The complete blood count and cardiovascular disease: analyses across six cohorts of 23,370 adults

Abstract

Background: The complete blood count (CBC) is one of the most performed laboratory studies. However, the CBC and its components are not commonly used to understand and quantify cardiovascular disease (CVD) risk.

Objective: We sought to define the relationships between the CBC, traditional CVD risk factors, and common CVD biomarkers and their joint association with all-cause mortality and CVD.

Methods: We examined the relationships between the CBC, traditional CVD risk factors, and mortality in NHANES (n=7843). We validated and extended these findings to more refined CVD endpoints in five additional cohorts (n=15,527).

Results: We first examined the variance accounted for by common laboratory studies (lipid panel, HbA1c, hs-CRP, and basic metabolic panel) by traditional risk factors in NHANES. Except for hemoglobin (Hb) components, we found that traditional risk factors accounted for less than 20% of the variance in the CBC, suggesting that the CBC provides unique information beyond traditional risk factors and CVD biomarkers. Additionally, the CBC was strongly associated with all-cause mortality (p<0.0001), even more than traditional CVD biomarkers (lipid panel, HbA1c, and CRP). We validated and extended these findings across five additional cohorts with a mean follow-up of 16 years and more refined CVD endpoints. In the fully adjusted analyses, several CBC components, including the white blood cell (WBC) count, neutrophil (PMN) count, Hb level, and an integrated immune cell score, were associated with individual CVD endpoints (incident stroke, MI, or revascularization) and a composite CV endpoint (MACE3) with standardized hazard ratios of 1.13 (p=0.002), 1.15 (p=0.0006), 0.82 (p<0.0001), and 2.16 (p<0.0001) respectively.

Conclusion: This study represents the first systematic examination of the relationship between the CBC, all-cause mortality, and CVD in a diverse cohort of 23,370 adults. These findings underscore the added value of the CBC over traditional risk factors and common CVD biomarkers for CVD risk assessment. Future studies should explore the integration of CBC parameters into predictive models to enhance our understanding, early identification, and prevention strategies for CVD.

Keywords: atherosclerosis; cardiovascular disease; complete blood count; hemoglobin; inflammation; lipoproteins; neutrophil; platelets.

Figure 1.. The variance explained of common laboratory studies by traditional risk factors and information content for all-cause mortality.

A) We leveraged data from the 2003–2010 cycles of NHANES to assess the variance explained of common laboratory measures by traditional risk factors using 6,162 subjects with complete data and fasting status. Type I analysis of variance accounting for survey weights was performed to estimate percent variance explained after hyperbolic arcsine transformation. These results illustrate that values from the CBC had comparable independent information content beyond traditional risk factors and comparable to that of lipid and basic metabolic panels. B) Using incremental chi-squared, we examined the survey-weighted association between common laboratory panels and all-cause mortality. Adjustments were made for age, sex, race, smoking, BMI, blood pressure, use of antihypertensive medications, and self-reported diabetes. Higher χ² indicates greater information content.

Figure 2.. Dimension reduction of the CBC and information content for all-cause mortality.

A) Principal component analysis (PCA) of the CBC with differential demonstrated that four components explain 62.9% of the total variance in the CBC with differential study. PCA was performed with varimax post-rotation and accounted for survey sampling weights for the fasting subpopulation in NHANES. The first component (explaining 20.5% of the total variance) was positively loaded on total WBC count, neutrophils, and NLR and negatively loaded on lymphocytes. The second component (explaining for 14.3% of the total variance) was loaded on red blood cell quantity measures, negatively loaded on RBC count, hematocrit, and hemoglobin. The third component (explaining 13.7% of total variance) was negatively loaded with multiple inflammatory cell parameters including total WBC count, lymphocytes, and eosinophils. Finally, the fourth component (explaining 14.1% of total variance) was loaded on red blood cell characteristics with positive loadings for MCH, MCHC, and MCV and negative loading for RDW. B) Analogously to Figure 2, we computed incremental χ² for Cox proportional hazards models for all-cause mortality, adding rotated principal components of common laboratory panels to clinical variables (age, sex, race, smoking, body mass index, systolic and diastolic blood pressure, use of antihypertensive medications, self-reported diabetes status). Higher χ² indicates greater information content. Models accounted for survey sampling weights for the fasting subpopulation in NHANES.

Figure 3.. CBC components are associated with refined CV endpoints in five diverse, longitudinal cohorts.

Heatmap of standardized β-coefficients from Cox proportional hazards models relating CBC components to CV endpoints. Values of standardized β-coefficients directionality (color), magnitude (color gradient), and statistical significance (color of circle) are depicted. All models are fully adjusted for age, sex, race, smoking, body mass index, systolic and diastolic blood pressure, use of antihypertensive medications, diabetes status, total cholesterol, HDL, LDL, and triglycerides.

Figure 4.. Key CBC-derived scores for all-cause mortality and MACE3 in five longitudinal cohorts.

Forest plots of standardized β-coefficients of key CBC-derived scores for all-cause mortality and MACE3. Tables on the left indicate the number of total subjects and events in each cohort. Hazard ratios (circles) and their 95% confidence intervals (bars) are indicated in different colors for individual cohort estimates (coral), summary meta-analysis (blue), and leave one cohort analysis sensitivity analyses (green). Cox models in each cohort were adjusted for age, sex, race, smoking, body mass index, systolic and diastolic blood pressure, use of antihypertensive medications, diabetes status, total cholesterol, HDL, LDL, and triglycerides.