When is it OK to Stop Anti-Programmed Death 1 Receptor (PD-1) Therapy in Metastatic Melanoma?

Abstract

Systemic therapy for metastatic melanoma has been revolutionized over the past decade with the development of highly effective immune checkpoint inhibition, specifically anti-Programmed Death 1 receptor (PD-1) therapy. However, even though one-third of patients will have durable response to single-agent or combination therapy, the optimal duration of therapy is unknown. Identifying the optimal duration of therapy is important, as exposure to anti-PD-1 therapy increases the risk of developing immune-mediated toxicities that can have significant morbidity and are, at times, fatal. It has long been understood that patients with complete responses to high-dose interleukin-2 and ipilimumab typically maintain their responses after a brief treatment course; thus, it is important to better understand the data to help understand the optimal management of melanoma patients treated with anti-PD-1 therapy. The clinical data with anti-PD-1-based therapy and published data on the duration of therapy suggest that patients may not require a full 2 years of anti-PD-1 therapy and that the risk of toxicity may be mitigated by further understanding the mechanisms and kinetics of response to therapy. Although novel markers to help guide therapeutic decision making are under investigation, there is an ongoing need to improve our tools to monitor response to therapy and disease activity.

Introduction

The treatment of metastatic melanoma has been transformed by the use of immune checkpoint inhibition, particularly monoclonal antibodies against the Programmed Death 1 receptor (PD-1). PD-1 is expressed on T cells and binds to its ligand, PD-L1, on the surface of both antigen-presenting cells and non-immune cells. The PD-1/PD-L1 interaction inhibits T cell activation and serves to downregulate the immune response [1]. Disrupting PD-1/PD-L1 has been shown to promote the anti-tumor immune response [2]. In 2014, the US Food and Drug Administration (FDA) approved two anti-PD-1 antibodies, pembrolizumab and nivolumab, for the treatment of advanced (stage III unresectable or metastatic) melanoma. These agents have led to unprecedented survival of patients diagnosed with metastatic melanoma; however, they are also associated with toxicities mimicking autoimmune disease of virtually every organ. While only up to 20% or so of patients will develop severe and/or life-threatening toxicity, these adverse effects can have significant morbidity, are occasionally fatal, and regularly require immunosuppressive therapy to reverse. Furthermore, these can occur with ongoing therapy and even for a limited time after treatment stops. Some patients even experience multiple immune-related adverse events (irAEs) in a cumulative manner while on therapy and with significant toxicities developing in the months following treatment discontinuation [3, 4]. These occurrences highlight the risk of continued exposure to anti-PD-1 treatment and set forth an argument to minimize time on therapy. Thus, an important next step in the field is to determine the optimal duration of therapy for those with partial and complete responses (CRs) to treatment.

The exact duration of anti-PD-1 therapy in patients with metastatic melanoma that provides the maximum anti-tumor effect and avoidance of undertreatment but minimizes time on treatment, and therefore the risk of toxicity, remains elusive. Preliminary data have been presented from CheckMate 153, an ongoing phase IIIb/IV trial specifically designed to compare outcomes in patients with non-small cell lung cancer stopping nivolumab after 12 months versus continuous therapy [5]. These data show a trend towards a benefit of continuous nivolumab treatment over stopping treatment at 1 year. However, it is not clear that these results can be extrapolated to melanoma, a disease with different biology and disease course and longer duration of responses (DORs) to anti-PD-1 therapy. As such, we await the completion of a similar study in patients with metastatic melanoma, DANTE (A randomised phase III trial to evaluate the Duration of ANti-PD1 monoclonal antibody Treatment in patients with metastatic mElanoma; ISRCTN15837212), which is currently ongoing and expected to mature in 2027. This is a multicenter phase III trial in the UK that randomizes patients to either discontinue anti-PD-1 therapy at 12 months or continue until disease progression or unacceptable toxicity, or a minimum of 2 years should neither occur. This will be the first study to date to specifically evaluate duration of therapy and will have significant clinical implications if indeed patients can stop treatment after half the amount of time on and exposure to therapy. However, in the meantime, those in the field will need to extrapolate from other datasets to help make the right clinical decision for individual patients.

