White blood cell count and all-cause and cause-specific mortality in the Guangzhou biobank cohort study

In clinical practice, the circulating white blood cell (WBC) count is readily available and widely used as an indicator of systemic inflammation. Numerous studies have shown positive associations between higher WBC count and deaths from all-causes [1‐4], coronary heart disease (CHD) [5], stroke [6] and cancer [7]. Most of these studies were conducted in high-income countries of Asia [3, 4] or in the West [1, 2, 5‐7]. Ethnic differences in WBC count exist, whereby reference intervals (RIs) for both Americans of Chinese and African origin are lower than those of European and Mexican origin [8]. Moreover, populations with different ethnicities may underlie different patterns of confounding factors (i.e., lifestyle and demographic factors), and an association consistently observed in different settings with different sets of confounders suggests that such association may be real or causal. Given this, and that no study describing the association between WBC count and mortality in Chinese populations was reported, we studied this association for all-cause, CHD, cancer, and respiratory disease mortality using prospective data from the Guangzhou Biobank Cohort Study.

As previous studies suggested an adverse association between pulmonary function and WBC count [9, 10], we hypothesized that this may account for any association between increased WBC count and higher risk of respiratory disease mortality. To examine whether such association, if any, could be attributable to pulmonary function we adjusted for pulmonary function and self-reported pulmonary disease in the analysis of respiratory disease mortality. We also examined the association for sub-types of WBC.

Of 30,430 participants at baseline, 391 were excluded because of loss to follow-up with unknown vital status and 202 excluded because of incomplete information on WBC count, giving 29,925 participants (21,648 women and 8277 men) in the present analysis. The mean age at baseline was 62 (standard deviation (SD) 7.1) years. During the average follow-up of 9.3 (SD 1.9) years, 2979 deaths (women 1581 (7.3%); men 1398 (16.9%)) were recorded.

Table 1 shows that the proportion of current smokers increased with increasing deciles of WBC count. Higher BMI, triglycerides, haemoglobin and platelet count were positively, while HDL-cholesterol, FEV1 and FVC was inversely associated with WBC count decile, but current alcohol drinking, physical activity and self-rated health, or age (as a continuous variable) were not.

For both all-cause and CHD mortality, the 4th decile group (i.e., WBC count of 5.4–5.6 × 10⁹/L in women and 5.8–6.0 × 10⁹/L in men) had the lowest mortality rate. Hence, the 4th decile group was used as the reference. Table 2 shows that greater WBC count was associated with both all-cause and CHD mortality in multivariable adjusted models, and the associations were slightly attenuated but remained statistically significant when the deaths in the first 2 years of follow-up were excluded. Relative to the 4th decile, the adjusted HRs for both all-cause and CHD mortality increased substantially in the highest decile (women > 8.2 × 10⁹/L; men > 8.8 × 10⁹/L; HR 1.83 (95% CI 1.54, 2.17) and 3.02 (1.84, 4.98), respectively).

For cancer mortality and respiratory disease mortality, the lowest mortality rate was observed in the 3rd decile group (i.e., WBC count of 4.9–5.3 × 10⁹/L in women and 5.3–5.7 × 10⁹/L in men), which was used as the reference group in the subsequent analyses. Table 3 shows that, relative to the 3rd decile group, participants in the lower decile (1st or 2nd) showed a higher risk for cancer mortality, and the HR was statistically significant for the 2nd decile (HR 1.54, 95% CI 1.17, 2.03). Greater WBC count showed no association with cancer mortality. We found positive associations between higher deciles of WBC count and respiratory disease mortality in the multivariable adjusted models, and the associations were unchanged after excluding deaths within the first 2 years of follow-up (Table 3). After further adjustment for pulmonary function and respiratory disease history including COPD, chronic bronchitis, emphysema, asthma, tuberculosis, pneumonia, the associations were attenuated slightly but remained statistically significant (HR 1.90, 95% CI 1.08, 3.33). After excluding current and former smokers, the results were similar (Table 4).

