https://orcid.org/0000-0002-1662-3523
Abstract
We aim to provide survival scenario estimates for patients with advanced melanoma starting targeted therapies and immunotherapies.
We sought randomized trials of targeted therapies and immunotherapies for advanced melanoma and recorded the following percentiles (represented survival scenario) from each overall survival (OS) curve: 90th (worst-case), 75th (lower-typical), 50th (median), 25th (upper-typical), and 10th (best-case). We tested whether these scenarios can be estimated for each OS curve by multiplying its median by 4 multiples: 0.25 (worst-case), 0.5 (lower-typical), 2 (upper-typical), and 3 (best-case).
We identified 15 trials with 8025 patients. For first-line combination targeted therapy treatment groups, the median (interquartile range, IQR) in months for each percentile was: 90th, 6.2 (6.0-6.5); 75th, 11.3 (11.3-11.4); and median, 24.4 (23.5-25.3). For the first-line combination immunotherapy treatment group, the percentiles in months were: 90th, 3.9 (2.8-4.5); 75th, 13.4 (10.1-15.4), median 73 (not applicable). In targeted therapy groups, simple multiples of the median OS were accurate for estimating the 90th percentile in 80%; 75th percentile in 40%; 25th percentile in 100%. In immunotherapy groups, these multiples were accurate at 0% for the 90th percentile, and 43% for the 75th percentile. The 90th percentile (worst-case scenario) was better estimated as 1/6× median OS, and the 75th percentile (lower-typical) as 1/3× median OS.
Simple multiples of the median OS are a useful framework to estimate scenarios for survival for patients receiving targeted therapies, not immunotherapy. Longer follow-up is required to estimate upper-typical and best-case scenarios.
Using simple multiples of the median OS to estimate worst-case and typical scenarios for survival remains a helpful framework to explain prognosis to patients with advanced melanoma treated with targeted therapies. It is possible that our simple multiples method may more accurately estimate the prognosis for patients being treated with immunotherapy for other tumor types. Therefore, there is an opportunity to test these multiples for trials of immunotherapy among other malignancies. For patients with advanced melanoma receiving immunotherapy, survival time may be best described as the percentage of patients alive at 1, 2, and 5 years. There is insufficient data on the longest-surviving trial participants. Extended follow-up is required to better understand long-term survival outcomes.
Introduction
Most people with advanced cancer want information about their prognosis.1,2 Accurate information on expected survival time helps patients and doctors make treatment decisions, plan for the future, and prepare for the end of life.3,4
Unfortunately, many patients have a limited understanding of their prognosis, often overestimating likely survival time.4,5 Inadequate communication between oncologists and patients contributes to this misunderstanding.6 Many oncologists find it difficult to estimate and explain survival time in a way that is accurate yet conveys hope.2
We have previously proposed that worst-case, typical, and best-case scenarios are a useful framework to communicate expected survival time to patients with advanced cancer, and are preferable to providing a single-point estimate of the median survival.5,7-10 We have shown that the percentiles of an overall survival (OS) curve can be used to define ranges representing these scenarios.11 The 90th percentile, which represents the time at which 90% of the study cohort are still alive and 10% have died, can be used to represent the upper limit of the range for a worst-case scenario (shortest 10% of survival times); the interval between the 75th and 25th percentiles (middle 50%) can represent a range for the typical scenario and; the 10th percentile, where 10% are alive and 90% have died, can represent the lower limit of a range for the best-case scenario (longest 10%; Figure 1). We have also shown that simple multiples (0.25, 0.5, 2, and 3) of an OS curve’s median can be used to accurately estimate these percentiles in clinical trials of patients with advanced cancers treated with chemotherapy5,12 and targeted therapy.9,10
Survival curve percentiles and their corresponding scenarios.
Historically, advanced melanoma had a poor prognosis. Recently, treatment has changed dramatically with the advent of targeted and immunotherapies. Combining targeted therapies dabrafenib and trametinib in first-line BRAF mutant advanced melanoma has shown 5-year survival of 34%.13 Similarly, single-agent first-line pembrolizumab has shown a 5-year survival of 38%.13 Estimating survival for patients with advanced melanoma has become challenging as these therapies improve survival compared to their chemotherapeutic predecessors. Rather than presenting a single-point estimate of the median OS or likelihood of surviving 5 years, providing ranges for typical, best-case, and worst-case scenarios can convey expected survival to patients in a more realistic and meaningful way.
We aimed to determine whether our method of estimating scenarios for survival can be applied to patients with advanced melanoma receiving targeted therapies and immunotherapies, providing clinicians with a framework to estimate and explain survival time to these patients.
Materials and methods
We searched Medline, EMBASE, and the Cochrane Central Register of Controlled Trials for phase II or III randomized controlled trials of immunotherapy or targeted therapy for patients with advanced (unresectable stage IIIC and stage IV) melanoma.
We included trials published from January 2001 to June 2023. Trials were only included if there were greater than 90 patients in each treatment arm and a Kaplan-Meier curve for OS. Immunotherapy or targeted therapy must have been given in at least one arm of the trial and not in combination with chemotherapy. Trial arms were only included if they were consistent with current standards of care. The inclusion criteria and search terms are summarized in Supplementary Tables S1 and S2. For each trial, we recorded the year of publication, number of treatment arms, line of treatment, and median age. For each treatment arm, we recorded the treatments used, median OS, median progression-free survival (PFS), ECOG performance status, serum lactate dehydrogenase (LDH), V600 mutation status, and stage (unresectable stage III, stage IV, and M0-M1c disease).
Two authors independently traced each OS curve using the UN-SCAN-IT graph digitizing software (Silk Scientific, Orem, UT).11 The median and following percentiles (represented scenarios) were extracted from each curve: 90th (worst-case), 75th (lower-typical), 25th (upper-typical), and 10th (best-case). Discrepancies between the 2 measurements were resolved by repeated measurement.
Based on our previous work, we hypothesized that simple multiples of the median of each OS curve would approximate its percentiles representing the ranges for 3 scenarios as follows: 0.25 times the median for the 90th percentile (upper limit of worst-case scenario), 0.5 times for the 75th (lower limit of typical scenario), 2 times for the 25th (upper limit of typical scenario), and 3 times for the 10th (lower limit of the best-case scenario). For consistency with our previous work, we classified each estimate as “accurate” if it was within 0.66 to 1.33 times the actual value from the OS curve.
We assessed associations between survival and baseline characteristics in each trial arm by calculating Pearson’s correlation coefficients between the median survival and the summary measures of the baseline characteristics from each trial. These measures included ECOG performance status, age (on a continuous scale), LDH, sites of disease, and V600 mutation.
We first tested the accuracy of the scenarios in targeted therapies and immunotherapies regardless of the line of treatment. We then stratified the trial arms into first-line (>85% of patients receiving initial treatment) versus second- and subsequent-line (≤85% receiving initial treatment) for advanced disease.14-30 Arms were further grouped into single-agent versus combination therapy: first-line single-agent immunotherapy18,31-35 (Group 1), first-line combination immunotherapy32,34-37 (Group 2), first-line combination targeted therapy13,38 (Group 3), second-line immunotherapy14-23 (Group 4), and second-line targeted therapy24-30 (Group 5). Data regarding treatment that is no longer standard of care, including first-line single-agent targeted therapy, and single-agent tremelimumab, were extracted but not included in this manuscript.
