Of the 4926 individuals who were interviewed and examined at the baseline exam, individuals were excluded from the analyses if there was no measurement for CRAE or CRVE in either eye (N=114), if data were missing for any of the laboratory values used in the analysis (N=14), or if the individual had a condition or disease such as leukemia, thrombocythemia or polycythemia (N=68), leaving 4730 individuals eligible for analysis.
Individuals who were excluded were more likely to be older, female, hypertensive, and more likely to have lower diastolic blood pressure, higher levels of glycosylated hemoglobin and all other blood elements except platelet count than those included. There were no differences in systolic blood pressure, pack-years smoked, platelet count, or current smoking between the two groups ().
Characteristics of Those Included and Excluded from Analysis.
An increase of one sex-specific quintile of WBC, red blood cell (RBC) count, hemoglobin, and hematocrit is significantly associated with larger CRAE and CRVE (). An increase of one sex-specific quintile of platelet count was also significantly associated with larger CRVE. Additional adjustment for age does not change the significance of these relationships. When the elements are considered as continuous variables the relationships are unchanged from when they are considered as quintiles.
Relationships of Blood Elements with Central Retinal Arteriole Equivalent and Central Retinal Venule Equivalent.
In order to investigate the relative strengths of the relationships of the blood elements to the vessel diameters, we developed models in which we sequentially included other known correlates of those diameters (). For all models, each element except platelet count was significantly associated with CRAE. The models indicate that including more of the additional variables increases the informativeness of the model measured by the change in R2. For all models, each blood element was significantly associated with CRVE. Additional adjustments for additional variables increased the R2 of the model. The presence of diabetes has a small incremental effect on the fit of the models.
Multivariate Relationships Between Blood Elements and Central Retinal Arteriole and Venule Equivalents
We next performed the modeling use a stepwise approach including as possible variables all those considered in model 4 of ( and ). For CRAE, systolic blood pressure had the highest partial R2, followed by smoking, the blood element, and diabetes for each blood element. In the model assessing the contribution of platelets to the other variables associated with CRAE, platelets no longer contributed a significant amount of information. For CRVE, smoking history has the greatest partial R2, followed by the blood element and then sex and systolic blood pressure. Because of the importance of relative diameter, we repeated the analyses in , this time including CRVE in the models for CRAE and CRAE in models for CRVE (). These additions markedly improved the R2 of all models. In all models, the relative importance of smoking was diminished. The blood elements are the third most informative variable for CRVE or fourth most informative variable for CRAE when controlling for the other vessel measurement.
Stepwise Selection for Variables with Central Retinal Arteriole Equivalent and Central Retinal Venule Equivalent.
Stepwise Selection for Variables with Central Retinal Arteriole Equivalent (CRAE) and Central Retinal Venule Equivalent (CRVE), with CRVE considered for CRAE models and CRAE considered for CRVE models.
While the previous analyses consider each blood element individually, we next developed models when they are included together. We chose only one of the three measures of red blood cell status (RBC, hemoglobin and hematocrit) for the purposes of these analyses. We modeled the outcomes as was done in and . In models where CRAE is the outcome, and CRVE is not included, each of the blood element components adds significantly to the models with the red blood cell variable being most informative and WBC and platelet count adding less information. When CRVE is included the WBC is no longer significant in the models. In models where CRVE is the outcome, and CRAE is not included, each of the blood element components adds significantly to the models and, again, the red blood cell variables being most informative and WBC and platelets adding less information. When we add CRAE, all of the three different blood components remain significant in most models. The relative importance of most of the other variables that were used in the models in and remain unchanged.