Methodological standards for using the patient-reported outcomes version of the common terminology criteria for adverse events (PRO-CTCAE) in cancer clinical trials

It is well established that clinical investigators miss up to 50% of the symptomatic adverse events experienced by patients participating in cancer clinical trials, leading to potential underestimation of the risks of drug products.^1–4 Moreover, clinician reporting of symptomatic adverse events has been found to have limited inter-rater reliability. ⁵

To address these problems, the National Cancer Institute (NCI) contracted development of the Patient-Reported Outcomes version of the Common Terminology for Adverse Events (PRO-CTCAE). ⁶ The PRO-CTCAE is an item library containing 124 questions (“items”) for patent self-report about symptomatic adverse events, such as nausea, sensory neuropathy, or pain. Rigorous assessment of the measurement properties of the PRO-CTCAE has established its validity, reliability, sensitivity, recall period, and mode equivalence. ⁷ The PRO-CTCAE is the intellectual property of the NCI, is freely available for use in research, is available in numerous languages, and is now in wide use in oncology drug development programs (available at: http://healthcaredelivery.cancer.gov/pro-ctcae).

Methodological standards and best practices for using the PRO-CTCAE have evolved over time, based on foundational development work for the items, ongoing methodological research, and real-world experience. ⁸ Key best practices include:

Selecting specific PRO-CTCAE items for a given trial based on common and expected reactions in a given context of use, based on literature review, known properties of products, and prior qualitative/mixed-methods evaluations;

Administering the same items across all arms of a trial if multi-arm;

Including an open-ended question for collecting unsolicited patient-reported adverse events (“free text” item);

Collecting baseline PRO-CTCAE data and adjusting summary adverse event data at the patient level to account for baseline symptomatology (i.e. if the worst score during treatment is the same or better than the participant’s baseline score, the worst score is adjusted to zero in order to avoid misattribution of a pre-existing symptom to study treatment, but if the worst score during treatment is worse than the participant’s baseline score, the worst score is preserved at its reported level to assure that the full impact of treatment on the patient experience is captured);

Collecting PRO-CTCAE via an electronic platform;

Standardizing the schedule for PRO-CTCAE reporting, with the identical schedule in all study arms;

Monitoring patient adherence with scheduled PRO-CTCAE completion and reminding patients who miss reports.

When the PRO-CTCAE was developed, a default 7-day recall period was used for the items. A dedicated recall period study was conducted, showing that 1-day and 7-day recall perform well, while 2-week and 3-week recall periods also perform well—although with a small amount of information loss compared to 7-day recall. ⁹ Longer than a 3-week recall period did not perform satisfactorily. Therefore, 7-day recall was selected as a standard, with an option to use 2 week or 3 week recall periods when necessary to accommodate a given trial design. Moreover, it was recommended that the frequency of administration of the PRO-CTCAE match the recall period, that is, if a 7-day recall period is used, then the PRO-CTCAE should be administered on a weekly basis.

However, there has not been prior evidence demonstrating whether spacing out frequency of administration of 7-day recall PRO items to be longer than weekly results in loss of information. The paper in the current issue of Clinical Trials by King-Kallimanis et al. ¹⁰ from the U.S. Food & Drug Administration (FDA) addresses this question eloquently. In this study, data submitted to the FDA for the AURA3 randomized controlled trial of an oral targeted agent versus standard chemotherapy for lung cancer was evaluated, to compare symptom worsening between trial arms for items administered either weekly or every 6 weeks for nausea, vomiting, fatigue, diarrhea, constipation, and appetite loss. Weekly symptoms were reported via the PRO-CTCAE, and every-6-week symptoms were reported via the EORTC QLQ-C30. Items in both the PRO-CTCAE and QLQ-C30 employed a 7-day recall period.

The authors found that between-group differences for each of the symptoms were substantially less with every-6-week PRO reporting than with weekly reporting, indicating loss of information with every-6-week reporting. In addition, detection of high-grade adverse events was lower with every-6-week reporting, and the proportion of patients concluded “never” to have had a symptom was lower in the every-6-week arm, suggesting that every-6-week reporting missed detecting incident high-grade and any-grade adverse events. This finding makes sense intuitively. The purpose of adverse event reporting is to capture continuous data throughout a trial to detect all incident adverse events. If the frequency of eliciting adverse events from patients does not match the recall period, then there will be periods of missing data, as demonstrated in the study by King-Kallimanis et al. ¹⁰ This study provides highly compelling evidence that the frequency of PRO-CTCAE reports should match the recall period of items and reinforces the value of weekly PRO-CTCAE collection in cancer clinical trials.

It has been suggested that in clinical trials, weekly PRO-CTCAE collection might be used during initial cycles of treatment, with spacing out later in the trial. The rationale for such an approach is that most incident adverse events of interest in analyses will occur initially, and longer-term treatment tends to be characterized by greater stability in symptomatology. This is likely to vary from context to context. For drug products where such a pattern is anticipated, this strategy is reasonable.

In conclusion, it should be added to the above list of best practices for employing the PRO-CTCAE in cancer clinical trials that the frequency of administration should match the recall period of items, with both ideally being 7 days.