Results
Trial characteristics
Our search yielded 951 references from which 15 trials were included. The most common reasons for elimination were small sample size (≤90), interventions that were not immunotherapy or targeted therapy, and the absence of a Kaplan-Meier OS curve (Figure 2). The 15 trials included 8025 patients with 21 treatment arms (Table 1; Supplementary Tables S3 and S4). Median follow-up ranged from 9 to 58 months with a median of 18 months.
Characteristics of included studies.
| Reference | Phase | Total patient for OS | Number of arms | Intervention type (regimen) | First line (%) | Second line (%)+ | ECOG 0-1 (%) | Median agea (years) | Hazard ratio (95% CI)a | Median follow-up (months)a |
|---|---|---|---|---|---|---|---|---|---|---|
| Immunotherapy | ||||||||||
| Ascierto et al31 Checkmate 066 NCT01721772 | 3 | 418 | 2 | Nivolumab vs dacarbazine | 100 | 0 | 100 | 64 | 0.46 (0.36-0.59), P < .001. (favors nivolumab vs dacarbazine) | Min 38.4 |
| Hamid et al17 KEYNOTE-002 NCT01704287 | 2 | 540 | 3 | Pembro 2 mg/kg vs pembro 10 mg/kg vs chemotherapy | 0 | 100 | 99 | 60 (10 mg/kg) 62 (2 mg/kg) | 0.74 (0.57-0.96) P = .0106 (favors pembrolizumab 10 mg/kg vs chemotherapy) 0.86 (0.67-1.10) P = .1173 (favors pembrolizumab 2 mg/kg vs chemotherapy) | 17 |
| Hodi et al36 CheckMate 069 NCT01927419 | 2 | 142 | 2 | Nivolumab + ipilimumab vs ipilimumab | 0 | 100 | 100 | 64 | 0.74 (0.43-1.26) P = 0.26 (favors nivolumab + ipilimumab vs ipilimumab) | 24.5 |
| Hodi et al35 Checkmate 067 NCT04540705. | 3 | 945 | 3 | Nivolumab + ipilimumab vs nivolumab vs ipilimumab | 100 | 0 | 100 | 62 (ipilimumab) 60 (nivolumab) 61 (ipilimumab + nivolumab) | Not reported | Min 90 |
| Larkin et al14 Checkmate 037 NCT01721746 | 3 | 272 | 2 | Nivolumab vs. chemotherapy | 0 | 100 | 100 | 59 | 0.95 (0.73-1.24) (nivolumab vs chemotherapy) | 24 |
| Lebbe et al37 CheckMate 511 NCT02714218 | 3 and 4 | 360 | 2 | Nivolumab 1 mg/kg + ipilimumab 3 mg/kg vs nivolumab 3 mg/kg + ipilimumab 1 mg/kg | 100 | 0 | 100 | 58 (nivolumab 1 mg/kg + ipilimumab 3 mg/kg) 58 (nivolumab 3 mg/kg + ipilimumab 1 mg/kg) | 1.09 (0.73-1.62) (nivolumab 3 mg/kg + ipilimumab 1 mg/kg vs nivolumab 1 mg/kg + ipilimumab 3 mg/kg) | 18.6 (nivolumab 1 mg/kg + ipilimumab 3 mg/kg) 18.8 (nivolumab 3 mg/kg + ipilimumab 1 mg/kg) |
| Long et al16 ECHO-301/KEYNOTE-252 NCT02752074 | 3 | 706 | 2 | Epacadostat + pembrolizumab vs pembrolizumab | 88 | 12 | 100 | 64 (epacadostat + pembrolizumab) 63 (pembrolizumab) | 1.13(0.86-1.49) (epacadostat + pembrolizumab vs pembrolizumab + placebo) | 12.4 |
| Nathan et al15 Checkmate 172 NCT02156804 | 2 | 1008 | 1 | Nivolumab | 0 | 100 | 93 | 61 | Not reported | 16.8 |
| Robert et al18 KEYNOTE 006 NCT01866319 | 3 | 834 | 3 | Pembrolizumab vs ipilimumab | 58 | 42 | 100 | 62 (combined pembrolizumab) 62 (combined ipilimumab) | 0.73 (0.61-0.88), P = 00049 (favors pembrolizumab vs ipilimumab) 0.73 (0.57-0.92), P = .0036 (favors first-line pembrolizumab vs ipilimumab) | 57.7 |
| Tawbi et al34 RELATIVITY-047 NCT03470922 | 2 and 3 | 714 | 2 | Relatlimab + nivolumab vs nivolumab | 100 | 0 | 100 | 62 (nivolumab) 63 (nivolumab + relatlimab) | 0.81 (0.64-1.01) (nivolumab + relatlimab vs nivolumab) | Min 13.2 |
| Wolchok 2022 NCT01844505 Checkmate 067 | 3 | 945 | 3 | Nivolumab + ipilimumab then nivolumab vs nivolumab vs ipilimumab | 100 | 0 | 100 | 61% <65 39% 65+ | 0.52 (0.43-0.64) (nivolumab + ipilimumab) 0.63 (0.52-0.76) (nivolumab) NA (ipilimumab) | Min 77 |
| Targeted therapy | ||||||||||
| Algazi et al27 S1320 NCT02196181 | 2 | 206 | 2 | Continuous dabrafenib + trametinib vs intermittent dabrafenib + trametinib | 70 | 30 | 100b | 61 | 1.03 (80% Cl 0.78–1.33), P = .93 (continuous vs intermittent) | 24 |
| Ascierto et al25 COLUMBUS NCT01909453 | 3 | 577 | 3 | Combo (encorafenib + binimetinib) vs encorafenib or vemurafenib | c | c | 100% | 57 (combo) 54 (encorafenib) 56 (vemurafenib) | 0.61(0.48-0.79) (favors combo vs vemurafenib) | 48.80 |
| Ascierto et al38 COBRIM NCT01689519 | 3 | 495 | 2 | Cobimetinib + vemurafenib vs vemurafenib + placebo | 100 | 0 | 100% | 55 (vemurafenib + placebo) 56 (vemurafenib + cobimetinib) | 0.8 (0.64-0.99) (favors cobimetinib + vemurafenib vs placebo + vemurafenib) | 16.6 (vemurafenib + placebo) 21.2 (vemurafenib + cobimetinib) |
| Robert et al13 COMBI-d NCT01584648 COMBI-v NCT01597908 | 3 | 563 | 2 | Dabrafenib + trametinib | 100 | 0 | 73 | 55 | Not reported | 22 |
