Implications for Design and Analyses of Oncology Clinical Trials During the COVID-19 Pandemic

Current enforcement of quarantines for patients and staff, site closures, postponement of nonurgent medical appointments, and interruptions to supply chains for investigational products are leading to challenges in meeting protocol-specified procedures required for assessment of clinical trial end points. The recent report put forth by the US Food and Drug Administration recognizes that the conduct of clinical trials may be affected during this time.¹ Published recommendations by McDermott and Newman² in JAMA suggest several solutions for maintaining trial integrity across a broad spectrum of clinical trials. However, the direct implications for the design and analyses of end points specifically from ongoing oncology clinical trials that critically depend on serial data collection have not yet been addressed.

Modifications and/or delays in assessment schedules for patient monitoring will directly affect estimation of clinical trial end points. To permit evaluation of the usability and interpretability of results, it is imperative that coronavirus disease 2019 (COVID-19)–relevant patient data be collected now and analytic plans be put in place to facilitate comparisons with prepandemic participants.³

In single-arm early-phase studies, delaying or skipping early scans may result in missing responses for therapies with a short duration of response affecting estimation of both the objective and best overall response rates. Failure to capture these responses will result in underestimation of the response rate and may lead to overlooking potentially promising therapies. Trials using a 2-stage design may be stopped erroneously. Alternatively, for therapies with a longer duration of response, less rigorous adherence to the assessment schedule may have an attenuated effect on the primary end point.

Skipping assessments can also result in overestimating progression-free survival (PFS). Although the effect on trial conclusions may be worse in single-arm studies focusing on progression status by a specific time point because of an increased likelihood of finding a therapy promising, results from randomized comparative trials will also need to be viewed cautiously. Furthermore, estimates obtained from these larger studies will not reliably characterize efficacy. This pertains not just to PFS but to other time-to-event outcomes requiring in-person visits that are frequently included as secondary end points such as time-to-tumor progression, duration of response, and disease-free survival.⁴

As telemedicine options are being implemented at lightning speed, adverse events such as assessment of pruritus can adequately be evaluated using remote video consultation. Patient self-reporting may also be acceptable if necessary. Mode of reporting these events should be documented.¹ However, adverse events that require testing for organ function or laboratory abnormalities will be affected, complicating interpretation of safety data from these trials. The window in which dose-limiting toxic effects are assessed should not be changed to accommodate these challenges.

When presenting results on response rates, PFS, or adverse events, we recommend reporting results separately for the time preceding the COVID-19 outbreak and the current time period when the effect on data collection is expected to be greatest. Future results collected after the pandemic has waned should also be presented separately and may be combined with prepandemic data.

Aside from complexities introduced by the inability to complete necessary in-person visits, complications may arise if trial participants cannot get access to the treatment regimens outlined in the protocol. Estimates of effect sizes quantifying therapeutic benefit may be attenuated. Although we maintain that primary analyses should be consistent with the intent-to-treat principle and use all enrolled patients, we recommend that secondary/sensitivity analyses also be performed excluding or censoring patients affected by clinical trial restrictions during the pandemic. Reasons for excluding patients from the analysis should be documented and clearly reported.

It is too early to understand how COVID-19 will affect mortality in the cancer population. Cause of death attributable to COVID-19 should be carefully tracked for clinical trial participants. Although the primary analysis should be consistent with what was previously specified in the protocol, there should also be sensitivity analyses that may include censoring patients at the time of death due to COVID-19 or a competing risks analysis where death from COVID-19 is treated as a competing event.⁵ In this case, as in the analyses recommended herein, there is the possibility that the primary and sensitivity analyses might be dissonant and some trials will never yield clear conclusions because of this disruption.

In the midst of this pandemic, ensuring the safety of patients, trial investigators, and staff should certainly be prioritized over obtaining serial measurements on patients for a research study, even if those measurements are important for the trial outcome. That being said, every effort should be made to obtain the final image or data point corresponding to when the primary end point is to be assessed. If this is not possible, we recommend that thought be given to expanding enrollment, with careful consideration of safety data, such that the number of patients with data evaluable for the primary end point is consistent with the original design specified in the protocol.

Our goal is to raise awareness of potential issues regarding the analysis and interpretation of data arising from ongoing oncology clinical trials and to draw attention to the need for further guidance from regulatory agencies and review boards. The oncology community will need to think carefully about the sweeping implications the pandemic will have not only for current studies, but for future clinical trial design as well. Sample size calculations and decision rules over what constitutes a promising effect rely on results from previously conducted clinical trials. Using currently open clinical trials as that benchmark for future studies might be difficult. Even estimates of overall survival, which as an end point is not prone to the same challenges as end points requiring in-person visits, may not be generalizable or comparable to estimates from previously published and future studies. Maintaining the integrity of our clinical trials as best we can during this difficult time is our responsibility to patients and the scientific community.

Corresponding Author: Chaya S. Moskowitz, PhD, Memorial Sloan Kettering Cancer Center, Department of Epidemiology and Biostatistics, 485 Lexington Ave, 2nd Flr, New York, NY 10017 (moskowc1@mskcc.org).

Published Online: June 18, 2020. doi:10.1001/jamaoncol.2020.2370

Conflict of Interest Disclosures: Dr Panageas has stock ownership in Johnson and Johnson, Pfizer, Viking Therapeutics, and Catalyst Biotec. Dr Panageas and Dr Moskowitz reported grants from the National Institutes of Health during the conduct of the study. No other disclosures were reported.

Funding/Support: Support for this work was provided to Memorial Sloan Kettering Cancer Center by a core grant from the National Cancer Institute (P30 CA008748).

Role of the Funder/Sponsor: The National Cancer Institute and Memorial Sloan Kettering Cancer Center had no role in the preparation, review, or approval of the manuscript, nor the decision to submit the manuscript for publication.