Neoadjuvant relatlimab and nivolumab in resectable melanoma

Abstract

Relatlimab and nivolumab combination immunotherapy improves progression-free survival over nivolumab monotherapy in patients with unresectable advanced melanoma¹. We investigated this regimen in patients with resectable clinical stage III or oligometastatic stage IV melanoma (NCT02519322). Patients received two neoadjuvant doses (nivolumab 480 mg and relatlimab 160 mg intravenously every 4 weeks) followed by surgery, and then ten doses of adjuvant combination therapy. The primary end point was pathologic complete response (pCR) rate². The combination resulted in 57% pCR rate and 70% overall pathologic response rate among 30 patients treated. The radiographic response rate using Response Evaluation Criteria in Solid Tumors 1.1 was 57%. No grade 3-4 immune-related adverse events were observed in the neoadjuvant setting. The 1- and 2-year recurrence-free survival rate was 100% and 92% for patients with any pathologic response, compared to 88% and 55% for patients who did not have a pathologic response (P = 0.005). Increased immune cell infiltration at baseline, and decrease in M2 macrophages during treatment, were associated with pathologic response. Our results indicate that neoadjuvant relatlimab and nivolumab induces a high pCR rate. Safety during neoadjuvant therapy is favourable compared to other combination immunotherapy regimens. These data, in combination with the results of the RELATIVITY-047 trial¹, provide further confirmation of the efficacy and safety of this new immunotherapy regimen.

Fig. 1. Study design.

Eligible patients receive two doses of relatlimab 160 mg with nivolumab 480 mg intravenously every 4 weeks (Q4W) in the neoadjuvant setting and then have repeat imaging for calculation of RECIST response. Surgery takes place at week 9 for evaluation of pathologic response. Patients receive up to ten doses of relatlimab 160 mg and nivolumab 480 mg every 4 weeks in the adjuvant setting and are followed for 2 years for evidence of recurrence. Blood and tumour are collected during screening, at weeks 3, 5 and at time of surgery at week 9. Blood is collected every 12 weeks (Q12W) in the adjuvant setting. ECOG PS, Eastern Cooperative Oncology Group Performance Status; RELA, relatlimab; NIVO, nivolumab; ORR, objective response rate; RECIST, Response Evaluation Criteria in Solid Tumors.

Fig. 2. Consort diagram and patient disposition.

A total of 41 patients were screened for protocol and there were 11 screen failures and 30 patients were eligible to initiate therapy. After completion of neoadjuvant therapy, one patient developed distant metastases and did not proceed to surgery. Twenty-nine patients proceeded to surgery and 17 patients (57%) achieved a pCR. Twenty-seven patients initiated adjuvant therapy and 15 went on to complete entire duration of treatment. path, pathologic.

Fig. 3. Response data and long-term outcomes.

a, Breakdown of pathologic responses for the 29 patients who underwent surgery as interpreted by the guidelines of the INMC. Result details (values in chart rounded): no operation, 1 of 30 patients (3.33%);pCR, 17 of 30 patients (56.67%); near pCR, 2 of 30 patients (6.67%);pPR, 2 of 30 patients (6.67%);pNR, 8 of 30 patients (26.67%). b, Waterfall plot of neoadjuvant response as per RECIST 1.1 criteria with colour coding indicating pathologic response. pCR indicates lack of viable tumour. Near pCR indicates greater than 0% but less than or equal to 10% viable tumour, pPR is greater than 10% to less than or equal to 50% viable tumour and pNR is greater than 50% viable tumour. c, Probability of being relapse-free based on any pathologic response versus no pathologic response. d, Overall survival curves for the entire cohort.

Fig. 4. Correlative analyses in tumour specimens.

Tumour tissue samples harvested from patients at baseline, and post relatlimab and nivolumab treatment were analysed in a single experiment by CyTOF (a–c). a, Frequency of CD45⁺ cells was assessed through manual gating. b, Frequency of an effector CD8⁺ T cell subset (CD3⁺CD8⁺CD45RO^low) in unsupervised clustering is shown. c, Frequency of a memory CD4⁺ subset (CD45RO⁺ICOS⁺ TCF7⁺BTLA⁺CD28⁺TIGIT⁺) was determined by unsupervised clustering. Data shown in a–c are mean ± s.d., and n values are indicated in the figure. P values shown in each graph were calculated by two-tailed unpaired t-test, with no multiple comparisons. Red indicates pathologic responders; blue, non-responders. CyTOF, mass cytometry; NR, non-responder; R, responder; wk, week. Source data.

Extended Data Fig. 1

Probability of being event-free for all patients who received study treatment.

Extended Data Fig. 2. Probability of being recurrence-free.

A) Probability of being recurrence-free for all patients who underwent surgery. B) Probability of being recurrence-free based on pathologic complete response versus non-pathologic complete response (P = 0.10). Source data

Extended Data Fig. 3. PD-1 and LAG-3 levels in baseline tumour.

Tumour infiltrating immune cells were assayed via CyTOF and analysed by manual gating for frequency of A) PD-1 and B) LAG-3 levels in T cells prior to treatment. Red, pathologic responders; blue, pathologic non-responders. Data are mean +/− SD; P values where shown were determined by two-tailed unpaired t-test, with no multiple comparisons. n values for each group/timepoint are indicated in each graph. Source data

Extended Data Fig. 4. M2-like macrophages in tumour and EOMES+ CD8+ T cells in blood.

A) Frequency of an M2-like macrophage subset (CD68+ HLA-DR+ CD14+ VISTA+ CD163+ CD45RO+ PD-L1+) was determined by unsupervised clustering of CyTOF data from a single experiment. B) Frequency of EOMES+ CD8+ T cells. PBMCs isolated from blood samples were analysed by flow cytometry from a single experiment. Data are mean +/− SD; P value was determined by two-tailed unpaired t-test, with no multiple comparisons. n values for each group/timepoint are indicated in each graph. Red indicates pathologic responders; blue, non-responders.

Extended Data Fig. 5. Gating schema for manual analysis of CyTOF data from tumour and blood specimens.

A) Tumours were mechanically dissociated and cells were stained with immune cell-specific antibodies. Specimens were assayed on the Helios mass cytometer via CyTOF Software. Cytometer data were then prepared for manual and unsupervised analyses via FlowJo software. Major cell populations were identified manually and reported. An example of one patient specimen is shown above for reference. B) Gating schema for flow cytometric analysis of blood specimens. Peripheral blood mononuclear cells from patient specimens were stained along with FMO (fluorescence minus one) controls and assayed via a BD LSRFortessa cytometer and BD FACSDiva acquisition software. Data were analysed via FlowJo software as described above. Briefly, live CD3+ singlets were identified and gated into T cell lineages, and those lineages analysed for frequency of each of eight phenotypic markers (BCL6, BLIMP1, CD27, CD28, cMYC, EOMES, ICOS, Ki67) as defined by FMO (fluorescence minus one) specimens. An example of one phenotypic marker (EOMES) in one patient specimen is shown above for reference.