Site-specific response patterns, pseudoprogression, and acquired resistance in patients with melanoma treated with ipilimumab combined with anti–PD-1 therapy
Abstract
Background
Patients with metastatic melanoma have variable responses to combination ipilimumab and nivolumab. The objectives of this study were to examine the patterns of response and survival in patients treated with combination ipilimumab and anti–PD-1 therapy (IPI + PD1) and to explore the nature of pseudoprogression and acquired resistance.
Methods
Patients with metastatic melanoma who received treatment with first-line IPI + PD1 had all metastases ≥5 mm measured on computed tomography/magnetic resonance imaging studies. The lesional response rate (LRR) and the overall response rate (ORR) were determined according to Response Evaluation Criteria in Solid Tumors, version 1.1.
Results
In total, 140 patients who had 833 metastases were studied. The ORR and the overall complete response (CR) rate decreased as tumor burden or the number of metastases increased. Metastases that had a CR (49%) were smaller than metastases without a CR (median, 13 vs 17 mm; P < .0001). Soft-tissue and lung metastases had the highest LRR (79% and 77%, respectively), whereas liver metastases had the lowest (46%). In multivariate analysis, patients with lung metastases had superior ORR (odds ratio [OR], 2.75; P = .02) and progression-free survival (hazard ratio [HR], 0.46; P = .02), whereas those with liver metastases had inferior ORR (OR, 0.33; P = .02), progression-free survival (HR, 4.03; P < .01), and overall survival (HR, 3.17; P = .01). Pseudoprogression occurred in one-third of patients who had progressive disease as their best response, with an overall survival that was comparable to that of patients without disease progression. Acquired resistance occurred in 12% of responders after a median of 9.6 months, with an overall survival rate of 83% at 1 year from progression.
Conclusions
Metastases in different anatomical locations display distinct response patterns and also are associated with overall response and survival with combination immunotherapy. Specific sites of disease may hold unique mechanisms of resistance and should allow for more personalized treatment.
Introduction
Great advances have been made in the treatment of melanoma in recent years, with checkpoint immunotherapies providing durable responses and prolonged survival in those with metastatic melanoma.1-3 The anti–CTLA-4 antibody (ipilimumab) was the first drug to show a survival advantage in metastatic melanoma, with durable activity in a small subset of patients.1, 4 Subsequently, anti–PD-1 antibodies (nivolumab and pembrolizumab) were shown to be more active and less toxic than ipilimumab; and, most recently, combination therapy with ipilimumab and nivolumab has demonstrated the highest efficacy, albeit at the cost of significant toxicity.2, 5, 6
Although immunotherapy is a vast improvement over previous therapies, still only one-half of patients respond up-front to the most active treatment—combination ipilimumab plus anti–PD-1—and many who initially respond subsequently progress.7 We previously have described the clinical heterogeneity of response and resistance with both targeted therapy8, 9 and anti–PD-1 monotherapy,10 have shown how the degree of heterogeneity affects survival (and thus may be a potential biomarker), and have demonstrated how this clinical heterogeneity reflects heterogeneity on a genomic and transcriptomic level.11 Whether such clinical heterogeneity exists with combination immunotherapy is unknown, nor is it known whether specific patterns of response occur and whether these impact survival.
In the current clinical study, we sought to examine associations between the site and size of individual metastases with patterns of response, progression, and survival in a large cohort of patients who received treatment with ipilimumab combined with anti–PD-1 therapy (nivolumab or pembrolizumab) by examining all metastases in all patients at baseline and throughout treatment. The nature and management of pseudoprogression and acquired resistance were explored in detail.
