Capturing Hyperprogressive Disease with Immune-Checkpoint Inhibitors Using RECIST 1.1 Criteria

This article is featured in Highlights of This Issue, p. 1779

The introduction of immune-checkpoint inhibitors (ICIs) in the cancer therapeutic arsenal has been one of the most important advances in medicine in recent years. Several ICIs have already been approved for the treatment of different tumors. Most of these approvals have been based on phase III clinical trials that compared ICIs with chemotherapy or targeted therapies as standard treatments in a wide range of malignancies such as melanoma (1–3), non–small-cell lung cancer (NSCLC) (4–8), head and neck squamous-cell carcinoma (HNSCC; refs. 9, 10), urothelial carcinoma (11, 12), and renal cell carcinoma (RCC; refs. 13, 14), among others. In orphan tumors such as Merkel cell carcinoma, uterine cervical carcinoma, or microsatellite instability-high (MSI-H) tumors, their approval has been based on single-arm phase II clinical trials (15–19). Despite the widespread use of ICIs, response assessment to these drugs continues to be a challenge. New patterns of response such as pseudoprogression, with an estimated incidence of around 5% to 10% of cases (20), prompted revision of the classically used Response Evaluation Criteria in Solid Tumors (RECIST). Since then, disease progression needs to be confirmed in a second tumor assessment performed at least 4 weeks apart in clinically stable patients by means of new response criteria such as immune-related response criteria (irRC), immune-related RECIST (irRECIST), or immune RECIST (iRECIST; refs. 21–23).

Hyperprogressive disease (HPD) is a phenomenon described more recently with ICI treatment (24, 25). The actual concept of HPD, whereby patients experience significant increase of their disease burden right after ICI exposure, remains currently under discussion. In some phase III studies comparing ICIs versus chemotherapy, there is inferior progression-free survival (PFS) and/or overall survival (OS) during the first 3 to 6 months in the ICI arm, with later overcrossing of survival curves, with improved long-term outcomes with anti–PD-1 antibodies (4, 7, 9–11, 26). These results in favor of the control arms in the first months have mainly been attributed to two potential explanations: (i) total lack of efficacy and (ii) HPD and early deaths with ICIs. Its incidence varies from 5% to 29% of patients, depending on the tumor type and definition (25, 27–29). Given the fact HPD implies a worsening due to a specific treatment, most classifications define it by using tumor growth rate (TGR), which compares the tumor growth of the target lesions during ICI treatment and a reference period immediately prior to ICIs (3 months to 2 weeks; refs. 24, 28–31). These methods have some clinically important limitations, such as (i) requiring a prior computed tomography (CT) scan, which can sometimes be difficult to retrieve, is not available or has been performed outside the established reference period; and (ii) the assessment of new lesions and nonmeasurable disease is not accounted in the definition of TGR; however, in certain tumor types, over half the patients with progressive disease have new lesions and progression due to an increase in the nontarget lesions (32, 33).

Given these difficulties, HPD still remains more of an intuition of the treating physician for an individual patient than a phenomenon that is evaluated and measured in a standardized way. To overcome these limitations, we explored if HPD could be identified using an adapted RECIST criterion in patients treated with ICIs in phase I clinical trials, in order to assess the concept in a more applicable way in the clinic. We also cross-compared patients with HPD captured by RECIST definition and by the original TGR criteria (increase in TGR of more than two times during ICI treatment vs. reference period; ref. 24). To determine if this phenomenon is exclusively seen with ICIs, we analyzed a control cohort of patients treated with targeted agents (TA) in phase I clinical trials (34). Finally, we investigated clinical, laboratory, and radiology parameters that could understand and predict HPD according to RECIST criteria.