There is some precedence for discontinuing cancer immunotherapy in melanoma. Specifically, high-dose interleukin (IL)-2 (HD-IL-2) and the anti-cytotoxic T lymphocyte associated antigen 4 (CTLA-4) agent ipilimumab are associated with complete and durable responses that can occur with as little as 3–4 months of therapy, in line with the recommended duration of these therapies as per their respective FDA labels. However, the FDA label for nivolumab and pembrolizumab allows for indefinite therapy, which for many patients represents overtreatment. The purpose of this review is to explore the optimal duration of anti-PD-1 therapy in light of these considerations and in the context of contemporary clinical experience.

1. Historical Precedent for Durable Responses to Immunotherapy

HD-IL-2 is a potent stimulant of T cell proliferation and therefore contributes to activation of the T cell response [6]. Given the immunogenicity of melanoma, treatment with IL-2 had been standard for the management of metastatic melanoma for patients with a high performance status. The regimen, which was developed at the US National Cancer Institute, included high-dose therapy (600,000 international units/kg every 8 h for up to 14 doses per week as tolerated) given every other week for 4 weeks with a potential second course of therapy 3 months later followed by observation [7, 8]. The dataset leading to FDA approval in melanoma included 270 patients prospectively treated and followed at a number of centers [9]. Responses were seen in 16% of patients but, most importantly, these responses were complete and durable in 6% of patients. High-grade toxicity, including capillary leak, hypotension, acute kidney injury, and neurotoxicity, limits which patients are eligible for therapy (e.g., younger, healthier patients) and requires highly specialized centers that are able to manage these patients with intensive care unit (ICU)-level care during inpatient hospitalizations. However, despite these limitations, HD-IL-2 was the only available, standard immunotherapy for patients with metastatic melanoma for nearly two decades. These early data and more contemporary cohorts have provided the evidence that persistent response off active therapy is a possibility in melanoma patients treated with immunotherapy [9–11].

A second such therapy is ipilimumab, the first of the checkpoint inhibitors currently approved for treatment of metastatic melanoma based on data demonstrating an overall survival (OS) advantage [12]. Ipilimumab is a blocking monoclonal antibody directed against CTLA-4, a negative regulator of T cell activation. The current FDA-approved dosing schedule for ipilimumab for patients with unresectable and metastatic melanoma is 3 mg/kg intravenously every 3 weeks for a total of four doses. This corresponds to a treatment duration of 12 weeks. Long-term survival data have been reported through analysis of the CA184–25 trial, a companion trial to the initial phase II studies demonstrating efficacy of ipilimumab as a treatment for metastatic melanoma [13]. Follow-up survival data from 1861 patients in phase II and III studies treated with the aforementioned regimen or the higher dose of 10 mg/kg demonstrated a 3-year OS of 22% and that the survival curve seemed to plateau at this timepoint (3 years) and extending to follow-up of 10 years [14]. In the most recent (and probably last) upfront ipilimumab data from the Checkmate 067 study, which randomized patients to frontline ipilimumab, nivolumab, or combination of the two, the 4- and 5-year OS is 30% and 26%, respectively [15]. This suggests long-term survival benefit of ipilimumab despite only 12 weeks of exposure to treatment.

The examples of IL-2 treatment and ipilimumab, both therapies that affect and modulate the T cell response, provide examples of sustained responses despite short treatment courses. They support the hypothesis that patients receiving anti-PD-1 therapy may be adequately treated with less than 2 years of therapy.

2. Data with Anti-Programmed Death 1 Receptor (PD-1)

3.1. Initial Anti-PD-1 Trials

The landmark trial leading to FDA approval of pembrolizumab for metastatic melanoma was the KEYNOTE-001 trial, a phase I study designed to interrogate the safety and efficacy of pembrolizumab in patients with solid tumors [3]. Study participants received varying doses and frequencies of pembrolizumab and were maintained on therapy for the duration of the study unless they developed treatment-limiting toxicity, had disease progression, or withdrew consent. No standard treatment duration was assigned, but the trial allowed for treatment discontinuation in the setting of CR per investigator and patient preference. The initial analysis of 135 patients with advanced melanoma in March 2013 was promising, demonstrating an overall response rate (ORR) of 38% by RECIST (Response Evaluation Criteria In Solid Tumors). Similar to the results of pembrolizumab in KEYNOTE-001, the phase I study of nivolumab enrolled 107 patients with advanced melanoma treated with nivolumab [4]. The ORR was 31%, and the estimated 1- and 2-year OS rates of 62% and 43%, respectively. These were unprecedented survival rates for immunotherapy at the time of data reporting.

