Measuring Patient Reported Outcomes in Multiple Myeloma: Are Legacy Instruments Fit for Purpose

Introduction:

Measuring patient-reported outcomes (PROs) in oncology clinical trials is important to capture the patient perspective on quality of survival and the outcomes of health-related quality of life (HRQoL), symptoms, and functional status. In multiple myeloma (MM) randomized clinical trials (RCTs) the common PRO instruments include the EORTC QLQ-C30 and the EQ-5D as generic measures of a patient's health status. While these legacy PRO instruments have been widely used in the MM population, it is important to continue to assess the instruments psychometric properties for the specific context of use and target population.

The purpose of this research was to investigate the strengths and limitations of the PRO instruments when analyzed free from developer construct and scoring recommendations to assess fit for purpose in various MM clinical trial populations.

Methods:

Data for the analysis included 3 MM RCTs (n=1,773) (Dimopoulos 2016, Palumbo 2016, Mateos 2018). The PRO analysis set consisted of 33 items from the EORTC QLQ-C30 and EQ-5D-5L collected at baseline and during the treatment phase of the RCTs. Item analysis and modern psychometric methods were applied to the PRO item pool as a 4-step process: (1) descriptive analysis of the items, (2) assessment of factor structure, (3) performance of factors across RCT populations (treatment naive, relapsed/refractory MM), and (4) assessment of meaningful change.

Items were characterized using polychoric correlations to examine the inter-item association and relationships before determining appropriate factor structures. An exploratory factor analysis (EFA) followed by multidimensional item response theory (MIRT) and bifactor IRT models were conducted to determine how items fit into factors and to test a total symptom and total impact scoring option. Differential item functioning (DIF) was conducted to evaluation item and domain performance across study populations (treatment naïve vs. relapsed/refractory) and longitudinally (baseline vs. week 24). Meaningful change in these new domain constructs was evaluated using anchor-based techniques based on the EORTC-QLQ-C30 and EQ-5D-5L global items. A novel provisional latent class model approach creating responder groups was also developed and applied to assess meaningful change.

Results:

The 33 items from the EORTC QLQ-C30 and EQ-5D-5L demonstrated strong inter-item correlations supportive of pre-specified domain structures and consistent across RCT study populations. The EFA supported 3-domain structures for both symptoms and impacts separately. These domains were broadly consistent with the PRO instrument developer specifications but unique in several regards. These domains were statistically identical across RCTs. Model fit showed that a proposed bi-factor IRT (Symptom domain and Impact domain) fit better than the MIRT suggesting that subdomains are less useful than total domain scores. An extensive DIF analysis within this IRT framework found no evidence of substantial DIF effects. All evidence indicated that the items performed similarly across sample populations and time points.

Anchor-based meaningful change calculations for the newly created total symptom score and the total impact score demonstrated separation between global health anchor groups but the meaningful change estimates were small. The latent class models demonstrated that treatment arms differed significant in responder rates within 1 RCT.

Conclusions:

The analyses confirmed psychometrically robust EORTC QLQ-C30 and EQ-5D-5L items in the MM population under newly specified factor structures. In addition, performance of the items allows for calculation of a total symptom score and total impacts score. The newly developed total domains performed identically across MM populations (treatment naïve versus relapsed/refractory). The results only focus on the on-treatment phase of the RCT so future research is needed to determine how psychometric methods can be adjusted to better accommodate survival status and censoring for progression in oncology RCTs. The psychometric properties and consistency of the findings across the newly created symptom and impact total domain structures support the results that daratumumab maintains HRQoL in patients with MM.