Patient-reported outcome (PRO) data help to determine evidence on risks, benefits, safety and tolerability of treatment in randomized-controlled clinical trials (RCTs), according to a report published in The Lancet.
In the report, called the Setting International Standards in Analyzing Patient-Reported Outcomes and Quality of Life (SISAQOL) Endpoints Data Consortium, detailed recommendations for PROs and ongoing developments using critical literature reviews and a structured collaborative process with diverse international stakeholders.
“PRO data, such as symptoms, functioning, and other [health-related quality of life; HRQOL] endpoints are increasingly assessed in cancer RCTs to provide valuable evidence on risks, benefits, safety, and tolerability of treatment,” the authors wrote. “The current SISAQOL recommendations represent an important first step towards generating international consensus-based standards for PRO analysis in cancer RCTs.”
Four issues were prioritized by the group, including developing a taxonomy of research objectives that can be matched with appropriate statistical methods, identifying appropriate statistical methods for PRO analysis, standardizing statistical terminology related to missing data, and determining appropriate ways to manage missing data.
In regard to the taxonomy of research objectives, the consortium decided that essentially 4 key attributes should be considered a priority for each PRO domain. They recommended including a broad PRO research objective comprising treatment efficacy and clinical benefit (confirmatory) or a description of the patient perspective (exploratory or descriptive), a between-group PRO objective consisting of superiority or equivalence or non-inferiority, within-treatment group PRO assumptions for the treatment or control group such as worsening, stable state, improvement, or overall effect, and within-patient or within-treatment PRO objectives consisting of time to event, magnitude of event, a time t, proportion of responders at time t, overall PRO score time, or response patterns or profiles.
Additionally, a set of necessary and “highly desirable” statistical criteria for defining appropriate statistical methods to PRO analysis were recommended. If a method did not meet a fundamental criterion, then it was not recommended as appropriate for PRO analysis. Two crucial statistical properties were identified. One was the ability to do a comparative test (statistical significance), and the other the ability to produce interpretable treatment effect estimates (clinical relevance).
Both the Cox proportional hazards model and linear mixed models were recommended as statistical methods. No agreement was reached in regard to the appropriate statistical method for evaluating longitudinal data, and no agreement was reached on a recommended summary measure for PRO data over time.
“Further investigation is needed for whether it is appropriate to analyze ordinal data as continuous; discussions on this issue revolved around statistical approximation, complexity of the model, and ease of interpretation,” the authors wrote.
The missing PRO data definition was recommended to be defined as data that would be meaningful for the analysis of a given research objective, but were not collected. This indicates that not all unobserved assessments are considered as missing data, however, depending on the method of analysis, all unobserved assessments may implicitly be treated similarly as missing data.
The review stressed the relevance of differentiating missing observations relative to a reference set of expected data. This decision resulted in 2 definitions, one being the so-called available data has a fixed denominator, defined as the number of patients in the PRO study population. The other decided definition was that the completion rate has a variable denominator, defined as the number of patients on PRO assessments at the designated timepoint. The numerator of both rates was characterized as the number of patients submitting a valid PRO assessment at the designated timepoint.
Though no agreement was reached on the threshold to define substantial missing data, it was recommended that collecting reasons for missing data is essential to assessing the effect of missing data for PRO findings. However, a case-report form to collect reasons for missing data in a standardized method is necessary and the consortium indicated that they intend to develop one in the future. General guidelines on how to handle missing data were proposed that were consistent with existing regulations.
“Generating robust PRO conclusions from cancer clinical trials is not only about agreeing on and using standardized research objectives and analysis standards,” the authors wrote. “We believe this set of recommendations will support clinical researchers, trialists, and statisticians to improve the conceptualization and design of PRO studies, the quality of statistical analysis, and the clinical interpretation of PROs in cancer clinical methods.”
Coens C, Pe M, Dueck AC, et al. International standards for the analysis of quality-of-life and patient-reported outcome endpoints in cancer randomised controlled trials: recommendations of the SISAQOL Consortium. The Lancet. doi:10.1016/S1470-2045(19)30790-9.