The first randomized study of pembrolizumab was KEYNOTE-002, which compared pembrolizumab to standard chemotherapy (carboplatin, paclitaxel, dacarbazine, or temozolomide) in patients with advanced melanoma [16]. This study was a double-blind randomized controlled trial that enrolled a total of 540 patients who had progressed after treatment with ipilimumab. Patients received one of two doses of pembrolizumab (2 or 10 mg/kg every 3 weeks) or standard chemotherapy (investigator choice). Treatment was continued until consent withdrawal, development of treatment-limiting toxicity, or disease progression. Patients receiving 2 or 10 mg/kg of pembrolizumab every 3 weeks had a 6-month progression-free survival (PFS) rate of 34% and 38%, respectively, with median DOR not reached in either group. Patients treated with chemotherapy had a 6-month PFS rate of 16% and a median DOR of 37 weeks. This established the superiority of pembrolizumab over chemotherapy in patients with ipilimumab treatment-refractory metastatic melanoma.

The counterpart to KEYNOTE-002 comparing nivolumab to chemotherapy was CheckMate 037. This was a phase III trial that compared the efficacy of nivolumab to investigator’s choice chemotherapy (dacarbazine or carboplatin + paclitaxel) [17]. As of the first interim analysis, the ORR was 31.7% of patients in the nivolumab and 10.6% the chemotherapy arm; median time to response in the nivolumab and chemotherapy arms was 2.1 and 3.5 months, respectively, with a median time on treatment of 5.3 and 2 months, respectively. Importantly, the DOR was not reached in the nivolumab arm at the time of the first interim analysis, with 87% of responses ongoing at that time compared with a 3.5-month DOR and 31% of responses ongoing in the chemotherapy arm. Similar to pembrolizumab, this trial established nivolumab as superior to chemotherapy as a treatment in patients with metastatic melanoma that has progressed on ipilimumab. Additionally, it highlighted the durable responses to anti-PD-1 therapy despite a relatively short time on therapy.

A separate phase III trial, CheckMate 066, compared nivolumab to chemotherapy (dacarbazine) in previously untreated advanced melanoma patients [18]. Not surprisingly, nivolumab emerged superior to chemotherapy with an ORR of 40.0% compared with 13.9% in patients treated with dacarbazine. This translated into a survival benefit, as median OS was not reached in the nivolumab arm and was only 10.8 months in the dacarbazine arm. Further, the 1-year OS was 72.9% and 42.1% for the nivolumab and dacarbazine arms, respectively.

Although KEYNOTE-002 and CheckMate 037 and 066 established that anti-PD-1 therapy was superior to chemotherapy, KEYNOTE-006 and CheckMate 067 were the landmark trials that established anti-PD-1 therapy as first-line treatment for patients with metastatic melanoma over ipilimumab.

KEYNOTE-006 was a phase III randomized controlled trial comparing the efficacy of pembrolizumab with ipilimumab and helped to establish anti-PD-1 therapy as first line in the treatment of metastatic melanoma [19]. Critically, this was also the first trial to have a specified pembrolizumab treatment duration, which was 24 months. The trial also allowed for treatment discontinuation in patients who were complete responders after treatment for at least 6 months and at least an additional two doses after determination of CR. The trial included two pembrolizumab treatment arms that differed by dosing schedule (dosing at 2-week intervals vs. 3-week intervals), which ultimately showed similar efficacy. Estimated 6-month PFS for the 2- and 3-week dose schedule pembrolizumab arms was 47.3% and 46.4%, respectively. Estimated 12-month OS rates were 74.1% and 68.4%, respectively. Estimated 6-month PFS and 12-month OS rates for the ipilimumab arm were significantly lower at 26.5% and 58.2%, respectively. Therefore, KEYNOTE-006 established that pembrolizumab was superior to ipilimumab as a frontline treatment for melanoma.

