To evaluate the baseline predictors of long-term all-cause- and cardiovascular (CV)-mortality amongst high-risk hypertensive patients, and to develop an integer-based risk score.
In the Anglo-Scandinavian Cardiovascular Outcomes Trial (ASCOT) legacy study, 8580 UK-based hypertensive patients were followed-up for mortality. All deaths (after the trial closure) were independently adjudicated by a team of 2 physicians.
Cox-proportional hazards models were developed separately for 2 outcomes: all-cause- and CV- mortality using backwards stepwise-selection. All variables were considered for inclusion in the model; but, pre-specified covariates (age, sex, race, socio-economic status [SES], body-mass index [BMI], systolic blood pressure [SBP], total cholesterol, presence of diabetes, smoking history) were included even if statistically insignificant. Continuous variables were categorised if non-linear, or pre-specified based on clinical utility. Goodness-of-fit and C-statistics were calculated.
During median follow-up of 15.7 years, 3282 (38.3%) of 8580 hypertensive patients (mean age at baseline, 64.1 year) died with about a third (n, 1210) from CV-related causes. Figure 1 lists the baseline predictors significantly associated with all-cause- and CV-mortality.
Aside from increasing age and smoking status, increasing baseline levels of fasting plasma glucose and serum creatinine, presence of atrial fibrillation and previous history of peripheral-or cerebrovascular- disease were significant risk factors for 2 outcomes. Notably, increase in pulse pressure (PP) (and not SBP) was an independent risk factor, with 47% and 53% significantly higher risk for all-cause and CV-mortality amongst those with PP> = 100 (vs. PP <60), respectively. BMI and alcohol intake were significantly associated with all-cause-(but not CV-) mortality: BMI < 20 kg/m2, and alcohol intake >28 units/week were associated with 79% and 20% increased risk respectively. Higher SES was associated with a significantly reduced risk for both outcomes.
Both models had an excellent discriminative ability (c-statistics, 0.72 and 0.73, respectively).
Our findings suggest that a few baseline determinants in our 2 models are different from those that are routinely used for CV-risk prediction. Particularly, increase in baseline PP (and not SBP) was found to be an independent and better predictor for both all-cause and CV-mortality in this high-risk hypertensive population.
1William Harvey Research Institute, Queen Mary University of London, London, United Kingdom
2Imperial College, London, United Kingdom
3London School of Hygiene and Tropical Medicine, London, United Kingdom