Patients treated with anti–PD-1/PD-L-1 monoclonal antibodies as monotherapy or in combination with other immunotherapeutic agents exclusively, in phase I clinical trials at Vall d'Hebrón Institute of Oncology (VHIO) included between January 2012 and June 2018, were analyzed (n = 287). For the control cohort, we collected data from patients treated in phase I clinical trials with TAs in monotherapy or combination with other targeted drugs considered Tier 2 to 3 in the ESMO scale of clinical actionability for molecular targets (ESCAT; ref. 34) included between January 2011 and January 2018 (n = 280). We excluded from both cohorts patients without a complete first tumor response assessment or not evaluable by RECIST (Supplementary Figs. S1 and S2). There was no patient overlap in both cohorts and no patients in the TAs cohort were treated with ICIs subsequently.

All procedures followed were in accordance with the ethical standards of the responsible committee on human research (institutional and national) and with the Helsinki Declaration of 1964 and later versions. This study was approved by the institutional review board of Vall d'Hebrón University Hospital, and informed consent from participants was not required because of the retrospective nature.

CT scans were acquired at baseline and every 6 to 8 weeks as per clinical trial specifications. The baseline CT scan (performed up to 7 days before treatment initiation) and the first CT scan for response assessment (week 6 or 8 ± 7 days) were reviewed. In order to evaluate HPD by TGR criteria, CT scans performed between 3 months and 2 weeks previous to the baseline CT scan were also reviewed. The TGR was assessed according to the definition by Ferte and colleagues tumor growth (TG) is equal to TG = 3 Log(Dt/D0)/t, where D is tumor size defined as the sum of the longest diameters of the target lesions as per RECIST 1.1. TGR were expressed as a percentage increase in tumor volume during 1 month using the following transformation formula: TGR = 100 (exp(TG) −1), where exp(TG) represents the exponential of TG (31).

The specifications of the image acquisition protocol are summarized in Supplementary Table S1. Previous-to-baseline CT scans were considered for evaluation only when they fulfilled the image acquisition requirements as per RECIST, with at least 5-mm contiguous slice thickness.

Target lesions defined by RECIST for every patient were collected at baseline and after treatment. In the ICI group, patients with progressive disease (PD) as best response who were clinically stable could continue treatment until PD confirmation in a second assessment at least 4 weeks apart, as per clinical trial specifications. In order to assess the rate of HPD by TGR, the same target lesions were identified and measured at the previous-to-baseline CT scan by an experienced radiologist (R. Perez-Lopez). Diameters of the target lesions before starting treatment with the ICIs were collected. The dates of the CT scans were also collected to calculate time between scans. All CT scans were centrally reviewed by two independent radiologists from two different centers (J. Mateos and R. Perez-Lopez).

We defined HPD based on RECIST as PD in the first 8 weeks after treatment initiation and minimum increase in the measurable lesions of 10 mm plus: (i) increase of ≥40% in sum of target lesions compared with baseline [which represents doubling in unidimensional target lesions compared with classic RECIST PD criterion (20%)]; and/or (ii) increase of ≥20% in sum of target lesions compared with baseline (the classic RECIST PD criterion) plus the appearance of new lesions in at least two different organs (Fig. 1). The reason to include both criteria (i and ii) as HP alternatives is that we noticed that some patients experienced progression that is largely driven by multiple new metastatic sites and not major target lesion growth. Importantly, the RECIST-based criteria to capture HPD were defined before the statistical analysis was performed, to avoid bias.

Survival analysis was calculated using the Kaplan–Meier method, and log-rank test was used for statistical comparison. Overall survival (defined as time from treatment initiation to death or last follow-up) and time from initial progression by RECIST to study discontinuation [in patients with treatment beyond progression (TBP)] of patients with PD as best response were evaluated according to their classification as HPD using RECIST and TGR criteria. The median follow-up was calculated using Kaplan–Meier reverse method. We performed a landmark analyses with OS calculated from the initial date of progression by RECIST to death or last follow-up. RECIST and TGR criteria were compared and analyzed to assess their performance using Fisher exact tests and concordance using kappa value. Univariate Cox proportional-hazard models were used to obtain hazard ratios (HR) with 95% confidence interval (CI). Statistical comparisons were performed using Student t test or Mann–Whitney–Wilcoxon test for continuous parametrical and nonparametrical variables. Chi-square or Fisher exact tests were used for categorical variables. All P values were two-sided, and values less than 0.05 were considered statistically significant. Statistical analyses were performed using a software program (SPSS for MAC, version 25.1; IBM Inc.).

