Modernizing Clinical Trial Eligibility Criteria: Recommendations of the ASCO-Friends of Cancer Research Performance Status Work Group

An important goal of the American Society of Clinical Oncology, Friends of Cancer Research, and the oncology community at large is broadening clinical trial eligibility criteria to enhance trial access and accrual, and to ensure trial populations better reflect patients with the disease (1). Performance status (PS) is one of the most common inclusion/exclusion criteria in oncology trials. Many trials are limited to high-functioning participants (i.e., “good” PS) and exclude low-functioning patients (i.e., “poor” PS; ref. 2).

Two main PS scales are utilized in oncology clinical trials: Eastern Cooperative Oncology Group (ECOG; ref. 3) and Karnofsky (KPS) scales (4). Multiple trials in various tumor types and settings have demonstrated that low-functioning PS (i.e., ECOG PS, 2–4 and KPS ≤ 70) is correlated with lower overall survival and progression-free survival compared with high-functioning PS (ECOG PS, 0–1 and KPS, 80–100; refs. 5–13). Because of this, PS is included as a common eligibility criteria and stratification factor. However, this practice prevents trial enrollment for many patients and limits generalizability of trial results. Select trials that have focused exclusively on participants with low-functioning PS demonstrated patient and clinician interest and enrollment (14–17). The underlying etiology for low-functioning PS is also important; for patients whose low-functioning PS is due to disease burden, investigational treatment may result in improved PS with tumor control and symptom alleviation, especially with highly effective treatments. However, current PS scales do not differentiate causes of low-functioning PS.

In addition, there are limitations to PS assessments. PS is inherently subjective, which can affect interrater reliability (18) and invite potential bias particularly for patients at the borderline between values. For example, studies have demonstrated that clinicians assign patients aged >65 years higher numeric PS scores than younger patients, despite no difference in objectively measured physical activity (19). In addition, PS is less predictive of cancer-related outcomes for older adults (20, 21).

Because clinical trials frequently exclude PS2 patients, the working group chose to focus on this category. To understand the potential effect of including PS2 patients, the working group conducted a simulation study, where randomized trials of a hypothetical agent were simulated under various conditions. We also examined the literature to identify the potential risks and benefits of including PS2 patients on therapeutic clinical trials and evidence of the effectiveness of PS2 as a prognostic factor, reviewed past and current clinical trials to determine how often PS2 was included in inclusion/exclusion criteria, and developed consensus recommendations on how PS eligibility criteria could be revised while ensuring the safety of participants and integrity of the trial, and additional areas for research.

Small, mainly single institution studies have demonstrated that of patients deemed ineligible for a clinical trial, exclusion was related to poor PS in a significant proportion of patients, with variability across disease type, investigational therapy, and therapy line (22, 23). Even if other objective eligibility measures can be addressed, PS may remain a broad factor that excludes many patients (Table 1).

Most patients with cancer are aged ≥65 years, however, existing PS scales are inadequate in this population (20). Restrictive PS eligibility criteria contribute to the pervasive age disparity between trial participants and the overall cancer population, raising concerns about whether PS is unjustly limiting older populations' ability to participate in trials (24–26). Multiple studies have demonstrated that other tools, such as the geriatric assessment, are better than PS at evaluating older adults' overall health status (27) and at predicting chemotherapy toxicity (20). While a full geriatric assessment may not be practical due to length, subcomponents may provide a better functional assessment, such as instrumental activities of daily living that measure functional independence.

Benefits for patients with high-functioning PS may not reflect outcomes for patients with low-functioning PS (28, 29). Many eligibility restrictions from registration trials, such as line of therapy or cancer stage, are incorporated explicitly into the labeled indications with the exception of PS limitations. Therefore, therapies tested only in participants with high-functioning PS are administered to patients with lower functioning PS. This extrapolation may occur more readily with targeted and immunotherapies given greater efficacy (30). Therefore, evaluation of an investigational agent in participants reflective of the patient population is important. More inclusive PS eligibility will also likely increase enrollment of older adults (24, 31) and address the lack of evidence noted above (32, 33).

Rates of adverse events (AEs) may be greater in PS2 participants as compared with PS0 and PS1 participants, and this may influence patient's outcomes and ability to comply with study procedures. As a result, investigators and sponsors may be reluctant to consider trial enrollment. PS2 patients risk AEs with standard therapy options as well, and thus participation on a trial may not necessarily pose a greater risk of AEs compared with standard therapy for a particular patient. Because targeted therapies often have higher response rates, PS2 patients may experience a greater therapeutic index in a targeted therapy trial than standard of care (e.g., cytotoxic chemotherapy), even if their absolute rate of AEs is higher than in patients with PS0 and PS1. Where the comparative tolerability between an investigational agent and standard therapy is less clear, including PS2 patients (who may be more sensitive to toxicity) may unmask subtle differences. Importantly, having a subset of PS2 patients will add important safety data to facilitate decision-making for patients in the post-approval setting (Table 1). Determining appropriate timing for including PS2 participants is challenging. When possible, inclusion of a small number of PS2 participants in early-phase trials is recommended to guide separate expansion cohorts for phase II or broader inclusion into registration trials.