Materials and Methods
Patients and Treatment
One hundred forty consecutive patients who were treated at Melanoma Institute Australia between December 2013 and March 2018 with combination anti–CTLA-4 (ipilimumab) and anti–PD-1 (nivolumab or pembrolizumab) antibodies in the first-line setting for stage IV melanoma were included. Patients were treated with different combination doses and schedules as part of a clinical trial (Phase 3 Study of Nivolumab or Nivolumab Plus Ipilimumab Versus Ipilimumab Alone in Previously Untreated Advanced Melanoma [CheckMate-0675]; Combination nivolumab and ipilimumab or nivolumab alone in melanoma brain metastases: a multicentre randomised phase 2 study [ABC – Anti-PD1 Brain Collaboration Study]12; Nivolumab Combined With Ipilimumab Followed by Nivolumab Monotherapy as First-Line Treatment for Patients With Advanced Melanoma [CheckMate-40113]; A Study of Two Different Dose Combinations of Nivolumab in Combination With Ipilimumab in Subjects With Previously Untreated, Unresectable, or Metastatic Melanoma [CheckMate-51114]; A Safety and Efficacy Study of Multiple Administration Regimens for Nivolumab Plus Ipilimumab in Subjects With Melanoma [CheckMate-74215]; and Safety and Tolerability of Pembrolizumab + Pegylated Interferon Alfa-2b and Pembrolizumab + Ipilimumab in Participants With Advanced Melanoma or Renal Cell Carcinoma [KEYNOTE 02916]) or were treated off trial with ipilimumab and nivolumab according to the standard, approved schedule. This project had institutional ethics review board approval, and written informed consent was obtained from each patient. Patient demographics and disease characteristics at the time of starting treatment were collected.
Disease Assessment
Computed tomography of the chest, abdomen, and pelvis (3-mm slices) and magnetic resonance imaging of the brain (1-mm slices) were obtained at baseline (before starting treatment), then every 12 weeks or more frequently, according to the specific clinical trial protocol. All metastases that measured ≥5 mm in the long axis (brain metastases ≥2 mm in the long axis, lymph node [LN] metastases ≥15 mm in the short axis) on computed tomography or magnetic resonance imaging before and during treatment were measured.
Tumor burden was defined as the sum of the long axis for all non-LN metastases plus the short axis of all LN metastases measured. Overall response was determined based on Response Evaluation Criteria in Solid Tumors, version 1.1, but included all measured lesions. In addition, response was recorded for each individually measured metastasis (lesional response) at each time point, and the best lesional response was classified as a complete response (CR) (disappearance or reduction to <10 mm in the short axis for an LN metastasis), a partial response (PR) (>30% reduction), stable disease (SD) (neither a CR, PR, nor progressive disease [PD]), or PD (>20% growth).
Similar to a previous study,10 we defined a homogeneous response as a decrease ≥30% in all individual lesions, whereas a heterogeneous response was defined as a decrease ≥30% in the overall tumor burden (the sum of the long axis for all non-LN metastases plus the short axis of all LN metastases measured), but not in all individual lesions, using the scan with the best overall response (see Supporting Table 1). Similarly, homogeneous progression was defined as an increase ≥20% in the long axis of all lesions with or without new lesions, whereas heterogeneous progression was defined as an increase ≥20% in the sum of the long axis of all lesions, but not of all individual lesions, or new lesions that emerged while others decreased in size. Finally, early progression (primary resistance) was defined as PD at the first scan, whereas late progression (acquired resistance) was defined as PD after an initial response (PR or CR) or SD for at least 6 months.
Statistical Analysis
Descriptive analysis was used to describe patients and disease characteristics, including medians and ranges for continuous variables and proportions for categorical variables. A nonparametric test (Mann-Whitney U test) was used to compare continuous variables between 2 independent groups. Univariate and multivariate logistic regression analyses were used to identify factors associated with the overall response rate (ORR), defined as the sum of CRs and PRs. Overall survival (OS) and progression-free survival (PFS) were calculated from the date of commencement of treatment to the date of an event (either death [OS] or progression [PFS]). Patients without a clinical event were censored at the last follow-up date. Univariate and multivariate time-to-event analyses (OS and PFS) were conducted using the Cox proportional hazards method. All statistical analyses were performed using GraphPad PRISM (Prism 8.0.2; GraphPad Software) and the Statistical Analysis System (SAS version 9.4; SAS Institute).
