| dc.description.abstract | Background
Several prognostic factors are associated with predicting mortality in patients with peripheral
artery disease (PAD), and some are included in prediction models for mortality risk. Still, there
is no literature about the prognostic models for mortality in PAD in the last decade. We present
a summary of the available models, make a comparison of the performance between them, and
assess their risk of bias.
Objective
We aim to identify prognostic models for mortality in patients with PAD, give an overview of
the model, to present a comparative discussion to establish common predictors factors for
mortality and appraise the risk of bias of the model.
Methods
We searched PubMed, EMBASE, and the Cochrane Library to identify studies developing or
internally/externally validating prognostic models for mortality in patients with PAD. We
extracted information on study design, population characteristics, and model characteristics,
and used the Risk Of Bias ASsessment Tool for prediction model (PROBAST) to assess the
risk of bias of the identified models.
Results
In total, four studies (three models) met the inclusion criteria. Two studies developed and
externally validated a model, namely the CORPAT and BOA-RC2 models, one study
developed a model for predicting mortality in females, and the last was an external validation
study of the CORPAT model. The identified models predicted mortality or the combined
outcome of mortality and non-fatal cardiovascular events at different moments (1, 2, 5 and 10
years). Age was included as a predictor in all models. Another frequently used predictor was kidney
function but even so, there was no agreement in the age categories nor in the measures of
kidney function used. Discrimination performance was comparable across studies and the risk of
bias was high in all models.
Conclusion
Despite the high risk of bias, the validated prognostic models demonstrate optimal performance
in predicting mortality among patients with PAD. Existing models need to be validated more
often and if necessary, the authors should consider updating by adding new predictors not
contained in existing models. | |