Even when clinical trial eligibility allows PS2 patients to enroll, relatively few PS2 participants are actually enrolled (34, 35). This may relate to clinicians' lack of familiarity with the investigational agent and concerns about the tolerability and safety. Enhanced information about safety, tolerability, and efficacy from earlier phase trials with the agent may help to counteract this. In addition, when clinically appropriate, allowing physician discretion in the treatment approach as a component of the clinical trial may help to mitigate this issue (36, 37).

In trials of novel therapies including PS2 participants, data suggest that outcomes may be inferior compared with participants with PS 0–1, even though low proportions of PS2 participants were included (38–40). This information alone should not be used as a justification for excluding PS2 patients. Instead, similar to other high-risk prognostic markers identified in oncology, PS information could be considered as a stratification factor. When safe, inclusion of participants with low-functioning PS provides valuable evidence to guide clinical care for most patients. Outcomes in low-functioning PS participants can also better inform statistical considerations for future trials.

The risk of inferior outcomes from low-functioning PS participants is a potential concern to sponsors, especially if compared with historical cohorts including high-functioning PS. The FDA has addressed a similar concern in a March 2019 final guidance on enrollment of patients with brain metastases stating, “to mitigate uncertainties about including patients with brain metastases in clinical trials, consider enrolling these patients in a separate subgroup within the trial” (41). In addition, FDA commentary has further indicated a willingness to restrict primary efficacy analysis to the participant subset who meet more conventional eligibility criteria when a sponsor enrolls a broader range of participants (42). FDA also notes that including a broader group of participants could offer benefits, such as additional information in drug labeling and/or reduced postmarketing commitments.

To explore the effects on inferences comparing trials that include versus exclude participants with PS2, simulations were conducted under a variety of trial settings with three levels of PS: PS0, PS1, and PS2. Figure 1 presents results based on: (i) total sample size of 500 participants, (ii) 1:1 randomization to two treatment groups, (iii) accrual time of 24 months, (iv) a time-to-event endpoint, and (v) follow-up until 283 events are observed, achieving power of 85% based on an HR of 0.70 versus a null hypothesis of 1.0 and a two-sided alpha of 0.05. Participants were assumed to vary in their median survival: 12-, 9-, and 6-month median survival in PS0, PS1, and PS2 participants, respectively. Differences in drop-outs due to AEs or other factors varied: 5%, 10%, and 20% of PS0, PS1, and PS2, respectively, and AEs were assumed to have censored event times within the first 4 months. Simulations assumed 45% PS0, 45% PS1, and 10% PS2 participants, and the true HRs reflecting treatment benefit were varied across PS groups. Scenario 1 assumes all three PS groups have the same treatment effect, HR = 0.7. Scenario 2 assumes PS0 and PS1 participants derive benefit, but PS2 participants do not (PS0 and PS1 HR, 0.7 and PS2 HR, 1.0). Scenario 3 assumes PS2 participants derive greater benefit compared with PS0 and PS1 participants (PS0 and PS1 HR, 0.7 and PS2 HR, 0.5). Outcome measures that were assessed to determine the differences in inferences due to the variability in HRs across the groups were (i) the estimated HR, (ii) power, and (iii) time to complete the study because fewer patients would be excluded (measured as the time from the first enrolled participant to the last event required for analysis). Inferences from simulated trials (10,000/scenario) were analyzed under two different approaches: (i) excluding PS2 participants (N = 450 PS0 and PS1 patients included in analysis) and (ii) including the PS2 participants (N = 500 for analysis). When excluding PS2 participants, the analysis was undertaken when there were 283 events among the PS0 and PS1 participants.

The simulation study demonstrated the following conclusions for including PS2 participants:

• (i) when the number of PS2 participants is relatively small (e.g., 10%), the effect on the estimated HR and power is relatively modest, even when the PS2 participants do not have a true treatment benefit (Fig. 1A and B).

• (ii) including PS2 participants is likely to shorten duration of the trial by increasing the number of potentially eligible trial participants (Fig. 1C) and due to the higher event rate in PS2 participants relative to PS0–1 participants.

These conclusions may not be generalized to all trial settings. Single-arm trials need attention given that previous trial results (to which the study results will be compared) may not have included PS2 participants. Similarly, trials with smaller (or larger) sample size may have more dramatic or muted effects depending on other trial parameters, such as the fraction of PS2 participants.

• Assessing safety concerns should take into account the potential increased risk in AE rates between standard-of-care and experimental intervention, rather than the absolute rate of expected AEs.

