Pembrolizumab in the adjuvant treatment of melanoma: efficacy and safety

Abstract

Introduction:

Regional or distant metastases from melanoma may be surgically resected, but remain at high-risk of recurrence. Over the last few years, several treatments have been approved to mitigate this risk. These include anti-PD-1 agents, specifically pembrolizumab and nivolumab.

Areas covered:

Herein, we will discuss the landscape of pembrolizumab safety and efficacy used in the adjuvant setting for high-risk, resected melanoma. We place this in context with other available adjuvant therapies, and discuss subgroup analyses.

Expert opinion:

Anti-PD-1 therapy with either pembrolizumab or nivolumab has become a standard of care for patients with resected stage III or IV melanoma. In our practice, we generally offer these agents (which have comparable safety and efficacy profiles) to patients with resected stage IIIb – IV melanoma regardless of BRAF mutation status.

1. Introduction

Melanoma is the most deadly form of skin cancer with approximately 90,000 new cases diagnosed annually in the United States. Early stage disease without lymph node involvement (stage I-II) portends excellent outcome, with five-year survival rates of 90+% [1–3]. Prior to the age of checkpoint inhibitors, melanoma with lymph node involvement patients diagnosed with stage III disease had heterogeneous outcomes based on risk level (i.e. depth and ulceration of primary, number of lymph nodes involved), and could expect five-year relapse-free survival rate of 11–63% [4, 5].

1.1. Historical perspective

First approved in 1996, high-dose interferon-alpha was the standard-of-care option for adjuvant in stage IIB/III patients based on ECOG 1684. This study demonstrated improved RFS and OS with maximum-tolerated dose of INF-alpha-2b for one month followed by weekly subcutaneous dosing for 48 weeks [6]. Adjuvant interferon improved the five-year relapse-free survival (RFS) of stage III melanoma from 26 to 37%, with an overall survival benefit of 45% with interferon versus 37% with observation alone. However, other studies did not demonstrate an OS benefit with interferon [7, 8]. This treatment was associated with significant toxicity, particularly with fevers and myalgias, requiring dosing delays or reductions for half of patients on study. Along with the questionable OS benefit and severe toxicity profile, many clinicians were very hesitant to offer this therapy, and only younger patients with good performance status were candidates.

1.2. Pembrolizumab

Pembrolizumab (trade name Keytruda, previously MK-3475, lambrolizumab; Merck) is a humanized IgG4 monoclonal antibody which binds to the programmed cell death-1 (PD-1) receptor on T cells and structurally inhibits the binding of PD-L1 and PD-L2 to functionally deactivate a negative regulatory signal of T cells, thereby furthering their antitumor activity [9, 10]. It was initially approved for use in metastatic melanoma in September 2014. In February 2019, based on positive results from the KEYNOTE-054 trial, pembrolizumab gained FDA approval in the adjuvant setting for melanoma with lymph nodal involvement following resection [11, 12].

2. Efficacy of adjuvant pembrolizumab in melanoma

2.1. KEYNOTE-054

Published in May 2018, the KEYNOTE-054 study serves as the cornerstone clinical trial evaluating pembrolizumab in the adjuvant setting for high-risk melanoma [12]. This study enrolled 1019 patients with completely resected Stage III melanoma, assigning them to receive either placebo or pembrolizumab every 3 weeks for a total of 18 doses (~1 year) or until disease recurrence or unacceptable toxicity. Patients with resected stage IV melanoma were not included in this study. Patients were followed with serial CT or MRI imaging completed every 12 weeks for the first two years, then every 6 months thereafter. Overall recurrence-free survival (RFS) and RFS in the PD-L1 positive subgroup were followed as primary endpoints. Crossover from placebo to pembrolizumab was permitted in patients who experienced a relapse.

