Real-world outcomes of patients receiving salvage therapies for immune checkpoint inhibitor-resistant Merkel cell carcinoma: a rationale for future clinical trials

Abstract

Background

Merkel cell carcinoma is an aggressive skin cancer that progresses to advanced/metastatic disease in ~40% of patients. First-line immune checkpoint inhibitors (ICI) that block the programmed cell death protein-1/programmed death-ligand 1 axis provide 3-year progression-free responses in only ~40% of patients. The relative efficacy of salvage therapies in this setting is unclear.

Methods

In a prospectively enrolled single-center cohort, 106 patients had disease progression during or shortly after ICI and received at least one local or systemic salvage therapy. Baseline disease characteristics, treatments, and outcomes data were collected. Patients were stratified by primary resistance (no response to initial ICI) or acquired resistance (loss of ICI response after initial benefit). Primary outcomes were progression-free survival (PFS) and disease-specific survival (DSS). Associations between salvage therapies and outcomes were evaluated using Cox models with time-varying covariates for treatments and adjustments for disease burden and ICI resistance type.

Results

In this cohort, 44 patients (42%) met criteria for primary resistance and 31 (29%) had acquired resistance. Median PFS from salvage initiation was more than double for patients with acquired versus primary resistance (9.5 vs 4.7 months; p=0.006). Median DSS was not reached for acquired resistance and 14.3 months for primary resistance (p=0.006). A minority of patients (n=14) survived ≥3 years after salvage initiation, typically following customized, multimodal salvage strategies. Among salvage regimens (ICI alone, ICI+radiation therapy (RT), chemotherapy, chemotherapy+ICI), only ICI+RT had a statistically significant association with improved DSS relative to ICI alone (after adjustment, including disease burden and ICI resistance type: adjusted HR 0.35, 95% CI 0.14 to 0.91).

Conclusions

Patients with acquired resistance receiving salvage therapy have improved survival compared with those with primary resistance. While the addition of radiation to ICI was clearly associated with improved DSS, there continues to be a major need for new approaches to address ICI-resistant disease. Nevertheless, a durable benefit in select patients is possible via sequential, individualized, multidisciplinary treatments. We anticipate these data will be relevant for the design of clinical trials for this challenging ICI-resistant setting.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• While first-line immune checkpoint inhibitors (ICI) targeting the programmed death-(ligand) 1 pathway lead to durable responses in ~40% of patients with advanced Merkel cell carcinoma, the remaining ~60% of patients ultimately experience primary or secondary disease progression.

WHAT THIS STUDY ADDS

• We observed significantly improved survival outcomes in patients with acquired versus primary resistance. A minority of patients (n=14) with sufficient follow-up data survived ≥3 years after salvage therapy initiation. Among the salvage therapy regimens, radiation when used with ICI in the salvage setting was associated with improved survival over immunotherapy alone.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• This study highlights the need for new salvage therapies, showcases the possibility of long-term survival in select patients, and supports the use of radiation with systemic therapies in the salvage setting when appropriate.

Introduction

Merkel cell carcinoma (MCC) is a rare and aggressive neuroendocrine skin cancer associated with exposure to extensive ultraviolet light, advanced age, and immunosuppression. There are approximately 3,200 new cases per year in the USA, and the incidence of MCC is rising.¹ Approximately 40% of patients with MCC experience disease recurrence and patients with advanced disease typically require systemic therapy.²

Prior to 2017, cytotoxic chemotherapy was the only option for treating advanced MCC (aMCC), which might consist of advanced locoregional or distant metastatic disease. While an objective response to chemotherapy occurred in about 60% of patients, durability was disappointing with a median progression-free survival (PFS) of only 90 days after initiation of therapy and only approximately 5% of patients having durable benefit for more than 1–2 years.³

After 2017, immune checkpoint inhibitor (ICI) therapy with anti-programmed death-(ligand) 1 (PD-(L)1) agents became the first-line treatment for aMCC.⁴ With an overall response rate (ORR) of approximately 60%, similar to chemotherapy, the chance of durable benefit from ICI was at least 10-fold higher, with ~80% of responders having disease control for more than 1–2 years.⁵ At 3 years, around 40% of all patients with MCC who receive ICI will not have experienced progression.⁵ Multiple US Food and Drug Administration (FDA) approvals arose from these data, with avelumab (anti-PD-L1 in 2017), pembrolizumab (anti-programmed cell death protein-1 (PD-1) in 2018), and retifanlimab (anti-PD-1 in 2023).⁶ Nivolumab (anti-PD-1) is also often used in the first-line setting.7,9 Despite this extensive progress, there is a remaining 60% of patients with aMCC that do not derive persistent benefit from ICI.⁵

