Special Issue PRO-Analysis of Clinically Meaningful Change on Patient-Reported Outcomes: Renewed Insights About Covariate Adjustment.

Determining clinically meaningful change (CMC) in a patient-reported (PRO) measure is central to its existence in gauging how patients feel and function, especially for evaluating a treatment effect. Anchor-based approaches are recommended to estimate a CMC threshold on a PRO measure. Determination of CMC involves linking changes or differences in the target PRO measure to that in an external (anchor) measure that is easier to interpret than and appreciably associated with the PRO measure. One type of anchor-based approach for CMC is the "mean change method" where the mean change in score of the target PRO measure within a particular anchor transition level (e.g. one-category improvement) is subtracted from the mean change in score of within an adjacent anchor category (e.g. no change category). In the literature, the mean change method has been applied with and without an adjustment for the baseline scores for the PRO of interest. This article provides the analytic rationale and conceptual justification for keeping the analysis unadjusted and not controlling for baseline PRO scores. Two illustrative examples are highlighted. The current research is essentially a variation of Lord's paradox (where whether to adjust for a baseline variable depends on the research question) placed in a new context. Once the adjustment is made, the resulting CMC estimate reflects an artificial case where the anchor transition levels are forced to have the same average baseline PRO score. The unadjusted estimate acknowledges that the anchor transition levels are naturally occurring (not randomized) groups and thus maintains external validity.

Meaningful within-patient change for clinical outcome assessments: model-based approach versus cumulative distribution functions.

OBJECTIVES: The FDA recommends the use of anchor-based methods and empirical cumulative distribution function (eCDF) curves to establish a meaningful within-patient change (MWPC) for a clinical outcome assessment (COA). In practice, the estimates obtained from model-based methods and eCDF curves may not closely align, although an anchor is used with both. To help interpret their results, we investigated and compared these approaches. METHODS: Both repeated measures model (RMM) and eCDF approaches were used to estimate an MWPC on a target COA. We used both real-life (ClinicalTrials.gov: NCT02697773) and simulated data sets that included 688 patients with up to six visits per patient, target COA (range 0 to 10), and an anchor measure on patient global assessment of osteoarthritis from 1 (very good) to 5 (very poor). Ninety-five percent confidence intervals for the MWPC were calculated by the bootstrap method. RESULTS: The distribution of the COA score changes affected the degree of concordance between RMM and eCDF estimates. The COA score changes from simulated normally distributed data led to greater concordance between the two approaches than did COA score changes from the actual clinical data. The confidence intervals of MWPC estimate based on eCDF methods were much wider than that by RMM methods, and the point estimate of eCDF methods varied noticeably across visits. CONCLUSIONS: Our data explored the differences of model-based methods over eCDF approaches, finding that the former integrates more information across a diverse range of COA and anchor scores and provides more precise estimates for the MWPC.

Finite Mixture Models, a Flexible Alternative to Standard Modeling Techniques for Extrapolated Mean Survival Times Needed for Cost-Effectiveness Analyses.

OBJECTIVES: To compare finite mixture models with common survival models with respect to how well they fit heterogenous data used to estimate mean survival times required for cost-effectiveness analysis. METHODS: Publicly available overall survival (OS) and progression-free survival (PFS) curves were digitized to produce nonproprietary data. Regression models based on the following distributions were fit to the data: Weibull, lognormal, log-logistic, generalized F, generalized gamma, Gompertz, mixture of 2 Weibulls, and mixture of 3 Weibulls. A second set of analyses was performed based on data in which patients who had not experienced an event by 30 months were censored. Model performance was compared based on the Akaike information criterion (AIC). RESULTS: For PFS, the 3-Weibull mixture (AIC = 479.94) and 2-Weibull mixture (AIC = 488.24) models outperformed other models by more than 40 points and produced the most accurate estimates of mean survival times. For OS, the AIC values for all models were similar (all within 4 points). The means for the mixture 3-Weibulls mixture model (17.60 months) and the 2-Weibull mixture model (17.59 months) were the closest to the Kaplan-Meier mean estimate of (17.58 months). The results and conclusions from the censored analysis of PFS were similar to the uncensored PFS analysis. On the basis of extrapolated mean OS, all models produced estimates within 10% of the Kaplan-Meier mean survival time. CONCLUSIONS: Finite mixture models offer a flexible modeling approach that has benefits over standard parametric models when analyzing heterogenous data for estimating survival times needed for cost-effectiveness analysis.