| Reference | Phase | Total patient for OS | Number of arms | Intervention type (regimen) | First line (%) | Second line (%)+ | ECOG 0-1 (%) | Median agea (years) | Hazard ratio (95% CI)a | Median follow-up (months)a |
|---|---|---|---|---|---|---|---|---|---|---|
| Immunotherapy | ||||||||||
| Ascierto et al31 Checkmate 066 NCT01721772 | 3 | 418 | 2 | Nivolumab vs dacarbazine | 100 | 0 | 100 | 64 | 0.46 (0.36-0.59), P < .001. (favors nivolumab vs dacarbazine) | Min 38.4 |
| Hamid et al17 KEYNOTE-002 NCT01704287 | 2 | 540 | 3 | Pembro 2 mg/kg vs pembro 10 mg/kg vs chemotherapy | 0 | 100 | 99 | 60 (10 mg/kg) 62 (2 mg/kg) | 0.74 (0.57-0.96) P = .0106 (favors pembrolizumab 10 mg/kg vs chemotherapy) 0.86 (0.67-1.10) P = .1173 (favors pembrolizumab 2 mg/kg vs chemotherapy) | 17 |
| Hodi et al36 CheckMate 069 NCT01927419 | 2 | 142 | 2 | Nivolumab + ipilimumab vs ipilimumab | 0 | 100 | 100 | 64 | 0.74 (0.43-1.26) P = 0.26 (favors nivolumab + ipilimumab vs ipilimumab) | 24.5 |
| Hodi et al35 Checkmate 067 NCT04540705. | 3 | 945 | 3 | Nivolumab + ipilimumab vs nivolumab vs ipilimumab | 100 | 0 | 100 | 62 (ipilimumab) 60 (nivolumab) 61 (ipilimumab + nivolumab) | Not reported | Min 90 |
| Larkin et al14 Checkmate 037 NCT01721746 | 3 | 272 | 2 | Nivolumab vs. chemotherapy | 0 | 100 | 100 | 59 | 0.95 (0.73-1.24) (nivolumab vs chemotherapy) | 24 |
| Lebbe et al37 CheckMate 511 NCT02714218 | 3 and 4 | 360 | 2 | Nivolumab 1 mg/kg + ipilimumab 3 mg/kg vs nivolumab 3 mg/kg + ipilimumab 1 mg/kg | 100 | 0 | 100 | 58 (nivolumab 1 mg/kg + ipilimumab 3 mg/kg) 58 (nivolumab 3 mg/kg + ipilimumab 1 mg/kg) | 1.09 (0.73-1.62) (nivolumab 3 mg/kg + ipilimumab 1 mg/kg vs nivolumab 1 mg/kg + ipilimumab 3 mg/kg) | 18.6 (nivolumab 1 mg/kg + ipilimumab 3 mg/kg) 18.8 (nivolumab 3 mg/kg + ipilimumab 1 mg/kg) |
| Long et al16 ECHO-301/KEYNOTE-252 NCT02752074 | 3 | 706 | 2 | Epacadostat + pembrolizumab vs pembrolizumab | 88 | 12 | 100 | 64 (epacadostat + pembrolizumab) 63 (pembrolizumab) | 1.13(0.86-1.49) (epacadostat + pembrolizumab vs pembrolizumab + placebo) | 12.4 |
| Nathan et al15 Checkmate 172 NCT02156804 | 2 | 1008 | 1 | Nivolumab | 0 | 100 | 93 | 61 | Not reported | 16.8 |
| Robert et al18 KEYNOTE 006 NCT01866319 | 3 | 834 | 3 | Pembrolizumab vs ipilimumab | 58 | 42 | 100 | 62 (combined pembrolizumab) 62 (combined ipilimumab) | 0.73 (0.61-0.88), P = 00049 (favors pembrolizumab vs ipilimumab) 0.73 (0.57-0.92), P = .0036 (favors first-line pembrolizumab vs ipilimumab) | 57.7 |
| Tawbi et al34 RELATIVITY-047 NCT03470922 | 2 and 3 | 714 | 2 | Relatlimab + nivolumab vs nivolumab | 100 | 0 | 100 | 62 (nivolumab) 63 (nivolumab + relatlimab) | 0.81 (0.64-1.01) (nivolumab + relatlimab vs nivolumab) | Min 13.2 |
| Wolchok 2022 NCT01844505 Checkmate 067 | 3 | 945 | 3 | Nivolumab + ipilimumab then nivolumab vs nivolumab vs ipilimumab | 100 | 0 | 100 | 61% <65 39% 65+ | 0.52 (0.43-0.64) (nivolumab + ipilimumab) 0.63 (0.52-0.76) (nivolumab) NA (ipilimumab) | Min 77 |
| Targeted therapy | ||||||||||
| Algazi et al27 S1320 NCT02196181 | 2 | 206 | 2 | Continuous dabrafenib + trametinib vs intermittent dabrafenib + trametinib | 70 | 30 | 100b | 61 | 1.03 (80% Cl 0.78–1.33), P = .93 (continuous vs intermittent) | 24 |
| Ascierto et al25 COLUMBUS NCT01909453 | 3 | 577 | 3 | Combo (encorafenib + binimetinib) vs encorafenib or vemurafenib | c | c | 100% | 57 (combo) 54 (encorafenib) 56 (vemurafenib) | 0.61(0.48-0.79) (favors combo vs vemurafenib) | 48.80 |
| Ascierto et al38 COBRIM NCT01689519 | 3 | 495 | 2 | Cobimetinib + vemurafenib vs vemurafenib + placebo | 100 | 0 | 100% | 55 (vemurafenib + placebo) 56 (vemurafenib + cobimetinib) | 0.8 (0.64-0.99) (favors cobimetinib + vemurafenib vs placebo + vemurafenib) | 16.6 (vemurafenib + placebo) 21.2 (vemurafenib + cobimetinib) |
| Robert et al13 COMBI-d NCT01584648 COMBI-v NCT01597908 | 3 | 563 | 2 | Dabrafenib + trametinib | 100 | 0 | 73 | 55 | Not reported | 22 |
aArms are labeled where more than one immunotherapy or targeted therapy arm is included. Where unlabeled, data refers to the sole immunotherapy or targeted therapy arm included in the study (eg, not chemotherapy).
bAll ECOG 0-2, further breakdown not specified by authors.
cAll patients were either untreated or progressed on first-line treatment.
Characteristics of included studies.