There is long-standing debate regarding the use of surrogate endpoints in clinical trials, particularly in the case of adjuvant studies in which overall survival is generally accepted as the natural gold standard for analysis, but often limited due to the time burden of such analyses. In the adjuvant setting, several disease-specific meta-analyses in colon, breast, pancreatic, lung cancer as well as melanoma have generally found that improvements seen in surrogate endpoints such as recurrence-free survival, disease-free survival and progression-free survival correlate well with overall survival benefits [13–17]. In other adjuvant studies of checkpoint inhibitors in melanoma, surrogate endpoints of recurrence-free survival, disease-free survival and progression-free survival have demonstrated relatively good fidelity to overall survival [18, 19]. For example, EORTC 18071 compared the use of adjuvant ipilimumab in completed resected stage III cutaneous melanoma; as a primary endpoint, this study found 5-year RFS to be improved to 41% with adjuvant ipilimumab compared to 30% with placebo, comparable to their secondary endpoint of 5-year OS 65% in the treatment group versus 54% in the control group [20].

KEYNOTE-054 was designed as an international double-blinded stage III randomized controlled trial in which patients were assigned 1:1 to receive pembrolizumab or placebo. Eligible patients were diagnosed with stage IIIA, IIIB or IIIC melanoma and had completed surgical excision and complete lymphadenectomy within 13 weeks of trial initiation. To avoid diluting the population with Stage IIIA patients with low relapse risk, stage IIIA patients were required to meet the threshold of the Rotterdam criteria of sentinel node tumor load with micrometastatic disease of at least 1mm in the positive node [21, 22]. Importantly, patients were staged based on the Seventh Edition of the American Joint Committee on Cancer staging guidelines, though at the time of publication, the Eight Edition had recently been released [23].

KEYNOTE-054 established a significant improvement in its established primary endpoint of relapse-free survival among resected stage III melanoma patients treated with pembrolizumab as compared to placebo. Analysis of the overall intention-to-treat population revealed a significant benefit to adjuvant pembrolizumab treatment over placebo, with 12-month RFS rate of 75.4% versus 61.0% (hazard ratio for recurrence or death (HR), 0.57). This effect was similar to that seen in the per-protocol analysis (HR 0.56). This effect was maintained at 18-month analysis, with RFS rate 71.4% versus 53.2% (Table 1). Prolonged analysis revealed a 3-year RFS rate 63.7% among pembrolizumab cohort compared to 44.1% among placebo cohort [24].

Table 1:

Key outcomes from KEYNOTE-054

Outcome	Pembrolizumab	Placebo	HR, p value
12-month RFS	75.4%	61%	HR 0.57, p<0.001
18-month RFS	71.4%	53.2%
18-month cumulative incidence of distanct metastasis as first type of recurrence	16.7%	29.7%	HR 0.53, p<0.001
Any TRAE	77.8%	66.1%
Grade 3–5 TRAE	14.7%	3.4%

RFS, relapse-free survival; TRAE, treatment-related adverse event

Interestingly, data from KEYNOTE-054 as originally published suggested that adjuvant pembrolizumab is more effective in preventing distant metastases than in preventing locoregional recurrence, with 27.3% of placebo-treated patients developing distant metastases at relapse compared with 15.2% of pembrolizumab-treated patients. In response to this observation Robert and Eggermont published a careful analysis of relapse patterns among KEYNOTE-054 subjects which revealed a similar hazard ratios of 0.69 for locoregional-only recurrence and 0.58 for distant metastases only [25]. This finding suggests that adjuvant systemic treatment with pembrolizumab is effective in preventing both loco-regional and distant metastatic relapsed disease, offer an advantage over local adjuvant therapies such as radiation.

