Clinical Predictors of Lung Function in Patients Recovering From Mild COVID-19

Arturo Cortes-Telles; Esperanza Figueroa-Hurtado; Diana Lizbeth Ortiz-Farias; Gerald Stanley Zavorsky

Disclosures

BMC Pulm Med. 2022;22(294) 

In This Article

Results

One hundred and fifty subjects were recruited from the Long-term follow-up COVID-19 clinic. Four patients were removed due to the reference equations not fitting the age range of the subjects, leaving 146 patients for the analysis. The anthropometric characteristics are presented in Table 1. Approximately 50% of the subjects were obese (BMI ≥ 30 kg/m2). The gas exchange parameter, DLCO, was reduced compared to predicted values (Table 2). In fact, 30% of the sample had a DLCO value below the LLN, and the percentages were similar between males and females (Table 3). Twenty-one percent of the sample had a restrictive spirometric pattern (FVC < LLN and an FEV1/FVC ≥ LLN) (Table 3), and when coupled with a DLCO < LLN, about 13% of the patients had both impaired DLCO and lung restriction.

For the regression analysis, two subjects had missing data, and then data from an additional five subjects were removed from the analysis, with their data being outliers (standardised residuals ≥ 3.0). Thus 139 subjects remained for the binary logistic regression analysis. The analysis revealed an overall model of four predictors that were statistically reliable in distinguishing between those with a DLCO below the LLN and those with normal DLCO [-2 Log-Likelihood = 101.7, Nagelkerke R2 = 0.51; Omnibus tests of model coefficients χ 2 = 60.0, df = 4, p < 0.001, Akaike Information Criterion (AIC) = 111; Bayesian Information Criterion (BIC) = 126]. Increased age (from 25 to 83 years of age) and a restrictive spirometric pattern increased the probability of an impaired DLCO, while a blocked/runny nose and excessive sweating reduced the probability of having an impaired DLCO. The model was a good fit [Hosmer and Lemeshow Test, χ 2 = 6.7, df = 8, p = 0.57], correctly identifying 81% of the cases (Table 4).

Using mean data from 21 previous studies (including the current study, see Additional file 1: Table S1), multiple linear regression analysis was used to identify which factors would predict the proportion of previously infected SARS-CoV-2 patients with impaired DLCO at follow-up. Regression results indicate an overall model of four predictors (previous history of severe COVID-19, criteria used to define impaired DLCO, mean age of the group, and number of days between diagnosis of COVID-19 and testing) that significantly determined the percentage of previously infected COVID-19 patients with an impaired DLCO at follow-up [R2 adj = 0.46, F(4,31) = 8.4, residual standard deviation = 15.6%, p < 0.001, AIC = 306, BIC = 316]. The model accounted for 46% of the variance defining those with DLCO impairment. A summary of the regression model is presented in Table 5. Having a previous severe case of COVID-19 increases the proportion of those with impaired DLCO by 21% at follow-up. When using the more liberal definition of impaired DLCO (< 80% of predicted), the proportion of those with impaired DLCO increased by 13% in a given study. Finally, when the patient's mean age for a study increased by one year (from 41 to 69 years), the proportion of those with impaired DLCO increased by about 1%. Interestingly, the time of follow-up (20 to 180 days) did not seem to affect the percentage of those with impaired DLCO in a particular study.

processing....