Testing the "RCT augmentation" methodology: A trial simulation study to guide the broadening of trials eligibility criteria and inform on effectiveness.

Background

Exclusion criteria that are treatment effect modifiers (TEM) decrease RCTs results generalisability and the potentials of effectiveness estimation. In "augmented RCTs", a small proportion of otherwise-excluded patients are included to allow for effectiveness estimation. In Hodgkin Lymphoma (HL) RCTs, older age and comorbidity are common exclusion criteria, while also TEM. We simulated HL RCTs augmented with age or comorbidity, and explored in each scenario the impact of augmentation on effectiveness estimation accuracy.

Methods

Simulated data with a population of HL individuals initiating drug A or B was generated. There were drug-age and drug-comorbidity interactions in the simulated data, with a greater magnitude of the former compared to the latter. Multiple augmented RCTs were simulated by randomly selecting patients with increasing proportions of older, or comorbid patients. Treatment effect size was expressed using the between-group Restricted Mean Survival Time (RMST) difference at 3 years. For each augmentation proportion, a model estimating the "real-world" treatment effect (effectiveness) was fitted and the estimation error measured (Root Mean Square Error, RMSE).

Results

In simulated RCTs including none (0%), or the real-world proportion (30%) of older patients, the interquartile range of RMST difference was 0.4-0.5 years and 0.2-0.3 years, respectively, and RMSE were 0.198 years (highest possible error) and 0.056 years (lowest), respectively. Augmenting RCTs with 5% older patients decreased estimation error substantially (RMSE = 0.076 years). Augmentation with comorbid patients proved less useful for effectiveness estimation.

Conclusion

In augmented RCTs aiming to inform the effectiveness of drugs, augmentation should concern in priority those exclusion criteria of suspected important TEM magnitude, so as to minimie the proportion of augmentation necessary for good effectiveness estimations.

Informations de copyright

© 2023 Published by Elsevier Inc.

Déclaration de conflit d'intérêts

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.