| Reference | Phase | Total patient for OS | Number of arms | Intervention type (regimen) | First line (%) | Second line (%)+ | ECOG 0-1 (%) | Median agea (years) | Hazard ratio (95% CI)a | Median follow-up (months)a |
|---|---|---|---|---|---|---|---|---|---|---|
| Immunotherapy | ||||||||||
| Ascierto et al31 Checkmate 066 NCT01721772 | 3 | 418 | 2 | Nivolumab vs dacarbazine | 100 | 0 | 100 | 64 | 0.46 (0.36-0.59), P < .001. (favors nivolumab vs dacarbazine) | Min 38.4 |
| Hamid et al17 KEYNOTE-002 NCT01704287 | 2 | 540 | 3 | Pembro 2 mg/kg vs pembro 10 mg/kg vs chemotherapy | 0 | 100 | 99 | 60 (10 mg/kg) 62 (2 mg/kg) | 0.74 (0.57-0.96) P = .0106 (favors pembrolizumab 10 mg/kg vs chemotherapy) 0.86 (0.67-1.10) P = .1173 (favors pembrolizumab 2 mg/kg vs chemotherapy) | 17 |
| Hodi et al36 CheckMate 069 NCT01927419 | 2 | 142 | 2 | Nivolumab + ipilimumab vs ipilimumab | 0 | 100 | 100 | 64 | 0.74 (0.43-1.26) P = 0.26 (favors nivolumab + ipilimumab vs ipilimumab) | 24.5 |
| Hodi et al35 Checkmate 067 NCT04540705. | 3 | 945 | 3 | Nivolumab + ipilimumab vs nivolumab vs ipilimumab | 100 | 0 | 100 | 62 (ipilimumab) 60 (nivolumab) 61 (ipilimumab + nivolumab) | Not reported | Min 90 |
| Larkin et al14 Checkmate 037 NCT01721746 | 3 | 272 | 2 | Nivolumab vs. chemotherapy | 0 | 100 | 100 | 59 | 0.95 (0.73-1.24) (nivolumab vs chemotherapy) | 24 |
| Lebbe et al37 CheckMate 511 NCT02714218 | 3 and 4 | 360 | 2 | Nivolumab 1 mg/kg + ipilimumab 3 mg/kg vs nivolumab 3 mg/kg + ipilimumab 1 mg/kg | 100 | 0 | 100 | 58 (nivolumab 1 mg/kg + ipilimumab 3 mg/kg) 58 (nivolumab 3 mg/kg + ipilimumab 1 mg/kg) | 1.09 (0.73-1.62) (nivolumab 3 mg/kg + ipilimumab 1 mg/kg vs nivolumab 1 mg/kg + ipilimumab 3 mg/kg) | 18.6 (nivolumab 1 mg/kg + ipilimumab 3 mg/kg) 18.8 (nivolumab 3 mg/kg + ipilimumab 1 mg/kg) |
| Long et al16 ECHO-301/KEYNOTE-252 NCT02752074 | 3 | 706 | 2 | Epacadostat + pembrolizumab vs pembrolizumab | 88 | 12 | 100 | 64 (epacadostat + pembrolizumab) 63 (pembrolizumab) | 1.13(0.86-1.49) (epacadostat + pembrolizumab vs pembrolizumab + placebo) | 12.4 |
| Nathan et al15 Checkmate 172 NCT02156804 | 2 | 1008 | 1 | Nivolumab | 0 | 100 | 93 | 61 | Not reported | 16.8 |
| Robert et al18 KEYNOTE 006 NCT01866319 | 3 | 834 | 3 | Pembrolizumab vs ipilimumab | 58 | 42 | 100 | 62 (combined pembrolizumab) 62 (combined ipilimumab) | 0.73 (0.61-0.88), P = 00049 (favors pembrolizumab vs ipilimumab) 0.73 (0.57-0.92), P = .0036 (favors first-line pembrolizumab vs ipilimumab) | 57.7 |
| Tawbi et al34 RELATIVITY-047 NCT03470922 | 2 and 3 | 714 | 2 | Relatlimab + nivolumab vs nivolumab | 100 | 0 | 100 | 62 (nivolumab) 63 (nivolumab + relatlimab) | 0.81 (0.64-1.01) (nivolumab + relatlimab vs nivolumab) | Min 13.2 |
| Wolchok 2022 NCT01844505 Checkmate 067 | 3 | 945 | 3 | Nivolumab + ipilimumab then nivolumab vs nivolumab vs ipilimumab | 100 | 0 | 100 | 61% <65 39% 65+ | 0.52 (0.43-0.64) (nivolumab + ipilimumab) 0.63 (0.52-0.76) (nivolumab) NA (ipilimumab) | Min 77 |
| Targeted therapy | ||||||||||
| Algazi et al27 S1320 NCT02196181 | 2 | 206 | 2 | Continuous dabrafenib + trametinib vs intermittent dabrafenib + trametinib | 70 | 30 | 100b | 61 | 1.03 (80% Cl 0.78–1.33), P = .93 (continuous vs intermittent) | 24 |
| Ascierto et al25 COLUMBUS NCT01909453 | 3 | 577 | 3 | Combo (encorafenib + binimetinib) vs encorafenib or vemurafenib | c | c | 100% | 57 (combo) 54 (encorafenib) 56 (vemurafenib) | 0.61(0.48-0.79) (favors combo vs vemurafenib) | 48.80 |
| Ascierto et al38 COBRIM NCT01689519 | 3 | 495 | 2 | Cobimetinib + vemurafenib vs vemurafenib + placebo | 100 | 0 | 100% | 55 (vemurafenib + placebo) 56 (vemurafenib + cobimetinib) | 0.8 (0.64-0.99) (favors cobimetinib + vemurafenib vs placebo + vemurafenib) | 16.6 (vemurafenib + placebo) 21.2 (vemurafenib + cobimetinib) |
| Robert et al13 COMBI-d NCT01584648 COMBI-v NCT01597908 | 3 | 563 | 2 | Dabrafenib + trametinib | 100 | 0 | 73 | 55 | Not reported | 22 |
| Reference | Phase | Total patient for OS | Number of arms | Intervention type (regimen) | First line (%) | Second line (%)+ | ECOG 0-1 (%) | Median agea (years) | Hazard ratio (95% CI)a | Median follow-up (months)a |
|---|---|---|---|---|---|---|---|---|---|---|
| Immunotherapy | ||||||||||
| Ascierto et al31 Checkmate 066 NCT01721772 | 3 | 418 | 2 | Nivolumab vs dacarbazine | 100 | 0 | 100 | 64 | 0.46 (0.36-0.59), P < .001. (favors nivolumab vs dacarbazine) | Min 38.4 |
| Hamid et al17 KEYNOTE-002 NCT01704287 | 2 | 540 | 3 | Pembro 2 mg/kg vs pembro 10 mg/kg vs chemotherapy | 0 | 100 | 99 | 60 (10 mg/kg) 62 (2 mg/kg) | 0.74 (0.57-0.96) P = .0106 (favors pembrolizumab 10 mg/kg vs chemotherapy) 0.86 (0.67-1.10) P = .1173 (favors pembrolizumab 2 mg/kg vs chemotherapy) | 17 |
