Overall, taskforce members agreed that the goal of this effort was to generate consensus definitions about PD-(L)1 inhibitor clinical resistance that would be most applicable for aspects of drug development and clinical trial design. The taskforce agreed that their efforts should not be aimed at generating treatment recommendations, and that the best clinical judgment should supplant any of the following recommendations as necessary.

An initial discussion topic among members of the taskforce concerned separation of clinical PD-(L)1 resistance scenarios. The taskforce recognized that an increasing number of clinical trials are being designed to address PD-(L)1 inhibitor resistance, and that these studies generally allow enrollment of patients with progressive disease who have received prior PD-(L)1 inhibitors.7 Taskforce members generally agreed, however, that enrolling patients with different clinical PD-(L)1 resistance scenarios onto the same study assessing a novel therapy may confound activity analyses and adversely affect the clinical development of otherwise active agents. As such, the taskforce agreed that the clinical scenario of PD-(L)1 inhibitor resistance is relevant and should be taken into consideration while developing consensus definitions.

Building on this premise, the taskforce specifically discussed how a patient whose cancer progressed within the first few weeks of anti-PD-(L)1 therapy may or may not have a biologically distinct mechanism of resistance compared with a patient who experienced an initial response, but then had disease progression later during treatment. The taskforce ultimately concluded that these two clinical scenarios were similar to the biological concepts of primary (or de novo/innate) and secondary (or acquired) resistance definitions that have been well established in oncology for other systemic treatment modalities. Therefore, discussing both scenarios in relation to PD-(L)1 inhibitor therapy would be warranted.

The taskforce noted, however, that these two clinical scenarios may differ from a patient whose disease progresses or relapses after cessation of treatment following an initial response, including patients whose disease progresses following a finite course of adjuvant or neoadjuvant therapy.8 Therefore, the taskforce also agreed to discuss clinical definitions of PD-(L)1 inhibitor resistance in the context of disease progression after treatment discontinuation, and more specifically for four settings: adjuvant treatment, neoadjuvant treatment, and after halting therapy due to aspects unrelated or related to toxicity.

Taskforce members noted other challenges in defining clinical resistance to PD-(L)1 inhibitor therapy, primarily in those patients demonstrating initial progression on imaging followed by response with continued therapy.2 9 For example, it was apparent in the early clinical experience with the anti-CTLA-4 antibody ipilimumab in metastatic melanoma that patients could experience long-term clinical benefit despite apparent progression of disease on the first set or sets of imaging. While the majority of patients in these studies who derived clinical benefit after ipilimumab treatment had unequivocal tumor shrinkage on imaging (as defined formally by the conventional RECIST response criteria and/or as judged by the treating physician), tumor shrinkage in some patients occurred months after the original treatment had been stopped in the setting of initial radiographic progression.10 Recognition that this phenomenon, now commonly referred to as “pseudoprogression,” occurs in approximately 5%–10% of patients receiving anti-CTLA-4 or anti-PD-(L)1 therapy has resulted in the development of new imaging criteria to measure response/progression while simultaneously accounting for this scenario,11 12 and has altered patient management by promoting treatment beyond traditional definitions of progression to confirm ICI resistance.13 Based on these discussions, the taskforce decided to consider the appropriate timeframes and response criteria for confirming PD-(L)1 inhibitor resistance in the three agreed on clinical scenarios.

Finally, the taskforce made a few general assumptions about the patient population to which these consensus definitions could be applied.

It was agreed that the goal of providing uniform definitions of resistance to PD-(L)1 inhibitor therapy was to minimize the chance that late responses would be attributable to subsequent therapies, but the taskforce recognized that available data to confirm such a cut-off were both inconsistent and limited. As such, there was general consensus that definitions should aim to identify a population of patients with a≤5% chance of having subsequent response if treatment was continued past progression with single-agent PD-(L)1 inhibitor or if therapy with single-agent PD-(L)1 inhibitor was restarted irrespective of intervening therapy, a cut-off which they agreed should be validated against existing clinical data sets from stakeholders within the field, as well as future data generated using the taskforce’s current resistance definitions.

