potential patients. Attempts to develop other predictive biomarkers [e.g., PD-L1 expression, KIR (killer-cell immunoglobulin-like receptor)/KIR ligand mismatch] are ongoing to understand tumor and host factors that might predict for remissions following IL2 therapy (35). An improved model for IL2 patient selection may emerge from these efforts and improve its therapeutic index. The potential correlation between tumor expression of immune inhibitory molecules (e.g., PD-L1) and response to IL2 seen in this trial has been reported in clinical trials of PD-1/PD-L1 monoclonal antibodies (19, 36, 37). PD-L1 expression on tumor cells is thought to be induced by infiltrating CD8⁺ T cells in the tumor microenvironment (38, 39). Given their proposed mechanism of action, it is not surprising that PD-1/PD-L1 pathway blocking antibodies would display greater clinical activity in tumors that express PD-L1. The hypothesis that IL2 administration may be more effective in “inflamed” tumors that are infiltrated by CD8⁺ T cells and express immune inhibitory molecules (e.g., PD-L1 or B7-H3) needs to be confirmed in prospective trials (40).

Two decades of retrospective, correlative research designed to narrow the application of cytokine-based immunotherapy have culminated in the improved clinical outcomes seen in this cohort. Although HD IL2 remains a reasonable treatment option for appropriately selected patients with mRCC, this prospective trial was unable to validate proposed predictive markers of response and further improve the selection criteria for HD IL2. However, lessons from this work may guide the development and validation of predictive biomarkers for novel immunotherapies (e.g., CTLA-4, PD-1/PD-L1 antibodies) in solid tumors (41). As the list of effective therapies for metastatic kidney cancer grows, improvements in patient selection will be necessary to improve OS and the remission rate for patients with this disease.

D.F. McDermott is a consultant/advisory board member for Bristol-Myers Squibb, Genentech, Merck, Pfizer, and Prometheus Laboratories. J.P. Dutcher reports receiving speakers bureau honoraria from Novartis, Pfizer, and Prometheus and is a consultant/advisory board member for Bristol-Myers Squibb, Merck, and Prometheus. T.F. Logan reports receiving speakers' bureau honoraria from Bristol-Myers Squibb, GlaxoSmithKline, Novartis, Pfizer, Prometheus, and Wyeth and is a consultant/advisory board member for Argos, Aveo, Bristol-Myers Squibb, Celgene, Genentech, GlaxoSmithKline, Novartis, Pfizer, Prometheus and Wyeth. B.D. Curti reports receiving a commercial research grant from Prometheus, other commercial research support from Gallectin Therapeutics, and speakers bureau honoraria from Prometheus. N.I. Khushalani reports receiving speakers bureau honoraria from Prometheus and is a consultant/advisory board member for Amgen, Bristol-Myers Squibb, Genentech, and Provectus. U.N. Vaishampayan reports receiving speakers' bureau honoraria from Prometheus. E.D. Kwon has ownership interest (including patents) in Bristol-Myers Squibb and MedImmune. M.B. Atkins is a consultant/advisory board member for Novartis and Prometheus. No potential conflicts of interest were disclosed by the other authors.

The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.

Conception and design: D.F. McDermott, S. Signoretti, K.A. Margolin, J.A. Sosman, J.P. Dutcher, B.D. Curti, M.S. Ernstoff, E.D. Kwon, R.A. Figlin, A.J. Pantuck, M.M. Regan, M.B. Atkins

Development of methodology: D.F. McDermott, B.C. Leibovich, E.D. Kwon, R.A. Figlin, M.B. Atkins

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): D.F. McDermott, K.A. Margolin, J.I. Clark, J.A. Sosman, J.P. Dutcher, T.F. Logan, B.D. Curti, M.S. Ernstoff, L. Appleman, M.K.K. Wong, N.I. Khushalani, L. Oleksowicz, U.N. Vaishampayan, J.W. Mier, D.J. Panka, R.S. Bhatt, A.S. Bailey, B.C. Leibovich, E.D. Kwon, F.F. Kabbinavar, R.A. Figlin, A.J. Pantuck, M.B. Atkins

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): D.F. McDermott, S.-C. Cheng, S. Signoretti, K.A. Margolin, M.S. Ernstoff, U.N. Vaishampayan, A.S. Bailey, E.D. Kwon, A.S. Belldegrun, R.A. Figlin, A.J. Pantuck, M.M. Regan, M.B. Atkins

Writing, review, and/or revision of the manuscript: D.F. McDermott, S.-C. Cheng, S. Signoretti, K.A. Margolin, J.I. Clark, J.A. Sosman, J.P. Dutcher, T.F. Logan, B.D. Curti, M.S. Ernstoff, L. Appleman, M.K.K. Wong, N.I. Khushalani, L. Oleksowicz, U.N. Vaishampayan, A.S. Bailey, B.C. Leibovich, E.D. Kwon, F.F. Kabbinavar, A.S. Belldegrun, R.A. Figlin, A.J. Pantuck, M.M. Regan, M.B. Atkins

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M.K.K. Wong, A.S. Bailey, B.C. Leibovich, E.D. Kwon, A.S. Belldegrun

Study supervision: D.F. McDermott, M.K.K. Wong, L. Oleksowicz, U.N. Vaishampayan, F.F. Kabbinavar, R.A. Figlin

The authors thank the patients who participated in this study, clinical faculty, and personnel, including Meghan French, Gretchen McGrath, RN, Rachel Simonson, and Alexander Carlson.

Research reported in this article was supported by Prometheus Laboratories Inc., Novartis International AG, the National Cancer Institute of the NIH, and SPORE in Kidney Cancer under award number P50CA101942.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.