| Hodi et al36 CheckMate 069 NCT01927419 | 2 | 142 | 2 | Nivolumab + ipilimumab vs ipilimumab | 0 | 100 | 100 | 64 | 0.74 (0.43-1.26) P = 0.26 (favors nivolumab + ipilimumab vs ipilimumab) | 24.5 |
| Hodi et al35 Checkmate 067 NCT04540705. | 3 | 945 | 3 | Nivolumab + ipilimumab vs nivolumab vs ipilimumab | 100 | 0 | 100 | 62 (ipilimumab) 60 (nivolumab) 61 (ipilimumab + nivolumab) | Not reported | Min 90 |
| Larkin et al14 Checkmate 037 NCT01721746 | 3 | 272 | 2 | Nivolumab vs. chemotherapy | 0 | 100 | 100 | 59 | 0.95 (0.73-1.24) (nivolumab vs chemotherapy) | 24 |
| Lebbe et al37 CheckMate 511 NCT02714218 | 3 and 4 | 360 | 2 | Nivolumab 1 mg/kg + ipilimumab 3 mg/kg vs nivolumab 3 mg/kg + ipilimumab 1 mg/kg | 100 | 0 | 100 | 58 (nivolumab 1 mg/kg + ipilimumab 3 mg/kg) 58 (nivolumab 3 mg/kg + ipilimumab 1 mg/kg) | 1.09 (0.73-1.62) (nivolumab 3 mg/kg + ipilimumab 1 mg/kg vs nivolumab 1 mg/kg + ipilimumab 3 mg/kg) | 18.6 (nivolumab 1 mg/kg + ipilimumab 3 mg/kg) 18.8 (nivolumab 3 mg/kg + ipilimumab 1 mg/kg) |
| Long et al16 ECHO-301/KEYNOTE-252 NCT02752074 | 3 | 706 | 2 | Epacadostat + pembrolizumab vs pembrolizumab | 88 | 12 | 100 | 64 (epacadostat + pembrolizumab) 63 (pembrolizumab) | 1.13(0.86-1.49) (epacadostat + pembrolizumab vs pembrolizumab + placebo) | 12.4 |
| Nathan et al15 Checkmate 172 NCT02156804 | 2 | 1008 | 1 | Nivolumab | 0 | 100 | 93 | 61 | Not reported | 16.8 |
| Robert et al18 KEYNOTE 006 NCT01866319 | 3 | 834 | 3 | Pembrolizumab vs ipilimumab | 58 | 42 | 100 | 62 (combined pembrolizumab) 62 (combined ipilimumab) | 0.73 (0.61-0.88), P = 00049 (favors pembrolizumab vs ipilimumab) 0.73 (0.57-0.92), P = .0036 (favors first-line pembrolizumab vs ipilimumab) | 57.7 |
| Tawbi et al34 RELATIVITY-047 NCT03470922 | 2 and 3 | 714 | 2 | Relatlimab + nivolumab vs nivolumab | 100 | 0 | 100 | 62 (nivolumab) 63 (nivolumab + relatlimab) | 0.81 (0.64-1.01) (nivolumab + relatlimab vs nivolumab) | Min 13.2 |
| Wolchok 2022 NCT01844505 Checkmate 067 | 3 | 945 | 3 | Nivolumab + ipilimumab then nivolumab vs nivolumab vs ipilimumab | 100 | 0 | 100 | 61% <65 39% 65+ | 0.52 (0.43-0.64) (nivolumab + ipilimumab) 0.63 (0.52-0.76) (nivolumab) NA (ipilimumab) | Min 77 |
| Targeted therapy | ||||||||||
| Algazi et al27 S1320 NCT02196181 | 2 | 206 | 2 | Continuous dabrafenib + trametinib vs intermittent dabrafenib + trametinib | 70 | 30 | 100b | 61 | 1.03 (80% Cl 0.78–1.33), P = .93 (continuous vs intermittent) | 24 |
| Ascierto et al25 COLUMBUS NCT01909453 | 3 | 577 | 3 | Combo (encorafenib + binimetinib) vs encorafenib or vemurafenib | c | c | 100% | 57 (combo) 54 (encorafenib) 56 (vemurafenib) | 0.61(0.48-0.79) (favors combo vs vemurafenib) | 48.80 |
| Ascierto et al38 COBRIM NCT01689519 | 3 | 495 | 2 | Cobimetinib + vemurafenib vs vemurafenib + placebo | 100 | 0 | 100% | 55 (vemurafenib + placebo) 56 (vemurafenib + cobimetinib) | 0.8 (0.64-0.99) (favors cobimetinib + vemurafenib vs placebo + vemurafenib) | 16.6 (vemurafenib + placebo) 21.2 (vemurafenib + cobimetinib) |
| Robert et al13 COMBI-d NCT01584648 COMBI-v NCT01597908 | 3 | 563 | 2 | Dabrafenib + trametinib | 100 | 0 | 73 | 55 | Not reported | 22 |
aArms are labeled where more than one immunotherapy or targeted therapy arm is included. Where unlabeled, data refers to the sole immunotherapy or targeted therapy arm included in the study (eg, not chemotherapy).
bAll ECOG 0-2, further breakdown not specified by authors.
cAll patients were either untreated or progressed on first-line treatment.
Trial selection flowchart.
Eleven trials included immunotherapy (ipilimumab, nivolumab, pembrolizumab, and relatlimab) and 4 included targeted therapy (cobimetinib, dabrafenib, encorafenib, trametinib, and vemurafenib). Most patients had an ECOG performance status of 0 or 1.
Six of 21 (28%) treatment arms had not reached median OS at the time of reporting. Follow-up was sufficient for the 25th percentile in only one of the treatment arms. None of the treatment arms met the 10th percentile. All 5 targeted therapy arms met the median OS, which allowed for the calculation of the multiples. Across all 16 immunotherapy arms, only 10 curves met the median OS.
Survival outcomes
The mean values for the survival times of interest (as extracted from the survival curves) for each treatment group are shown in Table 2.
Summary of median overall survival, progression-free survival and estimated scenarios for survival by treatment group.