The general biological definition of primary resistance to PD-(L)1 inhibitors can be defined as the inability of immune cells to mount an antitumor response on initial drug exposure.9 Multiple potential biological mechanisms of primary resistance have been proposed, including but not limited to, ineffective priming of a T cell response, lack of tumor recognition due to defective antigen presentation, inability of T cells to traffic to or penetrate effectively into viable areas of the tumor, inability of T cells to eliminate tumor cells due to suppression via other checkpoints such as T cell immunoreceptor with Ig and ITIM domains (TIGIT) and lymphocyte-activation gene 3 (LAG-3), an abundance of immune inhibitory cells such as M2-macrophages in the tumor, and others.14 Clinical definitions of primary resistance to PD-(L)1 inhibitor therapy, however, lack consistency across the field. For the purposes of drug development in the anti-PD-(L)1 setting, the taskforce felt it would be critical to define primary resistance with the intent of identifying patients who would not benefit from initial and more prolonged exposure to PD-(L)1 inhibitor monotherapy.

It was agreed by the taskforce that an important feature of any clinical definition of primary resistance is that it must reflect patients who truly have not received “clinical benefit” from initial therapy. While the term clinical benefit can vary across the field, in this context it means either tumor response or prolonged SD, as per RECIST version 1.1, lasting 6 months or greater, although the timeframe may need to be rethought in indolent tumor types.3 True resistance, or lack of clinical benefit, implies that drug exposure has been adequate to induce the desired biological effect required for antitumor activity, and additional drug exposure will not be effective; thus, it was critical to define a minimal exposure for sufficient PD-(L)1 inhibition to derive any possible clinical benefit. The taskforce noted that studies of PD-(L)1 inhibitors, which have a long circulating half-life, have involved repeat dosing every 2–4 weeks, and have failed to identify a major difference in efficacy or toxicity at doses greater than 1 mg/kg, although certain diseases may behave differently (eg, non-small cell lung cancer in the preliminary nivolumab experience), and higher doses may be necessary for certain PD-(L)1 inhibitors to induce antitumor activity.15–20 Based on these data, workshop attendees generally agreed that two cycles of therapy would be sufficient for assessing primary resistance. As such, the taskforce reached general consensus that a patient who has disease progression after receiving at least 6 weeks of exposure to PD-(L)1 checkpoint inhibitors, generally correlating with two complete cycles of U.S. Food and Drug Administration (FDA) approved PD-(L)1 inhibitor therapy, but no more than 6 months can be considered to have primary resistant disease (table 1).

Given the possibility of delayed responses to PD-(L)1 inhibitors, the taskforce agreed that a confirmatory set of radiographic scans (or in the case of clinically measurable lesions, physical examination) should be conducted to confirm resistance in patients who have documented progression after at least 6 weeks of starting therapy. The working groups varied in opinion regarding the timeframe in which this confirmatory scan should occur, but generally agreed on a 4–12 week range after initial evidence of disease progression. In general, the taskforce felt that patients should continue on therapy, if deemed clinically safe, until this second evaluation is performed. Furthermore, there was general consensus that late responses after rapid and confirmed radiographic progression (eg, lack of shrinkage), even at short time intervals such as 4 weeks, likely occur in no more than 5% of patients, and are unlikely to significantly affect results of subsequent clinical trials in which a patient may enroll (table 1). The taskforce recognizes that this aspect will need to be further validated with emerging data.

Patients who experience early toxicity requiring steroids and cessation of PD-(L)1 inhibitors are difficult to assess for primary resistance. For the purposes of future trial eligibility, randomization stratification or a fuller description of baseline characteristics on subsequent studies, primary resistance should be documented while still on active therapy or within 12 weeks of the last dose (see section on secondary resistance). The scenario of patients discontinuing therapy for low-grade toxicities and ultimately progressing off therapy greater than or equal to 12 weeks after last dose is addressed next (see section on progression after discontinuation of ICIs).

One exception to the requirement for confirmatory scans concerns patients who experience clear clinical progression, defined as a decline in the pretreatment performance status directly attributable to disease or increased disease-related symptoms and radiographic progression. The taskforce recognized that continued PD-(L)1 inhibitor therapy in patients with rapidly progressive disease would be unsafe, and the group subsequently indicated that the clinical judgment of the physician/clinical investigator might dictate an immediate change in the course of treatment, including enrollment onto clinical trials without fulfilling confirmatory imaging.