For those with ICI-resistant disease, salvage therapies such as alternative ICIs not FDA-approved for MCC, chemotherapy, surgery, and radiation can be used in addition to clinical trials. In particular, the addition of ipilimumab, a cytotoxic T lymphocyte antigen-4 (CTLA-4) antibody, to PD-1 pathway blockade is a prominent salvage therapy with efficacy in 0–30% of patients by various reports.^{7 10 11} Limited data exist on outcomes following salvage therapy for ICI-resistant MCC, particularly regarding the significance of ICI resistance that was primary (no initial response) or acquired (loss of response after initial benefit). In a prior cohort of 16 patients with MCC and primary resistance to first-line pembrolizumab, 38% survived >12 months.⁵ Studying survival and salvage therapy outcomes is crucial for optimizing patient care and setting the stage for future clinical trials to improve outcomes for these patients. Here, we analyzed real-world outcomes among 106 patients with aMCC who progressed following first-line ICI therapy and received salvage therapy. These data establish survival outcomes between ICI-resistance types in MCC, reveal the relative efficacy of salvage approaches, and serve as a baseline comparator for future approaches and clinical trials.

Methods

Study design and patient eligibility

We queried a prospectively enrolled MCC observational database (data cut-off August 16, 2024). First-line ICI was defined as any anti-PD(L)1 agent used for the treatment of MCC. Eligible patients were those who had progressed during or after first-line ICI and received at least one salvage therapy for MCC. The study excluded patients who met any of the following criteria: never underwent ICI therapy, lack of disease progression on first-line ICI therapy, or absence of subsequent salvage treatments.

Disease characteristics

Patients were staged at diagnosis based on the American Joint Cancer Committee eighth edition Staging Manual for MCC.¹² Patients were considered immunosuppressed at diagnosis if they had chronic lymphocytic leukemia, an autoimmune disease or solid organ transplant requiring immunosuppressive medications, other hematologic malignancies, HIV, or AIDS. Responses to treatment were determined by clinician assessment and/or imaging based on Response Evaluation Criteria in Solid Tumors (RECIST) criteria for progressive disease (PD), stable disease (SD), partial response (PR), and complete response (CR).¹³ Tumor viral status was determined via immunohistochemistry and/or oncoprotein-antibody production (details in table 1 legend).¹⁴

Table 1. Baseline characteristics of patients with primary or acquired resistance to first-line immune checkpoint inhibitor therapy.

	Primary resistance (n=44)	Acquired resistance (n=31)	P value*
Age (mean (range))	69.1 (47–86)	70.0 (39–93)	0.744
Sex (n (%))			0.446
Female	18 (40.9)	10 (32.3)
Male	26 (59.1)	21 (67.7)
Race (n (%))
White	37 (100.0)	28 (100.0)
Unknown	7	3
Primary site (n (%))			0.834
Buttock	2 (4.5)	2 (6.5)
Head and neck	13 (29.5)	6 (19.4)
Lower limb	10 (22.7)	9 (29.0)
Trunk	4 (9.1)	2 (6.5)
Upper limb	6 (13.6)	3 (9.7)
Unknown primary	9 (20.5)	9 (29.0)
AJCC stage at diagnosis (n (%))			0.013
I	11 (25.0)	1 (3.2)
II	1 (2.3)	4 (12.9)
III	28 (63.6)	19 (61.3)
IV	4 (9.1)	7 (22.6)
Profound immunosuppression at diagnosis† (n (%))	11 (25.0)	7 (22.6)	0.809
MCPyV oncoprotein-antibody (AMERK) status‡ (n (%))			0.035
Yes	20 (46.5)	23 (76.7)
Non-producer	23 (53.5)	7 (23.3)
Unknown	1	1
MCPyV viral status§ (n (%))			0.072
Positive	21 (48.8)	23 (74.2)
Indeterminate	16 (37.2)	6 (19.4)
Negative	6 (14.0)	2 (6.5)
Unknown	1	0
Disease burden at initiation of salvage therapy
Extent of disease¶ (n (%))			0.258
Locoregional	5 (11.4)	7 (22.6)
Metastatic, 1 organ system only	16 (36.4)	13 (41.9)
Metastatic, ≥2 organ systems	23 (52.3)	11 (35.5)
Lymph node involvement (n (%))			0.455
No	13 (29.5)	13 (41.9)
Yes, 1 site only	5 (11.4)	4 (12.9)
Yes, 2 sites only	5 (11.4)	5 (16.1)
Yes, ≥3 sites	21 (47.7)	9 (29.0)

^*Comparing primary and acquired resistance groups using χ2 for categorical variables and the t-test for continuous variables.

^†Profound immunosuppression is defined as chronic lymphocytic leukemia, an autoimmune disease or solid organ transplant requiring immunosuppressive medications, other hematologic malignancies, or HIV/AIDS. Type of immunosuppression at diagnosis for all patients is detailed in online supplemental appendix A.

^‡Yes, indicates if AMERK test result was ever positive; non-producer indicates a negative result ≤3 months from original MCC diagnosis; unknown indicates either an individual that did not receive an AMERK test, or a negative AMERK result >3 months from original MCC diagnosis.

