Immune checkpoint inhibitors (ICIs) revolutionized the treatment of metastatic melanoma patients, contributing to significantly improved overall survival (OS) rates compared to other treatment modalities (1, 2). Importantly, responses on ICIs are considered durable and patients achieving complete or partial responses seem to remain on response, even after treatment cessation (3, 4). However, the optimal treatment duration for patients with melanoma without progression remains unknown (5), while the total duration of therapy with ICIs are determined arbitrary in treatment protocols, ranging from 2 years in some studies to treatment until progressive disease (PD) in others (1, 2, 6).

Careful consideration of treatment duration with ICIs lies in balancing response preservation after ICI cessation and toxicity avoidance, while reducing costs. Immune related adverse events (irAEs), and especially late-onset irAEs, remain a major issue, affecting quality of life of patients receiving ICI, which is particularly important in the context of complete response (CR) (7, 8). To answer this question, observational studies analyzed the impact of treatment discontinuation electively or due to AEs on relapse, but their results remain inconsistent (9–12). In addition, meta-analyses examining optimal duration of ICIs in solid tumors, including melanoma, did not report either a survival benefit of patients treated with fixed duration compared to treatment until disease progression, or explored factors influencing relapse risk comprehensively (13, 14).

Consequently, the main aim of our review was to systematically review all available evidence on disease relapse following ICI discontinuation for reasons other than progressive disease (PD), to evaluate the role of various factors, such as type of ICI or reason for discontinuation, in an extensive way, and to report post-relapse management of patients who experienced PD after ICI cessation.

Survival outcomes after ICI treatment cessation for reasons other than disease progression or relapse in patients with advanced melanoma remains an unanswered question in medical literature (53–55). Our analysis shows a 20.9% relapse rate after a mean of 14.26 months post cessation. In addition, lower rates were found in patients who stopped treatment electively compared to patients who stopped therapy due to TLTs. When patients had PD after therapy cessation, reintroducing ICI led to a substantial DCR, including CRs in approximately one third of cases.

The optimal duration of treatment with ICI remains a matter of great debate in the literature (5, 56). As far as we know, there is no clinical, biological or scientific rationale to continue therapy until PD or to stop after 2 years. Still, although RCTs employed different schedules for total treatment duration, a substantial number of patients who ceased treatment for reasons other than PD will not recur during follow-up (1, 2, 6, 57). According to our analysis, approximately four out of 5 patients who discontinue treatment will not recur and this rate is line with relapse rates reported in observational studies and RCTs. Patients included in KEYNOTE-001 and KEYNOTE-006 trials, ceased treatment for CR or after completing 2 years of pembrolizumab, resulting in a 9.2% and 26.2% relapse rates respectively (2, 4, 6). Similar results were drawn from multicenter retrospective registry-based studies (9, 11, 42). Those findings hold significant implications, especially in CR patients, as treatment cessation could be associated with lower rates of chronic or delayed irAEs and reduced financial toxicity from prolonged ICI treatment.

Several factors influencing relapse rates after therapy discontinuation have been studied, particularly in the lack of reliable predictive biomarkers for safe ICI cessation. Dose-response and exposure-response curves of anti-CTL4 anti-PD1 inhibitors provided useful insights regarding the effect of treatment duration on outcomes. Recently published long-term survival data from randomized phase III trials, highlighted that patient who completed ≥94 weeks of treatment in KEYNOTE-006 or remained progression free at 3 years in CheckMate-067, maintained durable clinical benefit from ICI treatment (3, 4). In concordance, pharmacokinetic studies of nivolumab have demonstrated that receptor occupancy can be achieved with low doses, which persist for a significantly longer duration than the antibody’s half-life at given (58, 59). This suggest that efficacy is maintained independently of dose and treatment duration, which was verified also by our results, where duration of ICI treatment was not a significant predictor of relapse or time to PD post cessation in meta-regression analysis. However, real world data of patients with CR or PR from the EUMelaReg registry proved that prolonged ICI treatment led to longer PFS, compared to early cessation (<6 months), and patients with PR were mainly benefited (60). On the contrary, Lodde et al., reported no difference in outcomes of patients who achieved early responses to ICI compared to late, and these were not associated with the duration on treatment (61).

Those findings underscore the importance of the depth of response to treatment rather than the duration of treatment, as a biomarker of long-term preservation of response post-cessation. As demonstrated in the final analysis of CheckMate 067, patients who achieved at least 80% reduction in tumor size experienced long-term survival regardless of treatment with anti-CTL4 and anti-PD1 or anti-PD1 monotherapy (3). Similar findings were evident from our meta-analysis, where patients with CR demonstrated the lowest relapse rates, while type of treatment used did not influence significantly the risk of relapse, in line with findings from Pala et al. (13) and observational studies (26). In addition, retrospective multicenter observational studies concluded to a lower risk of relapse in patients who achieved CR on primary treatment compared to patients who ceased therapy with BOR of PR/SD or due to TLTs (9, 10, 42). The above are in line with our results, and highlights, especially for patients on PR/SD, the importance of careful co-estimation of disease-related factors along with the response to treatment, when ICI cessation is considered.

