Pulse pressure and stroke risk: development and validation of a new stroke risk model

• Published in: Current medical research and opinion Vol. 30; no. 12; pp. 2453 - 2460
• Main Authors: Ayyagari, Rajeev, Vekeman, Francis, Lefebvre, Patrick, Ong, Siew Hwa, Faust, Elizabeth, Trahey, Alex, Machnicki, Gerardo, Duh, Mei Sheng
• Format: Journal Article
• Language: English
• Published: England Informa UK Ltd 01-12-2014; Taylor & Francis
• Subjects: Aged; Blood Pressure - physiology; Electronic Health Records; Female; Humans; Hypertension; Hypertension - complications; Hypertension - physiopathology; Male; Middle Aged; Proportional Hazards Models; Pulse pressure; Reproducibility of Results; Retrospective Studies; Risk Factors; Risk prediction; Stroke; Stroke - etiology; Stroke - physiopathology; Hypertension; Pulse pressure; Risk prediction; Stroke
• ISSN: 0300-7995; 1473-4877
• DOI: 10.1185/03007995.2014.971357

Abstract
Objective:
This study aims to develop and validate a stroke risk model incorporating pulse pressure (PP) as a potential risk factor. Recent evidence suggests that PP, defined as the difference between systolic blood pressure (SBP) and diastolic blood pressure (DBP), could be an incremental risk factor beyond SBP.
Methods:
Electronic health records (EHRs) of hypertensive patients from a US integrated health delivery system were analyzed (January 2004 to May 2012). Patients with 1 PP reading and 6 months of observation prior to the first diagnosis of hypertension were randomly split into development (two-thirds of sample) and validation (one-third of sample) datasets. Stroke events were identified using ICD-9-CM 433.xx-436.xx. Cox proportional hazards models assessed time to first stroke event within 3 years of first hypertension diagnosis based on baseline risk factors, including PP, age, gender, diabetes, and cardiac comorbidities. The optimal model was selected using the least absolute shrinkage and selection operator (LASSO); performance was evaluated by the c-statistic.
Results:
Among 34,797 patients selected (mean age 59.3 years, 48% male), 4272 patients (12.3%) had a stroke. PP was higher among patients who developed stroke (mean [SD] PP, stroke: 02.0 [15.3] mmHg; non-stroke: 58.1 [14.0] mmHg, p < 0.001). The best performing risk model (c-statistic, development: 0.730; validation: 0.729) included PP (hazard ratio per mmHg increase: 1.0037, p < 0.001) as a significant risk factor.
Limitations:
This study was subject to limitations similar to other studies using EHRs. Only patient encounters occurring within the single healthcare network were captured in the data source. Though the model was tested internally, external validation (using a separate data source) would help assess the model's generalizability and calibration.
Conclusions:
This stroke risk model shows that greater PP is a significant predictive factor for increased stroke risk, even in the presence of known risk factors. PP should be considered by practitioners along with established risk factors in stroke treatment strategies.