Application of a multiple linear regression model of FEV1 in pulmonary function test

Objective To construct a multiple linear regression model of forced expiratory volume in 1 second (FEV1) for estimating FEV1 in special populations unable to receive or uncooperative in pulmonary ventilation function tests. Methods The multiple linear regression model of FEV1 was constructed based on the data of 813 individuals undergoing pulmonary function tests in First Affiliated Hospital of Zhejiang Chinese Medical University between September, 2017 and September, 2019, and was validated using the data of another 94 individuals from the same hospital between January and July, 2020. FEV1 of the individuals was measured by pulmonary ventilation function test, and respiratory resistance (Rrs) was measured using forced oscillation technique (FOT). Pearson correlation analysis was used to assess the correlation between the factors, and the model equation was established by multiple stepwise regression analysis. The calculated FEV1 based on the model was compared with the measured FEV1 among both the individuals included for modeling and validation. Results FEV1 was not significantly correlated with BMI (r=-0.026, P=0.457), poorly correlated with body mass (r=0.382, P=0.000), positively correlated with height (r=0.723, P=0.000), and negatively correlated with Rrs (r=-0.503, P=0.000) with an obvious gender differences (t=18.517, P=0.000). FEV1 was positively correlated with age among individuals below 25 years of age (r=0.578, P=0.000) and was negatively correlated with age among those beyond or at the age of 25 (r=-0.589, P=0.000). For individuals beyond or at the age of 25 years, the variables of height, gender, age and Rrs were included in the model, and the calculated FEV1 did not differ significantly from the measured values in either the modeling sample (n=751; t=1.293, P=0.196) or the verification sample (n=83;t=-1.736, P=0.086), and the two values were well correlated in the verification sample (r=0.891, P=0.000). For individuals below 25 years, only height was included in the model, and the calculated FEV1 and the measured values showed no significant difference in the modeling sample (n=62; t=-0.009, P=0.993) or the verification sample (n=11; t=-0.635, P=0.540) with a good correlation in the verification sample (r=0.795, P=0.003). Conclusions The multiple linear regression model for calculating FEV1 constructed in this study is suitable for clinical application.