Using patient-reported outcome measures in clinical trials: perspectives for and against a modular approach

Abstract

There is growing acknowledgement among international stakeholders, including regulators, health technology assessment (HTA) bodies and professional societies, that the inclusion of patient-reported outcome measures (PROMs) in clinical trials should be guided by clear rationales.¹ In this regard, existing PROMs may overreach or fall short in measuring domains most relevant for a specific study's context, population, treatments and stakeholders.

In response to the need for flexibility in assessing specific health-related quality of life (HRQoL) domains and regulatory recommendations,¹ PROM developers and stakeholders have provided guidance on a modular approach.^2-4 Although most guidance has focused on cancer clinical trials, the same principles are applicable to any context where PROMs are used (eg, clinical practice). In trials, the modular approach has been defined as the selection and assessment of specific patient-relevant and clinically relevant domains of interest, purposefully selected from multidomain PROMs, and then independently scored and interpreted.^{3, 4}

Selection of the most relevant disease- and treatment-specific domains can be based on previous research, consumer advice and clinical experience. Stakeholders have also recommended using conceptual frameworks and core patient-reported outcome (PROs) sets to guide the selection of PROM domains.⁵ HRQoL concepts identified as primary or key secondary outcomes should ideally be assessed using the most valid, reliable and dedicated measures, which are often more in-depth than domains in multidomain PROMs used to provide a broad overview. For example, if pain is a primary outcome in a trial investigating an anti-cancer treatment, the Brief Pain Inventory might be a comprehensive substitution for the two-item domain of pain from the widely used cancer-specific HRQoL PROM, the European Organisation for Research and Treatment of Cancer Quality of Life Core Questionnaire (EORTC QLQ-C30).⁶ Secondary outcomes, such as appetite loss and constipation, might be adequately assessed by the QLQ-C30's single-item domains for these outcomes. Thus, we propose three possible applications of the modular approach (Box 1): (i) using a study-specific conceptual framework, incorporating dedicated PROMs or domains to measure relevant HRQoL concepts; (ii) using a full-length PROM, removing domains less relevant to the study context or not covered by a core PRO set; or (iii) using a full-length PROM, substituting domains that are primary or key secondary outcomes with dedicated PROMs or domains.

Box 2 presents the cases for and against the use of a modular approach. In this perspective article, we discuss some of these arguments in the context of assessing clinical efficacy and conducting economic evaluations.

A modular approach can address respondent burden associated with lengthy PROMs and the potential for missing or poor quality PRO data in clinical trials⁷ (Box 3, Example 1). However, the aim of a modular approach is to prioritise assessment of the most important domains in each context, rather than to reduce PROM length. Indeed, patients may be willing to complete longer measures if items are relevant to their experience.¹⁰ Additionally, administering domains relevant to the clinical trial context can improve sensitivity to clinically important changes in PROs (Box 3, Example 2).

The modular approach also enables flexibility in assessing different domains at various timepoints. Domains that are conceptually proximal to the disease and treatment (eg, nausea severity) are more likely to change over a short period of time than more distal domains (eg, emotional function), and may therefore require more frequent assessment.¹¹

To preserve the psychometric properties of domains resulting from a modular approach, working with PROM developers and following published recommendations is encouraged.² If selected domains are obtained from PROMs validated in the target population, they are likely to retain psychometric properties, such as content and construct validity. As additional psychometric testing may not be required, this can be beneficial for novel treatments, given faster evaluation and approval times,¹² or for patients with rare cancers, where developing or validating bespoke disease-specific PROMs may be challenging.

The assessment of a broad set of items promotes the integration of PROMs into early-phase trials (phase 1 or phase 2) for capturing potential symptomatic adverse events and HRQoL problems. However, once these issues are identified, investigators must decide which domains to assess in late-phase trials, and to what extent. It is unsurprising that many may default to using full-length PROMs to maintain comparability with existing studies, avoid missing unexpected effects and meet HTA agencies’ requirements for economic evaluation.

To complicate matters, the modular approach necessitates greater scrutiny in the balance between the depth and breadth of issues and respondent burden (ie, time taken). Although this choice can be aided by international guidance and patient and public involvement,^{13, 14} it becomes less clear for domains amidst an abundance of options from various PROMs for the same HRQoL concept. Studies comparing the psychometric properties of PROMs in various study contexts can assist in this selection¹⁵ but, traditionally, conclusions have been drawn by assessing full-length PROMs instead of their domains. Psychometric evidence for specific combinations of subscales and items obtained from the EORTC suite are emerging,¹⁶ but evidence for the combination of subscales across different PROMs is lacking. This raises several questions requiring further conceptual and empirical consideration: Should only domains that have demonstrated the best psychometric evidence be selected and combined? How will a mixture of recall periods and response options across PROMs impact face validity and response distributions? What strategies can avoid biased selection of domains that may favour certain effects while downplaying others?

Guidance on the modular approach suggests that item order can affect the validity of subscales because responses may be influenced by the preceding questions.^{2, 4} These potential biases are usually addressed during early development of a PROM, which supports the argument for keeping PROMs in their original format unless there is potential for retesting. However, item order effects are mostly theoretical, with limited research available that has tested these directly.¹⁷ Order effects have primarily been considered at the questionnaire level and conclusions regarding their presence have only been conservatively applied to the PROMs assessed.¹⁷

The modular approach involves replacing existing full-length PROMs with a focused selection of HRQoL domains most relevant to a particular clinical trial context. Several considerations must be addressed before integrating the modular approach into clinical trials, including achieving broader acceptance from regulatory bodies (eg, HTA agencies). These considerations include expanding and clarifying the definition of what constitutes a module and how to apply the modular approach, while ensuring its suitability for economic evaluations. Further evidence that alleviates concerns associated with a modular approach, particularly evidence supporting its psychometric validity, would improve its acceptability in the field.

No relevant disclosures.

Not commissioned; externally peer reviewed.