From January 2012 to June 2018, 287 patients with different tumor types received treatment in the context of a clinical trial with anti–PD-1/PD-L1 monotherapy (41.6%) or combination (58.4%) in the Early Drug Development Unit at VHIO. Seventeen (5.9%) patients experienced clinical progression or toxicity before the first tumor assessment as illustrated in the flowchart (Supplementary Fig. S1). A total of 270 patients were analyzed and evaluable for HPD. Clinical characteristics are described in Table 1. The median follow-up was 16.9 months (95% CI, 14.6–19.1 months). Median OS (mOS) was 10.93 months (95% CI, 8.52–13.34 months). No case of pseudoprogression as per iRECIST criteria was observed.

We independently analyzed patients with PD as best response (n = 107). Median time from cycle 1 day 1 to first CT scan was 6.5 weeks (95% CI, 6.2–6.9 weeks) in these patients. Overall, 29 patients (10.7%) were considered HPD by RECIST definition, representing 27.1% of the evaluable patients with PD as best response. In this population, patients with HPD had significantly lower OS as compared with the non-HPD group, with mOS of 5.23 months (95% CI, 3.97–6.45) versus 7.33 months (95% CI, 4.53–10.12), respectively (HR = 1.73; 95% CI, 1.05–2.85; P = 0.04; Fig. 2). Landmark analyses showed very similar results (HR = 1.73; 95% CI, 1.04–2.84; P = 0.03; Supplementary Fig. S3). When stratified by treatment line (second or third line), the association between HPD and worsened outcomes was maintained (Supplementary Fig. S4).

In the multivariable analysis for OS in patients with PD as best response, including classically well-described prognostic factors [age, LDH, albumin, liver metastasis, more than two metastasis sites, derived neutrophils/(leukocytes minus neutrophils) ratio (ref. 35; dNLR) and count platelets], HPD using RECIST definition remained a significant prognostic factor together with low albumin, high LDH, and liver metastases (Table 2).

Overall, 34 patients continued treatment beyond progression by RECIST criteria. Time from initial progression to study discontinuation was 0.9 m (95% CI, 0.65–1.15) in the HPD group (n = 10) versus 1.4 months (95% CI, 1.14–1.46) in the non-HPD progressors (HR = 1.45; 95% CI, 0.67–3.10; P = 0.34).

Overall, 239 patients treated with TAs were included in the HPD analysis. A total of 41 patients (17.2%) were excluded due to clinical progression or toxicity before the first tumor response assessment as indicated in the flowchart (Supplementary Fig. S2). Main characteristics of the 239 patients are listed in Supplementary Table S2. None of the patients had received prior ICI treatment. The median follow-up was 29.66 months (95% CI, 23.6–35.7 months) and mOS was 7.93 months (95% CI, 6.90–8.95 months). From 119 patients who had PD as best response, 26 (21.8%) were classified as HPD by RECIST criteria. No differences in OS were observed between HPD group (mOS 4.23 months; 95% CI, 3.42–5.04) and non-HPD progressor group (mOS 5.7; 95% CI, 4.99–6.4; HR 1.09, 95% CI, 0.7–1.7; P = 0.70), as seen in Fig. 2.

HPD using TGR assessment was evaluated for every patient from the previously described ICIs cohort whenever possible. From a total of 107 patients with PD as best response, 32 (30%) could not be assessed for the previous-to-baseline CT scan: 22 (21%) did not have a previous-to-baseline CT scan and 10 (9%) had no measurable disease. From the 75 remaining patients, 17 (19%) had no available previous-to-baseline CT scan during the established reference period (within 3 months and 2 weeks before baseline). A total of 58 patients were evaluable following both definitions. The time between previous-to-baseline and baseline CT scans was 45.36 ± 20.98 days (mean ± standard deviation). The time between baseline and follow-up CT scans was 58.79 ± 11.20 days (mean ± standard deviation).