| Group 1 (1L mono-immunotherapy) | Group 2 (1L combo-immunotherapy) | Group 3 (1L combo-targeted therapy) | Group 4 (2L immunotherapy) | Group 5 (2L targeted therapy) | |
|---|---|---|---|---|---|
| n arms | 6 | 5 | 2 | 5 | 3 |
| n arms reaching median OS | 4 | 1 | 2 | 5 | 3 |
| Mean (IQR) in months | |||||
| Worst-case (90th percentile) | 4.4 (3.9-4.9) | 3.9 (2.8-4.5) | 6.2 (6-6.5) | 2.7 (2.2-2.7) | 5.9 (5-6.6) |
| Lower-typical (75th percentile) | 10.9 (10.5-11.4) | 13.4 (10.1-15.4) | 11.3 (11.3-11.4) | 7.2 (6.2-7.3) | 11.3 (10.9-11.6) |
| Median OS | 36.5 (36-38) | 73 (NA) | 24.4 (23.5-25.3) | 20.6 (14.6-24.7) | 30.7 (29.2-31.4) |
| Upper-typical (25th percentile) | NR (NR-NR) | NR (NR-NR) | NR (NR-NR) | 32.3 (32.3-32.3) | NR (NR-NR) |
| Best-case (10th percentile) | NR (NR-NR) | NR (NR-NR) | NR (NR-NR) | NR (NR-NR) | NR (NR-NR) |
| Median PFS | 4.8 (2.9-11.6) | 10.1 (9.7-10.5) | 11.9 (11.5-12.2) | 3.0 (2.8-8.4) | 7.1 (5.1-8.6) |
| Mean % (IQR) [number of treatment arms assessable] | |||||
| 1-year survival | 73 (73-74) [n = 6] | 77 (73-80) [n = 5] | 74 (74-74) [n = 2] | 61 (56-66) [n = 5] | 72 (70-73) [n = 3] |
| 2-year survival | 58 (58-58) [n = 4] | 64 (63-65) [n = 4] | 51 (50%-52%) [n = 2] | 44 (38-51) [n = 5] | 57 (56-57) [n = 3] |
| 5-year survival | 43 (43-44) [n = 2] | 51 (NA) [n = 1] | 33 (32-34) [n = 2] | 40 (40-40) [n = 1] | NR (NR-NR) [n = 0] |
| Group 1 (1L mono-immunotherapy) | Group 2 (1L combo-immunotherapy) | Group 3 (1L combo-targeted therapy) | Group 4 (2L immunotherapy) | Group 5 (2L targeted therapy) | |
|---|---|---|---|---|---|
| n arms | 6 | 5 | 2 | 5 | 3 |
| n arms reaching median OS | 4 | 1 | 2 | 5 | 3 |
| Mean (IQR) in months | |||||
| Worst-case (90th percentile) | 4.4 (3.9-4.9) | 3.9 (2.8-4.5) | 6.2 (6-6.5) | 2.7 (2.2-2.7) | 5.9 (5-6.6) |
| Lower-typical (75th percentile) | 10.9 (10.5-11.4) | 13.4 (10.1-15.4) | 11.3 (11.3-11.4) | 7.2 (6.2-7.3) | 11.3 (10.9-11.6) |
| Median OS | 36.5 (36-38) | 73 (NA) | 24.4 (23.5-25.3) | 20.6 (14.6-24.7) | 30.7 (29.2-31.4) |
| Upper-typical (25th percentile) | NR (NR-NR) | NR (NR-NR) | NR (NR-NR) | 32.3 (32.3-32.3) | NR (NR-NR) |
| Best-case (10th percentile) | NR (NR-NR) | NR (NR-NR) | NR (NR-NR) | NR (NR-NR) | NR (NR-NR) |
| Median PFS | 4.8 (2.9-11.6) | 10.1 (9.7-10.5) | 11.9 (11.5-12.2) | 3.0 (2.8-8.4) | 7.1 (5.1-8.6) |
| Mean % (IQR) [number of treatment arms assessable] | |||||
| 1-year survival | 73 (73-74) [n = 6] | 77 (73-80) [n = 5] | 74 (74-74) [n = 2] | 61 (56-66) [n = 5] | 72 (70-73) [n = 3] |
| 2-year survival | 58 (58-58) [n = 4] | 64 (63-65) [n = 4] | 51 (50%-52%) [n = 2] | 44 (38-51) [n = 5] | 57 (56-57) [n = 3] |
| 5-year survival | 43 (43-44) [n = 2] | 51 (NA) [n = 1] | 33 (32-34) [n = 2] | 40 (40-40) [n = 1] | NR (NR-NR) [n = 0] |
Summary of median overall survival, progression-free survival and estimated scenarios for survival by treatment group.
| Group 1 (1L mono-immunotherapy) | Group 2 (1L combo-immunotherapy) | Group 3 (1L combo-targeted therapy) | Group 4 (2L immunotherapy) | Group 5 (2L targeted therapy) | |
|---|---|---|---|---|---|
| n arms | 6 | 5 | 2 | 5 | 3 |
| n arms reaching median OS | 4 | 1 | 2 | 5 | 3 |
| Mean (IQR) in months | |||||
| Worst-case (90th percentile) | 4.4 (3.9-4.9) | 3.9 (2.8-4.5) | 6.2 (6-6.5) | 2.7 (2.2-2.7) | 5.9 (5-6.6) |
| Lower-typical (75th percentile) | 10.9 (10.5-11.4) | 13.4 (10.1-15.4) | 11.3 (11.3-11.4) | 7.2 (6.2-7.3) | 11.3 (10.9-11.6) |
| Median OS | 36.5 (36-38) | 73 (NA) | 24.4 (23.5-25.3) | 20.6 (14.6-24.7) | 30.7 (29.2-31.4) |
| Upper-typical (25th percentile) | NR (NR-NR) | NR (NR-NR) | NR (NR-NR) | 32.3 (32.3-32.3) | NR (NR-NR) |
| Best-case (10th percentile) | NR (NR-NR) | NR (NR-NR) | NR (NR-NR) | NR (NR-NR) | NR (NR-NR) |
| Median PFS | 4.8 (2.9-11.6) | 10.1 (9.7-10.5) | 11.9 (11.5-12.2) | 3.0 (2.8-8.4) | 7.1 (5.1-8.6) |
| Mean % (IQR) [number of treatment arms assessable] | |||||
| 1-year survival | 73 (73-74) [n = 6] | 77 (73-80) [n = 5] | 74 (74-74) [n = 2] | 61 (56-66) [n = 5] | 72 (70-73) [n = 3] |
| 2-year survival | 58 (58-58) [n = 4] | 64 (63-65) [n = 4] | 51 (50%-52%) [n = 2] | 44 (38-51) [n = 5] | 57 (56-57) [n = 3] |
| 5-year survival | 43 (43-44) [n = 2] | 51 (NA) [n = 1] | 33 (32-34) [n = 2] | 40 (40-40) [n = 1] | NR (NR-NR) [n = 0] |
| Group 1 (1L mono-immunotherapy) | Group 2 (1L combo-immunotherapy) | Group 3 (1L combo-targeted therapy) | Group 4 (2L immunotherapy) | Group 5 (2L targeted therapy) | |
|---|---|---|---|---|---|
| n arms | 6 | 5 | 2 | 5 | 3 |
| n arms reaching median OS | 4 | 1 | 2 | 5 | 3 |
| Mean (IQR) in months | |||||
| Worst-case (90th percentile) | 4.4 (3.9-4.9) | 3.9 (2.8-4.5) | 6.2 (6-6.5) | 2.7 (2.2-2.7) | 5.9 (5-6.6) |
| Lower-typical (75th percentile) | 10.9 (10.5-11.4) | 13.4 (10.1-15.4) | 11.3 (11.3-11.4) | 7.2 (6.2-7.3) | 11.3 (10.9-11.6) |
| Median OS | 36.5 (36-38) | 73 (NA) | 24.4 (23.5-25.3) | 20.6 (14.6-24.7) | 30.7 (29.2-31.4) |
| Upper-typical (25th percentile) | NR (NR-NR) | NR (NR-NR) | NR (NR-NR) | 32.3 (32.3-32.3) | NR (NR-NR) |
| Best-case (10th percentile) | NR (NR-NR) | NR (NR-NR) | NR (NR-NR) | NR (NR-NR) | NR (NR-NR) |
| Median PFS | 4.8 (2.9-11.6) | 10.1 (9.7-10.5) | 11.9 (11.5-12.2) | 3.0 (2.8-8.4) | 7.1 (5.1-8.6) |
| Mean % (IQR) [number of treatment arms assessable] | |||||
| 1-year survival | 73 (73-74) [n = 6] | 77 (73-80) [n = 5] | 74 (74-74) [n = 2] | 61 (56-66) [n = 5] | 72 (70-73) [n = 3] |
| 2-year survival | 58 (58-58) [n = 4] | 64 (63-65) [n = 4] | 51 (50%-52%) [n = 2] | 44 (38-51) [n = 5] | 57 (56-57) [n = 3] |
| 5-year survival | 43 (43-44) [n = 2] | 51 (NA) [n = 1] | 33 (32-34) [n = 2] | 40 (40-40) [n = 1] | NR (NR-NR) [n = 0] |
Scenarios for survival in targeted therapy groups
In the 2 treatment arms of first-line combination targeted therapy (Group 3), the mean (IQR) for the worst-case scenario was 6.2 months (6-6.5); lower-typical scenario was 11.3 months (11.3-11.4); and median OS was 24.4 months (23.5-25.3). Upper-typical and best-case scenarios were not reached in any treatment arms. Mean 1-year and 2-year survival rates were 74% (74-74%) and 51% (50-52%), respectively.