^§Prior to 2020, tumor virus status was determined as follows: if three tests (Ab3, qPCR, CM2B4) were performed, the best two of three results determined virus status. If only one test was performed, its result determined the status. However, if the patient was an oncoprotein-antibody producer, their MCC tumor was considered positive. From 2020 to present, tumors were positive if patients were an oncoprotein-antibody producer or CM2B4 positive. To call a tumor virus negative, both tests needed to be negative. If only one test was performed, that result determined status. Unknown signifies that neither AMERK nor CM2B4 tests were completed. Patients who do not fall into a clear definition are labeled as indeterminate.

^¶Organ systems were classified based on a prior publication from Lewis et al and is further detailed in online supplemental appendix B. Distant lymph node involvement was classified as one organ system.

AJCC, American Joint Committee on Cancer; MCC, Merkel cell carcinoma; MCPyV, Merkel cell polyomavirus.

Resistance type definitions

Each patient was classified into one of four resistance types based on the Society of Immunotherapy for Cancer (SITC) consensus recommendations.¹⁵ Briefly, primary resistance (pR) was defined as patients who received at least 6 weeks of ICI, had a best response of PD (or SD for <3 months followed by PD), and had PD either (1) while on ICI or, (2) within 12 weeks of ICI discontinuation. Acquired resistance (aR) was defined as patients who received at least 3 months of ICI, had SD for at least 3 months or a PR/CR, and had PD either (1) while on ICI or, (2) within 12 weeks after ICI discontinuation.⁴

The original guidance from the SITC suggested defining aR as occurring in patients who receive at least 6 months of ICI therapy and experienced 6 months of objective response; in pR, patients had to have an SD duration of <6 months.¹⁵ Updated guidelines from SITC in 2023 suggest a 3-month instead of a 6-month window for more aggressive malignancies.¹⁶ Therefore, these definitions were adjusted specifically for MCC to reflect the 2023 recommendations given the aggressive nature of MCC, its quick response to ICI, and to maintain sufficient patient numbers. Late progression was defined as patients who had a prior PR/CR on ICI and then developed PD more than 12 weeks after the discontinuation of ICI. Patients who otherwise did not fit within these three prior definitions were classified as having an unknown resistance type.

Disease burden at salvage regimen initiation

Disease burden was classified using the most recent available imaging (CT, positron emission tomography/CT, or MRI) reports prior to the start of salvage therapy. The number of distinct organ systems involved was recorded based on radiology reports.¹⁷ Locoregional disease extent was defined as tumor limited between the original primary site and the closest draining lymph node bed, or tumor that was distal to the primary site on the same extremity. Distant metastatic disease extent was defined as tumor that extended beyond locoregional involvement and was graded as involving 1 organ system or ≥2 organ systems. The number of involved distinct lymph node sites was recorded and categorized into 1, 2, or ≥3 to enhance the reproducibility of this parameter given variability in radiology reports. Since imaging reports are variable across radiologists, these two variables were selected to be included in the final model to maximize accuracy while acknowledging the limits of the source data quality.

Treatment data

Details of interventional treatments starting from initiation of first-line ICI to last follow-up date were captured. Characteristics of systemic treatments, including start and end dates and objective responses with corresponding dates were extracted. Systemic salvage therapies were categorized into ICI, chemotherapy, or other. ICI was subsequently categorized as anti-PD(L)1, anti-PD(L)1+CTLA-4 inhibitor, or anti-PD(L)1+other. Dates of local salvage radiation therapy (RT) and salvage surgery for MCC were also captured.

Descriptive details of treatments are provided for pR and aR patients that survived ≥3 years from the initiation of salvage therapy, as well as for those who received an anti-PD(L)1+CTLA-4 inhibitor.

Outcomes

Time 0 for all analyses was defined as the initiation date of the first salvage therapy. PFS was defined as the duration from initiation of first salvage until next PD or death from any cause. Disease-specific survival (DSS) was defined as the duration from initiation of first salvage until death from MCC, with deaths not attributed to MCC treated as a competing risk. Attribution of death to MCC was determined using clinical records from Seattle and collaborating medical teams based on the treating physician’s determination of the cause of death. The cause of death was classified as due to MCC, non-MCC, or unknown. ORR was defined as the fraction of patients that experienced an objective response (PR or CR) after initiation of first salvage before next PD or death divided by the total number of patients. Duration of response (DoR) was defined as the duration from first objective response after initiation of first salvage until next PD or death. Overall survival (OS) is defined as the duration from initiation of first salvage until death.

Statistical analysis

All statistical analyses were conducted using R V.4.4.1 (R Foundation for Statistical Computing, Vienna, Austria). Visualization of figure 5 was created using GraphPad Prism V.10.4.2 (GraphPad Software, Boston, Massachusetts). Distributions of baseline characteristics were compared between resistance types (primary vs acquired) using the χ² test for categorical variables and the t-test or Mann-Whitney U test for numeric variables. PFS, OS, and DoR probabilities were estimated using the Kaplan-Meier estimator. DSS probabilities were estimated using the cumulative incidence function estimator to account for competing risks. Median PFS, DoR, OS, and DSS were estimated as the time corresponding to a 50% probability on the Kaplan-Meier or cumulative incidence curve. The PFS, OS, and DoR distributions were compared between pR and aR groups using the log-rank test while the DSS distributions were compared between pR and aR groups using Gray’s test to account for competing risks.

