Creation of a model to predict survival in patients with refractory coeliac disease using a multinational registry

A Rubio-Tapia, G Malamut, W H M Verbeek, R L J van Wanrooij, D A Leffler, S I Niveloni, C Arguelles-Grande, B D Lahr, A R Zinsmeister, J A Murray, C P Kelly, J C Bai, P H Green, S Daum, C J J Mulder, C Cellier, A Rubio-Tapia, G Malamut, W H M Verbeek, R L J van Wanrooij, D A Leffler, S I Niveloni, C Arguelles-Grande, B D Lahr, A R Zinsmeister, J A Murray, C P Kelly, J C Bai, P H Green, S Daum, C J J Mulder, C Cellier

Abstract

Background: Refractory coeliac disease is a severe complication of coeliac disease with heterogeneous outcome.

Aim: To create a prognostic model to estimate survival of patients with refractory coeliac disease.

Methods: We evaluated predictors of 5-year mortality using Cox proportional hazards regression on subjects from a multinational registry. Bootstrap resampling was used to internally validate the individual factors and overall model performance. The mean of the estimated regression coefficients from 400 bootstrap models was used to derive a risk score for 5-year mortality.

Results: The multinational cohort was composed of 232 patients diagnosed with refractory coeliac disease across seven centres (range of 11-63 cases per centre). The median age was 53 years and 150 (64%) were women. A total of 51 subjects died during a 5-year follow-up (cumulative 5-year all-cause mortality = 30%). From a multiple variable Cox proportional hazards model, the following variables were significantly associated with 5-year mortality: age at refractory coeliac disease diagnosis (per 20 year increase, hazard ratio = 2.21; 95% confidence interval, CI: 1.38-3.55), abnormal intraepithelial lymphocytes (hazard ratio = 2.85; 95% CI: 1.22-6.62), and albumin (per 0.5 unit increase, hazard ratio = 0.72; 95% CI: 0.61-0.85). A simple weighted three-factor risk score was created to estimate 5-year survival.

Conclusions: Using data from a multinational registry and previously reported risk factors, we create a prognostic model to predict 5-year mortality among patients with refractory coeliac disease. This new model may help clinicians to guide treatment and follow-up.

Conflict of interest statement

DECLARATION OF INTEREST Author’s declaration of personal interest: The authors declare no conflict of interest related to this manuscript

© 2016 John Wiley & Sons Ltd.

Figures

FIGURE 1
FIGURE 1
Model Calibration Plot Footnote Figure 1: predicted survival was captured both from the apparent model and the bootstrap (bias-corrected) model. Dots convey the apparent calibration from the original model, while x portray bias-corrected calibration from the bootstrap models. For a reference of perfect calibration, the Y=X line is displayed.
FIGURE 2
FIGURE 2
Agreement between Predicted Survival from Multivariable and Risk Score Models: Panel A) 5-Factor Risk Score and Panel B) 3-Factor Risk Score
FIGURE 2
FIGURE 2
Agreement between Predicted Survival from Multivariable and Risk Score Models: Panel A) 5-Factor Risk Score and Panel B) 3-Factor Risk Score
FIGURE 3
FIGURE 3
Kaplan-Meier 5-Year Survival by Quartiles of 3-Factor Risk Score

Source: PubMed

Sponsorit ja yhteistyökumppanit

Sairaudet

Huumeiden interventiot

3
Tilaa