Fourteen patients (24.13% of patients with PD as best response and evaluable by TGR criteria) were categorized as HPD using TGR criteria. Differences in mOS between this group (mOS 4.2 months; 95% IC, 2.07–6.33) and non-HPD progressors (n = 44) by TGR criteria (mOS 6.27 months; 95% CI, 3.88–8.67) were not statistically significant (HR 1.4; 95% IC, 0.70–2.77; P = 0.346; Fig. 3). A landmark analysis did not detect differences in OS in HPD versus non-HP progressors as per TGR (Supplementary Fig. S3). In the multivariable analysis for OS in patients with PD as best response, using the same previous clinical prognostic factors, HPD by TGR criteria was not an independent prognosis factor (Supplementary Table S3).

The overall concordance rate between the two criteria was 70.7% (Supplementary Fig. S5), with a significant association and concordance between both definitions of HPD (P = 0.047). HPD using RECIST and not by an increase of TGR was observed in 11 patients (19%). Eight of these patients had new lesions in the first tumor assessment. In three of them, PD was based on increase in target lesions (49%, 46%, and 55%). We then compared mOS between two groups [HPD by RECIST combined with non-HPD by TGR definition (n = 11) versus non-HPD progressors by RECIST definition (n = 39)]. mOS was 3.43 months (95% CI, 2.70–4.15 months) and 8.7 months (95% CI, 4.73–12.66 months), respectively, which was statistically significant (HR 4.53; 95% CI, 1.92–10.67; P = 0.001). Six (10.3%) patients were categorized as HPD when assessing TGR dynamics and not by RECIST, and no patient had new lesions in at least two different organs (two patients had no new lesions, two patients had a single new lesion, and two patients had new lesions in a single organ).

In addition, we analyzed an initial subgroup of patients from the TA cohort with PD as best response (n = 75) for TGR criteria. Overall, 48 patients were included in this analysis. The reasons for exclusion were the following: 12 patients with unavailable CT scans prior to baseline and 15 patients had no available previous-to-baseline CT within the reference period. Twelve patients (25% of patients with PD as best response) were categorized as HPD when assessing TGR increase. We found no significant differences in outcome between HPD group and non-HPD progressors, with mOS of 5.37 months (95% CI, 0.55–10.12 months) and 4.23 months (95% CI, 2.42–6.041 months; HR 0.68, 95% CI, 0.34–1.36; P = 0.27), respectively (Fig. 2). The concordance rate between RECIST and TGR criteria in the TA cohort was 66.7%, but not statistically significant (P = 0.695).

Next, we analyzed the association between HPD by each of the criteria and three different types of variables: (i) clinical variables, (ii) radiologic parameters, and (iii) mechanism of action of the ICIs administered.

No clinical variables were associated with HPD using RECIST criteria, as summarized in Table 1. However, some clinical prognostic factors were associated with HPD using TGR assessment, such as the presence of liver metastasis (P = 0.001) and having more than two metastatic sites before treatment with ICIs (P = 0.038).

Regarding radiologic parameters in the ICI cohort, we analyzed the distribution of TGR in the reference period and in the experimental period (Fig. 3), and the sum of target lesions in the baseline CT scan and the first tumor assessment, for patients meeting each of the HPD definitions. We observed a lower TGR value in the reference period in patients defined as HPD when assessing the dynamics of TGR compared with the non-HPD progressor group (P < 0.001). In the experimental ICI treatment period, this variable was significantly reversed with higher TGR values in HPD patients (P = 0.002). We did not find differences in the total sum of the diameter of target lesions in either cohort, neither in the baseline CT scan (P = 0.317) nor in the first tumor assessment (P = 0.158). When evaluating those patients classified as HPD following RECIST criteria, TGR values during the experimental period were higher in the HPD group than in the non-HPD progressor group (P < 0.001). Finally, TGR values during the reference period and the total sum of the diameter of target lesions during any period were not different in HPD group by RECIST as compared with the non-HPD progressor group.

