Comparative Analysis of Durable Responses on Immune Checkpoint Inhibitors Versus Other Systemic Therapies

A Pooled Analysis of Phase III Trials

Elvire Pons-Tostivint, MD; Aurélien Latouche, PhD; Pauline Vaflard, MD; Francesco Ricci, MD, PhD; Delphine Loirat, MD, PhD; Ségolène Hescot, MD, PhD; Marie-Paule Sablin, MD; Roman Rouzier, MD, PhD; Maud Kamal, PhD; Claire Morel, MSc; Charlotte Lecerf, MSc; Vincent Servois, MD; Xavier Paoletti, PhD; Christophe Le Tourneau, MD, PhD


JCO Precis Oncol. 2019;2019(3) 

In This Article


Our study confirms that a higher proportion of patients experience durable responses to ICIs than to other drug classes (25% v 11%). Durable responses were more frequent in patients treated with anti–PD-1/PD-L1 agents than in those treated with anti–CTLA-4 agents. In multivariable analysis, the effect of treatment with anti–PD-1/PD-L1 agents and the effect of first-line treatment were statistically associated with a higher mean proportion of durable responses. These results are of primary importance because it cannot be excluded that some patients who experience a durable response might be cured of their recurrent and/or metastatic cancer, which represents the overarching goal of cancer research. Only a longer follow-up will enable a drawing of robust conclusions.

Our definition of durable response can be challenged because no definition exists in the literature. Estimation of the proportion of patients with a PFS that exceeds three times the median PFS of the whole cohort of patients treated with the same drugs instead of arbitrarily defining a minimum PFS is an elegant way not to be biased by heterogeneity in terms of the natural history of the disease, clinical setting, and drugs used. We arbitrarily chose a threshold of at least three times the median PFS, which seemed clinically meaningful to us. We selected a threshold of at least two times the median for OS because the follow-up was insufficient in most of the trials to get the information for three times the median OS. The use of PFS and not OS to define durable response allowed us not to be biased by subsequent treatments received beyond progression, especially ICIs. Using durable response as we defined it provides additional information that is not given by any other classic end point, such as duration of response or PFS because it is calculated on the basis of the median PFS of the whole treatment arm. It gives an insight into the proportion of patients who experience durable responses and takes into account the natural history of the disease and the overall activity of the treatment. Whether this measure correlates with OS, which can be evaluated only with individual data, remains to be determined.

Twenty-five percent of patients treated with ICIs in the trials retrieved for our study had a durable response according to our definition. The proportion of patients who experienced a durable response was higher in patients treated with anti–PD-1/PD-L1 agents than in those treated with anti–CTLA-4 agents, which is in line with long-term survival data that have been reported for these drugs[15,16] and with the fact that anti–PD-1 agents were demonstrated to be more active than anti–CTLA-4 agents in patients with melanoma.[25,26] Of note, although the difference in the proportion of durable responses was substantial between patients treated with ICIs and the others, the median PFS was similar (3.8 v 3.5 months). This phenomenon corroborates that the majority of patients do not respond to ICIs and do not experience a prolonged PFS, which is commonly the case with chemotherapy. Also of note, the rates observed in nonmelanoma trials were in the same range as the melanoma trials. Although the proportion of durable responses is high and confirms one of the specificities of ICIs, 11% of patients treated with other drug classes, including chemotherapy and targeted therapy, also experienced durable responses, which means that this phenomenon is not unique to immunotherapy. In a recent meta-analysis of patients with epidermal growth factor receptor (EGFR)–mutated NSCLC treated with EGFR inhibitors, 8% of patients had a PFS that exceeded three times the median PFS.[27] This low figure needs to be interpreted with caution because these patients were molecularly selected and do derive benefit from the treatment, as suggested by the much longer median PFS for EGFR inhibitors than for chemotherapy (11 v 5.6 months). Our study did not involve any targeted therapy arm in molecularly selected patient populations. How the results of our study would have been affected by such patient populations remains to be determined.

The proportion of patients who had an OS that exceeded two times the median OS was also higher in patients treated with an ICI compared with those treated with other drug classes, although to a lesser extent (30% v 23%). This corresponds to a 1.3 ratio compared with a 2.3 ratio for PFS. This difference might be explained by two factors. First, the threshold chosen of two times the median was lower for OS than for PFS (for the reasons explained at the beginning of the Discussion). This lower threshold might hamper the revelation of a difference that might occur later during follow-up. Second, one cannot exclude that patients in the non-ICI arms have been treated eventually with an ICI, which potentially dilutes the effect when considering OS. Crossover was indeed allowed in five of the trials retrieved in our study. For example, 58% of patients crossed over in the nivolumab versus chemotherapy NSCLC trial.[19] Reporting of long-term OS data will be key to estimate the real effect on OS.

One of the limitations of our study is the small number of non-ICI arms, which hampered the separate evaluation of the proportion of durable responses in patients treated in the various control arms. We attempted as well to include in our study phase III trials that did not involve ICI arms to have a better representation of patients treated with chemotherapy and targeted therapy. The follow-up was too short in these studies to estimate the proportion of patients who experienced a PFS that exceeded three times the median PFS. However, the lower proportion of patients who experienced durable responses to other drugs than ICIs in our study confirms the usual short-lived responses reported with chemotherapy and targeted therapy (mainly those related to the occurrence of resistance mutations in this latter case). Another limitation is that our study was limited to published papers in PubMed. How the inclusion of unpublished, sometimes negative trials might have affected the results remains to be determined. Finally, we restricted our analysis to trials with sufficient follow-up to estimate the proportion of patients who experienced a durable response, as defined in Patients and Methods under Statistical Analyses, which means that the use of our end point cannot be generalized to all trials.

In conclusion, our pooled analysis of randomized phase III trials objectively confirms that durable responses are more frequent with ICIs than with other drug classes, especially with ICIs that target PD-1 and PD-L1. Our study also shows that durable responses according to our definition are not specific to ICIs because this phenomenon also exists with targeted therapies and chemotherapy, although to a lower extent. The identification of predictive biomarkers of efficacy, and especially durable response, is of upmost importance to select patients who will need only a single-agent ICI and patients for whom a drug combination will be necessary. The ongoing initiatives that aim to decipher the molecular landscape of exceptional responders to anticancer therapies (eg, identifier: NCT02243592) are essential in that aim.