Multivariable Cox proportional hazards (PH) models were used to compare the cause-specific hazards for each outcome between pR and aR groups while adjusting for age, sex, and disease burden, represented by extent of disease and number of distinct lymph node sites involved. The PH assumption was tested for each covariate in both models.¹⁸ As a sensitivity analysis, for any covariate with evidence of non-PH (p<0.05 in the PH test), another model was fit with a time interaction with that covariate added to assess the potential impact. Among these models, only the extent of disease in the DSS model showed some evidence of non-PH (p<0.05), but adding the time interaction had minimal impact on the model results. Viral status and AMERK status were not included in the final regression models due to missing data; separate analyses found that these variables were not associated with outcomes relating to salvage. Varying definitions of disease burden were also considered, with distinct organ systems and lymph nodes involved as the final model to maximize model stability while using reliable measures from clinical radiology reports to capture a broad range of disease burden. Associations were summarized using unadjusted HRs and adjusted HRs (aHRs), with corresponding 95% CIs and p values.

Cox PH models with time-varying covariates were used to evaluate the associations of different salvage treatments with PFS and DSS.¹⁹ Only patients who underwent a systemic salvage therapy (chemotherapy or ICI) were included, with the start of the first systemic therapy used as the entry time. To avoid immortal time bias, salvage therapies were modeled as time-varying covariates with four possible values: ICI alone, ICI and radiation, chemotherapy±radiation, and a combination of chemotherapy and ICI (sequential or concurrent)±radiation.²⁰ The chemotherapy and chemotherapy+ICI groups were not further separated by use of radiation due to low sample sizes. The value for each treatment covariate at any point in time was defined as by whether the patient had received the corresponding therapy or therapies by that time or earlier. Other therapy types, including surgery, were not included as additional covariates due to their low frequency in this cohort. Both adjusted and unadjusted models were created. The adjusted models included age, sex, ICI-resistance type (pR vs aR), and disease burden covariates as defined above. The PH assumption was tested as with the other multivariable Cox models. There was some evidence of non-PH with age in the DSS model (p<0.05), so a similar sensitivity analysis was performed by adding an age-time interaction to the model, which had minimal impact on the results. The primary and aR groups were also analyzed separately. However, due to sample size limitations, only unadjusted models were used for these stratified analyses.

Results

Among 1,848 patients with MCC in the Seattle MCC registry, 334 patients received ICI for MCC, 126 of whom experienced progression either during or after receiving first-line ICI. Of these 126 patients, 20 patients did not receive any salvage therapy. Among these 20, median time from progression date on or after first-line ICI therapy to death was 66 days (IQR: 34–117); 8 patients had pR, 2 had aR, and 10 had an unknown resistance type.

The remaining 106 patients underwent at least one salvage therapy and met study inclusion criteria (figure 1). Of these 106 patients, 44 (42%) had pR, 31 (29%) had aR, 22 (21%) had an unknown resistance type, and 9 (8%) had late progression following discontinuation of ICI. The primary analysis cohort consisted of the pR and aR groups (n=75). Descriptive analyses of the 31 patients in the late progression (n=9) and unknown resistance (n=22) types are summarized in the appendices as indicated.

Figure 1. Flow chart of study cohort.

Patient cohort characteristics

Clinical and disease characteristics of the primary analysis cohort are detailed in table 1, with comparisons between patients with pR and aR. Additional characteristics, including the late progression and unknown resistance types, are further detailed in online supplemental appendix A. Age at salvage therapy initiation, sex, race, initial primary site, and immunosuppression at diagnosis were not statistically different between pR and aR. Of patients with pR, 47% produced antibodies to the Merkel cell polyomavirus (MCPyV) compared with 77% in aR (p=0.035). Among patients with pR, 49% were MCPyV-positive while 74% of patients with aR were MCPyV-positive (p=0.072).

Median time between radiologic disease burden assessment via imaging and initiation of salvage therapy was 17 days (IQR: 12–33) and was similar between pR and aR (Mann-Whitney U test, p=0.23). At the time of salvage therapy initiation, 23/44 (52%) of patients with pR had metastatic disease involving ≥2 organ systems, compared with 11/31 (35%) for aR (p=0.258). Additionally, 21/44 (48%) of patients with pR had ≥3 sites of lymph node involvement, compared with 9/31 (29%) in patients with aR (p=0.455). Specific sites of metastasis are recorded in online supplemental appendix B. In both pR and aR, distant lymph nodes and distant skin/body wall were the two most common sites of metastasis. In pR, there was more bone (8/44, 18%) and liver (8/44, 18%) involvement compared with aR (2/31, 6% and 0/31, 0%), respectively.