Results
Patient Characteristics
One hundred forty patients with metastatic melanoma who received treatment with the combination of anti–CTLA-4 (ipilimumab) and anti–PD-1 (nivolumab or pembrolizumab) were studied (Table 1). This was a typical metastatic melanoma cohort; the median age was 63 years, 66% were male, 29% had an Eastern Cooperative Oncology Group (ECOG) performance status ≥1, 41% had a BRAF mutation, 29% had elevated lactate dehydrogenase (LDH) levels; and, according to the American Joint Committee on Cancer, Eighth Edition, 39% had M1c disease (visceral metastases), and 22% had M1d disease (brain metastases). In all 140 patients, 833 metastases were followed from baseline, with a median of 5 metastases per patient (range, 1-31 metastases per patient). The median diameter per metastasis was 15 mm (range, 2-101 mm), and the median sum of diameters per patient was 72 mm (range, 9-531 mm). The most frequent site of metastases was lung (69%) followed by LN (44%), and the adrenal glands (9%) and spleen (8%) were the least frequent sites of metastases. Of note, liver metastases were found in 28% of patients, and brain metastases in were found in 19% (in 4 patients, the brain metastases were resected before starting treatment). Two patients had pleural effusions, 1 had leptomeningeal disease, and 34 patients had bone metastases (without a soft-tissue component). The patients with pleural effusions had lung metastases, and the patient with leptomeningeal disease had concomitant brain metastases. The majority of patients received treatment as part of a clinical trial (89%), and the standard dose for the combination of ipilimumab (3 mg/kg every 3 weeks) plus nivolumab (1 mg/kg every 3 weeks) was the most commonly used (45%), followed by ipilimumab (1 mg/kg every 3 weeks) plus pembrolizumab (2 mg/kg every 3 weeks) in 29% of patients. The median follow-up was 16.1 months (95% CI, 12.1-17.9 months), and the 2-year PFS and OS rates were 60% and 82%, respectively (see Supporting Fig. 1).
| Characteristic | No. of Patients (%) |
|---|---|
| Total no. | 140 (100) |
| Age: Median [range], y | 63 [27-80] |
| Sex: Male | 93 (66) |
| ECOG PS: 1 | 41 (29) |
| Mutational status | |
| BRAF mutant WT | 58 (41) |
| NRAS mutant | 19 (14) |
| BRAF/NRAS WT | 63 (45) |
| LDHa | |
| Normal | 100 (71) |
| Elevated | 40 (29) |
| AJCC v8a | |
| Stage IIIC | 5 (4) |
| M1a | 12 (9) |
| M1b | 37 (26) |
| M1c | 55 (39) |
| M1d | 31 (22) |
| Site of metastases | |
| Lung | 96 (69) |
| Lymph node | 61 (44) |
| Subcutaneous | 40 (29) |
| Liver | 39 (28) |
| Bone | 34 (24) |
| Brainb | 27 (19) |
| Gastrointestinal | 26 (19) |
| Adrenal | 12 (9) |
| Spleen | 11 (8) |
| Othersc | 6 (4) |
| Treatment | |
| IPI 3 mg/kg q3w + NIVO 1 mg/kg q3w | 63 (45) |
| IPI 1 mg/kg q3w + PEMBRO 2 mg/kg q3w | 40 (29) |
| IPI 100 mg q12w + PEMBRO 200 mg q3w | 13 (9) |
| IPI 50 mg q6w + PEMBRO 200 mg q3w | 11 (8) |
| Unknownd | 13 (9) |
- Abbreviations: AJCC v8, American Joint Committee on Cancer, version 8 (anatomical staging); ECOG PS, Eastern Cooperative Oncology Group performance status; IPI, ipilimumab; LDH, lactate dehydrogenase; NIVO, nivolumab; PEMBRO, pembrolizumab; q3w, every 3 weeks; WT, wild type.
- a Staging excluded LDH.
- b Four patients had resected brain metastases.