Similarly to KEYNOTE-006, CheckMate 067 was a randomized phase III trial comparing the efficacy of nivolumab with ipilimumab in treatment-naïve patients with either unresectable stage III or metastatic melanoma [15, 20]. CheckMate 067 also included an arm of patients treated with combination therapy (ipilimumab and nivolumab). Patients treated with ipilimumab were administered a total of four doses. Patients treated with nivolumab were continued on therapy until disease progression, treatment-limiting toxicity, or consent withdrawal. Combination therapy had the longest median PFS of 11.5 months. However, the nivolumab group had a statistically significant longer median PFS than ipilimumab (6.9 vs. 2.9 months, respectively). The ORRs of patients treated with nivolumab and ipilimumab were 43.7% and 19.0%, respectively, and 57.6% with the combination. This trend was mirrored in the proportion of CRs, partial responders (PRs), and patients with stable disease (SD). Median time to response was consistent among the nivolumab, ipilimumab, and combination arms at 2.78, 2.79, and 2.76 months, respectively. The median DOR was not reached in any group. With 5 years of survival data now available, the median OS of the three arms has not been reached with the combination but is 36.9 months with nivolumab and 19.9 months with ipilimumab. The 5-year OS of the three groups was 52%, 44%, and 26%, respectively. Although combination therapy appeared more efficacious, it also had the highest rate of toxicity, with 55% of patients experiencing treatment-related grade 3 or 4 toxicity compared with 16.3% and 27.3% in nivolumab- and ipilimumab-treated patients, respectively. With the results of CheckMate 067, nivolumab, alone or in combination, was determined to be a more effective frontline therapy for the treatment of metastatic melanoma than ipilimumab.

As a result of the KEYNOTE and CheckMate trials, anti-PD-1 therapy, either as a single agent or in combination with ipilimumab, has become the standard first-line treatment for metastatic melanoma. Primarily based on the results of KEYNOTE-006 and the initial trial design from the multidose phase I trial of nivolumab, standard practice has been to treat patients for 2 years, providing they have not had disease progression and/or have not developed treatment-limiting toxicity. Notably, the choice of 2 years was completely arbitrary and not based on clinical data or an informed hypothesis. Yet with subsequent trials utilizing the ‘historical precedent’ of 2 years, it is hard to find data supporting a shorter duration of therapy without looking at subsets of patients on these trials who were treated with less therapy for reasons other than progression of disease.

3.2. Long-Term Follow-Up from Clinical Trials

Follow-up analysis from the aforementioned trials has shed some light on clinical outcomes of patients and the observed duration of anti-PD-1 therapy in practice. Five-year follow-up data from KEYNOTE-001 found that the median time on treatment for the 655 patients ultimately enrolled in the study was only 5.6 months [21]. Interestingly, the duration of treatment was much longer in patients who achieved a CR with a median time on treatment of 24 months. Of the 105 patients who achieved a CR in the KEYNOTE-001 study, 67 patients elected to stop pembrolizumab and received no further treatment for their cancer. The median time to response and CR were 3 and 13 months, respectively, and patients were treated for a median of 7 months after reaching CR. Ninety-one percent of patients remained in CR after a median time off treatment of 22 months. From this analysis, it is clear there is a high proportion of patients who had a CR and remained in CR for nearly 2 years after stopping pembrolizumab treatment. These data suggest that persistent exposure to drug, and therefore treatment until progression, may not be necessary to maintain a durable response to therapy.

Similarly, 5-year follow-up analysis of KEYNOTE-006 found patients had a median time on pembrolizumab of 6 months [22]. Only 19% of patients remained on pembrolizumab for the prespecified 2 years of therapy. Of these patients, best overall response was CR in 20% of patients, PR in 67% of patients, and SD in 13% of patients. The significant overrepresentation of response to therapy is expected given the criteria for continuing treatment as lack of progression of disease. Follow-up analysis from KEYNOTE-006 more pointedly addressed the length of treatment given the prespecified 2-year treatment duration in the trial protocol. The authors mention that 24-month PFS was similar between patients with CR who completed 2 years of therapy and those with CR who completed at least 6 months of therapy but less than 2 years. This calls into question the benefit of treating patients for 2 years if a shorter duration of therapy may yield similar results without clinical evidence of undertreatment.

Analysis of the 4-year outcomes of CheckMate 067 found that patients were treated with a median of 15 doses of nivolumab (~ 7.5 months with 2-week dosing) [23]. With a minimum follow up of 48 months, the rate of CR as best overall response for patients treated with nivolumab was 18%. The median DOR had not been reached. Overall, this follow-up analysis demonstrates ongoing response to anti-PD-1 therapy after cessation of treatment. Furthermore, these durable responses persist after a median of only 7.5 months of therapy.

