Coronary artery disease
Usefulness of a Complete Blood Count-Derived Risk Score to Predict Incident Mortality in Patients With Suspected Cardiovascular Disease

https://doi.org/10.1016/j.amjcard.2006.08.015Get rights and content

The complete blood cell (CBC) count is an inexpensive, frequently obtained blood test whose information content is potentially underused. We examined the predictive ability of the CBC count for incident death in 29,526 consecutive consenting patients who underwent coronary angiography. Subjects were randomly assigned to training (60%) and test (40%) groups and were followed for an average of 4.9 years. Computed and integer risk score models for all-cause death were developed for 30 days and 1, 5, and 10 years using multivariable logistic regressions applied to CBC metrics, age, and gender. The study cohort was an average age of 61 years, 62% were men, and had a 3.3% annual risk of mortality. An integer (scalar) risk score (range 0 to 18) successfully separated patient cohorts into subgroups at markedly different mortality risks (<1% to >14% at 30 days). Predictive fractions (area under risk curve) at 30 days for the CBC-only model and the age- and gender-adjusted CBC model were 0.76 and 0.78, respectively, in the training set and 0.71 and 0.75, respectively, in the test set (all p values <<0.001). The CBC model was markedly more informative than models based only on hematocrit, white blood cell count, or age and gender and was superior to models with all 7 traditional risk factors. In conclusion, in a large, prospectively assembled database, a CBC risk model had high predictive ability for risk of incident mortality. A total CBC score is an important new addition to risk prediction, and it can be easily generated by computer for clinical use at negligible incremental cost.

Section snippets

Study aims

Our primary aim was to develop, evaluate, and validate the utility of a cardiovascular risk model based on multiple CBC count parameters for prediction of incident all-cause mortality in a moderate- to high-risk cardiovascular population at 30 days (primary time point), 1 year (secondary time point), and 5 and 10 years (exploratory time points). A second aim was to compare the predictive ability of the total CBC risk models with simple models based on age, gender, and other standard risk

Characteristics of the study population

Characteristics, clinical presentation, angiographic status, and outcome of the study population (n = 29,526) are presented in Table 2. The average age was 61 years, and 62% were men. Almost 50% of study subjects had clinical diagnoses of hypertension and hyperlipidemia, 17% were diabetic, and 7% had a previous myocardial infarction. Coronary angiography was normal in 35% and indicated mild/moderate coronary artery disease in 8% and severe disease in 57%, of which 46%, 26%, and 28% had 1-, 2-,

Discussion

In this large population cohort (almost 30,000 patients at moderately high overall cardiovascular risk), we successfully developed, tested, and validated CBC count–based predictive models for incident all-cause mortality. Results clearly demonstrated the excellent predictive ability (AUC 0.712) of concurrently considered WBC count, platelet count, and 4 red blood cell–related metrics. Further, the CBC count model was markedly superior to hematocrit alone, WBC count alone, and gender and age

References (20)

There are more references available in the full text version of this article.

Cited by (96)

View all citing articles on Scopus

This study was supported by the Deseret Foundation, Salt Lake City, Utah.

View full text