When evaluating the association between HPD and mechanism of action of the ICIs administered using RECIST criteria, we observed a clear trend for increased risk with anti–PD-1 treatment as compared with anti–PD-L1 antibodies (P = 0.077), not detected when evaluating TGR change (P = 0.775). The association with HPD was very strong for patients receiving ICI combinations using both RECIST criteria (P < 0.01) and TGR change criteria (P = 0.019) as compared with monotherapies. Of note, a total of 129 patients were treated with anti–PD-1 based combinations (representing 82.7% of the total patients treated in combination).

The concept of HPD, where patients present clinical deterioration secondary to ICI treatment, is a new and controversial phenomenon, which has important clinical implications. Currently, there are no easily applicable criteria for routine use in the clinic. In this retrospective study of patients receiving ICIs in the context of phase I clinical trials, HPD was observed in 10.7% of the cases (29 of 270) using an adapted RECIST definition and in 6.3% (14 of 229) using TGR evaluation. If we analyze only those patients who obtain PD as best response in the first tumor assessment, the prevalence is quite similar with 27.1% (29 of 107) and 24.1% (14 of 58) of cases, respectively. These results are comparable with studies published by Champiat and colleagues, with 24.4% of HPD in the cases with PD as best response (12 of 49) in phase I patients receiving PD-1/PD-L1 ICIs (24) and by Ferrara and colleagues, observing a 32.9% (56 of 170) of HPD exclusively in non–small cell lung cancer patients (29).

When comparing both ways to evaluate HPD, the main discordance is seen in those patients who are categorized as HPD when assessed by RECIST and non-HPD using TGR. This is mainly due to the criterion that takes into account the appearance of multiple new metastases, which are not captured with TGR evaluation. Arguably this is an important limitation, as in our study most patients with PD as best response (72 of 107, 67.3%) had multiple new metastases, and importantly these patients had a similar poor prognosis independently of the TGR, with a clear negative impact in OS. Other clinical trials published have shown that around 40% to 50% of patients with PD had new lesions (32, 33). The slight increase of our series could be explained given the nature of phase I population.

Understanding the concept of HPD as tumor burden increase due to ICI exposure, one limitation of using RECIST could be overestimating the HPD phenomenon in cases with intrinsically rapid aggressive disease. It is worth noticing that despite this being potentially true, we did not find differences in the TGR values during the reference period between the HPD group and the non-HPD progressor group using RECIST, or with the baseline tumor burden, both expected to be higher in aggressive tumors. On the other hand, we did observe higher TGR values during the experimental period in the HDP group by RECIST, consistent with a rapid progression while receiving ICIs therapy.

Independent of the fact that assessing HPD by RECIST criteria could be capturing patients with a rapid progression (irrespective of it being caused by the treatment or by the intrinsic behavior of the disease), it is able to identify a subgroup of patients with lower OS when treated with ICIs, and this could already have a relevant impact in the clinic. Importantly, HPD by RECIST criteria remained an independent poor survival factor in a multivariable model adjusted for known prognostic factors. Furthermore, 38 patients with PD as best response in our ICI cohort continued treatment to confirm PD; however, the median time from initial progression to study discontinuation was less than 1 month. This implies that if a patient meets RECIST criteria of HPD in his first tumor assessment, PD should not be confirmed with a second tumor assessment and an alternative rescue therapy without anti–PD-1/PD-L1 should be recommended. Our cohort was based on patients receiving ICIs in the context of phase I clinical trials, and this could represent a bias, as there is a smaller prevalence of patients with very bad prognosis, being one of the most common inclusion criteria having a life expectancy of more than 3 months. However our definition has been assessed in independent cohorts of patients with gynecological and genitourinary tumors treated with ICIs at our institution, with similar results (36, 37).