The working groups differed on which radiographic response evaluation criteria should be utilized to measure disease progression on PD-(L)1 inhibitors. The FDA typically uses RECIST1.1 as a primary trial endpoint, although alternative response criteria are used for solid tumors with disease sites/patterns of progression that are not well characterized by RECIST1.1. Additionally, immune RECIST (iRECIST) was specifically developed to address issues of mixed responses and pseudoprogression, but has yet to be the sole response criterion used to validate efficacy of anti-PD-(L)1 therapy in registration trials.12 One working group showed no preference between RECIST1.1 and iRECIST, while one working group each preferred the use of RECIST1.1 or iRECIST, respectively. The general consensus of the taskforce was that iRECIST was not yet standard and required additional validation. Efforts are ongoing to develop response criteria using positron emission tomography (PET) scans to measure changes in flurodeoxyglucose (FDG) uptake,21 yet the taskforce consensus was that this modality has not yet been sufficiently developed or validated. All three working groups agreed that FDG/PET may be an indicator of response to PD-(L)1 inhibitors and that interpretation may be confounded by metabolically active immune infiltrates that can accompany regressing tumor lesions. All three groups also stated that they were primarily concerned with ensuring that no more than 5% of patients enrolled on clinical trials are incorrectly determined to have primary resistant disease and felt that either RECIST1.1 with required confirmatory scans described above or iRECIST would accomplish this.

Patients who have initial SD by RECIST1.1 or iRECIST pose a challenge. There was considerable discussion regarding a duration of SD that would denote responsiveness rather than primary resistance. The general taskforce consensus was that patients who initially have SD on imaging but meet criteria for disease progression within less than 6 months of initiation of PD-(L)1 inhibitors are considered to have primary resistance.

All working groups identified important caveats and exceptions to the general definitions for primary resistance:

In general terms, the taskforce considered secondary resistance as that which arises when a patient is treated with antineoplastic therapy, has a documented, confirmed objective response or prolonged SD (>6 months), and then has disease progression in the setting of ongoing treatment. Biologically, this type of resistance may occur through adaptation of tumor cells, and may involve one or more epigenetic, transcriptomic and/or proteomic changes. Secondary resistance may also arise through the clonal selection or clonal evolution of tumor cells with outgrowth of clones containing genetic changes imparting resistance to therapy.2 9 14 25 Otherwise known as “acquired resistance,” secondary resistance mechanisms have been described with every type of antineoplastic therapy and, with respect to PD-(L)1 inhibitors, investigation into these mechanisms has only recently begun. To date, resistance-defining mechanisms include mutations in the interferon gamma response genes JAK1 and JAK2,25–27 and alterations in antigen presentation pathways, including downregulation and/or loss of beta-2-microglobulin.26 Additional potential mechanisms of secondary resistance include upregulation of proteins involved in alternative immune checkpoints, such as LAG-3 and T cell immunoglobulin and mucin domain 3 (TIM-3).25 28 29 Another general mechanism of secondary resistance is an increase in immunosuppressive cellular subsets in the tumor microenvironment such as T-regulatory cells (T-regs), myeloid-derived suppressor cells (MDSCs), and M2 macrophages (M2).30 31

As described above, the taskforce discussed whether distinguishing patients with primary from secondary resistance was necessary and, ultimately, came to the consensus that since the kinetics and/or biological mechanisms may be potentially distinct, it would be useful to do so. It was also agreed that the generated definitions herein would not aim to address these biological aspects at this time, but would rather be used to ensure that patients are being appropriately triaged for clinical trials testing new drugs or drug regimens, and are not being excluded from potential treatments and/or clinical trials. Studies addressing the treatment of anti-PD-1/PD-L1 resistant tumors can potentially use the generated definitions for:

For patients who derived clear clinical benefit (complete response (CR) or partial response (PR)) from PD-(L)1 inhibitors, the taskforce generally defined secondary resistance as the development of disease progression ≥6 months from initiation of PD-(L)1 inhibitors while still on active therapy. Scenarios involving patients with SD (defined by either RECIST1.1 or iRECIST), however, were not as easily classified within the secondary resistance definition and were further discussed in each of the three working groups, and among the taskforce at large. Some members of the taskforce expressed the opinion that patients with slow, steady disease growth that took 6 months to qualify as disease progression represented a different group of patients than those with any degree of disease regression, although not enough to meet objective criteria for a PR. Ultimately, the taskforce consensus was that patients with SD for ≥6 months who then had overall disease progression, independent of the target lesion measurements (ie, overall tumor regression of <30% or tumor growth <20%), would be considered as having secondary resistance (table 1). The taskforce noted that this definition might not be applicable to indolent tumors (see caveats below).

Two of three working groups recommended a confirmatory scan for validation of secondary resistance. Similar to the recommendations for primary resistance, the timeframe recommended by both working groups in favor of confirmation for such a scan was within 4–12 weeks after evidence of initial disease progression. In addition, these two working groups agreed that disease progression should be verified at ≥two metastatic sites/lesions in patients with multiple metastases, and that asymptomatic nodal-only disease progression was insufficient for classifying a patient as having secondary resistance without pathological confirmation. The third working group indicated no preference for a requirement of a confirmatory scan. The rationale given by this group was that the false-positive rate for defining progressive disease on therapy following clinical benefit, particularly in the setting of a CR or PR, is less than the agreed 5% threshold based on a single scan alone.