In the 3 treatment arms of second- and subsequent-line targeted therapy (Group 5), the mean (IQR) for the worst-case scenario was 5.9 months (5-6.6); lower-typical scenario was 8.9 months (6.5-10.9); median OS was 20.5 months (15.9-27.8); upper-typical was 33.2 months (19.4-46.6); and best-case scenario was 25.7 months (24.9-26.4). Mean 1-year and 2-year survival rates were 63% (IQR 59-71%) and 41% (32-56%), respectively.
Scenarios for survival in immunotherapy groups
In the 6 first-line single-agent immunotherapy treatment arms (Group 1),18,31-35 the mean (IQR) of the worst-case scenario was 4.4 months (3.9-4.9); lower-typical was 10.9 months (10.5-11.4); median OS was 36.5 months (36-38). The upper-typical (25th percentile) and best-case scenarios (10th percentile) were not reached in any curves. The mean (IQR) 1-year and 2-year survival rates were 73% (73%-74%) and 58% (43%-44%), respectively.
In the 5 first-line combination immunotherapy treatment arms (Group 2), the mean (IQR) in months for the worst-case scenario was 3.9 (2.8-4.5); lower-typical was 13.4 (10.1-15.4); median OS was 73 (NA). The upper-typical and best-case scenarios were not reached in any curves. The mean 1-year and 2-year survival rates were 77% (73-80%) and 64 (63-65%), respectively.
In the 5 second- and subsequent-line immunotherapy treatment arms (Group 4), the mean (IQR) in months for the worst-case scenario was 2.7 (2.2-2.7); lower-typical scenario 7.2 (6.2-7.3); the median OS was 20.6 (14.6-24.7); upper-typical was 32.3 (32.3-32.3). The best-case scenario was not reached. Mean 1-year and 2-year survival rates were 61% (IQR 56-66%) and 44% (38-51%), respectively.
Accuracy of survival scenarios
The proportion of curves where simple multiples were accurate for estimating scenarios is outlined in Table 3. Across both targeted therapy treatment groups (Groups 3 and 5), simple multiples of the median OS accurately estimated the worst-case scenario in 80% (4 of 5 treatment arms that met the median OS), and the lower-typical scenario in 40% (2 of 5 treatment arms). Accuracy for estimating the upper-typical scenario and best-case scenario could not be determined because the follow-up was too short. The simple multiples method overestimated the actual worst-case and lower-typical scenarios, meaning that the proportions of participants who died early in these trials were higher than we predicted (1 worst-case estimate and 3 lower-typical estimates were overestimated by our multiples).
Proportions of OS curves for each treatment group where simple multiples were accurate for estimating scenarios (n = number of OS treatment arms that met the percentile).
| Group 1 (1L mono-immunotherapy) | Group 2 (1L combo-immunotherapy) | Group 3 (1L combo-targeted therapy) | Group 4 (2L all immunotherapy) | Group 5 (2L all targeted therapy) | |
|---|---|---|---|---|---|
| 0.25× median est. 90th percentile (worst-case) | 0% (n = 4) | 0% (n = 1) | 100% (n = 2) | 0% (n = 5) | 66.7% (n = 3) |
| 0.5× median est. 75% percentile (lower-typical) | 0% (n = 4) | 0% (n = 1) | 100% (n = 2) | 60% (n = 5) | 0% (n = 3) |
| 2× median est. 25% percentile (upper-typical) | NR (n = 0) | NR (n = 0) | 100% (n = 2) | 100% (n = 1) | NR (n = 0) |
| 3× median est. 10th percentile (best-case) | NR (n = 0) | NR (n = 0) | NR (n = 0) | NR (n = 0) | NR (n = 0) |
| Group 1 (1L mono-immunotherapy) | Group 2 (1L combo-immunotherapy) | Group 3 (1L combo-targeted therapy) | Group 4 (2L all immunotherapy) | Group 5 (2L all targeted therapy) | |
|---|---|---|---|---|---|
| 0.25× median est. 90th percentile (worst-case) | 0% (n = 4) | 0% (n = 1) | 100% (n = 2) | 0% (n = 5) | 66.7% (n = 3) |
| 0.5× median est. 75% percentile (lower-typical) | 0% (n = 4) | 0% (n = 1) | 100% (n = 2) | 60% (n = 5) | 0% (n = 3) |
| 2× median est. 25% percentile (upper-typical) | NR (n = 0) | NR (n = 0) | 100% (n = 2) | 100% (n = 1) | NR (n = 0) |
| 3× median est. 10th percentile (best-case) | NR (n = 0) | NR (n = 0) | NR (n = 0) | NR (n = 0) | NR (n = 0) |
Proportions of OS curves for each treatment group where simple multiples were accurate for estimating scenarios (n = number of OS treatment arms that met the percentile).
| Group 1 (1L mono-immunotherapy) | Group 2 (1L combo-immunotherapy) | Group 3 (1L combo-targeted therapy) | Group 4 (2L all immunotherapy) | Group 5 (2L all targeted therapy) | |
|---|---|---|---|---|---|
| 0.25× median est. 90th percentile (worst-case) | 0% (n = 4) | 0% (n = 1) | 100% (n = 2) | 0% (n = 5) | 66.7% (n = 3) |
| 0.5× median est. 75% percentile (lower-typical) | 0% (n = 4) | 0% (n = 1) | 100% (n = 2) | 60% (n = 5) | 0% (n = 3) |
| 2× median est. 25% percentile (upper-typical) | NR (n = 0) | NR (n = 0) | 100% (n = 2) | 100% (n = 1) | NR (n = 0) |
| 3× median est. 10th percentile (best-case) | NR (n = 0) | NR (n = 0) | NR (n = 0) | NR (n = 0) | NR (n = 0) |
| Group 1 (1L mono-immunotherapy) | Group 2 (1L combo-immunotherapy) | Group 3 (1L combo-targeted therapy) | Group 4 (2L all immunotherapy) | Group 5 (2L all targeted therapy) | |
|---|---|---|---|---|---|
| 0.25× median est. 90th percentile (worst-case) | 0% (n = 4) | 0% (n = 1) | 100% (n = 2) | 0% (n = 5) | 66.7% (n = 3) |
| 0.5× median est. 75% percentile (lower-typical) | 0% (n = 4) | 0% (n = 1) | 100% (n = 2) | 60% (n = 5) | 0% (n = 3) |
| 2× median est. 25% percentile (upper-typical) | NR (n = 0) | NR (n = 0) | 100% (n = 2) | 100% (n = 1) | NR (n = 0) |
| 3× median est. 10th percentile (best-case) | NR (n = 0) | NR (n = 0) | NR (n = 0) | NR (n = 0) | NR (n = 0) |
Across all 3 immunotherapy treatment groups (Groups 1, 2, and 4), simple multiples accurately estimated the worst-case scenario in 0% (0 of 10 treatment arms that met the median OS), the lower-typical scenario in 30% (3 of 10 treatment arms), and the upper-typical scenario in 100% (1 treatment arm). Follow-up was too short to evaluate the best-case scenario. Our multiples overestimated the worst-case and lower-typical scenarios in 100% of estimates.
