Establishment And Validation Of The Estimation Model For Lung Function For The Primary Screening Of Chronic Obstructive Pulmonary Disease In China

ObjectivesChronic obstructive pulmonary disease(COPD)is one of the most important causes of disability and mortality around the world.It also poses a great threat to the public health in China.COPD is known as a preventable and partially treatable disease.Early detection,appropriate intervention,and secondary prevention can not only reduce patients' medical costs,but also improve their quality of life and prolong their life expectancy.Hence,accurate diagnosis of COPD is essential.The gold standard for diagnosing COPD is spirometry test.However,spirometry test in China faces problems such as low popularity,poor quality control,and non-standard measurement,which may lead to serious misdiagnosis or missed diagnosis.Meanwhile,the lung function reference equation used for COPD severity assessment is not unified in China.The applicability of foreign reference equations and the representativeness of domestic reference equations need to be further verified.To improve the current status of spirometry test in China,we aim to establish an estimation model for forced expiratory volume in 1s(FEV1)and forced vital capacity(FVC)to indicate COPD(determined by the ratio of estimated FEV1 to estimated FVC:FEV1/FVC),and to update the reference value for lung function in healthy population to assess airflow limitation severity(determined by the ratio of estimated FEV1 to reference value of FEV1:FEV1/pred).The estimation model and the reference equation can help to diagnose COPD without spirometry test,and thus promote the early screening for COPD in the community,and address the potential limitations of spirometry test.MethodsWe made use of the baseline data from the Prospective Urban and Rural Epidemiological(PURE)Study in China,which enrolled 46 285 individuals with 35-70 years of age from 115 communities(45 urban and 70 rural)across 12 provinces between 2005 and 2009.Participants were recruited using multi-stage stratified cluster sampling method.For every consenting participant,questionnaire-based interviews and basic physical examinations were conducted by trained local staff at accessible clinics or at home.Information on demographic profiles,socioeconomic status,lifestyle behaviors,and medical histories were obtained at individual level.Data for living environment,house structure,and indoor air pollution were collected at household or community level.Height,weight,and pre-bronchodilator spirometry were measured through standardized procedures.The estimation model for lung function was fitted by a multi-level nonlinear multiplicative function.The study population consisted of all participants who had acceptable quality of spirometry tests and without missing values in spirometry measurements.Independent variables were height,age,sex,and COPD risk factors(including smoking history,unhealthy diet,sedentary,low body mass index,dust-related job,and poor ventilation in kitchen);dependent variables were FEVi and FVC.The reference equation for lung function was fitted by a single-level linear function,a quadratic function,an exponential function,and a nonlinear multiplicative function respectively,and the one with the highest goodness of fit(evaluated by the determination coefficient R2)was chosen for internal and external validation.The study population focused on healthy non-smoking participants.Independent variables were height,age,sex,and ethnicity;dependent variables were normal values of FEVi and FVC.Stepwise regression method was used to select significant variables,with the significant level of 0.05 for both entry and staying in the model or equation.Participants of the study were divided into the development dataset(enrolled from Jan 1,2005 to Sept 18,2007,accounting for about 75%of the total population)and the validation dataset(enrolled from Sept 19,2007 to Dec 31,2009,accounting for about25%of the total population)according to the baseline recruitment date.Establishment ofthe estimation model and the reference equation was done in the development dataset.Validation of the estimation model and the reference equation was accomplished throughthe Bootstrap simulation in the development dataset(internal validation),and through thedirectly application in the validation dataset(external validation).ResultsThe estimation model for lung function reflected the nonlinear relationship betweenindependent variables and dependent variables at three levels of individual,community,and center(province × urban or rural)data.For the fixed effect:FEV1[L]=(0.52/100)× height[cm]1.55 × age[yr]-0.43 ×(1 + 12.59/100 × male)×(1-1.54/100 × former smoking)×(1 + 0.83/100 × current smoking)×(1 + 2.22/100 × regular intake of fruit and veget)×(1+ 1.31/100 × dust-related job)FVC[L]=(0.37/100)× height[cm]1.59 x age[yr]-0.36 ×(1 + 12.38/100 × male)×(1-0.41/100 × former smoking)×(1 + 1.20/100 x current smoking)×(1 + 2.15/100 × regular intake of fruit and veget)x(1 + 1.10/100 × dust-related job)For the random effect:we further considered the influence of individual-level residuals,community-level intercept,smoking status,fruit and vegetable intake,occupational exposure,and central-level intercept on lung function.The estimated values of FEV1 and FVC were calculated by the empirical Bayesian method.The goodness of fit of the model(FEV1:AIC =-2340.9,BIC =-2352.9;FVC:AIC =-2914.8,BIC = 2926.8)was much higher than that of the multi-level linear model(FEV1:AIC = 47136.8,BIC = 47124.8;FVC:AIC = 55769.3,BIC = 55757.3)or the single-level nonlinear model(FEV1:AIC = 9349.9,BIC = 9358.3;FVC:AIC = 12357.7,BIC =12366.1)under the same conditions.The estimation accuracy of the model reached the repeatability standard of spirometry test(FEVi:absolute deviation-0.05 L,relative deviation 2.91%;FVC:absolute deviation-0.06 L,relative deviation 2.81%)?and the model can be well-applied in both urban and rural areas,and in different geographic regions and ethnic populations in China.The model has a good discrimination in the diagnosis of COPD(area under the receiver operating characteristic curve:0.82).With FEV,/FVC<80.7%as the diagnostic criteria,the model can be used as a preliminary screening tool for COPD in the community(sensitivity:70.0%,specificity:78.5%).The reference equation for lung function was best-fitted by a linear function(FEVi:R2 =0.37;FVC:R2 = 0.34):FEV1 =-0.84-0.015×age[yr]+0.024×height[cm]+0.37×male+0.20×Mongol-0.049×Dai FVC =-1.03-0.016×age[yr]+0.029×height[cm]+0.45×male+0.24×Mongol+0.16×UygurBased on the parameter estimation,we found significant differences in reference values of FEV1 and FVC among ethnicities,which were not explained by height,age,and sex.Compared with existing reference equations for lung function,our new equation had less absolute deviation and relative deviation(FEV1:absolute deviation-0.04 L,relative deviation 1.71%;FVC:absolute deviation 0.08 L,relative deviation 6.55%).ConclusionsThe estimation model and the reference equation for lung function can estimate FEVi and FVC based on baseline characteristics with no need for spirometry test.Therefore,they solve the problems of spirometry test in China,and do help in the diagnosis of COPD and the assessment of airflow limitation severity for COPD patients,in the sense of personalized medicine.Moreover,the reference equation provides information on normal values of lung function among different populations,that may guide better medical resource allocation and more rational decision making,in the sense of public health.