Association Of Peripheral White Blood Cell Count With Lung Function And Its Mediation Effect In The Smoking Related Lung Function Reduction

Chronic obstructive pulmonary disease（COPD）is an important chronic non-infectious disease influencing human heath in the worldwide.Findings from Global Burden of Disease（GBD）dataset showed that about 3.2 million people died of COPD in the 2017,which was third cause of death and ranking behind cardiovascular disease and cancer.Lung function reflects respiratory health,and it is used in the diagnosis of COPD.The existing evidence shows that inflammation plays an important role in the process of lung function deterioration.Peripheral white blood cell（WBC）count is commonly used to reflect systemic inflammation in the clinical practice,and it contains five subtypes,including neutrophil,lymphocyte,monocytes,eosinophil,and basophil,and there are correlations between different WBC subtypes.Previous studies have separately explored the association between WBC subtype and lung function level.However,the independent effects of different WBC subtypes on lung function were under-investigated in the condition of co-exposure.Besides,the relationship of peripheral WBC with lung function may be modified by sex or smoking status.In the recent years,Mendelian randomization（MR）is becoming a popular method for causal inference.In the MR analysis,genetic variants are used as instrumental variables.The well-designed MR study provides reliable evidence for causal inference as the randomized controlled trial.In addition,smoking is the established environmental risk factor for lung function,and inflammatory reaction plays a key role in the process of smoking related lung function impairment.However,the mediation effect of systemic inflammation in the smoking related lung function alteration has not been reported in the population study.The present study aimed to explore the relationship of peripheral WBC count with lung function and its mediating effect in the smoking related lung function reduction.In the Part 1,we aimed to evaluate the epidemiological associations of total and differential WBC counts with forced vital capacity（FVC）and forced expiratory volume in one second（FEV₁）,and whether the above associations were modified by sex or smoking status.The 19,159 subjects from Dongfeng-Tongji cohort（DFTJ,n=13,827）,coke-oven workers cohort（COW,n=1762）and National Health and Nutrition Examination Survey 2011-2012（NHANES 2011-2012,n=3570）were included in the analysis.In the Part 2,single nucleotide polymorphisms（SNP）,which was reported in the genome wide association study of WBC among Asian,was used as the instrumental variable.The genetic associations with FVC and FEV₁ were explored among 4012 subjects with genotyping results from DFTJ（n=2886）and COW（n=1126）studies.These summary statistics were used to investigate the causal association between peripheral WBC and lung function.In the Part 3,we evaluated the mediation effects of total and differential WBC in the relationship of smoking with lung function among 13,827 subjects from DFTJ cohort.Owing to the complex design of NHANES 2011-2012 and incomplete measurement of WBC subtypes in the COW study,the mediation analysis was only performed in the DFTJ cohort.Part 1 Association of peripherial white blood cell count with lung function alterationObjective: To evaluate the associations of total and differential WBC counts with FVC and FEV₁,and whether the above associations were modified by sex or smoking status.Methods: The 38,295,1886,and 9756 subjects were enrolled in the cross-sectional surveys of DFTJ,COW,and NHANES 2011-2012,respectively.The standardized questionnaire was used to collect demographic information,disease history,drug use,life style factors,and so on.Spirometer was used to monitor prebronchodilator lung function,and hematology analyzer was used to determine total and differential WBC counts.After excluding participants with lung cancer,tuberculosis,silicosis,or leukemia,missing of covariates（age,sex,race,height,cigarette smoking,alcohol drinking,and physical activity）,WBC counts,or lung function,or total WBC count exceeded