Shortly after initial publication, the authors re-analyzed the KEYNOTE-054 data according to the AJCC-8 melanoma staging system [26]. The original randomization based on AJCC-7 staging distributed patients among four groups: IIIA, IIIB, IIIC with 1–3 positive lymph nodes and IIIIC with ≥4 positive lymph nodes. Upon reanalysis based on AJCC-8 staging, the population was shifted to a higher substage (AJCC-7 to −8, IIIA 153 (15.0%) versus 82 (8.0%), IIIB 472 (46.3%) versus 354 (34.7%), IIIC 394 (38.7%) versus 506 (49.7%), IIID 0 (0%) versus 38 (3.7%), not evaluable 0 (0%) versus 39 (3.8%)) due to the classification of patients with a large Breslow thickness and/or presence of several clinically involved lymph nodes in higher stages. This reanalysis concluded that the benefit of adjuvant pembrolizumab was maintained across AJCC-8 subgroups in resected high-risk stage III melanoma patients. As a primary endpoint of the KEYNOTE-054 study, 1-year relapse-free survival rate for the AJCC-8 subgroups is as follows: IIIA, 92.7% versus 92.5%, IIIB 79.0% versus 65.5%, IIIC 73.6% versus 53.9%, IIID 50.0% versus 33.3% (HR 0.75, 99% CI 0.11–5.43). Interestingly, with prolonged follow up (median of three years), the hazard ratios for RFS were maintained across subgroups.. This is of particular interest among stage IIIA patients, for whom the 3-year RFS 82.6% among the pembrolizumab cohort versus 67.4% in the placebo cohort (HR 0.43, 99% CI 0.13–1.43) [24].

The efficacy of pembrolizumab in the adjuvant setting is similar to that seen with the other PD-1 inhibitor with FDA approval in this setting. The CheckMate-238 trial of adjuvant nivolumab versus ipilimumab in high-risk melanoma patients undergoing complete resection demonstrated a 12-month RFS of 70.5% for nivolumab compared with 60.8% for ipilimumab (HR 0.65) [27]. This statistically significant benefit was sustained with four years of follow up (4-year RFS 51.7% with nivolumab compared to 41.2% with ipilimumab) [28]. While slightly lower than the 12-month RFS of 75.4% seen with pembrolizumab, this is likely due to inclusion of higher risk patients; stage IIIA (by AJCC 7) patients accounted for 15% of the total enrollment on KEYNOTE-054 but were excluded from CheckMate-238, which enrolled patients with IIIB, IIIC and resectable stage IV disease only; 18% of patients had resected stage IV disease. Further, this study compared nivolumab with an active treatment (ipilimumab) whereas Keynote-054 had a placebo comparator. Indirect comparisons suggest similar benefit for nivolumab vs. placebo as pembrolizumab vs. placebo [29].Similar efficacy benefits across subgroups were seen with nivolumab; thus both pembrolizumab and nivolumab are considered standard therapy for patients with either resected stage III or IV melanoma.

Another adjuvant study to consider is the 167 patient, randomized phase II IMMUNED study that compared combination ipilimumab plus nivolumab versus nivolumab monotherapy versus placebo as adjuvant therapy in completely resected stage IV melanoma patients with no evidence of disease. Here, a 12-month RFS rate of 52% was estimated, again likely related to highest baseline relapse risk seen with resected stage IV patients [30]. However, the combination of ipilimumab and nivolumab was associated with superior RFS compared with nivolumab or placebo (75% vs. 52% vs. 32%, logrank p <0.001) suggesting this could be a more active regimen in the highest risk patients. The larger randomized Checkmate-915 study will address this question in a larger population of resected stage III and IV patients.

2.2. Considerations for adjuvant treatment of BRAF mutant melanoma

Importantly, patients were eligible for KEYNOTE-054 with regardless of BRAF mutation status. The treatment arm enrolled 40.9% BRAF-mutant patients and 45.3% BRAF-wild type patients (with the remainder of unknown BRAF status), roughly similar to the control arm. There was no significant difference seen in RFS at one year with adjuvant pembrolizumab (HR 0.54 versus 0.61). At median follow-up of three years, this effect was maintained (HR 0.51 versus 0.66). As a comparison, the COMBI-AD trial is a Phase 3 study which compared adjuvant dabrafenib plus trametinib versus placebo in patients whose tumors harbor BRAF V600E/K mutations. Estimated one-year RFS among the dabrafenib + trametinib treated patients was 88% as compared to 56% in the placebo group; this effect was maintained at five-year follow up, with 52% of dabrafenib + trametinib treated patients alive without relapse compared to 36% of placebo-treated patients [31]. Consensus guidelines do not offer clear recommendations regarding selection of adjuvant treatment for patients with BRAF V600-mutant tumors, but suggest that toxicity profile may help drive the decision [32].

