Data from three clinical trials were pooled. All patients were treated with 7 days of lymphodepleting chemotherapy with cyclophosphamide and fludarabine, infusion of autologous TILs and interleukin-2 (IL-2). Study I was performed in two parts including the first 6 patients treated with low-dose IL-2 subcutaneously and the next 25 consecutive patients treated with an intermediate continuous intravenous IL-2 dosing regimen (decrescendo regimen) (ClinicalTrials.gov ID. In both study II and study III), 12 patients were treated with the decrescendo IL-2 regimen. Further information on clinical trial design and treatment details is available in previously published reports for study I6 11 and study II,5 and manuscript for study III is in preparation.12 Data cut-off for this analysis was 14 December 2018.

Importantly, in these trials all TILs were produced at a single institution and following similar procedures ensuring comparability of treatment products.

Survival and progression-free survival curves were computed according to the Kaplan-Meier method using GraphPad Prism V.5 software. P values of comparisons between survival curves were performed using the log-rank Mantel-Cox test. RR comparisons of subgroups were tested using Fisher’s exact test. A p value of <0.05 was considered to be statistically significant.

A total of 75 patients were enrolled with an intention-to-treat and had a tumor lesion surgically removed between August 2009 and May 2018. For two patients (2.7%), TIL growth was not successful. In one patient, the removed tumor was found to be a sarcoma and 17 patients (22.7%) deteriorated clinically and were excluded before treatment with TILs. Online supplementary figure S1 shows an outline of patient enrollment and number of dropouts. Thus, a total of 55 patients were treated with TILs. For those patients, the TIL product was ready for infusion in a median of 5.5 weeks after surgery (data not shown).

At baseline, all patients had progressive MM, unresectable stage IIIB or higher with 78% being stage M1c (according to the Melanoma American Joint Commission of Cancer, AJCC staging seventh edition; table 1). All patients had received prior systemic therapy (median 2; range 1–4) for unresectable or metastatic disease, and 84% had received two or more prior lines of therapy (table 1). Most patients had widespread metastatic disease, with a median of four organ systems involved (range 1–8) including 33% having liver metastasis and/or a history of brain metastasis (these had to be small, with no edema and asymptomatic, or treated and stable), and 62% had elevated lactate dehydrogenase (LDH). At inclusion, all patients were in good performance status (PS 0–1), however some patients clinically deteriorated during TIL manufacturing (typically 4–6 weeks) and at admission three patients had PS≥2.

TIL therapy comprised preconditioning chemotherapy, infusion of autologous TILs and subsequent interleukin-2 is an intense treatment requiring admission for monitoring and managing of side effects. We observed no new safety signals, although two patients died within days after receiving TIL infusion; one patient due to an intratumoral hemorrhage in a previously irradiated brain metastasis and one patient from multiorgan failure attributed to the patient’s fast progressing cancer and very large tumor burden described in more detail in Andersen et al.5 The two deaths underline that patient selection is important in the context of TIL therapy.

The median time in hospital was 19 days (online supplementary table T1). All patients experienced bone marrow suppression with the majority needing multiple red blood cell and platelet transfusions. Furthermore, most patients developed well-known side effects of IL-2 treatment (eg, electrolyte derangement, hypotension, fluid retention, fever and dyspnea). Toxicities were transient (except vitiligo observed in 3 of 55 patients) and responded to standard interventions or cessation of treatment. On discharge, all toxicities had resolved or was grade 1 or 2.

Of the 55 patients treated, 53 patients were evaluable according to Response Evaluation Criteria in Solid Tumors criteria. Of 53 evaluable patients, 6 patients achieved a complete response (CR) and 14 patients achieved a partial response (PR), hence the overall RR was 38% (online supplementary table T2). Two patients died within days after receiving TIL infusion and were, thus, not evaluable.

With a median follow-up time of 60 months for the entire cohort, the median overall survival (mOS) was 15.9 months, with the survival curve reaching a plateau after 30 months, and a 3-year survival rate at 29% (figure 1A).

Median progression-free survival (mPFS) was 3.7 months, nevertheless only few patients progressed later than 12 months after treatment indicating that most patients with disease control were still progression free many years later (figure 1B). Indeed, seven patients (13%) had not progressed at data cut-off, and another patient died of an unrelated infectious disease almost 3 years after treatment.

Of the 20 patients with a CR or PR, 13 (65%) are still alive. Hence, mOS was not reached for this subgroup and mPFS was 19.1 months (see figure 2A and B).