- c These include 3 patients with kidney metastases and 3 patients with nonmeasurable metastases (2 with pleural effusions and 1 with leptomeningeal disease).
- d CheckMate-511 trial (n = 13), patients were blinded and randomized to receive either IPI (3 mg/kg) plus NIVO (1 mg/kg) or IPI (1 mg/kg) plus NIVO (3 mg/kg), followed by a maintenance phase of NIVO alone in both arms.
Overall Patient Response
When all lesions were included as targets, the ORR was 66% (93 of 140 patients), and 31 patients (22%) had a CR. Approximately one-third of patients (n = 39; 28%) had PD as their best response, whereas 8 patients (6%) had SD, including 5 patients who had a mixed response (Fig. 1A). The median change in the sum total of lesions from baseline was −72%, and 76% of patients had some degree of tumor regression (Fig. 1B). Patients who achieved an overall CR (n = 31) had significantly smaller tumor burden at baseline than those who did not have an overall CR (n = 109; median sum of diameter, 43 mm and 88 mm, respectively; P < .0001) (Fig. 1C) and had significantly fewer lesions (median number of metastases, 3 and 5 metastases, respectively; P < .0001) (Fig. 1D).
Individual Metastasis Response
On a lesional level, 71% (n = 591) of metastases had a response (n = 404 metastases; CR, 49%), whereas 15% (n = 129 metastases) had SD, and 14% (n = 113 metastases) progressed (Fig. 2A). Similar to the overall response findings, metastases that achieved a CR were significantly smaller than non-CR metastases (n = 429), with a median diameter of 13 mm and 17 mm, respectively (P < .0001) (Fig. 2B). Brain metastases were significantly smaller than metastases in all other sites of disease (P < .0001), whereas lung and subcutaneous metastases were smaller than LN and liver metastases (P < .0001) (Fig. 2C). The best change in tumor size from baseline varied according to the site of disease; the greatest regression was observed in subcutaneous (median −100%; range −100% to 138%), gastrointestinal (median −100%; range −100% to 117%), soft-tissue (median −90%; range −100% to 364%), and lung (median −77%; range −100% to 465%) sites. Liver metastases had the lowest tumor regression (median −3%; range −100% to 1713%) (Fig. 2D). Similarly, the likelihood of a response (CR or PR) also differed by site of metastasis; the highest lesional response rates were observed in soft-tissue (79% response), lung (77% response), and gastrointestinal (77% response) sites, and the lowest was observed in liver sites (46%) (Fig. 2E).
Sites of Disease Impact, Overall Response, and Survival
Observing that metastasis size and site appeared to influence lesional response to therapy, we first explored how site, along with other characteristics (sex, age, ECOG performance status, mutation status, LDH, and tumor burden), affected overall response (ORR) and survival (PFS and OS).
On univariate analysis, the presence of lung metastases (odds ratio [OR], 2.4; P = .0177) was associated with a higher ORR, whereas elevated LDH (OR, 0.27; P = .0008) and the presence of liver (OR, 0.25; P = .0006), spleen (OR, 0.26; P = .0380), or bone metastases (OR, 0.33; P = .0072) were associated with a lower ORR (see Supporting Table 2).
The presence of lung metastases (HR, 0.49; P = .0109) also was associated with longer PFS; whereas an ECOG performance status of 1 (HR, 1.79; P = .0431), elevated LDH (HR, 4.06; P < .0001), tumor burden greater than or equal to the median (HR, 2.06; P = .0137), and the presence of liver (HR, 3.81; P < .0001), spleen (HR, 2.97; P = .0077), or bone metastases (HR, 2.04; P = .0167) were associated with shorter PFS (univariate analysis) (Fig. 3A-C; see Supporting Table 3).
An ECOG performance status of 1 (HR, 3.75; P = .0028), elevated LDH (HR, 4.65; P = .0008), tumor burden greater than or equal to the median (HR, 2.66; P = .0433), and the presence of liver (HR, 6.11; P < .0001), brain (HR, 3.99; P = .0017), and spleen metastases (HR, 4.76; P = .0023) were associated with shorter OS (univariate analysis) (Fig. 3D; see Supporting Table 4).
