Sargramostim plus Ipilimumab vs Ipilimumab Alone for Treatment of Metastatic Melanoma: A Randomized Clinical Trial

Patients

Patients were eligible with histologic diagnosis of metastatic melanoma, measureable disease, no more than one prior therapy, no CNS metastases, ECOG performance status of 0 or 1, at least 18 years of age, greater than four weeks from prior therapy, adequate hematologic, hepatic, and renal function, no autoimmune disease, no infection with HIV, hepatitis B, or hepatitis C, and no prior CTLA-4 blockade or CD137 agonist. To assess for disparities in enrollment of minorities on NCI-sponsored clinical trials, it is required to collect data and to analyze differences in treatment effect by gender, race, and ethnicity. Given the nature of melanoma population, however, differences in treatment effect by race/ethnicity could not be fully assessed in this study.

Study Design and Treatment

The primary objective of the study was to evaluate the overall survival for the combination of sargramostim plus ipilimumab and ipilimumab alone in patients with advanced melanoma. Secondary objectives were to evaluate the progression-free survival, response rate, safety and tolerability of study treatments. Patients were stratified according to metastatic (M) stage and whether they had received a prior therapy or not. The final protocol is available in Appendix A. The planned accrual goal was 220, but due to rapid accrual a total of 245 patients were enrolled. Patients were equally randomized to receive ipilimumab 10 mg/kg every three weeks intravenously for four doses then every twelve weeks plus sargramostim (yeast-derived, rhu GM-CSF) 250 μg total dose subcutaneously on days 1-14 of 21 day cycles (group A; patients received no treatment days 15-21 of each cycle) or ipilimumab 10 mg/kg alone (group B). Stratified (AJCC stage and prior therapy) randomization based on permuted blocks within strata with dynamic institution balancing was used. Treatment assignments were obtained from the Central Randomization Desk at the ECOG Coordinating Center. No maximum number of cycles was planned. Patients could continue treatment without confirmed continued progression of disease or toxicity as outlined in the protocol. Due to known inflammatory effects of treatments which may appear as disease progression¹⁴, patients were permitted to continue with up to 100% increase in the sum of the product of the diameter of target lesions and the development of up to four new lesions in the absence of declining performance status and at the discretion of the treating physician.

Safety and tolerability were evaluated according to the NCI Common Terminology Criteria for Adverse Events (AEs), version 4.0. Protocol guidelines were utilized for management of immune-related AEs. Ipilimumab delay was permitted for any ≥grade 2 related AE except lab abnormalities, any ≥grade 3 skin related AEs (resolution to ≤grade 1 was required before continuing treatment), and any treatment-related ≥grade 3 laboratory abnormality. Patients who developed hypophysitis were able to continue treatment once receiving stable hormone replacement. Patients who experienced elevated liver function tests could continue once laboratory values returned to pretreatment grade. When AEs resolved to ≤grade 1 or returned to baseline within two weeks of ipilimumab administration, continued ipilimumab treatment was permitted at the next scheduled time point. If the AE did not resolve in this timeframe then that dose was omitted. GM-CSF was reduced by 50% for documented WBC greater than 60,000/mm³ and discontinued for any significant toxicity believed due to GM-CSF by the investigator.

This study was powered to compare an improvement in median survival from .87 to 1.23 years, including one interim analysis at 50% information time. This design provided an overall one-sided type I error rate of 0.10 and 80% power. Information time refers to a percentage of deaths observed at the time of an interim analysis in comparison to the total number of planned deaths in the study (i.e. 50% information time = 75 deaths/149 planned deaths). To preserve the overall type I error rate, critical values at the analyses were determined using the O'Brien and Fleming boundary¹⁵. The repeated confidence interval (RCI) of Jennison-Turnbull¹⁶ was constructed using the nominal type I error rate corresponding to the O'Brien and Fleming boundary. The RCI is an interval which provides a measure of accuracy for the estimated hazard ratio while adjusting for the fact that the hazard ratio was evaluated at the interim analysis during the course of study. Since the primary comparison is based on overall survival using the one-sided type I error rate of .10, one-sided 90% RCI of the hazard ratio for overall survival accurately describes the hazard ratio with an adjusted coverage probability for this design.

Assessments

Patients were required to complete eligibility assessments within 28 days of randomization. Computed tomography of the chest, abdomen, and pelvis and magnetic resonance imaging of the brain were performed prior to treatment and subsequently every 12 weeks following the initiation of treatment. If restaging scans indicated partial or complete response, a repeat confirmatory scan was performed four weeks later. Tumor responses were determined by the investigators using RECIST criteria and are audited as a part of ECOG-ACRIN standard procedures.