2.3. PD-L1 as a biomarker

The quest for validated predictive biomarkers in melanoma has been a major focus of immunotherapy research, but there is no validated biomarker currently used in the adjuvant setting [33]. Most work in the field of biomarker research has occurred in the metastatic setting. The KEYNOTE-054 did evaluate tumor PD-L1 expression as a potential biomarker for response to pembrolizumab in the adjuvant setting as a primary endpoint. Most patients enrolled (83.3%) were PD-L1 positive, defined as 22C3 antibody assay score ≥2. This assay is commonly reported in clinical studies of melanoma and other solid tumors [34]. Adjuvant treatment with pembrolizumab was associated with a significantly prolonged RFS among patients with PD-L1-high tumors as compared to placebo, but this effect was also seen among PD-L1-low tumors. The CheckMate-238 trial of adjuvant nivolumab versus ipilimumab enrolled triple the number of PD-L1 negative patients as did KEYNOTE-054 (33.6% versus 11.5%, albeit using a different assay), but analysis of RFS based on PD-L1 expression demonstrated a similar prolonged RFS with PD-1 inhibitor nivolumab regardless of PD-L1 expression.

2.4. Other biomarkers

In addition to PD-L1, recent data suggests that other markers may have predictive value in the adjuvant setting. CheckMate-238, which evaluated adjuvant nivolumab versus ipilimumab in high-risk melanoma patients undergoing complete resection, included evaluation of potential biomarkers as an exploratory endpoint. These included interferon-gamma gene expression signature, tumor mutational burden (TMB), and CD8+ T-cell infiltration by immunohistochemistry). Data presented at ESMO in 2019 demonstrated all three potential biomarkers correlate with improved RFS in both the nivolumab and ipilimumab treated cohorts [35].

2.5. Dosing considerations

Still ongoing, the KEYNOTE-555 study (NCT03665597) is exploring novel dosing strategies for pembrolizumab in melanoma patients, including subcutaneous injection and prolonged treatment cycles. Data from one arm (Arm B) of this trial was presented at the 2020 virtual meeting of the American Association for Cancer Research evaluated the alternative dosing regimen of 400mg every six weeks. The data suggested comparable pharmacokinetic profile and comparable response rates and toxicity between the two treatment strategies [36]. On April 28, 2020, new extended dosing of 400mg every six weeks was approved across all indications, including adjuvant therapy for melanoma [37]. Pressures to minimize patient visits in order to reduce infection risk during the time of the COVID-19 pandemic make this prolonged dosing schedule particularly desirable [38]. Governing bodies have outlined considerations for cancer care delivery during the COVID-19 pandemic; the National Comprehensive Care Networks specifically mentions consideration for prolonged dosing schedules for melanoma patients on treatment with checkpoint inhibitors [39]. In practice, many oncologists initiate treatment at standard dosing of 200mg every three weeks for the first few treatment cycles to allow for careful monitoring for toxicity.

3. Toxicity

Pembrolizumab, like all checkpoint inhibitors, is associated with a novel set of toxicities termed immune-mediated adverse events (irAE) [40]. Treatment-related toxicities are of particular importance in the adjuvant setting, as these patients have no evidence of disease, and may have been cured by surgery alone. Slightly over one-third of pembrolizumab-treated patients in KEYNOTE-054 experienced any immune-related adverse event, compared to less than one-tenth of patients in the control group. The most common irAEs associated with pembrolizumab treatment included endocrine disorders (hypothyroidism or hyperthyroidism); all but one case was mild and did not require discontinuation of therapy. One death was observed in the pembrolizumab cohort, secondary to myositis. Other common irAEs observed with adjuvant pembrolizumab included pneumonitis, vitiligo, dermatitis, arthritis, and colitis. In KEYNOTE-054, pembrolizumab treatment was associated with rate of grade 3–5 treatment-related AEs of 14.7% in the pembrolizumab cohort compared to 3.4% in the placebo group (7.1% vs. 0.6% grade 3–5 irAEs). The rate of treatment discontinuation was 14.8% in the pembrolizumab cohort compared to 2.2% in the placebo cohort. In the pembrolizumab cohort, most patients (34 of 43 total patients, 79%) who suffered G3–4 irAE experienced resolution of toxicity, most commonly within 2 months after the last dose of pembrolizumab.