Multivariable analysis including all significant findings on univariate analysis confirmed that the presence of lung metastases was associated with a higher ORR (OR, 2.68; P = .0309) and longer PFS (HR, 0.46; P = .0212). Elevated LDH and the presence of liver metastases were associated with a lower ORR (LDH: OR, 0.32; P = .0172; liver metastases: OR, 0.33; P = .0199) and shorter PFS (LDH: HR, 2.99; P = .0013; liver metastases: HR, 3.17; P = .0010). Interestingly, the presence of liver metastases was the only factor significantly associated with OS (HR, 4.03; P = .0143) (Table 2). Therefore, patients who had liver metastases had a lower ORR and shorter PFS and OS compared with those who had no liver metastases (ORR, 43.6% vs 76.8%, respectively; 6-month PFS rate, 43% vs 80%, respectively; 1-year OS rate, 65% vs 94%, respectively).
| Variable | ORR | PFS | OS | |||
|---|---|---|---|---|---|---|
| OR (95% CI) | P | HR (95% CI) | P | HR (95% CI) | P | |
| LDH | ||||||
| Normal | 1.00 | .0172 | 1.00 | .0013 | ||
| Elevated | 0.32 (0.12-0.82) | 2.99 (1.53-5.84) | ||||
| Lung metastases | ||||||
| No | 1.00 | .0309 | 1.00 | .0212 | ||
| Yes | 2.68 (1.09-6.55) | 0.46 (0.24-0.89) | ||||
| Liver metastases | ||||||
| No | 1.00 | .0199 | 1.00 | .0010 | 1.00 | .0143 |
| Yes | 0.33 (0.13-0.84) | 3.17 (1.59-6.29) | 4.03 (1.32-12.29) | |||
- Abbreviations: LDH, lactate dehydrogenase; OR, odds ratio; ORR, overall response rate; OS, overall survival; PFS, progression-free survival.
Patients with lung metastases and without liver metastases (lung+ liver−) had the longest PFS, followed by patients without lung or liver metastases (lung− liver−), and patients with both lung and liver metastases (lung+ liver+), whereas patients who had liver metastases without lung metastases (lung− liver+) had the shortest PFS, with 6-month PFS rates of 88%, 62%, 52%, and 22%, respectively (see Supporting Fig. 2). Logistic regression analysis for patients with lung or liver metastases, including the median size of lesions, showed no significant effect of size on the ORR (P > .05), suggesting that the site associations were independent of size (see Supporting Table 5). Therefore, patients with liver metastases have worse clinical outcomes independent of the size of their lesions.
Homogeneous and Heterogeneous Response and Progression
Within the cohort of 93 patients who had an overall response, 73 (78%) had homogeneous responses (see Supporting Fig. 3A), and 20 (22%) had heterogeneous responses (see Supporting Fig. 3B). There was no difference in PFS between these 2 groups; however, patients, who had homogeneous responses had longer OS (2-year rate, 100% vs 81%; P = .001) (Fig. 4A,B).
In patients who had PD as their best response, a minority (31%) had homogeneous progression (see Supporting Fig. 3C), whereas most (69%) had heterogeneous progression (see Supporting Fig. 3D). Although PFS was expectedly similar between these groups, interestingly, there was also no difference in OS (Fig. 4C,D).
Brain Metastases
Of the 27 patients who had brain metastases, 1 had a CR, 15 had PRs, 2 had SD, and 9 had PD in the brain. These responses were concordant with the extracranial response in all but 2 patients; 1 had extracranial PD but a PR in the brain, and the other had an extracranial PR but PD in the brain. Moreover, there was concordance between the type of response or progression (homogeneous vs heterogeneous) in the majority of patients (16 of 23; 64%). However, in the remaining 7 patients, the extracranial response/progression was heterogeneous, whereas the brain response/progression was homogeneous.