Of the six complete responders, two patients progressed after 13.2 and 48.1 months, but had solitary relapses removed and one is still without evidence of disease (NED) 76+ months after surgery. The 3-year survival rate for complete responders was 100% (data not shown). Of interest, the two patients progressing after CR both received low-dose IL-2 in the first part of study I, whereas the four remaining complete responders (ongoing between 41+ and 83+ months) received an intermediate IL-2 dosing regimen.

Three patients with PR have not progressed. Two of these had residual disease removed surgically and have NED as previously described,11 now 73+ and 76+ months after treatment.

Patients with tumors harboring certain immunohistopathological features (ie, dense immune cell infiltration, high mutational burden and interferon (IFN)-γ signature) have a higher chance of responding to CPI.13 14 This implies a selection towards enrichment of less favorable characteristics in patients progressing on or after treatment with CPI. In the present cohort, most patients had received and progressed on prior treatment with CPI, and to elucidate whether resistance to checkpoint inhibition affects clinical outcome after TIL therapy, we performed subgroup analyzes stratifying for prior treatment with either anti-CTLA-4 only, anti-PD-1 only, sequential anti-CTLA-4 and anti-PD-1, or no prior CPI. Patients were only stratified for prior CPIs, but the majority had received other systemic lines of treatment in addition. The number of prior lines of therapy were well balanced except for the group of patients treated with anti-PD-1 only (n=4), as they all received TIL therapy as a second-line therapy (data not shown).

In our cohort, a total of 41 patients had received prior treatment with anti-CTLA-4. Of those, 22 patients had received no other prior CPI and this subgroup appeared to have a longer mOS compared with patients who were not previously treated with CPI (21.8 vs 11.2 months), but survival curves were not significantly different (p=0.5) and plateau was reached at the same time and level (online supplementary figure S2A). The same pattern was observed for PFS (online supplementary figure S2B). RR was numerically higher but not statistically significant in the prior anti-CTLA-4-treated group (45.5% vs 33.3%; p=0.48).

Results from published studies indicated that response to TIL therapy after progression on anti-PD-1 is possible, although RR appeared low.10 15 In our cohort, a total of 23 patients had received prior treatment with anti-PD-1 either as the only CPI (n=4) or sequentially to anti-CTLA-4 (n=19). To evaluate specifically the effect of anti-PD-1, the group of patients previously treated with anti-CTLA-4 alone was compared with patients treated with anti-PD-1 and anti-CTLA-4 sequentially. Clinical features and prognostic variables (PS, stage, LDH, metastatic sites, prior lines of therapy) were similar (data not shown). PFS curves for the two groups were significantly different favoring the patients previously treated with anti-CTLA-4 alone (p=0.03; mPFS 4.9 vs 2.8 months), but no significant difference between the groups was observed for OS (p=0.56) and RR (p=0.52) (online supplementary figure S3 and table T2).

As prior treatment with anti-CTLA-4 did not significantly affect clinical response after TIL therapy, we divided our data set in patients previously treated with anti-PD-1 regardless of anti-CTLA-4 treatment (PD-1 progressors; n=23), or not (PD-1 naïve; n=32). Baseline clinical characteristics were similar between the two groups (online supplementary table T3). No statistical difference was observed in OS and RR (figure 3A and table 2), but PFS curves started to separate at around 4 months and are significantly different favoring the anti-PD-1-naïve group (p=0.046, figure 3B). The difference appeared to be driven by a longer median duration of response (DOR) in the anti-PD-1-naïve group as a whole (32.2 vs 7.6 months) and especially for partial responders (table 2).

Although not systematically assessed, it is noteworthy that durable responses to anti-PD-1 were observed in two anti-PD-1-naïve patients who progressed after TIL therapy (one patient had PR and one had PD on TIL).

Experimental treatments like TIL therapy are often employed only after progression on approved therapies. Since multiple therapies have been approved for MM during the last decade, TIL therapy has been moved to third or fourth line of treatment depending on BRAF mutational status. To assess the impact of moving TIL therapy to later lines, we analyzed treatment responses according to line of treatment. In our cohort, all patients were treated in second line or later. We found that patients treated in second line responded favorably with a RR of 62.5% compared with 33.3% for later lines and had numerically higher median OS and PFS, although not statistically significant (online supplementary table T4 and figure S4). CR rates were significantly impacted by the line of treatment as 33.3% (3/9) achieved CR if treated in second line compared with 6.5% (3/46) treated in third or later line (p=0.049).

Median DOR decreased incrementally with increasing number of prior lines of therapy and, notably, only one of 18 patients treated in fourth line or later have an ongoing response (PR (NED); (online supplementary table T4).