Follow-up data from KEYNOTE-006 and CheckMate 067 suggest that indefinite exposure to anti-PD-1 therapy is likely unnecessary for lasting responses to therapy. They also highlight the need to better understand the kinetics of response to anti-PD-1 therapy outside the limitations of drug administration. Unlike chemotherapy, checkpoint inhibition in part reprograms T cells, and it is not clear how long the changes to the immune system persist after drug exposure. It is likely the duration of activity of therapy and immune activation surpasses active administration of the drug. For example, in the 5-year follow-up of KEYNOTE-006, 8% of patients with prior best overall response of PR converted to CR after stopping pembrolizumab [22]. Similarly, 18% of patients receiving nivolumab monotherapy in CheckMate 067 reached CR after a minimum of 48 months’ follow-up. This is double the rate of CRs in this population after a minimum of 9 months’ follow-up, suggesting best overall response for patients can be reached after therapy is complete [23]. These data support the idea of an ongoing effect of treatment without continuous exposure to anti-PD-1 agents, adding yet another layer of complexity to defining the optimal duration of treatment and an area in need of further exploration.

3.3. Clinical Outcomes of Anti-PD-1 Therapy Outside of Trials

There have yet to be clinical trials specifically addressing the optimal duration of anti-PD-1 therapy. Analysis of clinical outcomes of patients treated outside the confines of a clinical trial protocol can be helpful in guiding the discussion. One such analysis by Jansen and colleagues [24] describes clinical outcomes of 185 patients from across Europe and Australia with advanced melanoma treated with anti-PD-1 therapy (either pembrolizumab or nivolumab) who electively discontinued treatment. The median time on treatment for these patients was 12 months. Progression of disease occurred in 22% of patients after a mean follow-up of 18 months, with a median PFS that was not reached. Interestingly, among patients with a CR, treatment duration of less than 6 months did correlate with a significantly higher risk of disease progression, with a median PFS of 18.9 months compared with a PFS that was not reached in patients with a CR who were treated for more than 6 months. However, there was no difference in the risk of progressive disease for patients with a CR who were treated for 6–12, 12–18, 18–24, or greater than 24 months. These data suggest that patients who achieve a CR should receive 6 months of treatment but that further treatment beyond 6 months may not be necessary; more studies are needed to confirm this finding.

A retrospective study of patients treated at the Oulu University Hospital in Oulu, Finland also investigated the duration of treatment in 59 patients with metastatic cancer (genitourinary, lung, or melanoma) [25]. Within the study cohort, 23 patients had metastatic melanoma, with a median PFS and OS of 8 and 27 months, respectively. Among the 59 patients in the study, the authors reported a median time on therapy of 3 months. Although the median duration of therapy was not reported by cancer type, this does allude to a relatively short time on therapy with a significant survival benefit.

3. Future Directions and Unmet Needs

The ability to analyze clinical outcomes in the context of treatment duration becomes more powerful as the number of patients treated with anti-PD-1 therapy rapidly grows. Retrospective analyses will continue to provide insight into the duration of therapy. One can infer from the KEYNOTE and CheckMate trials, as well as analyses of clinical outcomes of patients treated with pembrolizumab outside of clinical trials, that patients can have durable responses to anti-PD-1 therapy when treated for less than 2 years. Yet the optimal duration of treatment, although likely to be less than 2 years, is unlikely to be uniform. Factors such as time to response and time to best overall response likely play important roles in an individual’s optimal duration of therapy. Ongoing post hoc analyses with larger cohorts of patients than those to date will help to further dissect the roles these factors play in determining appropriate duration of treatment.

In addition to more robust retrospective analyses, it will be important to identify methods and biomarkers for tracking response while on therapy as well as DOR after stopping treatment. This will not only allow for much more precise measurement of the kinetics and durability of response, but also allow for the anticipation of recurrence. Melanoma lacks the classic tumor markers of other solid tumors, such as carcinoembryonic antigen (CEA), carbohydrate antigen 19–9 (CA19–9), or prostate-specific antigen (PSA). Presently, response to therapy is typically based on anatomic radiographic response as measured by RECIST. Additionally, serum lactate dehydrogenase (LDH) is typically followed in patients as an increase in LDH, although somewhat non-specific, does correlate with disease activity. Accordingly, finding better biomarkers of disease activity is an area of focus in solid tumor research. One such method is the use of liquid biopsies, which involves analyzing parts of tumors (whether circulating tumor cells [CTCs] or tumor DNA) released into systemic circulation.