Survival outcomes of primary and acquired resistance types

In aggregate across the pR and aR groups, 46 progressed after initiating salvage therapy and an additional 19 died of any cause without documented clinical or radiographic progression (65 total events of either progression or death). A total of 56 patients had died of any cause by the data cut-off, with MCC being the proximal cause of death in 41 of these patients.

Shorter survival in pR compared with aR was observed across all survival measures from initiation of first salvage therapy to next progression (figure 2). Median PFS was 9.5 months for aR and 4.7 months for pR (p=0.006, unadjusted HR 0.49 (95% CI 0.29 to 0.82), figure 2A); median PFS was 5.3 months for the combined pR+aR cohort (figure 2B). Median DSS was not reached for aR and 14.3 months for pR (p=0.006, unadjusted HR 0.37 (95% CI 0.18 to 0.73), figure 2C); median DSS was 22.9 months for the combined group (figure 2D). After adjustment for risk factors, including age, sex, and disease burden, PFS was still numerically better for aR than pR (aHR 0.61, 95% CI 0.35 to 1.09; p=0.097) (table 2). DSS remained significantly better for aR than pR (aHR 0.41, 95% CI 0.19 to 0.87; p=0.020). ORR was 43.2% (19/44) in pR and 64.5% (20/31) in aR. Among responders, median DoR was 5.7 months for aR and 3.1 months for pR (p=0.027, figure 2E); median DoR was 4.5 months for the aggregate pR+aR group (figure 2F). Median OS was 12.9 months (online supplemental appendix C). Survival rates at the 1-year, 2-year, and 3-year marks for these metrics are summarized in online supplemental appendix D.

Figure 2. Progression-free survival (A, B), disease-specific survival (C, D), and duration of response (E, F) of primary and acquired ICI-resistant MCC from initiation of first salvage therapy. (A) Progression-free survival for primary versus acquired ICI-resistant MCC. Median PFS was 9.5 months for acquired resistance, and 4.7 months for primary resistance (p=0.006). (B) Progression-free survival for primary and acquired ICI-resistant MCC. Median PFS was 5.3 months. (C) Disease-specific survival for primary versus acquired ICI-resistant MCC. Median DSS was not reached** for acquired resistance, and 14.3 months for primary resistance (p=0.006). **The median disease-specific survival time is not able to be estimated for the acquired resistance subgroup as the survival probability did not fall below 50% during the observation period. At the end of 5 years (60 months), the survival rate remained around 65%, indicating favorable survival outcomes. (D) Disease-specific survival for primary and acquired ICI-resistant MCC. Median DSS was 22.9 months. (E) Duration of response for primary versus acquired ICI-resistant MCC. Overall response rate was 20/31 (65%) for acquired resistance, and 19/44 (43%) for primary resistance. Median DoR was 5.7 months for acquired resistance and 3.1 months for primary resistance (p=0.027). (F) Duration of response for primary and acquired ICI-resistant MCC. Overall response rate was 39/75 (52%). Median DoR was 4.5 months. DoR, duration of response; DSS, disease-specific survival; ICC, immune checkpoint inhibitor; MCC, Merkel cell carcinoma; PFS, progression-free survival.

Table 2. Multivariable Cox proportional hazards model with adjustment based on ICI resistance type.

	PFS, aHR (95% CI) pR+aR (n=75)	DSS, aHR (95% CI) pR+aR (n=75)
Age, per 10-year increase	0.84 (0.65 to 1.90)	0.94 (0.69 to 1.28)
Sex
Female	Reference	Reference
Male	0.82 (0.47 to 1.44)	0.84 (0.42 to 1.69)
ICI resistance type
Primary	Reference	Reference
Acquired	0.61 (0.35 to 1.09)	0.41 (0.19 to 0.87)
Extent of disease
Locoregional	Reference	Reference
Metastatic, 1 organ system	0.96 (0.39 to 2.37)	1.63 (0.54 to 4.93)
Metastatic, ≥2 organ systems	1.48 (0.60 to 3.63)	3.24 (1.02 to 10.27)
Number of distinct lymph nodes (0–3+)	1.20 (0.95 to 1.51)	0.98 (0.74 to 1.31)

aHR, adjusted HR; aR, acquired resistance; DSS, disease-specific survival; ICI, immune checkpoint inhibitor; PFS, progression-free survival; pR, primary resistance.

Treatment courses by primary and acquired resistance types

Among pR and aR patients, median time from date of progression on first-line ICI to initiation of salvage therapy was 35 days (IQR: 15.5–62.5). There was no significant difference in time to salvage therapy initiation between pR and aR (Mann-Whitney U test, p=0.21). Swimmer plots of all treatments, responses, and outcomes for pR and aR are shown in figure 3. Patients with late progression or unknown resistance types are detailed in online supplemental appendix E. An interactive swimmer plot with all patients is available at https://dx.doi.org/10.17632/wsxgxcwg9h.1.

