Premium
Development and validation of a multivariable risk prediction model for serious infection in patients with psoriasis receiving systemic therapy
Author(s) -
Yiu Z.Z.N.,
Sorbe C.,
Lunt M.,
Rustenbach S.J.,
Kühl L.,
Augustin M.,
Mason K.J.,
Ashcroft D.M.,
Griffiths C.E.M.,
Warren R.B.
Publication year - 2019
Publication title -
british journal of dermatology
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 2.304
H-Index - 179
eISSN - 1365-2133
pISSN - 0007-0963
DOI - 10.1111/bjd.17421
Subject(s) - psoriasis , systemic therapy , medicine , multivariable calculus , systemic risk , intensive care medicine , risk of infection , dermatology , engineering , economics , macroeconomics , financial crisis , cancer , control engineering , breast cancer , genetics , biology
Summary Background Patients with psoriasis are often concerned about the risk of serious infection associated with systemic psoriasis treatments. Objectives To develop and externally validate a prediction model for serious infection in patients with psoriasis within 1 year of starting systemic therapies. Methods The risk prediction model was developed using the British Association of Dermatologists Biologic Interventions Register ( BADBIR ), and the German Psoriasis Registry PsoBest was used as the validation dataset. Model discrimination and calibration were assessed internally and externally using the C ‐statistic, the calibration slope and the calibration in the large. Results Overall 175 (1·7%) out of 10 033 participants from BADBIR and 41 (1·7%) out of 2423 participants from PsoBest developed a serious infection within 1 year of therapy initiation. Selected predictors in a multiple logistic regression model included nine baseline covariates, and starting infliximab was the strongest predictor. Evaluation of model performance showed a bootstrap optimism‐corrected C ‐statistic of 0·64 [95% confidence interval ( CI ) 0·60–0·69], calibration in the large of 0·02 (95% CI −0·14 to 0·17) and a calibration slope of 0·88 (95% CI 0·70–1·07), while external validation performance was poor, with C ‐statistic 0·52 (95% CI 0·42–0·62), calibration in the large 0·06 (95% CI −0·25 to 0·37) and calibration slope 0·36 (95% CI −0·24 to 0·97). Conclusions We present the first results of the development of a multivariable prediction model. This model may help patients and dermatologists in the U.K. and the Republic of Ireland to identify modifiable risk factors and inform therapy choice in a shared decision‐making process.