4.1. Circulating Biomarkers

In melanoma, the identification of CTCs has relied on the use of reverse transcriptase–polymerase chain reaction (RT-PCR)-based assays. These assays are aimed at detecting the expression of messenger RNA (mRNA) transcripts specific to melanoma cells [26]. A prospective phase III trial in 2012 investigated the use of an RT-PCR based assay to detect the burden of CTCs in patients with metastatic melanoma after complete metastasectomy [27]. Investigators identified this to be prognostic, but it is not clear how well this method would allow for clinical monitoring at time of discontinuation decision, based on difficulty with quantification. A more recent study examined the change in CTC burden in patients with metastatic melanoma in response to checkpoint inhibition. The authors developed a “CTC score” that entailed quantitative analysis of 19 genes associated with circulating melanoma cells, isolated via a microfluidics device, to create an “RNA signature” [28]. The sensitivity of this score to identify CTCs was established using melanoma cell lines and validated in a test cohort of melanoma patients. Investigators next examined the CTC score in a prospective cohort of 49 patients who were followed longitudinally during treatment. They found that it was not the baseline CTC score but the change in CTC score after treatment that correlated with PFS and OS. Sixty-four percent of patients with an increase in their CTC score after starting treatment had progression of disease at 12 months compared with 15% of patients who experienced a reduction in their CTC score from baseline. This suggests prognostic utility of CTCs and the detection of early responses, both of which are essential to rigorous study and determination of the optimal duration anti-PD-1 therapy.

Another form of liquid biopsy is circulating tumor DNA (ctDNA). There is cell-free DNA detectable in plasma, a fraction of which can be attributed to tumor DNA in patients with cancer, presumably shed during tumor cell turnover. ctDNA can be detected and quantified through the application of known tumor-specific somatic genetic alterations. In a study of 136 patients with metastatic cancer of various types, including melanoma, 82% of patients had detectable ctDNA [29]. The utility of ctDNA in melanoma was further investigated in a study monitoring treatment response in melanoma [30]. In this study, the authors measured baseline ctDNA prior to treatment, which was detectable in the plasma of 35 of 48 patients (73%). They then compared baseline levels of ctDNA with response to treatment with BRAF/MEK inhibitors and found that all patients with fewer than ten copies of ctDNA/mL responded to targeted therapy. Similarly, patients receiving immunotherapy who responded to treatment had a significantly lower baseline level of ctDNA than non-responders. This trend was mirrored in PFS.

Supporting the idea that ctDNA levels might correlate with disease course, a larger study examined longitudinal ctDNA analysis in 76 patients with metastatic melanoma treated with anti-PD-1 therapy [31]. The authors of the study assigned patients to one of three groups: Group A (those with undetectable baseline ctDNA), Group B (patients whose ctDNA level became undetectable within 8 weeks after starting treatment), or Group C (patients whose ctDNA was detectable in all samples collected). Patients in Group A and Group B had an ORR of 72% and 77%, respectively, while Group C patients had an ORR of only 6%. Similarly, PFS was significantly longer in patients in Groups A and B than in Group C. Interestingly, the median PFS was not reached in Groups A or B. The authors found the sensitivity and specificity of undetectable baseline ctDNA levels or ctDNA levels that became undetectable with treatment as a predictor of response to be 79% and 94%, respectively. Overall, this study demonstrates the potential utility of ctDNA in predicting response to treatment but did not address the use of ctDNA to guide therapy decisions such as when to discontinue anti-PD-1.

Moving one step closer to incorporating ctDNA into clinical decisions was a study that sought to differentiate patients with pseudoprogression from those with true progression of disease after starting therapy [32]. Pseudoprogression, a phenomenon where patients have radiographic evidence of increased tumor burden shortly after starting immunotherapy, at present can only be confirmed retrospectively after continued therapy and subsequent imaging. After analyzing the first restaging imaging in 125 patients after starting checkpoint inhibitor therapy, the authors of this study found 29 patients who had radiographic evidence of disease progression, nine (31%) of whom were ultimately found to have pseudoprogression. All nine of these patients had a favorable ctDNA profile, with a favorable profile defined as ctDNA undetectable at baseline, conversion from detectable to undetectable at 6 weeks, or a ten-fold decrease in ctDNA level at 12 weeks. Furthermore, 18 of 20 patients with disease progression had an unfavorable ctDNA profile. With their data, the authors determined that the sensitivity and specificity of a ctDNA profile predicting pseudoprogression were 90% and 100%, respectively. This study serves as a proof of concept that ctDNA can be used as a biomarker to characterize patients’ current responses to therapy and may support further study into ctDNA analysis to help clinical decision-making, such as whether or not to continue with current therapy.