three standard deviation from the mean,19,159 participants（13,827 from DFTJ,1762 from COW,and 3570 from NHANES 2011-2012）were included in the following analyses.The continuous variables,including age,height,weight,body mass index（BMI）,FVC,FEV₁,and counts of total WBC,neutrophil,monocyte,and lymphocyte,were in normal distribution.The distributions of eosinophil and basophil counts were right skewed,and they were log10-transformed to approximate normal distribution.In the single-marker multiple linear regression model,the lung function measurements FVC and FEV₁ were separately included as a dependent variable,and the counts of total and differential WBC subtypes were separately entered as an independent variable,with adjustment for age,sex,race（only in NHANES 2011-2012 population）,height,smoking status,alcohol use,and exercise.Restricted cubic spline plots were used to explore the shapes of the above associations,with three knots at 5th,50 th,and 95 th percentiles of total or differential WBC,and the lower limits of clinical threshold were used as the reference.Furthermore,sensitivity analysis was also carried out among subjects without other cancers or without using anti-infectives.For the different WBC subtypes those had significant associations with lung function in the single-marker model（P<0.05）,we further used the LASSO regression with 10-fold cross-validation to select the most significant WBC components at minimum mean squared error.All the covariates were included in the LASSO regression without penalization.For the significant WBC subtypes selected by the LASSO regression,they were simultaneously included in the linear regression analysis（multiple-marker model）,with adjustment for the same covariates used in the single-marker model.Then,subjects were stratified by sex and smoking status,the associations of total and differential WBC（P<0.05 in the multiple-marker model）with lung function were investigated in each strata,and Pint was obtained by entering a cross-product term in the model（WBC×sex or WBC×smoking status）.Results: Among participants from three studies,we observed consistently negative relationships of the total WBC count with FVC [β（95% CI）=-40.73（-47.67,-33.79）,-29.39（-47.95,-10.84）,and-24.01（-37.62,-10.39）in the DFTJ,COW,and NHANES 2011-2012,respectively,all P<0.005] and FEV₁ [β（95% CI）=-22.88（-28.88,-16.88）,-31.66（-47.35,-15.98）,and-34.54（-51.11,-17.98）in the DFTJ,COW,and NHANES 2011-2012 respectively,all P<0.001] in the single-marker model.The Meta-analysis revealed that per 1×109 cells/L increase in the total WBC was associated with a separate 36.13（95% CI: 30.35,41.91）m L and 25.23（95% CI: 19.97,30.50）m L decrease in FVC and FEV₁（both P<0.001;heterogeneity P=0.057 and 0.266,respectively）.In the single-marker model,the differential WBC counts,including neutrophil,lymphocyte,monocyte,eosinophil,and basophil,were individually and negatively associated with FVC and FEV₁（all P<0.05）.Restricted cubic spline plots showed that the associations of total and differential WBC counts with FVC and FEV₁ were in a linear manner.The sensitivity analysis among subjects without other cancers or using anti-infective drugs also confirmed the concordant directions and similar effect sizes in NHANES 2011-2012 and DFTJ cohort.The positive WBC subtypes selected by the LASSO regression analyses were simultaneously included in the multiple-marker model,and we observed the independent associations of neutrophil,monocyte,eosinophil,and basophil with FVC and FEV₁（all P<0.05）,except lymphocyte（all P>0.05）.FVC and FEV₁ decreased with the increase in the peripheral neutrophil count [β（95% CI）=-32.69（-42.34,-23.05）and-19.25（-26.99,-11.52）,respectively;both P<0.001,heterogeneity P=0.608 and 0.697,respectively].Per 1×109 cells/L increase in the monocyte was associated with a separate 196.03（95% CI:-3.89,395.94）m L and 118.72（95% CI: 48.13,189.32）m L decrease in FVC and FEV₁.The counts of eosinophil and basophil were also negatively associated with FVC and FEV₁ [eosinophil: β（95% CI）=-80.44（-194.78,33.91）and-108.55（-210.07,-7.04）,respectively;basophil: β（95% CI）=-143.55（-173.34,-113.77）and-28.56（-54.30,-2.82）,respectively].In