Figure 3. Salvage therapy treatment courses and outcomes of patients with Merkel cell carcinoma with primary or acquired resistance to first-line ICI therapy. Each primary resistance and acquired resistance subgroup is ordered by longest follow-up time or time to death. First-line ICI therapy is plotted prior to the time 0 mark. Time 0 represents the start of the first salvage therapy following progression on first-line ICI. Chemotherapy, ICI (anti-PD(L)1, anti-PD(L)1+CTLA-4, anti-PD(L)1+other), and “Other” treatments are mapped as bars based on start and end dates, while dates of surgery and radiation are mapped with their corresponding icons. Objective responses to each treatment are marked. CR, complete response; CTLA-4, cytotoxic T lymphocyte antigen-4; ICI, immune checkpoint inhibitor; PD, progressive disease; PD-L1, programmed death-(ligand) 1; PR, partial response; SD, stable disease.

Survival outcomes by type of salvage treatment

Salvage therapy regimens from initiation of first salvage therapy to next progression are detailed in online supplemental appendix F. For both pR and aR, ICI with radiation was the most common first salvage regimen received (pR 15/44, 34%; aR 14/31, 45%). Among patients with pR, chemotherapy was used more often (17/44, 39%) than for patients with aR (5/31, 16%). Among patients with aR, a single or combined ICI was more common for salvage (26/31, 84%) than for patients with pR (30/44, 68%). Among the 75 patients with either pR or aR, 7 received local salvage treatment without any systemic therapy (6 RT, 1 surgery).

Among these 75 patients, only 7 had surgery as any part of their salvage regimen. Of these, 2/7 (29%) had locoregional disease only, 1/7 (14%) had a craniotomy, and the remainder had excision of individual lesions.

A forest plot (figure 4) is used to depict outcomes following systemic salvage therapies after adjustment for risk factors. Unadjusted associations, with and without stratification by resistance type, are detailed in online supplemental appendix G.

Figure 4. Forest plot of time-varying Cox proportional hazards models for progression-free survival and disease-specific survival among patients who received chemotherapy, immune checkpoint inhibitor or both as part of the first salvage therapy. (A) Progression-free survival. (B) Disease-specific survival. n represents the number of patients in each group. Salvage therapy was included as a time-varying covariate in the model and in this case, n corresponds to the number of patients by the final salvage therapy category reached during the first salvage therapy period (between initiation of first salvage therapy and date of next progression/death or last follow-up).

For PFS of systemic salvage therapies with salvage ICI alone as the reference group, ICI+RT had an aHR of 0.66 (95% CI 0.28 to 1.57; p=0.346), chemotherapy±RT had an aHR of 1.03 (95% CI 0.39 to 2.73; p=0.956), and a combination of chemotherapy+ICI±RT demonstrated an aHR of 2.37 (95% CI 0.94 to 5.98; p=0.067).

For DSS, ICI+RT had an aHR of 0.35 (95% 0.14 to 0.91; p=0.030), chemotherapy±RT had an aHR of 1.24 (95% CI 0.48 to 3.19; p=0.660) and chemotherapy+ICI±RT had an aHR of 1.20 (95% CI 0.38 to 3.80; p=0.751) compared with salvage ICI only. Compared with patients with locoregional disease extent only, DSS was poorer for those with metastatic disease involving 1 organ system (aHR 3.57, 95% CI 1.24 to 10.28; p=0.018) and for metastatic disease involving ≥2 organ systems (aHR 7.71, 95% CI 2.36 to 25.13; p=0.001).

Among the 29 patients who received ICI+RT, 15 (52%) had metastatic disease involving 1 organ system, 8 (28%) had metastatic disease involving ≥2 organ systems, and 6 (21%) had only locoregional disease. Of these patients, 20 (69%) received a hypofractionated course of radiation, 4 (14%) received a conventional course of radiation, and 5 (17%) had an unknown radiation dosage. Disease burden stratified by salvage therapy regimen is summarized in figure 5.

Figure 5. Disease burden at time of indicated first salvage regimen. (A) Number of involved metastatic sites. (B) Number of involved lymph node beds. (C) Disease burden stratified by first salvage regimen.

The addition of anti-CTLA-4 was the first salvage approach attempted in 24 patients. An objective response was observed in 27% of these patients. Fewer patients with pR (only 8%) benefited from the addition of anti-CTLA-4 compared with aR (44% with an objective response). Online supplemental appendix H depicts a per-patient summary of best objective response, response duration, survival time, and first salvage treatment course.

Patients surviving ≥3 years since salvage initiation

For patients with sufficient follow-up who survived ≥3 years after initiation of salvage therapy (n=14), a per-patient view of specific salvage therapies received is available in online supplemental appendix I. The patient with the longest survival times for each ICI resistance type is highlighted.