Pseudoprogression
Thirteen of 39 patients (33%) who had PD as their best response continued on the same treatment and had a subsequent response, defined as pseudoprogression. Representing 9% of the total population, these patients had a 2-year OS rate of 100%, similar to patients who had a CR or PR as their best response (95%) (see Supporting Fig. 4A).
Ten patients with pseudoprogression had initial heterogeneous progression. All patients had new lesions; 7 had a mixed response in the remaining lesions, 2 had SD, and 1 had regression of the remaining lesions.
Three patients had initial homogeneous progression, including 2 with progressing existing lesions and no new lesions and 1 with new lesions while existing lesions also increased in size.
In the 11 patients who developed new lesions, the median number of new lesions was 3 (range, 1-13 new lesions), and the median size of these new lesions was 15 mm (range, 2-39 mm). The majority of these patients (55%) had only 1 site involved with new lesions, and LN was the most frequent site of new lesions (7 patients, including 3 with a mediastinal/sarcoid-like pattern), followed by lung (3 patients), liver and adrenal (2 patients), and brain, gastrointestinal tract, soft tissue, and subcutaneous tissue (1 patient) (Fig. 5A).
Acquired Resistance
Late (acquired) resistance was not rare (12 of 140 patients in the total cohort [9%]; 12 of 99 responders [12%]) but occurred less frequently than early (primary) resistance (28%) (see Supporting Fig. 4B,C). The median time to progression in those with acquired resistance was 9.6 months (range, 4.5-27.2 months). No patients (0 of 31) developed acquired resistance after a prior CR, but acquired resistance was observed in 6 of 62 patients (10%) who had a prior PR (4 with homogenous responses, 2 with heterogeneous responses), in 4 of 8 patients (50%) who previously had SD, and in 2 of 13 patients (15%) who had initial pseudoprogression (in the initial pseudoprogression lesion in 1 patient and in a new lesion in the other).
At the time of acquired resistance, patients had a median of 5 progressing lesions (range, 1-10 progressing lesions), and progression occurred either in new lesions (4 of 12 patients; 33%), existing/baseline lesions (2 of 12 patients; 17%), or concurrently in both new and existing lesions (6 of 12 patients; 50%) (Fig. 5B). The most frequent site of new lesions was the gastrointestinal tract (5 patients; involving the small bowel, stomach, and gallbladder), followed by LNs (4 patients; 1 with a mediastinal/sarcoid-like pattern) and soft tissue (4 patients; all intra-abdominal lesions, 3 in the retroperitoneal space), lung (3 patients), brain (2 patients), and liver, bone, and subcutaneous tissue (1 patient). The most frequent sites of progressive preexisting lesions were LNs and soft tissue (4 patients) followed by subcutaneous tissue, liver, gastrointestinal tract, and lung (1 patient).
Of the patients who progressed while on therapy (6 of 12 patients; 50%), 5 (83%) received further systemic and/or local therapy (targeted therapy in 2 patients, continued PD-1 with concurrent radiotherapy in 1 patient, re-challenge with ipilimumab and anti–PD-1 therapy plus surgery in 1 patient, and radiotherapy and surgery in 1 patient), and 1 patient was not suitable for more treatment. The 1-year (from progression) OS rate was 83%. Of those who progressed while off therapy (6 of 12 patients; 50%), all received subsequent systemic therapy (targeted therapy in 2 patients, PD-1 in 2 patients, ipilimumab in 1 patient, and combined nivolumab and ipilimumab in 1 patient), and one-half also received local therapy (surgery in 2 patients and radiotherapy in 1 patient), with a 1-year (from progression) OS rate of 83%.