A critical potential limitation of ctDNA is its ability to reflect CNS disease activity in the aforementioned studies. Fewer patients with primary brain tumors, such as medulloblastoma or gliomas, have detectable ctDNA levels [33]. Additionally, a patient with metastatic melanoma with initial treatment response but recurrence in the brain only did not have the rise in ctDNA level seen in patients with peripherally recurrent disease [32]. Therefore, although promising, it is likely that ctDNA alone will not serve as a sufficient biomarker for all patients, including for those in whom a decision is to be made about therapy discontinuation.

4.2. Imaging Biomarkers

In addition to liquid biopsies, imaging biomarkers are currently under investigation as tools to track and predict patients’ responses to immune checkpoint inhibitor (ICI) therapy. Specifically, fluorodeoxyglucose (FDG)–positron emission tomography (PET) imaging has recently emerged as such a biomarker. A small prospective study of 20 patients sought to address the role of FDG-PET imaging in the prognostication of patients on ICI therapy [34]. Sixteen of these patients were treated with anti-CTLA4 while the remaining four were treated with anti-PD-L1 or anti-PD-1 treatment. Patients in the study had imaging at three timepoints: prior to starting treatment, 21–28 days after starting treatment, and 4 months after starting treatment. Computed tomography (CT) imaging was analyzed based on RECIST1.1 (RECIST version 1.1.) and irRECIST (Immune-related RECIST), and PET/CT imaging was analyzed by PERCIST (PET Response Criteria In Solid Tumors) and EORTC (European Organisation for Research and Treatment of Cancer) criteria. Ultimately, the result (CR, PR, SD, or progressive disease) by these criteria on imaging at 21–28 days after starting therapy was correlated with the best overall response on imaging 4 months after starting treatment. Of the four criteria, RECIST1.1 had the highest sensitivity for predicting best overall response at 4 months. Interestingly, however, increased FDG uptake at 21–28 days seemed to correlate with clinical benefit as defined by CR or PR at 4 months or SD at 6 months.

A number of retrospective studies have since examined PFS and OS as they relate to response on FDG-PET in patients treated with ICI therapy. A retrospective analysis of 37 patients with cutaneous melanoma treated with either ipilimumab, pembrolizumab, or nivolumab found that results of imaging 12 weeks after starting treatment correlated with both PFS and OS [35]. Patients with brain metastases only were excluded. Patients were defined as responders if they had complete metabolic response (CMR), partial metabolic response (PMR), or stable metabolic disease (SMD). Patients with progressive metabolic disease (PMD) were identified as non-responders. The median PFS for responders and non-responders was 23.8 and 6.14 months, respectively. The median OS for responders and non-responders was 26.15 and 11.5 months, respectively. Analysis of CT imaging by iRECIST and RECIST mirrored the results for both median PFS and OS.

Specifically examining patients treated with anti-PD-1 therapy, a retrospective analysis of 105 patients treated with either pembrolizumab or nivolumab (31% had combined treatment with ipilimumab) sought to further characterize patients without CR on CT [36]. Acknowledging that this is a heterogenous population of patients with both sustained and short-lived responses, the initial hypothesis was that CMR among patients with a partial response is associated with a better outcome. Indeed, patients with a CMR at 1 year had a median PFS that was not reached, while patients without a CMR had a mean PFS of 12.8 months with a hazard ratio (HR) of 0.06. The results were nearly identical when patients with a PR on CT were similarly stratified between CMR and non-CMR, with an HR of 0.07.

Yet another retrospective study of 55 patients treated with anti-PD-1 therapy used different parameters noted on a baseline scan to predict PFS and OS [37]. Based on metabolic tumor volume and bone marrow to liver standard uptake value (SUV) ratio, patients were stratified into low-risk, intermediate-risk, and high-risk categories. These categories correlated with OS such that low-, intermediate-, and high-risk classifications had median OS of 52.4, 36.7, and 13.9 months, respectively. These results suggest that predicting a response to anti-PD-1 therapy may be possible prior to treatment initiation using baseline pre-treatment FDG-PET imaging. Again, it should be noted that patients with brain metastases only were not included in this study.