In contrast to better survival outcomes in patients who ceased ICI due to CR, our analysis found a survival disadvantage in patients with TLTs, characterized by higher relapse rates and shorter time to PD. Results about the effect of irAEs on outcomes of melanoma patients remain contradictory (62–65). While observational studies reported a prolonged PFS in patients who developed irAEs, implicating the presence of a strong immunogenic reaction against the tumor, our meta-analysis, along with other studies, did not verify that benefit after discontinuation (34, 62, 65). From a pathophysiologic perspective, the durability of ICI response is based on the generation of tissue-resident memory (TRM) CD8-T- cells (13). However, the possible use of high dose of corticosteroids and immunosuppressants for the management of high grade irAEs, especially in the early phase of ICI administration (66), an cause inhibitory effect on TCR affinity and decrease the production of effective memory cells, which may justify the increased relapse rate post-cessation (67–69).

In our analysis the mean time to PD after ICI cessation was estimated to be 14.26 months, with studies in the literature reporting similar times from 3.9 (37) to 33.3 months (6). The estimation of that time is considered clinically relevant for two reasons. First, a numerically longer time to PD in patients who ceased for CR compared to TLTs, possibly corroborate the aforementioned pathophysiologic mechanisms associated with relapse post cessation. In concordance, disease-related factors, such as histologic subtype, could influence the time to PD, as shown by our quantitative synthesis, where mucosal melanomas demonstrated shorter time to PD compared to other histologies. This could be explained by the distinct biologic characteristics between subtypes, with mucosal lesions exhibiting more aggressive biologic course, absence of UV signature and lower mutational burden, which could explain the lower efficacy of ICIs in this subtype and the shorter time to PD compared to cutaneous lesions (70, 71). Moreover, accurate estimation of the interval between ICI cessation and PD is crucial for implementing effective surveillance strategies in these patients, especially considering the importance of early detection of recurrence in patients where consensus on the follow-up scheme has not been reached. Simultaneously, the emergence of imaging- (i.e. FDG PET/CT) and blood-based biomarkers (i.e. ctDNA), combined with the knowledge of time to relapse, could provide valuable tools not only for patient monitoring, but also for decision making regarding treatment continuation or intensification in high-risk patients (31, 32, 48).

Following progression after cessation, a second course of ICI could be considered a reasonable option, especially in the context of limited therapeutic choices in patients with BRAF-wild type melanomas. Our analysis showed a 57.9% DCR and 28.9% after ICI (of any type) rechallenge, while in the literature, re-introducing ICI after PD resulted in a various ORRs, ranged from 20% (39) to 40% (72) and more than 60% (73, 74) and influenced by factors, such as the time of rechallenge after PD or the concomitant use of other treatments, such as radiotherapy. Our analysis included only patients who ceased treatment without progression, which means that, by definition, patients with primary resistance to treatment were excluded. Gang et al. summarized the evidence regarding rechallenge strategies in patients with NSCLC, pinpointing differences in pathophysiologic mechanisms according to reason for discontinuation (75). Consistent results were evident for melanoma also, reinforcing the concept of T-cell revitalization after rechallenge. This could potentially improve the ability of immune cells to recognize and eliminate tumor cells that escape from memory-T-cells surveillance (76). Although results from our analyses are promising, supporting rechallenge with ICI after PD, more sophisticated analyses on factors affecting ORRs, such as reason for cessation (including further analyses for patients who discontinue treatment on CR or for TLTs), type and sequence of ICI in rechallenge and BOR at first course are important, but were out of the scope of our analysis.

Our meta-analysis has several strengths. We reported the pooled RRs after ICI cessation, DCR after ICI rechallenge and the estimation of time to PD post discontinuation, which have not been reported before. Some limitations need to be mentioned. There was a high heterogeneity among studies both in primary and secondary outcomes, which is mainly attributed to different definitions of reason for discontinuation, the type of ICI investigated in each study and the different treatment duration. To minimize this heterogeneity, we conducted several subgroup and sensitivity analyses. However, we didn’t see significant changes in our primary results. Another limitation of our study was that uveal melanoma was an exclusion criterion. There are distinct biologic characteristics between histologic subtypes and, because of that, most of clinical trials and observational studies published in the literature included only patients with cutaneous melanoma, with uveal melanoma being actively excluded, especially in clinical trials. For those reasons, we decided to focus on cutaneous melanoma only, aiming to synthesize the evidence for the most reported subtype and simultaneously aiming to avoid higher heterogeneity in our results.

Discontinuation of immunotherapy for reasons other than PD could be a reasonable option for patients diagnosed with stage IV melanoma, possibly mitigating the risk of chronic irAEs and the economic burden of prolonged ICI therapy. Time and risk of PD after therapy discontinuation seem to be affected mostly by disease-related factors. A second course of ICI at time of PD remains considerable. Future clinical trials supported by more real-world evidence may help to answer the question about the optimal duration of ICI treatment in this setting (77).