The toxicity profile of pembrolizumab is largely similar in the adjuvant and metastatic settings. The two pivotal studies of pembrolizumab in the advanced/metastatic setting include KEYNOTE-002 and KEYNOTE-006. The phase 2 KEYNOTE-002 study demonstrated pembrolizumab’s efficacy in ipilimumab-refractory melanoma. Patients treated with 2mg/kg pembrolizumab (most equivalent to the current 200mg flat dosing) experienced an 11% rate of G3–4 toxicity compared to 14% G3–4 toxicity with the 10mg/kg dosing [41]. KEYNOTE-006 compared ipilimumab to pembrolizumab in the setting of advanced disease. Rate of G3–5 toxicity at the 10mg/kg dosing q3weeks was 10.1%, with most common toxicities included fatigue, endocrinopathies, colitis, rash and vitiligo [42]. As expected, toxicity associated with adjuvant pembrolizumab in melanoma is similar to that seen with adjuvant nivolumab. CheckMate-238 found a rate of grade≥3 irAE of 14.4% among nivolumab-treated patients, on par with the 14.7% rate seen with pembrolizumab adjuvant treatment in KEYNOTE-054. There were no deaths in the nivolumab-treated patients of CheckMate-238.

Our group has estimated the rate of fatal toxicities with single agent anti-PD-1 therapy is approximately 0.35% [43]. It is critical to consider the long-term ramifications of toxicity, particularly in patients who may have already been cured by surgery alone. While most severe acute events resolve with therapy discontinuation and glucocorticoids, other events such as endocrinopathies and arthritis may persist [44–47]. The rates of chronic events have not been well-defined, and is a topic of investigation by our group currently.

Similar to patterns seen with treatment in the unresectable/metastatic setting, both PD-1 inhibitors boast a favorable toxicity profile when compared to single-agent CTLA-4 inhibition with ipilimumab. Adjuvant treatment with ipilimumab monotherapy in CheckMate-238 demonstrated grade≥3 irAE in nearly half of patients treated with ipilimumab; two deaths occurred in this group. This is consistent with higher rates of irAE seen with ipilimumab monotherapy compared to PD-1 inhibitor monotherapy, even at the lower dosing regimen of ipilimumab 3mg/kg every three weeks [42]. Similarly, the Phase 3 EORTC 18071 trial evaluated adjuvant ipilimumab versus placebo in completely-resected stage III melanoma. Among the ipilimumab-treated cohort, 44% of patients experienced G3–5 toxicity, requiring treatment cessation.

A secondary analysis of KEYNOTE-054 evaluated whether development of irAE portends a greater response to treatment. This phenomenon has been well-described with PD-1 inhibitors used in the unresectable/metastatic setting across a variety of solid tumors, where objective response rates and progression-free survival are significantly greater among patients who experience irAEs than those who do not [48, 49]. In melanoma, this effect is somewhat more nebulous. Two large analyses of melanoma patients receiving PD-1 inhibition in the unresectable/metastatic setting; the first of these demonstrated that the presence of an irAE was significantly associated with improved overall survival, while the second and larger study demonstrated an improved overall response rate among melanoma patients who experienced any-grade irAE without significant improvement in PFS [50, 51]. Among melanoma patients treated with adjuvant pembrolizumab on KEYNOTE-054, secondary analysis revealed a longer recurrence-free survival among patients who experienced any irAE. The hazard ratio for recurrence-free survival or death was significantly reduced after the onset of irAE among the prembrolizumab group as compared to the placebo group (HR 0.37, 05% CU, 0.24–0.57) [52]. As suggested by previous studies in melanoma, the presence of vitiligo and likely other skin toxicities was associated with the greatest association with clinical benefit [53].