Overall, there is a clearly established correlation between responses on FDG-PET and response to ICI therapy. While this alludes to the potential utility of FDG-PET imaging to track response to therapy, it is not yet actionable in guiding duration of therapy. As an example, it is not known for certain that the low-risk patients defined by baseline imaging will still have better outcomes than high-risk patients if treated for only 6 months. Additionally, while it is important to further understand the clinical trajectory of PRs, it is not known when it is safe to stop therapy such that patients with PR and CMR continue to have a better outcome than patients with PR and non-CMR. Therefore, to further vet the utility of FDG-PET imaging as a biomarker of disease activity, it will need to be included as a parameter analyzed in future studies on the duration of anti-PD-1 therapy. Perhaps most importantly, the results of FDG-PET at time of treatment discontinuation will need to be included into the analysis of these patients to understand if CMR on PET/CT is associated with non-progression after treatment discontinuation and whether it can be used to help decide which patients can safely stop therapy.

One important limitation of relying solely on FDG-PET for guiding therapy duration is the fact that FDG-PET imaging is less useful in detecting uveal, mucosal, or intracranial disease. Indeed, most of the studies discussed here excluded patients with uveal, mucosal, or disease with intracranial metastases only. Although uveal and mucosal melanomas are uncommon, intracranial metastases are common and responsible for significant morbidity associated with metastatic melanoma when present. In the study by Tan et al. [36], one of the three patients with PR and CMR who progressed did so intracranially only. Furthermore, a recent retrospective study of 60 patients treated with ipilimumab comparing the relationship between FDG-PET and survival found that patients with active brain metastases had worse OS than those without conferring an HR of 2.6 [38]. Therefore, while FDG-PET imaging does seem to have the potential to become a valuable tool tracking response to anti-PD-1 therapy, similar to ctDNA, it is unlikely to stand alone as a biomarker of disease activity because of its inability to reliably evaluate intracranial disease. Additionally, it is expected that a brain MRI would have to be performed in concert with a PET/CT and/or circulating biomarker analysis before making a decision to discontinue therapy.

4. Conclusion/Current Opinion

Anti-PD-1 therapy has significantly altered the treatment landscape for and prognosis of metastatic melanoma. Patients have had ongoing responses to pembrolizumab and nivolumab that lead to CR in a significant proportion. However, these therapies have also exposed patients to at times devastating toxicities, which makes identifying the minimal amount of time needed on treatment for a durable response to therapy increasingly important. Integral to determining optimal treatment duration is a better understanding of the precise biology and mechanism behind disease control and the kinetics of response to anti-PD-1 therapy, as highlighted by observed experience with pembrolizumab and nivolumab thus far. Future work, both in the form of ongoing detailed outcome analysis as well as identifying precise ways of tracking response to therapy, are needed to identify the optimal duration of anti-PD-1 therapy to maximize response and minimize toxicity.

Retrospective and long-term follow-up analysis of clinical outcomes with anti-PD-1 therapy suggests that patients can have durable and clinically meaningful responses to treatment without requiring indefinite therapy (and likely less than 2 years of therapy). Based on these data, it is our opinion that patients with CRs should complete at least 6 months of therapy prior to treatment discontinuation. More generally, patients who have CRs or PRs without substantial residual tumor burden at the 1-year timepoint may safely discontinue therapy after a discussion about what is and is not known about the risks of continued therapy and the chance of disease progression in the setting of treatment discontinuation. For patients with PRs with substantial residual tumor burden or patients with SD, treatment for a total of 2 years is recommended, although in some patients indefinite therapy may be indicated based on the amount of disease evident on imaging.

Key Points.

• Anti-Programmed Death 1 receptor (PD-1) therapy is the standard of care for patients with metastatic melanoma and the current guideline-directed duration of therapy is 2 years based on landmark trial design. However, this exposes patients to significant, and at times life-threatening, toxicity.

• While a significant minority of patients will have durable benefit, which often persists after treatment discontinuation, the optimal duration of therapy is unknown.

• Clinical data and experience supports the notion that patients may not require a full 2 years of therapy for maximum benefit of anti-PD-1 therapy while highlighting the ongoing need to improve our tools to monitor response to therapy and disease activity.