Among immunotherapy treatment groups, the worst-case scenario was more accurately estimated using 1/6× the median OS which was accurate in 60% of treatment groups. The lower-typical scenario was more accurately estimated using the multiple 1/3× the median OS, which was accurate in 90% of treatment arms.
Distribution of survival and characteristics correlated with survival
The distribution of survival times for each of the percentiles (scenarios) is shown in Figure 3. Each was skewed toward longer survival times. The range of the distribution broadened as the scenarios increased from worst-case to best-case scenario.
Distribution of scenarios for survival.
Supplementary Table S5 summarizes the correlation between the characteristics of a trial and its median OS, calculated with the Pearson correlation. Longer median OS was significantly associated with trials where there was a higher percentage of participants with grade 3+ adverse events. There were insufficient data to perform a reliable multivariable analysis.
Discussion
Targeted therapies and immunotherapies have greatly improved survival for patients with advanced melanoma, with the median OS for first-line combination immunotherapy being over 6 years.35 Simple multiples of the median OS accurately estimated the worst-case (90th percentile), and lower-typical (75th percentile) scenarios in the targeted therapy treatment groups but were inaccurate in the immunotherapy groups where they underestimated the frequency of early deaths. Few trials had follow-up beyond the median OS (only one treatment arm reached the 25th percentile and none reached the 10th percentile) and therefore we have minimal information on long-term outcomes for these patients.
The targeted therapy results are similar to the findings in our previous work in metastatic HER2-positive breast cancer treated with dual HER2-targeted therapies,9 where the simple multiples method was accurate for the short-term outcomes but follow-up was insufficient to quantify the best-case scenario.
The survival curves for targeted therapy in advanced melanoma follow a similar shape to the survival curves for chemotherapy5,10 and other targeted therapy trials.9 Therefore, the same simple multiples that accurately estimated survival in previous studies of chemotherapy and targeted therapy were similarly accurate here. Assuming that the same shape is followed beyond the median, we hypothesize that the best-case scenario for patients being treated with targeted therapy for advanced melanoma can be estimated as longer than 3 times the median OS.
For example, to estimate survival times for a patient starting combination targeted therapies, a median OS time of 24 months is a useful starting point. The 3 scenarios could then be calculated as a worst-case scenario of less than 6 months (<0.25 times 24 months), a typical scenario of 12-48 months (0.5-2 times 24 months), and a best-case scenario of more than 6 years (>3 times 24 months). These scenarios for survival can then be explained to a patient as: “If we imagine 100 people in the same situation, then we would expect the 5-10 who did worst would die within 6 months; the middle 50 would live for 1-4 years, and the 5-10 who did best would live longer than 6 years.” Patients should be informed that the best-case scenario remains uncertain due to minimal long-term follow-up of the clinical trials.
There is limited long-term survival information available for patients with advanced melanoma receiving first-line combination immunotherapy. To date, only one treatment arm (Checkmate 06735) of combination immunotherapy has reached the median OS and 4 of 6 treatment arms of single-agent immunotherapy have reached the median OS. With this limitation, we found that our simple multiples of the median did not accurately estimate the worst-case (90th percentile) or lower-typical scenarios (75th percentile). Across all 16 immunotherapy arms (Groups 1, 2, and 4), multiples of 1/4 and 1/2 the median, based on chemotherapy trials, tended to overestimate the worst-case and lower-typical scenarios, underestimating the number of early deaths for patients in these groups. These scenarios were more accurately estimated using 1/6× and 1/3× the median OS. For patients being treated with immunotherapy, the worst-case and lower-typical scenarios may be better estimated and explained as 1/6 and 1/3 of the median OS. Follow-up duration was insufficient to calculate the upper-typical and best-case scenarios in this group.
The shape of these OS curves from trials of immunotherapy for melanoma differed subtly from the shape of those in trials of chemotherapy and targeted therapies. Higher proportions of participants died earlier than 1/4 of the median and 1/2 the median in trials of immunotherapy for melanoma than in trials of chemotherapy or targeted therapies, meaning that the predicted worst-case and lower-typical scenarios should be shorter. Objective tumor responses and survival may be more durable with immunotherapy than with chemotherapy or targeted therapies, and often extended beyond the maximum reported follow-up in these trials. In addition, patients are living longer, and trial follow-up is not yet in-line with this increase in survival. It may be more accurate and helpful to estimate and explain survival time to these patients using percentages expected to be alive at 1, 2, and 5 years.
This paper provides a framework for clinicians to estimate and explain survival times in patients with advanced melanoma receiving targeted therapies. There has been a paucity of data on how to communicate survival to patients treated with targeted and immunotherapies, whereby previous estimates of survival have relied on trials of chemotherapy, which are now not relevant in the management of advanced melanoma. This study is the most comprehensive summary, to date, of prognostic information in patients with advanced melanoma receiving targeted and immunotherapies.
One of the main limitations of this study is that follow-up was insufficient in most trials to accurately obtain the long-term outcomes, especially the best-case scenario, as trials were published either before, at, or soon after, the median OS was reached. This limits what clinicians can tell their patients about the potential best-case scenario. The small number of trials included in this paper was quite heterogenous, particularly in the combination immunotherapy group. This limits our ability to make conclusions that encompass all trials within the group. Another limitation is that the survival times reported in this review are of a clinical trial population, who are generally fitter than a real-world cohort.
Given the method of using 0.25×, 0.5×, 2×, and 3× the median was not accurate for estimating scenarios for survival for patients receiving immunotherapy, further research in patients with melanoma and other advanced cancers receiving immunotherapy is needed. Different multiples may be more accurate, such as 1/6 and 1/3, or a different approach entirely. It is important to ask patients in this situation for their preferences regarding prognostic information, in particular quantitative information.
Supplementary material
Supplementary material is available at The Oncologist online.
Author contributions
Megan Smith-Uffen (Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Writing—original draft, Writing—review & editing), John Park (Conceptualization, Writing—review & editing), Andrew Parsonson (Data curation, Formal analysis, Writing—review & editing), Belinda.E. Kiely (Conceptualization, Supervision, Writing—review & editing), and Anuradha Vasista (Conceptualization, Methodology, Supervision, Writing—original draft, Writing—review & editing).
Funding
This work did not receive funding.
Conflicts of interest
B.E.K. reported honoraria from MSD Oncology, Eisai, Novartis, Gilead, and AstraZeneca; Scientific Advisory Board for Gilead and Novartis; and support for virtual meeting registration from Pfizer, MSD Oncology, and Novartis. The other authors indicated no financial relationships.
Data availability
No new data were generated or analyzed in support of this research.