Sensitivity analysis showed that, after excluding participants with abnormal WBC count, or restricting data analysis to current smokers only, the results showing that greater WBC count was associated with increased respiratory disease mortality were similar (Additional file 1: Tables S1-S4). Another sensitivity analysis examining the association of subtypes of WBC (i.e., granulocyte or lymphocyte count) with mortality showed that participants with higher granulocyte count (> 5.44 × 10⁹/L) were associated with 45%, 69% and 1.6-fold higher risk of mortality from all-cause, CHD and respiratory disease, respectively, but not associated with cancer mortality (Additional file 1: Tables S5-S6). However, no association between lymphocyte count and mortality was found (Additional file 1: Tables S7-S8).

In this understudied population, high WBC count was robustly associated with all-cause, CHD and respiratory disease mortality after adjusting for potential confounders. The significant association between high WBC count and respiratory disease mortality supports our hypothesis that an increased in WBC count predicts a higher risk of respiratory disease mortality, independently of smoking and pulmonary function.

WBC count is a standardized and accurate measurement that has been used in epidemiologic studies as a predictor of mortality from all- and specific causes in the general population [1, 3, 4, 16, 17], or a predictor of specific outcomes in hospital-based studies of patients [18‐21]. Our results on all-cause and CHD mortality are consistent with those from other settings [1, 3, 4]. The association of WBC count with cardiovascular risk or mortality has been well described in studies that have focused on western or economically-developed Asian populations [2‐5, 22]. Such association persists after adjustment for multiple cardiovascular risk factors including smoking, an important confounder of the WBC-mortality association [23]. We found no evidence for an interaction between smoking and WBC count, which is in agreement with previous studies [23, 24], supporting an independent association between WBC count and risk of CHD in smokers and in non-smokers. Elevated WBC count may affect CHD through a number of possible pathologic mechanisms, such as altering endothelial function [25], attenuating nitric oxide and prostacyclin production [26] and consequently reducing the vasodilatory, antithrombotic and anti-atherogenic properties of the vascular endothelium [27, 28]. Increased adherence of stimulated WBC to the vascular endothelium promotes capillary leukocytosis and subsequent increased vascular resistance [29].

Results of previous prospective cohort studies on the association of WBC count with cancer mortality have been less consistent. We found only 6 studies [1, 3, 7, 30‐32], and of them, three showed a positive association [7, 31, 32], two showed no association [1, 3] and one showed an inverse association [7]. In a subset of participants in the Multiple Risk Factor Intervention Trial (MRFIT), elevated WBC count was associated with a higher risk of cancer death [31]. A prospective study conducted in postmenopausal women also found the evidence of higher WBC counts associated with higher risk of breast, lung and overall cancer mortality [32]. The other study from Australia also showed a higher risk of cancer mortality in those with high (4th quartile) versus low WBC count (1st quartile group) [7]. However, a large Korean study found no association between WBC count and cancer mortality, although, as expected, it found a significant and robust association with CHD mortality [3]. In a more recent study from the Women’s Health Initiative (WHI), the association of WBC count with cancer mortality was weaker than those of all-cause or CHD mortality, and was attenuated toward the null when women with comorbidity and deaths within the first 6 years of follow-up were excluded [1]. Our results also indicate a possible association between lower WBC count and cancer mortality, which was observed in the WHI study and a Taiwan study [30]. Within the normal range, a relatively low WBC count may be an indicator of low immune response which could also increase the risk of cancer mortality, although the underlying mechanisms warrant further investigation. We did not find any evidence supporting an association of higher WBC count with cancer mortality. As some treatment for cancer may lower WBC counts in clinical practice and cancer patients may be more vulnerable to opportunistic pathogens, both are often associated with a higher risk of infection that requires treatment with antibiotics and may affect WBC counts subsequently. However, our sensitivity analysis excluding participants with self-reported cancer history showed similar results.