4. Considerations for treatment at recurrence

4.1. Surveillance imaging

The role of routine surveillance imaging for patients treated with adjuvant pembrolizumab is not well established. Guidelines published by the National Comprehensive Cancer Network (NCCN) suggest consideration of both routine imaging and only symptom-directed imaging [54]. This equivocal recommendation is influenced by pre-checkpoint inhibitor data suggesting from 2007 which suggested that most instances of first relapse are detected by patients or their partners or a physician’s physical examination (71.1% and 16.6%, respectively) rather than an imaging study (9.9%) [55]. More recent data suggests that among stage III melanoma patients treated with adjuvant PD-1 inhibitors, most (66%) locoregional recurrence is diagnosed by clinical exam, while most (78%) distant disease was diagnosed on imaging [56]. In clinical practice, patient-specific factors including personal choice most commonly influence an oncologist’s plan to follow physical examination with routine CT imaging. When routine surveillance imaging is pursued, oncologists often obtain CT imaging every six months for at least two years after diagnosis.

4.2. Treatment of recurrent disease after adjuvant pembrolizumab

While adjuvant pembrolizumab decreases rates of disease relapse, approximately one-third of patients will still experience recurrent disease [12]. Primary considerations for retreatment include 1) site of relapse and 2) chronology of adjuvant treatment and relapse. Among patients studied as part of KEYNOTE-054, a majority of patients (57.7%) experienced metastatic disease at first relapse (51.1% metastatic disease only, 6.6% with synchronous metastatic and locoregional recurrence) [25]. It is important to note that both the KEYNOTE-054 and Checkmate-238 studies were performed in an era where completion lymph node dissections (CLND) were the standard of care. Following the completion of the MSLT-II study, which demonstrated no survival benefit for CLND, regional recurrences may be more prevalent [57].

For the group of patients who generally require systemic treatment, the major consideration for therapy is whether they developed recurrence while on treatment with adjuvant pembrolizumab or after completion of one year of adjuvant pembrolizumab. Owen et al evaluated management patterns among early recurrence in patients treated with adjuvant PD-1 inhibitors. Their multicenter retrospective analysis included 850 patients with resected stage III/IV melanoma who were followed for at least two years from initiation of adjuvant therapy;. Of these patients, a 24-month recurrence rate of 17% was observed. Recurrence patterns were similar to that seen in KEYNOTE-054, with 43% of patients recurrent with locoregional-only disease, 40% experiencing metastatic recurrence, and 16% of patients experiencing both locoregional and metastatic disease at first recurrence. Upon progression, patients with resectable disease underwent surgical resection, either alone or in combination with adjuvant radiotherapy, BRAF/MEK inhibition, or PD-1 inhibition. While extrapolating from a small cohort (n=59), patients who were treated with resection and adjuvant BRAF/MEK inhibition in this setting fared best, with no recurrence at median follow-up of 6.5 months compared to 56% recurrence rate with median follow-up 8.3 months in the other groups. Among patients with distant metastatic disease at first recurrence, 33% of patients were treated with ipilimumab-based therapy (either monotherapy or in combination with PD-1 inhibition), 31% were treated with BRAF/MEK inhibitor, 15% were treated with PD-1 inhibitor alone or in combination with an investigational agent, and 21% were treated with local therapy at the metastatic site. Best outcomes were seen in the BRAF/MEKi treated patients, where an overall response rate of 82%. No patients who were treated with continued PD-1 monotherapy after recurrent disease on PD-1 inhibition experienced a response, but two patients who experienced recurrent disease after completion of adjuvant PD-1 inhibitor therapy enjoyed a second response that persisted after 10.3 and 5.4 months. From this data, options for first disease recurrence occurring during adjuvant pembrolizumab treatment include BRAF/MEKi (for BRAF mutant patients), ipilimumab (either alone or in combination with PD-1 inhibitor), or clinical trial. Treatment options for first disease recurrence after completion of one year adjuvant pembrolizumab include all aforementioned options as well as re-challenge with PD-1 inhibitor.