As in the setting of primary resistance, the taskforce was in agreement that patients who have clear clinical progression should not require a confirmatory scan to move onto subsequent treatment.

Similar to discussions concerning primary resistance, the working groups differed on which response criteria should be utilized for secondary resistance (RECIST1.1 vs iRECIST). While one group had indicated no preference, one preferred RECIST1.1 and the other iRECIST. Overall, the consensus of the taskforce was to ensure that the selected criteria would limit the false-positive rate of secondary resistance classification. In the absence of specific validation of either or both response evaluation criteria, it was felt to be reasonable to use either RECIST1.1 or iRECIST or both for clinical trial enrollment criteria.

The taskforce identified important caveats and exceptions to the general recommendations for defining secondary resistance.

Single agent PD-(L)1 inhibitors are effective for advanced disease in multiple tumor types, translating into overall survival benefit, and in some cases, leading to long-term responses off-treatment.2 Additionally, these agents are being increasingly utilized in the adjuvant and neoadjuvant settings, where a fixed treatment duration is employed. Anti-PD-1 therapies have been approved by the FDA as adjuvant treatment for patients with melanoma,32 33 and additional adjuvant/neoadjuvant indications are currently being tested in clinical trials.7 Based on the increased use of PD-(L)1 inhibitors in these settings, the taskforce decided to discuss adjuvant and neoadjuvant therapy in the context of progression after discontinuation of ICIs. In parallel, considering the fact that therapy could be discontinued in the metastatic setting either secondary to toxicity or after achieving maximal benefit, the taskforce also addressed this setting in its deliberations.

This form of resistance could have elements of either primary or secondary resistance as defined above. For instance, long-term disease-free survival following adjuvant therapy is achieved through either complete elimination of tumors or from ongoing antitumor immunological memory. Recurrence after discontinuation of therapy can be attributed to the kinetics of disease relapse with early relapse on adjuvant therapy resembling primary resistance, and late recurrence after therapy discontinuation and initial disease control resembling secondary resistance. The underlying mechanisms could thus be possibly related to the inability of immunotherapy to eradicate malignant cells because of lack of priming and activation of tumor antigen-specific immune cells, or the inability of cytolytic cells to infiltrate and/or kill tumor cells, a scenario consistent with the above described potential mechanisms of primary resistance. Alternatively, adjuvant immunotherapy might induce a new immune equilibrium that eventually results in sufficient immune editing of remaining tumor cells leading to the emergence of resistant clones, or immune cell exhaustion secondary to various mechanisms. The latter scenario would be more consistent with the definitions of secondary or acquired resistance as presented above. A noted difference between these two settings and resistance in the advanced setting, however, is that the bulk of the tumor is not present in the adjuvant setting.

In addition, there are pharmacokinetic and pharmacodynamic implications to the discontinuation of therapy. Available evidence indicates that receptor occupancy with anti-PD-1 nivolumab begins to decline 2–3 months after the single drug dose and is generally not identifiable at 6 months.15 While acknowledging that T cell activation may continue in the absence of receptor occupancy, those timeframes serve to provide guidance on defining resistance in this setting.

There was considerable discussion within the taskforce regarding the timeframe after which a patient could be labeled as having resistant disease in the adjuvant setting. Patients could be classified into two groups: “adequate treatment exposure” and “inadequate treatment exposure.” The adequate exposure group included patients who had disease progression within 6–12 weeks after the last dose of adjuvant therapy; after discussion all working groups generally agreed that this population of patients should be considered as having primary resistance to checkpoint inhibitors and subsequently follow the definitions/protocols as described above. The “inadequate exposure” population includes patients who had disease progression >12 weeks after their last dose of checkpoint inhibitor (table 2). Two of the three working groups suggested 6 months as a cut-off for the “relapse after cessation of treatment” population, but participants ultimately agreed on 12 weeks as consensus as to limit potential confusion in the event a patient has disease progression after adjuvant therapy in a timeframe between 12 weeks and 6 months (table 2).

Members of the taskforce generally agreed that patients who have progressive and/or recurrent disease after adjuvant therapy should undergo a biopsy to confirm recurrence. None of the working groups indicated that a confirmatory radiographic scan would be necessary (table 2).

All working groups identified exceptions to the general recommendations for resistance during adjuvant treatment.