Table 1 shows the distribution of peripheral leukocyte counts and smoking in the sample. The mean rate of respiratory symptom reporting and smoking status by age, race, and sex is shown in Table 2. Phlegm was more common than cough in this population, reflecting the way these questions were asked (see methods). In multiple logistic regression models, asthma was associated with pack-years of cigarettes smoked. Bronchitis was associated with pack-years as well, but dummy variables for current smoking and former smoker were also predictive. For phlegm, significant predictors of symptom status included current smoker, former smoker, cigarettes per day, and pack-years of smoking. For dyspnea, significant predictors were cigarettes per day, pack-years, and years since quitting. For persistent cough, predictive factors included current smoking, years since quitting, and pack-years.
Table 3 shows the regression coefficients and robust standard errors for eosinophil count and neutrophil count when added (singly) to the previous models. Asthma was only significantly associated with eosinophils. Bronchitis was significantly associated with neutrophils and marginally associated with eosinophils (p = 0.072). Inclusion of both neutrophil count and eosinophil count simultaneously in the model for bronchitis had no impact on the estimated coefficients and standard errors for neutrophil count ((3 = 0.0090; SE = ±0.000377) or for eosinophil count ((3 = 0.00076; SE= ±0.00043).
Phlegm was also associated with both the neutrophil count (p = 0.041) and the eosinophil count (p = 0.030). Again, inclusion of both differential cell counts simultaneously in the model had little impact on the size of their coefficients and their standard errors. In that model the coefficient of neutrophil count was 0.000048 ±0.000023, and the coefficient of eosinophil count was 0.00053 ±0.00024.
Dyspnea was only associated with eosinophil count, and persistent cough was associated with neutrophils. No symptoms were significantly associated with the lymphocyte count, although a negative trend was seen with dyspnea, asthma, and persistent cough.
Figure 1 shows the relative odds of asthma by quartile of eosinophil count, controlling for age, race, sex, and smoking. Figure 2 shows the relative odds of bronchitis by quartile of neutrophil count, again controlling for all covariates. Figures 3 to 5 similarly show the relative odds of phlegm and dyspnea by quartile of eosinophil count and the relative odds of persistent cough by quartile of neutrophil count.
Table 1—Distribution of Peripheral Leukocyte Counts and Smoking
Table 2—Means of Dichotomous Variables by Age, Race, and Sex (%)
Table 3—Regression Coefficients and Robust Standard Errors for Neutrophil and Eosinophil Counts (cells/mm1) in Logistic Regression models for respiratory symptoms, Controlling for Age, Race, Sex and Smoking
Figure 1. Relative risk of self-report of physicians diagnosis of asthma plotted versus mean eosinophil count among subjects aged 30 to 74 years seen in NHANES I (1971 to 1975). These results are adjusted by regression for age, race, sex, and smoking.
Figure 2. Relative risk of self-report of physicians diagnosis of bronchitis plotted versus mean neutrophil count among subjects aged 30 to 74 years seen in NHANES I (1971 to 1975). These results 6are adjusted by regression for age, race, sex, and smoking.
Figure 3. Relative risk of phlegm plotted versus mean eosinophil count among subjects aged 30 to 74 years seen in NHANES I (1971 to 1974). These results are adjusted by regression for age, race, sex, and smoking.
Figure 4. Relative risk of self-report of dyspnea plotted versus mean eosinophil count among subjects aged 30 to 74 years seen in NHANES I (1971 to 1975). These results are adjusted bv regression for age, race, sex, and smoking
Figure 5. Relative risk of self-report of persistent cough plotted versus mean neutrophil count among subjects aged 30 to 74 years seen in NHANES I (1971 to 1975). These results are adjusted by regression for age, race, sex, and smoking.0