Other than the already mentioned logistical difficulties to quantify TGR due to the need of a previous-to-baseline CT scan, assessing the kinetics through TGR can also have limitations to capture the biological impact of the treatment. Our results show that a lower TGR during the reference period is associated with higher chances of an HPD classification in the later period (both in the ICI and TA cohorts). This could underline the fact that in tumors with a lower TGR prior to treatment it will be easier to reach a doubling of the TGR independently of the kind of treatment received. For example, a tumor with a change from a 10% volume increase during the month prior to the experimental treatment to a 20% increase during ICI treatment would be classified as a HPD; however, with a change from 50% during the reference period to 70% during the experimental period, it would not be considered as an HPD, despite having a higher absolute increase in tumor burden than the previous situation.

One important question that remains to be answered is if HPD is really a novel way of response exclusive of ICIs (29, 38). In this regard, we also analyzed a similar cohort of patients treated with TAs. Targets that have already been approved, such as BRAF inhibitors in BRAF-mutated tumors, were excluded from the analysis. HPD by RECIST was observed in 10.8% (26 of 239) of all population and in a 21.8% (26 of 119) of patients with PD as best response. However, it had no impact in mOS. We observed similar results analyzing TGR with 25% (12 of 48) of patients having HPD without an impact in survival. Thus, in our cohort of patients, the radiologic definition of HPD is met with both types of treatments; however, the impact in patients' outcome appears to be exclusively seen in the ICI cohort. This is in line with the previously described 14.2% (3 of 21) early PD seen in a chemotherapy cohort by Ferrara and colleagues. Understanding the limitations of the analysis, such as a selected patient population, heterogeneous tumor types, and previous treatments, with 20.4% having received previous ICIs, our results suggest the phenomenon of HPD (where the type of progression is not only a radiologic description but also has a biological impact with a worsening of survival) could be exclusive of ICIs. Given the fact that the half-life of ICIs is around 25 days, longer than most TAs (39–42), the deleterious effect of ICIs in some patients could be continued after stopping treatment, even Osa and colleagues demonstrated that PD-1–blocking antibody nivolumab persists in patients several weeks (>20 weeks) after the last infusion, bound to its receptor (43), which could explain in part this detrimental impact in mOS. To fully determine if the negative impact of survival seen in a subset of patients is exclusive of ICIs, this observation should be further evaluated in bigger cohorts of patients. Our definition based on RECIST criteria could be easily tested in the previously performed prospective clinical trials.

Finally, we analyzed if HPD was related with a specific ICI in our cohort. Using RECIST criteria we observed a clear trend of HPD in patients treated with anti–PD-1 therapies, and this association was very robust in patients treated with combinations based on an anti–PD-1 backbone. In most cases, the combination agent was an antibody against a lymphocyte receptor made with a IgG4_(S228P) Fc subtype, similarly to the most common used anti–PD-1 agents (44, 45). Preclinical models have demonstrated that an anti–PD-1 IgG4_S228P antibody can mediate cross-linking between PD-1⁺ T cells and FcγRI⁺ macrophages, resulting in macrophage-mediated phagocytosis of PD-1⁺ T cells (46). This seems to be in line with the observation published by Lo Russo and colleagues, where they found tumor infiltration by M2-like CD163⁺CD33⁺PD-L1⁺ clustered epithelioid macrophages in patients with HPD. Additionally, in patient-derived xenograft models, TG was enhanced by treatment with anti–PD-1, but not by anti–PD-1 F(ab)2-fragments punctuating the importance of the Fc section to cause the growth phenomenon (47). Interestingly, when analyzing the data presented in phase III clinical trials comparing ICIs with other therapies, the overcrossing of survival curves in the first months is more prominent in anti–PD-1 agents (4, 7, 9–11), and to our knowledge has not been described so far in positive clinical trials with anti–PD-L1 agents (6, 8, 48, 49). A potential hypothesis to explain HPD in some patients could be the fact that lymphocytes marked with these antibodies could interact with the immune system and be destroyed, causing a deleterious effect. IgG4_(S228P) could activate macrophages through its FcγRI and FcγRIIb, producing antibody-dependent cell-mediated phagocytosis. This is only a hypothesis and should be further investigated in prospective studies.