Patient 1 (primary resistance, 107+ months survival since salvage initiation)

Patient 1 was in their late 50s at the time of salvage initiation. They initially presented with seropositive pathologic stage IV MCC involving the bilateral cervical neck and retroperitoneal lymph nodes (unknown primary) that was treated with six cycles of carboplatin/etoposide and resulted in a CR. Nine months after chemotherapy, the patient had recurrent disease in the left neck and retroperitoneal lymph nodes for which they were started on avelumab for 2 months and experienced PD. They subsequently received conventional radiation to both lymph node sites, along with 5 months of pembrolizumab, stopping treatment after experiencing a PR. After disease progression was noted, the patient was started on ipilimumab/nivolumab for 2 months with PD, followed by nivolumab maintenance for 43 months with PR. They stopped nivolumab due to SD involving the left axillary and right external iliac nodes only, out of concern for eventual possible side effects. Approximately 6 months later, the patient experienced PD and was restarted on nivolumab with a PR to the same sites of prior involvement. At the time of data cut-off, the patient continued nivolumab with sustained improvement in the prior involved lymph nodes.

Patient 45 (acquired resistance, 80+ months survival since salvage initiation)

Patient 45 was in their 70s at the time of salvage initiation. They initially presented with seropositive pathologic stage IIIB MCC involving the left buttock and left inguinal/external iliac lymph nodes. They were treated with surgery, lymph node dissection, and adjuvant radiation to both the primary and lymph node sites. They eventually developed recurrent distant metastatic MCC on the left thigh skin and multiple lymph node sites. The patient received nivolumab and experienced a PR before eventual progression of portacaval and left iliac adenopathy after 30 months on PD-1. Ipilimumab was added to nivolumab for 6 months with PD of the portacaval and new adrenal lesions. They received 1 fraction of 8 Gray radiation to a left para-aortic lymph node while on ipilimumab/nivolumab. The patient was then started on avelumab, during which they also received 1 fraction of 8 Gray to peripancreatic and left adrenal masses. At the time of data cut-off, the patient had received 54 months of avelumab so far with a PR (stable adrenal metastasis).

Discussion

For patients who have ICI-resistant MCC, there is currently no consensus regarding how to select among several major salvage therapy categories. Given the typical rapid progression of MCC, the prompt choice of a first salvage regimen is important for optimizing outcomes. In this cohort, outcomes were markedly better among patients who had acquired resistance rather than primary resistance. Despite extensive adjustment for disease burden, the addition of radiation to ICI was associated with significantly longer survival than other salvage approaches. It is hoped that this analysis will help guide management of ICI-resistant MCC and also highlight the need for additional salvage approaches.

Outcomes following salvage therapies in ICI-resistant MCC have been previously explored.511 21,24 In a previous Australian/European study conducted by Mo et al, no significant differences in survival were found between pR, aR, and late progression (≥3 months after discontinuation of ICI).²² However, Mo et al only included patients who received a subsequent systemic therapy, excluding local salvage modalities such as radiation and surgery. Their study described a series of 35 patients (23 pR, 12 aR), as compared with the present cohort of 75 cases (44 pR, 31 aR). Their finding of no difference between pR and aR may have been affected by their inclusion of patients with late progression, many of whom are still sensitive to ICI on rechallenge. For this reason, patients with late progression are not included in the survival analysis of the present cohort. Systemic salvage therapy usage patterns were similar between the present cohort and Mo et al. Both studies found that chemotherapy was more commonly used in pR than aR, and that continued ICI (sometimes with addition of anti-CTLA-4) was the most common systemic salvage approach for aR.

The addition of ipilimumab, a CTLA-4 antibody, to an anti-PD(L)1 agent has been investigated in the salvage setting. In a review of all published cases, this salvage regimen had a 31% chance of clinical benefit, consistent with findings in the present cohort (ORR 27%).¹⁰ This regimen is often thought of as the default salvage option for ICI-resistant MCC. However, a lack of benefit in the majority of patients and the risk of significant toxicity are important considerations for this regimen.

Our findings indicate that for patients with ICI-resistant MCC, those with aR have greatly improved survival compared with patients with pR. After controlling for demographic risk factors and proxies for disease burden, patients with acquired (vs primary) resistance had a 40% lower hazard of progression and a 60% lower in hazard of death from MCC following salvage. This is similar to survival patterns described for melanoma in which patients with aR have about half the risk of post-progression death relative to patients with pR.²⁵ It is hoped that these outcomes data will help clinicians provide more accurate prognostication for patients with ICI-resistant MCC. The aggregate survival metrics may be used as a reference point for the design of future clinical trials.