Endpoint Definition and Assessments

Overall survival was defined to be time from randomization to death from any cause, censoring cases who had not died at the date last known alive. Progression-free survival was defined to be time from randomization to disease progression or death (whichever occurred first), censoring cases without progression or death at the date of last disease assessment documenting the patient was free of progression. Patients without any post baseline tumor assessments were censored at random assignment. Overall survival and progression free survival were assessed per the standard ECOG-ACRIN follow up schedule, i.e., every three months if patient was less than two years from study entry, and every six months if patient was 2-5 years from study entry. Tumor assessments were made using RECIST v1.1 criteria. Complete and partial responses were considered as response. Response rate refers to a percentage of patients who achieved a response. The highest grade for adverse events recorded was assessed at the completion of each cycle and 30 days after the last dose of protocol therapy.

Peripheral Blood Analysis

Peripheral blood was collected before and during treatment. Of the patients with both pre- and post- samples, forty patients in each group (80 total), 20 patients with good survival outcome and 20 patients with poor survival outcome were selected by the study biostatistician. Due to limited resources only 80 patients could be included in this initial analysis. This method of sampling was used to balance patient groups representing a broad range of outcomes. Post hoc analyses were performed blinded to clinical outcomes or assigned treatment group. Inducible T cell costimulator (ICOS) T cell expression is necessary for optimal therapeutic effects of CTLA-4 blockade and is associated with improved survival with ipilimumab^17,18. Changes of ICOS expression in CD4⁺ and CD8⁺ T cells were assayed by multi-color flow cytometry.

Statistical Analysis

Efficacy analyses on overall survival, progression free survival, and response data were performed in 245 randomized patients. Two treatment groups were compared using the intent-to-treat (defining groups by assigned treatment), Distributions of overall survival and progression free survival were estimated using the Kaplan-Meier method¹⁹. One-year overall survival and 6 months progress free survival rates were assessed from Kaplan-Meier estimates. Treatment effect was assessed using the stratified log rank test for overall survival and progression free survival. The hazard ratio was estimated based on the Cox proportional hazard model while adjusting for the stratification factors²⁰. The stratification factors used in the randomization (AJCC stage and prior therapy) were used for the stratified log rank test and estimation of the hazard ratios. The distributions of categorical data (clinical response and toxicity) were compared using the Chi-square test or Fisher's exact test²¹ (for small samples).

There was one planned interim analysis on overall survival around 50% information time (75 deaths observed /149 planned deaths) using the O'Brien-Fleming boundary. Due to rapid accrual, information time on death accumulated slowly at the beginning but then rapidly reached 69.8% (104 deaths observed /149 planned deaths) in December 2012. The planned analysis was conducted using the O'Brien-Fleming boundary corresponding to 69.8% information time. The corresponding nominal significance level was .05. Using the corresponding significance level at this information time, repeated confidence interval (RCI)¹⁶ was estimated. Per study design, this represents a one-sided confidence interval for the hazard ratio of overall survival at this information time. Per study design, a one-sided p-value was presented for the primary endpoint (overall survival) comparison. All other p-values were based on two-sided tests (P < .05 considered significant). Analyses were conducted using SAS software version 9.2. (SAS Institute, Cary NC)

Patients and Treatment

A total of 245 patients were randomized between December 28, 2010 and July 28, 2011, 123 patients to the ipilimumab plus sargramostim group and 122 patients to the ipilimumab-only group. The baseline patient characteristics are shown in Table 1. Treatment groups were well balanced for demographics. The median time of follow up was 13.3 months (range: .03-19.9). Treatment summary by cycle and reasons for treatment termination are presented in eTable S1. Populations of patients screened, treated, and followed are represented in enrollment and outcomes diagram (Figure 1). Overall survival data as of December 2012 and other data as of March 2013 were used. Efficacy analyses were performed on the 245 ITT patients. Safety data analysis was performed on the 238 patients who received protocol therapy and submitted data as of March 2013.