5. Future directions

Applications for immunotherapies have ushered in the “golden age” of cancer medicine, with checkpoint inhibitors inducing dramatic survival benefits across tumor types. Pembrolizumab has earned tumor-agnostic approvals in the setting of microsatellite instable and high-tumor mutational burden disease. In the coming decade, the role of pembrolizumab is likely to expand even further as it undergoes clinical trials to evaluate its role in drug combinations and in expanded treatment settings.

5.1. BRAF mutant

Following in the footsteps of both IMSpire150 trial of targeted therapy plus PD-L1 inhibitor, the Phase II NeoTrio study (NCT02858921) will begin enrollment in November 2020. This study to evaluate the efficacy of targeted therapy (dabrafenib and trametinib) in combination with pembrolizumab in the adjuvant setting among patients with BRAF-mutant stage IIIB or IIIC melanoma. Arms of this study will enroll patients for sequential targeted therapy followed by pembrolizumab, concurrent targeted therapy plus pembrolizumab, and standard-of-care pembrolizumab only treatment.

5.2. Potential for neoadjuvant application

Use of pembrolizumab in the neoadjuvant setting is being explored in both preclinical models and ongoing clinical trials. Application in the pre-surgical setting may simplify surgical resection with debulking and/or reduce the need for adjuvant radiation [58]. It has long been postulated that the use of immunotherapeutics may have greater efficacy in the neoadjuvant (pre-resection) setting as compared to the adjuvant (post-resection) setting, as the greater burden of immune-stimulating antigen present before resection could potentially induce a greater immune response [59].

A number of preclinical studies have explored this approach. Liu et al used two mouse models of spontaneous breast metastatic breast cancer to compare the therapeutic power of neoadjuvant versus adjuvant immunotherapy [60]. The first looked at depletion of regulatory T cells with anti-CD25 monoclonal antibody to relieve tumor-induced immunosuppression; 40% of mice treated neoadjuvantly experience long-term survival (defined as >250 day) as compared to 10% of mice treated adjuvantly. Similarly, mice treated with combination anti-PD-1 and anti-CD137 (a costimulatory receptor expressed on T and NK cells that has shown efficacy in mouse models) showed a 50% long-term survival rate when treated neoadjuvantly, but 0% long-term survival rate when treated neoadjuvantly. As a control, mice were treated with paclitaxel; no significant benefit of neoadjuvant versus adjuvant therapy was found with cytotoxic chemotherapy.

The use of neoadjuvant PD-1 inhibitors has also been explored in clinical trials. This is an area of such great interest that consensus guidelines have been generated for the design of such adjuvant trials [61]. Rozeman et al published the OpACIN-NEO trial in 2018; this phase 2 study assigned patients with resectable stage III melanoma to one of three dosing scheduled of ipilimumab plus PD-1 inhibitor nivolumab. High rates of grade>3 immune-mediated adverse events were seen, forcing premature closure of one treatment arm, but rate of pathologic complete response impressively ranged from 65–85% [62]. Huang et al employed a neoadjuvant/adjuvant anti-PD-1 therapy stagey in resectable melanoma; patient received a single dose of neoadjuvant pembrolizumab (200mg) followed by complete resection three weeks later, then adjuvant pembrolizumab (200mg) every three weeks for one year. Correlative studies identified that changes in TIL infiltration observed on pre-treatment biopsy and resected tumor could predict clinical outcome. Excitingly, 30% of patients in their study experienced a complete or major pathologic response after one dose of pembrolizumab [63]. This pathologic complete response rate is similar to that seen in a Phase 2 study by Amaria et al, in which patients treated with either up to four doses of neoadjuvant nivolumab or up to three doses of neoadjuvant ipilimumab plus nivolumab followed by surgical resection, then six months of adjuvant nivolumab. A pathologic complete response rate of 25–45% was observed [64]. There are current ten clinical trials ongoing which explore pembrolizumab in the neoadjuvant setting. Many of these trials evaluate pembrolizumab in combination with targeted therapies (including KEYNOTE-022, NCT02130466) or other immunotherapies, such as cancer vaccines.