• Reassess and revise PS eligibility criteria at each phase of drug development, in accordance with growing knowledge about the investigational agent. Early-phase data (AE rates and durable objective responses) for PS2 participants can decrease uncertainty of subsequent randomized trials. For example, trials could:

  • (i) include an exploratory PS2 cohort in early-phase trials to collect data without compromising internal validity and to inform inclusion in later phase trials, incorporating early stopping rules for unacceptable toxicity, or

  • (ii) if tolerability/safety is acceptable during early phase for PS0–1 participants, expand to include PS2 participants in later phases.

• Consider alternate trial designs and settings. Examples may include:

  • (i) trials specifically for PS2 participants and, where appropriate, PS3 participants. This may be most ideal for studies of modified (“deintensified”) regimens where the overall goal is to develop a more tolerable therapy.

  • (ii) flexibility in the dosing schema, particularly for palliative trials. For example, enable investigator discretion to allow participants to initiate treatment at a reduced dosage with escalation to full dosage based on tolerability (37). This may be most appropriate for studies in advanced cancer where the goal of therapy is palliation.

  • (iii) Consider expansion cohorts to enhance enrollment of PS2 patients. This may be the most effective strategy for therapies with novel mechanisms or less well-defined AE profiles, whereby initial enrollment includes patients with high-functioning PS and once safety and tolerability are better understood, expansion to include PS2 patients occurs.

  • (iv) A postmarketing study that focuses on subgroups not well represented in premarket studies (43, 44). This may be most effective strategy for approved therapies where limited data currently exists for patients with low-functioning PS.

• Discuss study design and statistical analysis approaches for broader eligibility and implications for postmarketing research with FDA during trial design, where appropriate. This may include performing simulations under a variety of assumptions regarding fraction of PS2 patients and heterogeneity of efficacy and safety across PS groups.

Recommendations for inclusion of PS2 participants are included in Table 2. Although discussion has focused on inclusion of PS2 participants, PS3 participants should also be considered. With targeted therapies for rare alterations, inclusion of PS3 participants may be considered to expand the eligible patient population, if the agent has demonstrated favorable toxicity and efficacy signals.

Alternate methods for assessing physical function exist, such as patient-reported outcome measures (45), objective performance measures (e.g., gait speed; ref. 46), and activity monitoring devices (e.g., wearable devices; ref. 47). Further research is needed to understand how to incorporate and use these alternative methods in oncology trials. Enhancing the objectivity of PS assessments may more accurately characterize functional capacity and improve trial suitability assessment, particularly if low-functioning PS is related to disease burden versus other factors around the time of diagnosis. Incorporating these methods may also reduce bias of PS assessments.

The majority of newly approved investigational agents have targeted mechanisms of action, however, the safety and efficacy of many of these therapies remain unclear in the PS2 population given their underrepresentation on clinical trials leading to approval (48). Understanding safety and efficacy of novel therapies in PS2 patients, particularly for patients with low-functioning PS due to disease burden, is a critical area of need, as a targeted therapy or immunotherapy with a high objective response rate may afford improvement in PS by improving disease-related symptoms.

Broadening PS eligibility criteria to be more inclusive can increase the number and diversity of trial participants. More effective biomarker-driven therapies warrant reconsideration of this traditional approach. Trial sponsors should justify exclusion of PS2 patients and limit exclusions to those affecting patient safety and trial integrity. Several strategies can encourage broader inclusion of PS2, and in select cases PS3, participants. Implementation of these recommendations will require cooperation of multiple stakeholders and can result in incentives following FDA approval.

A. Magnuson reports receiving National Institute on Aging K76 Career Development Award, NCI Loan Repayment Award, and honoraria for educational lecture provided at the American Society of Radiation Oncology 2020 Annual meeting. K.D. Wilner reports other from Pfizer Inc outside the submitted work. H.D. Klepin reports personal fees from Pfizer outside the submitted work and is a contributor to UpToDate. M.E. Fleury reports grants from Genentech, Bristol Myers Squibb, Merck, and IQVIA outside the submitted work. A. Melemed reports full-time employment with Chimerix, a biopharmaceutical company, and was previously employed by Eli Lilly and Company as a full-time employee. F.I. Arnaldez reports other from MacroGenics, Inc. outside the submitted work. No disclosures were reported by the other authors.

The opinions expressed in this article are those of the authors and do not necessarily reflect the views or policies of the authors' affiliated institutions.

This article was developed as a consensus document of the Performance Status Working Group as part of a collaboration among ASCO, Friends, the FDA, and the NCI. The contents of this document were presented on August 2, 2019, as part of an invitation-only, collaborative workshop. The authors thank all the participants of the workshop, as well as the planning committee, ASCO, Friends, FDA, and NCI staff, for their expertise and thoughtful input on the project and article, and for logistical support. A. Magnuson was supported by a National Institute on Aging Beeson Career Development Award (NIA K76 AG064394). S.M. Lichtman was supported, in part, through the NIH/NCI Cancer Center Support Grant P30 CA008748.