Our study has several limitations. It is a retrospective study in a single institution, although the data were collected prospectively for patients participating in phase I clinical trials. Both the investigational and control cohorts include a multitude of tumor types and different treatment regimens both as monotherapy and combination; this heterogeneity might potentially influence the survival results, despite multivariate analysis. Given the population were all phase I participants, the nature of the disease could be more aggressive than in other settings such as first- or second-line treatment. Most importantly, as discussed previously, capturing HPD by RECIST criteria may overestimate the phenomenon in specific cases with an aggressive intrinsic behavior, and underestimate it in patients with a rapid clinical deterioration not allowing confirmatory scans to be performed. However this definition can be used easily by oncologists in current practice, helping in the decision-making process of discontinuing a treatment with ICIs and looking for alternative treatments (50).

Our study is the first to evaluate how HPD with ICIs therapy can be captured based on RECIST criteria. Using a control arm with patients treated with TAs and an investigational arm with patients receiving ICIs, we show that the radiologic definition of HPD following both TGR criteria and RECIST criteria occurs in both cohorts. However, in our series, significant OS difference was observed exclusively with RECIST criteria in the ICI cohort. We proposed new criteria to capture HPD that is intuitive and easy to use in daily clinical practice. Importantly, it correlates with poor mOS and represents a contraindication for treatment beyond progression.

I. Matos reports receiving other remuneration in the form of an ESMO Research Fellowship - Translational Focus, sponsored by Roche. J. Martin-Liberal reports receiving speakers bureau honoraria from Astellas, Bristol-Myers Squibb, MSD, Novartis, Pierre Fabre, Pfizer, and Roche, is an advisory board member/unpaid consultant for Bristol-Myers Squibb, Novartis, Pierre Fabre, and Roche, and reports receiving other remuneration from Bristol-Myers Squibb, MSD, Novartis, Pierre Fabre, Pfizer, Roche, and Ibsen. C. Hierro reports receiving commercial research grants from Bayer, reports receiving speakers bureau honoraria from Ignyta and Lilly, and reports receiving other remuneration from Amgen, Merck and Roche. M. Vieto is an advisory board member/unpaid consultant for Debiopharma and Roche. I. Brana is an employee/paid consultant for MSD, reports receiving commercial research grants from AstraZeneca, Bristol-Myers Squibb, Gliknik, GlaxoSmithKline, MSD, Novartis, Pfizer, Shattuck Labs and Roche, and reports receiving speakers bureau honoraria from Bristol-Myers Squibb, AstraZeneca, and Merck Serono. J. Ros reports receiving speakers bureau honoraria from Amgen and reports receiving other remuneration in the form of a travel grant from Amgen. G. Villacampa reports receiving speakers bureau honoraria from MSD. M. Oliveira is an employee/paid consultant for Roche, GlaxoSmithKline, PUMA Biotechnology, AstraZeneca, and reports receiving speakers bureau honoraria from Roche and Seattle Genetics. M. Alsina reports receiving speakers bureau honoraria from Lilly, BMS, MSD, and Servier. A. Oaknin reports receiving speakers bureau honoraria from Roche, AstraZeneca, PharmaMar, Clovis Oncology, Tesaro, and Immunogen. E. Felip is an employee/paid consultant for AbbVie, AstraZeneca, Bergenbio, Blueprint Medicines, Boehringer Ingelheim, Bristol-Myers Squibb, Celgene, Eli Lilly, Guardant Health, Janssen, Medscape, Merck KGaga, Merch Sharp & Dohme, Novartis, Pfizer, Prime Oncology, Roche, Samsung, Takeda and Touchtime, and reports receiving other commercial research support from Fundacion Merck Salud and a grant for oncology innovation EMD Serono. J. Rodon is an employee/paid consultant for Peptomyc, Merck Sharp Dome, Kelund Pharma, Spectrum Pharmaceuticals, Roche, Certera, Bayer, and reports receiving other commercial research support from Bayer. J. Tabernero is an employee/paid consultant for Array Biopharma, AstraZeneca, Bayer, BeiGene, Boehringer Ingelheim, Chugai, Genentech, Inc., GenMab A/S, Halozyme, Imugene Limited, Inflection Biosciences Limited, Ipsen, Kura Oncology, Lilly, MSD, Menarini, Merck Serono, Merrimack, Merus, Molecular Partners, Novartis, Peptomyc, Pfizer, Pharmacyclics, ProteoDesign SL, Rafael Pharmaceuticals, F. Hoffmann-La Roche Ltd, Sanofi, SeaGen, Seattle Genetics, Servier, Symphogen, Taiho, VCN Biosciences, Biocartis, Foundation Medicine, HalioDX SAS, and Roche Diagnostics. R. Dientsmann reports receiving speakers bureau honoraria from Roche, MSD, Servier, Boehringer Ingelheim, Ipsen, Amgen, and Sanofi. E. Garralda is an employee/paid consultant for Roche/Genentech, Ellipses Pharma, Boehringer Ingelheim, Janssen, AstraZeneca, Seattle Genetics, Alkermes, and Neomed Therapeutics1, reports receiving commercial research grants from Novartis and Roche, and reports receiving speakers bureau honoraria from Bristol-Myers Squibb, Merck, Roche, and Thermo Fisher. No potential conflicts of interest were disclosed by the other authors.

