Reply to T. Olivier et al

We appreciate the thoughtful comments from Drs Olivier and Prasad¹ regarding the primary manuscript² and editorial³ describing the CheckMate 067 6.5-year follow-up data. In general, we agree that there is no single standard of care for all patients with metastatic melanoma. The choice of primary therapy is one that requires careful balancing of antitumor activity with toxicities and is best decided through transparent discussions between patients and oncology clinicians. We also concur that such decisions are best informed by randomized trials and that only data from such trials would address the important point made regarding combination checkpoint blockade or monotherapy. However, the choice of a control arm for such a trial is challenging given the dynamic nature of available therapies for metastatic melanoma. As noted in the editorial, given the recently published RELATIVITY-047 results,⁴ which demonstrated superior progression-free survival with nivolumab plus relatlimab (a LAG-3–blocking antibody) versus nivolumab monotherapy, an ethical question would emerge when comparing nivolumab plus ipilimumab with nivolumab monotherapy.

Another point of agreement is that the novel end point of treatment-free interval should be examined in the context of the higher frequency of toxicity-driven treatment discontinuation in the combination arm of CheckMate 067. These data, along with detailed toxicity and clinical response metrics, were included in the primary manuscript (Data Supplement; Table A7)² and have been published and presented previously.^5,6 Nevertheless, descriptive analyses of the CheckMate 067 4-year data set suggested that health-related quality of life (HRQoL) was maintained during the treatment-free interval in patients treated with nivolumab plus ipilimumab, regardless of the reason for treatment discontinuation.⁷ In addition, an outcome measure that more holistically characterizes the quality of the time free of anticancer therapy is treatment-free survival, which quantifies this time both with and without toxicity. At a minimum of 5 years of follow-up in CheckMate 067, nivolumab plus ipilimumab was associated with sustained benefits in treatment-free survival, with the majority of that time being free from grade 2 to 4 treatment-related adverse events.⁸

The impact of cancer treatment on patients' overall HRQoL is also a topic of ongoing importance. In considering the high frequency of grade 3 or 4 adverse events with nivolumab plus ipilimumab, we note that symptom duration was variable, and most, but not all, patients discontinued immunomodulating therapies and were symptom-free within weeks of onset. To better quantify this, we are obtaining HRQoL data using the European Quality-of-Life 5-Dimensions questionnaire, which is being used every 6 months to assess survival quality as one of the protocol-defined secondary end points, as part of an ongoing follow-up of CheckMate 067; prior data were published in 2017.⁹ We also acknowledge the impact of postprogression therapy on overall survival, as presented in the primary manuscript.²

We agree that when considering appropriate statistical comparisons between the two experimental arms of CheckMate 067 (ie, nivolumab plus ipilimumab and nivolumab monotherapy), nonproportional hazards scenarios with delayed effects in the combination arm may have been relevant. Given that analyses of time-to-event outcomes are typically performed under the proportional hazards assumption (ie, that the hazard ratio of an outcome between two patient groups does not vary with time), we recognize that there may be some concerns with these analyses. A recent analysis comprehensively compared statistical tests, including weighted log-rank tests, Kaplan–Meier curve-based tests, and combination tests, for time-to-event end points under nonproportional hazards.¹⁰ In general, results of these simulations showed that all evaluated tests effectively controlled type I error compared with the unweighted log-rank test, although no single test was identified as most powerful across all scenarios.¹⁰ These findings suggest that some statistical tests may be robust in situations that violate the proportional hazards assumption¹⁰; however, we would like to reiterate that the CheckMate 067 analyses of nivolumab plus ipilimumab versus nivolumab monotherapy were descriptive in nature.

With regard to the limitation of presenting medians to demonstrate differences in overall survival and the use of hazard ratios to measure the magnitude of benefit, we agree that multiple efficacy measures are needed to accurately capture treatment effect in a scenario of nonproportional hazards (also discussed in Lin et al¹⁰).

We also note that if one were to design a comparative trial with a delayed treatment effect under nonproportional hazards, a sizable study with sufficiently long follow-up would be needed to compensate for the considerable losses in statistical power that could hinder observations regarding the true effects of treatment. Such considerations, as well as those discussed previously, should be applied to future research to further inform the clinical decision-making process.

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

Reply to T. Olivier et al

The following represents disclosure information provided by authors of this manuscript. All relationships are considered compensated unless otherwise noted. Relationships are self-held unless noted. I = Immediate Family Member, Inst = My Institution. Relationships may not relate to the subject matter of this manuscript. For more information about ASCO's conflict of interest policy, please refer to www.asco.org/rwc or ascopubs.org/jco/authors/author-center.

Open Payments is a public database containing information reported by companies about payments made to US-licensed physicians (Open Payments).