BACKGROUND: Absolute Psoriasis Area and Severity Index (PASI) is a key endpoint in psoriasis management. Petto et al. [Pharm Stat. 2019;18(1):4-21] developed a statistical method to estimate the proportion of patients reaching absolute PASI response given baseline PASI score and proportion of patients achieving relative improvements at predefined time points.
OBJECTIVES: To test this method on clinical data from two phase 3 tildrakizumab trials (reSURFACE 1/2) comparing estimated absolute PASI ≤1/≤2/≤3/≤5 responses with reference responses from clinical databases.
METHODS: Reference PASI responses of ≤1/≤2/≤3/≤5 were extracted from clinical databases. Estimation of absolute PASI ≤1/≤2/≤3/≤5 response rates at week (W) 12 and W28 by treatment and trial were performed. Differences between estimated and reference responses were analysed. Bland-Atman limits of agreement and Passing-Bablok regression to assess variations between estimated and reference responses were performed.
RESULTS: Differences between estimated and reference absolute PASI ≤1/≤2/≤3/≤5 responses at W12 and W28 by treatment and trial were of little clinical relevance with an overall mean difference in PASI response proportion of -2.2% (e.g., for the tildrakizumab 100-mg arm, original proportions of patients achieving PASI of ≤1/≤2/≤3/≤5 at W28 were 38.5%/52.2%/63.5%/73.9% and 39.8%/54.8%/63.6%/76.9% [reSURFACE 1 and 2, respectively] vs. estimated proportions of 33.2%/49.8%/62.5%/78.3% and 34.3%/51.6%/64.5%/79.9%). Limits of agreement were -7.1% to 1.4% at W12 and -6.8% to 4.3% at W28. Scatterplots revealed linearity that stood the cusum test in Passing-Bablok regression with slope 1.14 (95% confidence intervals: 1.06 to 1.20).
CONCLUSION: Good estimates of absolute PASI response rates were achieved with the application of the statistical method to tildrakizumab data reported in the phase 3 studies, in particular in the verum study arms. Our data support the method provided by Petto et al.  to estimate proportions of psoriasis patients reaching absolute PASI value thresholds using relative PASI improvements.