6. Expert opinion

One key challenge is the decision which anti-PD-1 agent to use. The efficacy and safety of these agents is essentially identical, therefore the choice largely comes down to schedule and convenience. Pembrolizumab is now approved every 6 weeks, which would result in only 8–9 treatments to complete a 1 year course. This compares favorably to the 13 treatments required for one year of nivolumab given every 4 weeks. However, we tend to give pembrolizumab every 3 weeks for the first 3 months on therapy to monitor more closely during the peak times of toxicity onset. This modification makes the difference in schedule more modest (~11 vs. ~13 treatments). Thus, we view that this decision is nearly a “toss-up” and comes down to patient and physician preference.

Another important decision for the future will be whether the adjuvant therapy mainstay will be anti-PD-1 monotherapy or the combination of PD-1 and CTLA-4 blockade. Current evidence is conflicting. The small randomized phase II IMMUNED study suggested a substantial benefit to the combination in resected stage IV patients, who are at higher risk for recurrence than most stage III patients. In contrast, preliminary evidence from the Checkmate-915 study did not suggest a benefit for RFS in the subset of stage III-IV patients without PD-L1 expression on a preliminary analyses [65]. Further follow-up and full study analyses remain ongoing. In our current practice, we at least consider the combination of ipilimumab and nivolumab in patients with resected stage IV melanoma (particularly in the subset with resected brain metastases) for patients with adequate performance status. However, for stage III patients, we largely opt for anti-PD-1 monotherapy.

Next, the choice between BRAF/MEK inhibition and anti-PD-1 in high-risk BRAF V600 mutated patients remains challenging. We usually elect for anti-PD-1 given the (admittedly subjective) view that anti-PD-1 is more easily tolerated compared with BRAF/MEK inhibition. Further, anti-PD-1 is usually associated with more durable response and benefit compared with targeted therapy; thus we favor offering the patient the best chance of cure upfront. However, some of our colleagues disagree with this approach, and argue that BRAF/MEK inhibition in this setting may allow for cure in some patients in this setting, and reserve anti-PD-1 therapy for progression. It is difficult to imagine that any prospective study will settle this debate; thus this question will likely remain contentious absent another superseding treatment.

The question of adjuvant vs. neoadjuvant treatment is also challenging. In our view, neoadjuvant treatment should be considered in patients with a clinically-obvious, surgically resectable lesion. It should be noted though, this is not the setting that most patients approach adjuvant therapy; most patients experience a positive sentinel lymph node and then consider adjuvant therapy. Many other questions remain as well, including: do patients with an excellent response to neoadjuvant therapy need surgery? Do patients without pathologic responses at surgery need a switch in therapy? Is neoadjuvant therapy more effective than adjuvant therapy on an absolute basis? The latter question is being addressed for pembrolizumab in a randomized phase II study (SWOG 1801).

Finally, one elusive factor to help determine many of these questions has been effective biomarkers. Markers of recurrence and/or anti-PD-1 effectiveness would allow for improved decision-making for treatment vs. withholding therapy, and anti-PD-1 vs. BRAF/MEK (or ipilimumab/nivolumab). However, early biomarker efforts have not yielded many promising candidates in the adjuvant setting.

Article highlights.

• Pembrolizumab as an adjuvant therapy improves relapse-free survival (RFS) in patients with resected stage III melanoma.

• The impact of overall survival for adjuvant anti-PD-1 has not yet been defined.

• Patients with immune-related adverse events (irAEs) experience superior RFS compared with patients not experiencing irAEs

• The choice of adjuvant therapies depends on dose and schedule considerations, patient co-morbidities, and for some physicians, BRAF mutation status.