A propensity-score weighting approach to compare registry and trial populations of patients with psoriasis on biologic therapies

Yiu, Zenas; Mason, Kayleigh; Barker, Jonathan; Hampton, Philip; McElhone, Kathleen; Smith, Catherine; Warren, Richard; Griffiths, Christopher; Lunt, Mark; Burden, David

Abstract

Background: Previous studies using incomplete trial inclusion and exclusion criteria suggest a difference between the reported effectiveness and safety of biologic therapies between randomised clinical trials (RCT) and real-world populations of patients with psoriasis.Objectives: To estimate the difference in safety and effectiveness outcomes in biologic therapies for psoriasis between the trial and real-world populations using a standardisation method.Methods: Data from patients on etanercept, adalimumab, or ustekinumab in the British Association of Dermatologists Biologic Interventions Register (BADBIR) were appended to individual participant-level data from two RCTs assessing ustekinumab in patients with psoriasis, PHOENIX 1 and 2. Baseline patient variables were assessed for association of being in a RCT using a multivariate logistic regression model. Propensity score based standardized mortality ratio weights were derived to reweigh the registry sample so that all variables had the distribution seen in the trial sample. The incidence rate of SAEs in the first year and the proportion of patients achieving an absolute Psoriasis Area and Severity Index (PASI) <1.5 at 6 months were calculated in the BADBIR cohort before and after weighting, with 95% confidence intervals calculated using bootstrapping.Results: 6790 registry and 2021 trial participants were included. There were marked differences in the baseline co-variates between the registry and trial participants, and the multivariate logistic model had a C-statistic of 0.84 (95% confidence interval 0.83, 0.84). All variables had a standardised difference between −0.1 and 0.1 after weighting. The incidence rate difference of SAEs between the registry and trial sample was 9.30 (95% CI −3.91, 22.50) per 1000 person-years; the incidence rate ratio was 1.14 (95% CI 0.91, 1.37). The difference in proportion achieving absolute PASI < 1.5 between the registry and trial sample was 1.09% (95% CI −1.98%, 4.15%); the relative risk ratio was 1.03 (95% CI 0.94, 1.11).Conclusions: Our results suggest that the difference between the baseline characteristics of a real-world and a trial psoriasis population leads to the underestimation of the true real-world incidence of SAEs, but not effectiveness, in psoriasis clinical trials. Real-world data are therefore needed to understand the true safety of biologic therapies in psoriasis.

Citation

Yiu, Z., Mason, K., Barker, J., Hampton, P., McElhone, K., Smith, C., …Burden, D. (2018, August). A propensity-score weighting approach to compare registry and trial populations of patients with psoriasis on biologic therapies. Presented at 34th International Conference on Pharmacoepidemiology & Therapeutic Risk Management, Prague Congress Centre, Prague, Czech Republic

• Presentation Conference Type: Speech
• Conference Name: 34th International Conference on Pharmacoepidemiology & Therapeutic Risk Management
• Conference Location: Prague Congress Centre, Prague, Czech Republic
• Start Date: Aug 22, 2018
• End Date: Aug 26, 2018
• Deposit Date: Jun 20, 2023