Efficacy

Efficacy data are summarized in Table 2. The median overall survival was 17.5 months (95% CI; 14.9, not reached) for the ipilimumab plus sargramostim group and 12.7 months (95% CI; 10.0, not reached) for the ipilimumab-only group. The one-year overall survival was 68.9% (95% CI; 60.6%, 85.5%) for the ipilimumab plus sargramostim group and 52.9% (95% CI; 43.6%, 62.2%) for the ipilimumab-only group. The overall survival was significantly improved with the addition of sargramostim to ipilimumab with a stratified logrank one-sided p-value of .01. The Kaplan-Meier curves for overall survival are represented in Figure 2A. The stratified hazard ratio for overall survival on sargramostim plus ipilimumab compared to ipilimumab alone was .64. The one-sided 90% RCI for hazard ratio was (not applicable, .90). The O'Brien-Fleming boundary for overall survival was crossed at 69.8% (104 deaths observed / 149 planned deaths) information time. The progression free survival for the ipilimumab plus sargramostim group was 3.1 months (95% CI; 2.9,4.6) and for the ipilimumab-only group was 3.1 months (95% CI; 2.9,4.0). The differences between the treatment groups for progression free survival were not statistically significant (P=.37). The Kaplan-Meier curves for progression free survival are presented in Figure 2B. The response rate for the ipilimumab plus sargramostim group was 15.5% (95% CI; 9.6%, 23.1%) and for the ipilimumab-only group was 14.8% (95% CI; 9.0%, 22.3%), not statistically significant (P=.88). The subgroup analyses of overall survival and progression free survival hazard ratios are presented for line of therapy(the number of treatment regimens a patient has received) in eFigure S1 and eligible patients (those whose eligibility status was centrally reviewed and confirmed) in eFigure S2. Forest plots for treatment effects in the intent-to-treat patients are presented in Figure 3A for overall survival and Figure 3B for progression free survival. Eastern Cooperative Oncology Group performance status of zero (hazard ratio=.45, 95% CI .25,.80) and male (hazard ratio= .44, 95% CI .25, .76) subgroups had significant benefit for overall survival favoring ipilimumab plus sargramostim. Otherwise there were no significant differences for overall survival and progression-free survival in each category.

Adverse Events

Toxicity/treatment data was compared using all cases who received treatment and have data submitted as of March 11, 2013. Of the 123 patients randomized to the ipilimumab plus sargramostim group, 4 patients did not start therapy (1 due to early death and 3 due to refusal) and 1 patient has no data submitted. Of the 122 patients randomized to the ipilimumab-only group, 1 patient did not start therapy due to early death and 1 patient has no data submitted. Thus 118 versus 120 patients in ipilimumab plus sargaramostim and ipilimumab groups were included in toxicity/treatment analyses.

The treatment-related AEs reported in the safety evaluation are listed in Tables 3A/3B; based on the worst degree, grade 3-5 AEs occurred in 44.9% (95% CI; 35.8%, 54.4%) of patients treated with ipilimumab plus sargramostim and in 58.3% (95% CI; 49.0%, 67.2%) of patients treated with ipilimumab alone. Toxicity was significantly lower in the sargramostim containing group than in the ipilimumab-only group (P=.04). The grade 3-5 AEs with incidence rate of >3% at least in one group are summarized in Table 3A and all grade 3-5 adverse events by grade in eTable S2. Table 3B summarizes grade 3-5 AEs by toxicity categories. Most notable were differences in gastrointestinal toxicities, 16.1% (95% CI: 9.9%, 24.0%) vs. 26.7% (95% CI: 19.0%, 35.5%) (P=.05) and pulmonary toxicities, 0% (95% CI: 0%, 3.1%) vs. 7.5% (95% CI: 3.5%, 13.8%) (P=.03) in patients treated with ipilimumab plus sargramostim vs. ipilimumab alone, respectively. Treatment-related lethal AEs occurred in 2 patients (1 cardiac arrest, 1 colonic perforation) in the ipilimumab plus sargramostim group and 7 patients (2 colonic failure, 2 multi-organ failure, 2 respiratory failure, 1 hepatic failure) in the ipilimumab-only group (eTable S3). To investigate whether the improved survival benefit with sargramostim was due to less toxicity, cases were censored with lethal AEs that included treatment-relations of “possibly”, “probably” and “definitely”, and all lethal AEs regardless of attributions. OS improvement was maintained with sargramostim when cases with lethal AEs were censored (eTable S4). In 25 cases with sargromostim and 39 with ipilimumab alone who terminated treatment due to AEs, the sargromostim group had better OS than ipilimumab alone (One-sided P= .04).

ICOS Expression

Post hoc analyses for changes in CD4⁺ and CD8⁺ ICOS T cells and correlation to treatment group are summarized in eFigure S3. ICOS (CD278) increases as a function of treatment were greater in the sargramostim containing group. The median change in CD4+ cells was 2.55 vs.1.85 (P=.11) and the median change in CD8+ cells was 0.5 vs. 0.4 ( P=.01).

Conclusions

Among patients with unresectable stage III or stage IV melanoma, treatment with a combination of sargramostim and ipilimumab, compared with ipilimumab alone, resulted in longer overall survival and lower toxicity, but no difference in progression free survival. These findings require confirmation in larger sample sizes and with longer follow up.