the stratification analysis by sex,the effects of total WBC on decreasing FVC and FEV₁ were enhanced among males than females（Pint=0.014 and <0.001,respectively）.For the association of total WBC with FVC,effect estimate and 95% CI was-39.83（-49.03,-30.64）and-33.96（-41.01,-26.91）in the males and females,respectively（both P<0.001）.For the association of total WBC with FEV₁,effect estimate and 95% CI was-33.63（-42.00,-25.25）and-18.55（-24.69,-12.41）in the males and females,respectively（both P<0.001）.The significant interactions with sex on FVC and FEV₁ were also observed for neutrophil（Pint=0.010 and 0.001,respectively）,but not for monocyte（Pint=0.396 and 0.081,respectively）.Besides,there was significant interaction of eosinophil with sex on FEV₁ [β（95% CI）=-191.28（-367.40,-15.16）and-50.17（-76.35,-23.99）in the males and females,respectively;Pint<0.001] and interaction of basophil with sex on FVC [β（95% CI）=-129.15（-277.83,19.53）and-56.53（-242.48,129.42）in the males and females,respectively;Pint=0.039].When stratified by the smoking status,the Meta-analysis indicated significant modifying effect of smoking on the association between total WBC and FEV₁ reduction,which was stronger in the smokers than that in the non-smokers [smokers: β（95% CI）=-33.26（-43.15,-23.37）vs.non-smokers: β（95% CI）=-20.30（-25.79,-14.82）;Pint=0.006].However,the relationship of total WBC with FVC in the smokers was smiliar to that in the nonsmokers（Pint=0.202）.As for the WBC subtypes,smoking could significantly modify the association of monocyte（Pint=0.002）and eosinophil（Pint=0.005）with FEV₁.Conclusions: Peripheral neutrophil,monocyte,eosinophil,and basophil counts were independently and negatively associated with FVC and FEV₁,but not lymphocyte.The above associations were modified by sex and smoking status,which is enhanced among the males and smokers.Part 2 Peripherial white blood cell count and lung function: A Mendelian randomization studyObjective: To evaluate the causal relationships of total WBC,neutrophil,monocyte,eosinophil,and basophil counts with and FVC and FEV₁ by using MR study.Methods: Among population of Asian ancestry,the largest genome-wide association study of peripheral WBC was performed in the 107,964 Japanese,and they have reported 36 SNPs associated with WBC count,21 SNPs associated with neutrophil count,31 SNPs associated with monocyte count,18 SNPs associated with eosinophil count,and 26 SNPs associated with basophil count.Summary statistics of these SNPs were free available,which enabled us to perform MR analysis to explore the causal relationship between WBC count and lung function among Asian.SNPs with F statistic <10 or without genotyping result in the DFTJ and COW studies were excluded.Among 4012 participants with genotyping results from DFTJ（n=2886）and COW（n=1126）studies,we evaluated the associations between SNPs associated with WBC count and confounders,including height,smoking status,drinking status,and exercise.Meta-analysis was used to combine the results from different populations,SNPs significantly associated with confounders at a Bonferroni-corrected level were excluded.Among 4012 participants with genotyping results from DFTJ and COW studies,genetic associations with FVC and FEV₁ was investigated by using multiple linear regression model,with adjustment for age,sex,height,smoking status,drinking status,and exercise.Results from different populations were pooled by using Meta-analysis. MR-PRESSO and MR-Egger regression were used to evaluate horizontal pleiotropy,potential outliers identified by MR-PRESSO were further excluded until MR-PRESSO global test P≥0.05.Finally,causal relationships between total and differential WBC counts and lung function were explored by using inverse variance weighted（IVW）method,two-sided P<0.05 was regarded as statistically significant.Results: In the MR analysis of total WBC count with FVC,33 SNPs,which were not associated with confounders,were included as candidate instrumental variables.MR-PRESSO global test did not detect the existence of horizontal pleiotropy（P=0.798）,and the intercept of MR-Egger regression did not deviate from 0 [β（95% CI）=7.31（-6.32,20.95）,P=0.293].IVW method showed that genetically determined total WBC count was significantly associated with FVC [β（95% CI）=-145.12（-263.81,-26.43）,P=0.017].In the MR analysis of total WBC count with FEV₁,33 SNPs,which were not associated with confounders and did not violate “no horizontal pleiotropy” assumption,were included as instrumental variables.The association between genetically determined total WBC count and FEV₁ was marginally significant [β（95% CI）=-103.87（-206.97,-0.77）,P=0.048].Similarly,we also found significant relationships of genetically determined neutrophil count with FVC and FEV₁ [β（95% CI）=-131.90（-243.90,-19.90）and-105.10（-207.11,-3.09）,respectively;P=0.021 and 0.043,respectively].However,we did not find any significant relationship of genetically determined monocyte count,eosinophil count,or basophil count with FVC（P=0.143,0.765,and 0.610,respectively）or FEV₁（P=0.080,0.966,and 0.475,respectively）.Conclusions: Findings from MR analyses showed that FVC and FEV₁ levels decreased with the increase in the genetically determined peripheral total WBC and neutrophil counts,suggesting that there were causal relationships of peripheral total WBC and neutrophil counts with FVC and FEV₁.However,no causal association was identified between other WBC subtypes and lung function.Part 3 Mediation effect of peripherial white blood cell in the smoking related lung function reductionObjective: To investigate the mediation effects of total WBC,neutrophil,monocyte,eosinophil,and basophil in the smoking related lung function reduction.Methods: Owing to the complex design of NHANES 2011-2012 and incomplete measurement of WBC subtypes in the COW study,the mediation analysis was only performed in the 13,827 participants from DFTJ cohort.Among the whole study population,males,and females,we investigated the relationships of smoking with FVC and FEV₁ by using linear regression model,with adjustment for age,sex,height,alcohol use,exercise,and total WBC count.The mediation effects of total WBC,neutrophil,monocyte,eosinophil,and basophil in the relationship of smoking with lung function were investigated by using R package “mediation”.The mediator model（relationships of smoking with total and differential WBC counts）was adjusted for age,sex,BMI,alcohol use,and exercise.The outcome model（relationships of smoking and WBC with lung function）was adjusted for age,sex,height,alcohol use,and exercise.The total effect of smoking on lung function includes two parts: direct effect and mediation effect,the proportion of WBC related mediation effect was calculated.The mediation proportions of total WBC,neutrophil,monocyte,eosinophil,and basophil in the relationship of smoking with lung function were also calculated in both males and females,and heterogeneity test was used to explore sex difference in mediation proportions.Results: We observed that smokers had lower FVC and FEV₁ levels than non-smokers（all P<0.05）.Mediation alalysis showed that 13.0% of the association between smoking and FVC was mediated by total WBC（P<0.001）,the direct effect was-99.04（95% CI:-131.35,-64.30）m L,and the mediation effect was-14.91（95% CI:-18.77,-11.30）m L;the mediation proportion was 11.6% and 20.5% in the males and females,respectively（P=0.486 for heterogeneity test）.The 6.4% of the effect of smoking on FEV₁ was driven by total WBC（P<0.001）,and the mediation proportion in the males was similar to that in the females（7.4% vs.8.1%,P=0.876 for heterogeneity test）.The 11.5%,3.9%,6.6%,and 6.9% of the effect of smoking on FVC was mediated by neutrophil,monocyte,eosinophil,and basophil,respectively（all P<0.001）;and 5.6%,2.5%,5.7%,and 1.9% of the effect of smoking on FEV₁ was mediated by neutrophil,monocyte,eosinophil,and basophil,respectively（all P<0.001）.The mediation proportion of WBC subtypes in the males was similar to that in the females（all P>0.05 for heterogeneity test）.Conclusions: Peripheral total WBC could mediate 13.0% and 6.4% of the effects of smoking on FVC and FEV₁,respectively,and neutrophil,monocyte,eosinophil,and basophil play mediating roles in the process of smoking related lung function impairment.