Our study showed that elevated WBC count was significantly associated with a higher risk of respiratory mortality. Previous large cohort studies suggested that such an association could be due to confounding from smoking [33], i.e., smoking leads to both an elevated WBC count and a higher risk of respiratory mortality [33‐35]. Using baseline data of the GBCS, we also found that relative to never smokers, current smokers were associated with a greater WBC count and demonstrated a dose response relationship [36]. However, in the present analysis, the association was still observed in our never smokers, suggesting that smoking was unlikely to account for the findings. One possible mechanism is that systemic inflammation-induced endothelial dysfunction may lead to pulmonary vascular filtration and lung tissue injury [9]. In addition, neutrophils, one type of granulocytes, secrete serine proteases including neutrophil elastase, cathepsin G and proteinase-3, as well as MMP-8 and MMP-9, which may lead to destruction of the alveolar tissue and increase the risk of lung disease [37]. The exact mechanisms underlying the association of WBC count with cause-specific mortality require investigation.

For the association with CVD/CHD, we found six studies [38‐43] showing a positive association with granulocyte count, and of them, four showed no association with lymphocyte or monocyte counts [38, 39, 41, 42], and one showed a positive association with monocyte count [40]. A meta-analysis showed that people in the highest, versus those in the lowest, granulocyte quartile had a risk ratio of 1.32 (95% CI 1.15–1.51) for CHD [44]. Our study is in accordance with this meta-analysis showing that participants with higher granulocyte count were associated with a 69% higher risk of CHD mortality. Our results suggest that the increased risk related to WBC count might be partly due to the granulocyte count. Further studies exploring the differentiation, production and function of granulocytes and the related health effects are warranted. Although in our present study, the number of deaths was not sufficient to support more analyses for the subtypes of cardiovascular events, we would follow up all participants to enable more in-depth analysis.

The strengths of this study included the large sample size and comprehensive measurement and adjustment of potential confounding factors. However, there were some limitations. First, as noted above, the number of deaths from cause-specific mortality (especially CHD or respiratory disease) was relatively small. More detailed analyses of specific types of cardiovascular or respiratory disease could not be performed. Moreover, participants in the lower decile groups appeared to have a higher risk of death from all-cause, CHD and respiratory disease, although the results were not statistically significant. As the number of deaths in the lower decile groups were small, further studies are warranted to clarify. Second, as baseline WBC count was measured only at one-time point, the WBC counts might not accurately indicate a lifelong exposure. Repeated WBC measurements may provide a more reliable measurement of WBC count and minimize potential misclassification of exposure. However, in a recent study in women, similar results were observed using a single baseline measurement and using the average of the measurements at baseline and at the third year [1]. Thus, the lack of repeated measurement in the current study could not have influenced the risk estimation substantially. Third, our results based on prospective observational data cannot confirm a causal relation between WBC count and deaths due to CHD or respiratory disease. But as a predictive biomarker, the results from our study are informative and complementary to the literature. Fourth, although we have adjusted for a range of potential confounding factors, elevated WBC count may reflect an inflammatory response due to unknown causes (i.e., underlying chronic diseases or specific exposures that would increase WBC count and mortality at the same time). An alternative newly developed method, Mendelian randomization, might provide a unique opportunity to investigate the direct association, if any, between WBC count and CHD or respiratory disease. Fifth, women were oversampled in this study, as other population-based elderly cohorts. However, we tested for sex interaction and found no evidence for interaction (P values for interaction ranged 0.33 to 0.91), indicating the association between WBC count and the risk of all-cause and cause-specific mortality did not vary by sex. Moreover, sex was also adjusted in all models to minimize its potential confounding effect. Thus, the unbalanced sex ratio might not be a major concern in the current study. Finally, as all participants in the GBCS were older Chinese from southern China, generalizability may be a concern, although within sex and age group, the participants had fairly similar prevalence of major chronic diseases (i.e., diabetes and hypertension) to nationally representative samples of urban Chinese [11].

In conclusion, our results provide robust prospective evidence supporting that WBC and specifically granulocyte count is associated with the risk of all-cause, CHD and respiratory mortality. Further investigation is warranted to clarify whether decreases in inflammation would have effects on WBC count associated mortality.