Regarding chemotherapy in the present cohort, whether given alone or together with ICI, there was no clear evidence of survival benefit relative to salvage ICI alone. Specifically, the combination of chemotherapy+ICI had numerically worse PFS compared with salvage ICI alone. Indeed, chemotherapy has been previously associated with decreased response to subsequent ICI.²⁶ However, it is possible that poorer outcomes for patients who received chemotherapy were due to baseline disease that was higher in burden and/or more aggressive, although we did adjust for disease burden as feasible. In our practice, chemotherapy in the setting of ICI-resistant MCC is typically used as a means to debulk a high amount of disease burden or for palliation, with the intent of transitioning to ICI quickly given the short-lived efficacy of chemotherapy.²⁷

Although this study investigated salvage for MCC in a heterogeneous, retrospective, real-world setting, we identified a potential synergy between radiation and immunotherapy associated with improved DSS, amidst careful adjustment. There was a similar trend for PFS among patients who received ICI+RT. Previous cases have described survival benefit from adding radiation while continuing immunotherapy.^{23 24} An important possibility for why ICI+RT patients had better outcomes could be that they had lower disease burden at the time of salvage. We therefore adjusted for the number of involved organ systems, number of lymph node sites, and the type of ICI resistance across the salvage groups. Specifically, as shown in figure 5, 80% of ICI+RT patients had advanced disease involving at least one metastatic site, and the proportion of patients with advanced disease was qualitatively similar across the four salvage therapy groups. Although we adjusted for age, sex, ICI resistance, and number of organ systems and nodal basins involved, these factors are unlikely to fully account for differences in disease aggressiveness between patients receiving ICI+RT versus other salvage modalities. As such, sources of selection bias we could not account for may have contributed to the more favorable outcomes observed in the ICI+RT group.

With the introduction of hypofractionated radiation in MCC, this accessible and lower-risk approach could be used in the salvage setting to provide local tumor control, reduce disease burden, and potentially stimulate the immune system.²⁸ Some data suggest that irradiating all clinically evident disease may lead to better outcomes.²⁹ Unfortunately, the limited availability of detailed radiation records in the present cohort precluded reliably characterizing the extent of radiation that each patient received. Nevertheless, these data suggest that radiation in combination with ICI should be considered for salvage when feasible.

Surgery was not included in our salvage analyses given the limited number of cases in which this approach was used (7 of 75) and restriction to cases with very low disease burden.

The fact that 14 of the 75 patients with sufficient follow-up (~20%) in the present cohort lived ≥3 years following salvage demonstrates that long-term survival is possible after developing ICI-refractory disease. These patients all received a flexible, multimodal approach (summarized in online supplemental appendix I) for the management of their ICI-resistant aMCC. As illustrated by the two patients whose courses are summarized in the results section, durable responses did not arise via a single salvage therapy, but rather through thoughtful sequencing and combination of multiple systemic and local modalities. These cases demonstrate that the ongoing use of immunotherapy — even after experiencing progression — combined with appropriate approaches to debulk tumor can provide meaningful clinical benefit. Radiation can serve as a bridge to further systemic therapy, including a re-challenge with immunotherapy or clinical trial enrollment.

Limitations of study

There are important limitations to this study. Since these data were collected from a single-center observational database, they are subject to selection bias. Some aspects of this cohort may not be representative of the general US population with MCC, given that some patients traveled long distances to a referral center which may skew the population towards a higher socioeconomic status compared with the national average.³⁰ Additionally, as a high-volume referral center that has ready access to highly specialized care, generalizing results beyond this setting should be done with caution. Such high-volume centers have been associated with improved survival outcomes in national database studies.³¹

Despite adjusting for disease burden, additional treatment allocation biases that we were unable to adjust for are likely to have affected results. Such factors include aggressiveness of disease, tumor metastasis location, comorbidities/functional status, and treatment preference by each patient. The number of progressing lesions was a characteristic that could have been useful in addressing disease aggressiveness, but these data could not be accurately captured due to lack of sufficient detail in clinical radiology reports. These issues are likely to have been confounding factors in the choice of salvage therapy.

Since this was a real-world cohort, the frequency of disease surveillance was not uniform across patients. While response assessments in this cohort were based on RECIST criteria applied to available radiographic reports, formal radiographic RECIST reporting may sometimes have resulted in different assessments. We attempted to mitigate this by reporting on regions of disease involvement rather than individual tumor measurements.

Exploratory biomarker data, such as PD-L1 expression and tumor mutational burden, were not available; these markers may offer potential explanations for the observed survival differences between primary and aR.

Conclusion

In summary, while the introduction of ICI has made significant strides in improving survival outcomes for advanced MCC, the 60% of patients who ultimately experience disease progression on an ICI typically have disappointing outcomes, regardless of salvage approach. Still, approximately 20% of patients experience long-term survival with a supportive clinical team that facilitates multidisciplinary salvage approaches. While the addition of radiation to ICI was the only salvage approach clearly associated with improved survival outcomes, its applicability may be limited by disease extent. The strikingly better survival of patients with acquired rather than primary resistance to first-line ICI suggests important differences in their underlying biological mechanisms. As clinical trials are designed to address this population, outcomes should be stratified by ICI resistance type and compared with current salvage approaches. These findings clearly highlight the need to develop additional approaches for this population and to establish benchmark survival metrics for new therapies to surpass.