Discussion
To our knowledge, this is the first study to comprehensively describe the nature of response to combination ipilimumab and anti–PD-1 therapy, including how metastasis location and size influence response, as well as pseudoprogression and acquired resistance. This study of 140 patients with 833 melanoma metastases reveals that there is significant interpatient and intrapatient heterogeneity of response and progression to combination immunotherapy, likely reflecting underlying molecular heterogeneity. Metastases in different anatomical sites have distinct response patterns to combination immunotherapy; lung metastases have a high response rate, and their presence is associated with improved overall response and survival, whereas liver metastases are more resistant to immunotherapy, and their presence is associated with inferior overall response and survival. Pseudoprogression occurs in one-third of primary progressors, usually in new lesions in LNs and the lung, with OS similar to that of patients without initial progression. Acquired resistance occurs in 12% of responders at a median of 9.6 months, often in new lesions, with a 1-year survival rate (from progression) of 83%.
Examination of all metastases in our large cohort of patients with metastatic melanoma revealed clear interpatient and intrapatient heterogeneity of response and resistance with combination immunotherapy, similar to that previously described with targeted therapy8 and anti–PD-1 monotherapy.10 Moreover, we were able to define distinct patterns of response (homogeneous vs heterogeneous) and progression (homogeneous vs heterogeneous, early vs late) that affect survival, with improved OS in patients with homogeneous versus heterogeneous responses. Putting together previous data along with our results, we hypothesize that clinical interheterogeneity and intraheterogeneity result from tumor biology, the host, and sites of disease.
As expected and as previously reported by our group and others, with PD-1 monotherapy,10, 17-19 melanoma lesions that achieve a CR with combination immunotherapy are smaller than non-CR lesions; and, overall, CR is more frequent in patients who have fewer metastases and smaller tumor burden at baseline. Although such information may improve pretreatment counselling of patients about the likelihood of response, it also may suggest that patients should be treated earlier, for example, for occult micrometastatic disease in the adjuvant setting, with smaller tumor burden, rather than waiting and treating when overt metastases develop.17, 18, 20 This concept needs confirmation in future studies to control for possible bias. The Study of Pembrolizumab Versus Placebo After Complete Resection of High-Risk Stage III Melanoma (KEYNOTE-054/EORTC-1325) currently in follow-up will clarify this issue.21 In addition to size, the site of metastases also influences response to treatment and survival. For many years, it has been known that metastasis location is prognostic because patients who have distant LN and soft-tissue metastases (M1a) or lung metastases (M1b) have OS that is superior to the OS of patients who have visceral or brain metastases (M1c/M1d).22 In the current study, we show that the site of metastasis is also predictive, with lung metastases having greater regression on treatment than other sites, and the presence of lung metastases is an independent predictive marker for an improved ORR and longer PFS, similar to data from PD-1–treated patients.10, 23
Interestingly, and on the contrary, liver metastases (unlike other visceral sites of metastasis) respond poorly to treatment, and the presence of liver metastases is an independent marker (eg, independent from the size of metastases) for a lower ORR and shorter PFS and OS. Similar findings were reported in patients with PD-1 monotherapy-treated melanoma and patients with lung cancer.18, 24 Comparing the ORR in patients with liver metastases who received anti–PD-1 alone versus the those who received the combination, it was slightly higher in the combination group (combination ORR, 43.6% vs 30.6%). Liver tolerance is a well established concept that may enclose several mechanisms, including nonprofessional antigen presentation by the endothelial cells or by immature dendritic cells to CD4-positive and CD8-positive T cells, inducing their transformation into regulatory T cells and partially activated T cells (which will undergo a passive cell death), respectively.25, 26 The liver microenvironment is unique and immunosuppressive; however, it remains unclear whether the presence of liver metastases induces a systemic immunosuppressive effect or, rather, whether more aggressive tumors preferentially metastasize to the liver, contributing to poor outcomes. Several studies are ongoing trying to clarify this issue.
