Development and Validation of Prediction Models for Hypertensive Nephropathy, the PANDORA Study

Abstract

ImportanceHypertension is a leading cause of end-stage renal disease (ESRD), but currently, those at risk are poorly identified.ObjectiveTo develop and validate a prediction model for the development of hypertensive nephropathy (HN).Design, Setting, and ParticipantsIndividual data of cohorts of hypertensive patients from Kailuan, China served to derive and validate a multivariable prediction model of HN from 12, 656 individuals enrolled from January 2006 to August 2007, with a median follow-up of 6.5 years. The developed model was subsequently tested in both derivation and external validation cohorts.VariablesDemographics, physical examination, laboratory, and comorbidity variables.Main Outcomes and MeasuresHypertensive nephropathy was defined as hypertension with an estimated glomerular filtration rate (eGFR) < 60 ml/min/1.73 m2 and/or proteinuria.ResultsAbout 8.5% of patients in the derivation cohort developed HN after a median follow-up of 6.5 years that was similar in the validation cohort. Eight variables in the derivation cohort were found to contribute to the risk of HN: salt intake, diabetes mellitus, stroke, serum low-density lipoprotein, pulse pressure, age, hypertension duration, and serum uric acid. The discrimination by concordance statistics (C-statistics) was 0.785 (IQR, 0.770-0.800); the calibration slope was 1.129, the intercept was –0.117; and the overall accuracy by adjusted R2 was 0.998 with similar results in the validation cohort. A simple points scale developed from these data (0, low to 40, high) detected a low morbidity of 7% in the low-risk group (0–10 points) compared with >40% in the high-risk group (>20 points).Conclusions and RelevanceA prediction model of HN over 8 years had high discrimination and calibration, but this model requires prospective evaluation in other cohorts, to confirm its potential to improve patient care.