Conception and design: I. Matos, J. Martin-Liberal, C. Hierro, A. Azaro, M. Vieito, J. Ros, J. Tabernero, R. Dienstmann, E. Garralda

Development of methodology: I. Matos, J. Martin-Liberal, C. Hierro, J. Ros, E. Elez, J. Tabernero, R. Dienstmann, R. Perez-Lopez, E. Garralda

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): I. Matos, J. Martin-Liberal, A. García-Ruiz, C. Hierro, M.O. de Olza, A. Azaro, M. Vieito, I. Braña, G. Mur, J. Mateos, M. Oliveira, M. Alsina, E. Elez, A. Oaknin, E. Felip, J. Rodón, J. Tabernero, R. Dienstmann, R. Perez-Lopez, E. Garralda

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): I. Matos, J. Martin-Liberal, A. García-Ruiz, C. Hierro, C. Viaplana, M. Vieito, I. Braña, G. Mur, J. Ros, J. Mateos, G. Villacampa, J. Rodón, R. Dienstmann, R. Perez-Lopez, E. Garralda

Writing, review, and/or revision of the manuscript: I. Matos, J. Martin-Liberal, A. García-Ruiz, C. Hierro, M.O. de Olza, C. Viaplana, A. Azaro, M. Vieito, I. Braña, G. Mur, J. Ros, J. Mateos, G. Villacampa, R. Berché, M. Oliveira, M. Alsina, E. Elez, A. Oaknin, E. Muñoz-Couselo, J. Carles, E. Felip, J. Rodón, J. Tabernero, R. Dienstmann, R. Perez-Lopez, E. Garralda

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): I. Matos, J. Martin-Liberal, A. García-Ruiz, C. Hierro, M.O. de Olza, C. Viaplana, A. Azaro, M. Vieito, I. Braña, G. Mur, J. Ros, M. Oliveira, M. Alsina, E. Elez, A. Oaknin, E. Muñoz-Couselo, J. Carles, E. Felip, J. Rodón, J. Tabernero, R. Dienstmann, R. Perez-Lopez, E. Garralda

Study supervision: A. García-Ruiz, C. Hierro, J. Ros, A. Oaknin, J. Carles, R. Dienstmann, E. Garralda

This research has been partially funded by the Comprehensive Program of Cancer Immunotherapy and Immunology (CAIMI) supported by the Banco Bilbao Vizcaya Argentaria Foundation (BBVA Foundation; grant 89/2017). The research leading to these results has received funding from “la Caixa” Foundation (LCF/PR/CE07/50610001). Cellex Foundation for providing research facilities and equipment.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.