Both pseudoprogression and acquired resistance remain clinical challenges, and to date they have been insufficiently described.27-31 To be able to differentiate pseudoprogression from true progression and to identify those responders who will have a shorter duration of benefit from immunotherapy are key issues for patient management. A recent study showed that circulating tumor DNA is able to accurately differentiate pseudoprogression from true progression, which, although promising, can only be applied to patients whose tumors carry an identifiable mutation, for example, BRAF.32 To date, no circulating tumor DNA studies have attempted to detect acquired resistance before overt disease progression on imaging. Data from this study suggest that pseudoprogression and acquired resistance are not uncommon, happening in approximately 10% of patients each. With pseudoprogression, most patients develop a few new lesions, often in LNs, with a sarcoid-like, bilateral, mediastinal pattern occurring in approximately one-fourth of these patients. In our study, pseudoprogression corresponded to one-third of all patients who had PD according to Response Evaluation Criteria in Solid Tumors, version 1.1, and these patients had similar OS compared with nonprogressors. The majority of patients with acquired resistance developed new lesions, often in LN, lung, soft-tissue, and gastrointestinal tract sites. In our cohort, one-half of the patients who developed acquired resistance were off treatment at the time of progression because of toxicity. The mechanisms responsible for resistance in this setting may be different from those in patients who develop acquired resistance while on treatment; however, response rates in those retreated with PD-1 immunotherapy who progress while off treatment are low.33-35 All patients but 1 received further systemic and/or local therapy, with a 1-year (from progression) OS rate of 83%.
We acknowledge the limitation of a single-institution, retrospective study in which patients received different doses of ipilimumab and anti–PD-1 antibodies; however, this detailed, longitudinal, lesion-by-lesion analysis of hundreds of metastases in a large cohort of treatment-naive patients is the largest study conducted to date with the most active immunotherapy available. The 3 clinical outcomes in this study—ORR (66%) and 2-year PFS (59%) and OS (81%) rates—were slightly higher than those described in the phase 3 CheckMate-067 trial (ORR, 58%; 2-year PFS, 43%, 2-year OS, 64%)2, 5; however, different treatment schedules and doses were used in our cohort, including some nontrial patients who continued anti–PD-1 alone after high-grade combination toxicity, all metastases were included as targets, and immunotherapy may have higher activity in Australian patients,36 likely as a result of higher cumulative ultraviolet exposure and resultant tumor mutation burden.37, 38
This detailed evaluation of response suggests that there is significant heterogeneity of response and resistance to treatment, likely reflecting underlying molecular heterogeneity, which represents a barrier to treatment efficacy and biomarker development. The way melanomas interact with host cells and respond to treatment may be influenced by different microenvironments in particular organ sites, with lung and liver sites appearing most distinct. Such information is critical to improve our understanding of tumor biology and biomarker development and is the foundation for the investigation of specific mechanisms of resistance crucially needed for the development of personalized immunotherapy to overcome resistance and improve patient outcomes.
Funding Support
Georgina V. Long is supported by a Practitioner Fellowship from the National Health and Medical Research Council and by the University of Sydney Medical Foundation. Alexander M. Menzies is supported by a Cancer Institute New South Wales Fellowship.
Conflict of Interest Disclosures
Matteo S. Carlino reports personal fees from Bristol-Myers Squibb, Merck Sharp & Dohme, Amgen, Roche, Novartis, and Pierre Fabre outside the submitted work. Georgina V. Long reports personal fees from Aduro, Amgen, Array, Bristol-Myers Squibb, Merck Sharp & Dohme, Novartis, Pierre-Fabre, Oncosec, Roche, and Incyte outside the submitted work. Alexander M. Menzies reports personal fees from Bristol-Myers Squibb, Merck Sharp & Dohme, Novartis, Roche, and Pierre-Fabre outside the submitted work. Ines Pires da Silva, Serigne Lo, Camelia Quek, and Maria Gonzalez made no disclosures.
Author Contributions
Ines Pires da Silva: Conceived and designed the study, collected the clinical data., analyzed the data, wrote the article, revised the article, and approved the submission. Serigne Lo, Camelia Quek, Maria Gonzalez, Matteo S. Carlino, Georgina V. Long: Analyzed the data, revised the article, and approved the submission. Alexander M. Menzies: Conceived and designed the study, analyzed the data, wrote the article, revised the article, and approved the submission.
