Explanation

For each outcome, including PROs, the trial protocol should define four components: the specific measurement variable, which corresponds to the data collected directly from trial participants (eg, Beck Depression Inventory score, all-cause mortality); the participant-level analysis metric, which corresponds to the format of the outcome data that will be used from each trial participant for analysis (eg, change from baseline, final value, time to event); the method of aggregation, which refers to the summary measure format for each study group (eg, mean, proportion with score >2) and the specific measurement timepoint of interest for analysis.¹ Many PRO questionnaires are multidimensional, assessing multiple facets of the impact of a disease and its treatment and usually include multiple assessments over the course of the trial. The multidimensional nature of PROs is most apparent in HRQL questionnaires, which often include various aspects of functioning and symptoms, which are often scored as distinct ‘domains’. These domains may not be affected equally by the trial interventions. The SPIRIT-7-PRO Extension encourages protocol writers to identify the domains that are most likely to be affected in the trial objectives and hypotheses, drawing on previous evidence (SPIRIT-6a-PRO Extension). The SPIRIT-12-PRO Extension item reinforces the statement of these key domains, and also the most important timepoints (ie, where greatest impact of interventions are expected), and develops that concept further by encouraging protocol contributors to think about how these PRO domains and timepoints will be analysed, that is, the analysis metric.³⁴ To ensure transparency and credibility of the analysis, it is recommended that there is prespecification of the PRO concepts/domains, analysis metric(s) and timepoint(s) of interest, whether the PRO is a primary, secondary or exploratory outcome. These should closely align with the study hypotheses/objectives and the nature and trajectory of the disease or condition under investigation.^{34 43} The selected key domains, timepoints and analysis metric should be used to specify the PRO endpoints, integrated in the full endpoint model of the trial.

A clearly defined endpoint model, organising all trial outcomes (PRO and non-PRO), typically in primary, secondary and exploratory endpoints, allows rigorous control of the evidence demonstration, especially the control of the statistical testing. Each PRO endpoint in the model should explicitly specify a single domain and a single time horizon. The endpoint model enables procedures for type I error control (risk of false positive finding) (see SPIRIT-20a-PRO Elaboration). Broadly, the concepts and domains (sub concepts) measured by a PRO may be ‘proximal’ in nature, that is, direct impact of the disease and treatment (eg, symptoms such as pain, fatigue, nausea, rash and anxiety) or more distal, ‘knock-on’ effects, (eg, functional status and global quality of life), as illustrated for ovarian cancer in (figure 1, inspired by the Wilson and Cleary model⁴⁴). Of note, the Food and Drug Administration (FDA) are increasingly focused on the individual measurement of well-defined concepts that impact on HRQL but are more proximal to a therapy’s effect on the patient and the patient’s disease: symptomatic adverse events, physical function and, where appropriate, a measure of the key symptoms of the disease.⁴⁵

Proximal and distal effects of therapy on patient symptoms and quality of life. Adapted from Wilson and Cleary.⁴⁴

Common analysis metrics may include magnitude of event at time t, proportion of responders at time t, overall PRO score over time or response patterns/profiles. These should be prespecified alongside the levels of statistical and clinical significance for the study and any responder definition in use.¹⁶ Timepoints for analysis should be chosen to best address the research question, while taking into account aspects such as the natural history of the disease/condition and its treatment, the PRO measurement properties and recall period and participant completion burden.^{16 46}

The example, idelalisib and rituximab improve PFS over rituximab alone in unfit patients with relapsed CLL: a phase III study, illustrates a ‘time to event’ PRO endpoint or analysis metric, where the event is definitive improvement or definitive deterioration in a PRO. This approach allows repeated PRO measurements to be converted to a single measure: time to definitive increment or decrement. This requires quite complex and specific criteria for degree and duration of change. Also, this particular example does not specify any key domains of the FACT-Leu, but rather applies this analysis metric to all HRQL domain and symptom scores. In contrast, the RATE-AF example identifies a single score (the SF-36 physical component score), and a specific timepoint (6 months) as the primary outcome, with other SF-36 domains, questionnaires and timepoints specified as secondary outcomes.