The long term effects of COVID-19 on Pulmonary diffusing capacity for nitric oxide (DLNO) and carbon monoxide (DLCO)
Description
Hypothesis: This study hypothesized that combining lung diffusing capacities for nitric oxide (DLNO) and carbon monoxide (DLCO) enhances detection of COVID-19-related lung pathologies over using DLCO alone, or spirometry, total lung capacity (TLC), or single measures. A combined DLNO+DLCO z-score was expected to better classify post-COVID lung function abnormalities, capturing a broader spectrum of pulmonary impairment. Data Overview: The filtered pooled dataset (Covid19_Controls_Filtered.sav) includes 572 COVID-19 survivors & 72 controls (644 total) who completed all pulmonary function tests (PFTs) across five countries: Italy, France, Spain, Denmark, and Australia. About 98% of data from these centers is published in 6 peer-reviewed journals & 2 conference abstracts. Controls and COVID-19 cases were tested with the same equipment and protocols. The full PFT battery included DLNO and DLCO (median breath-hold 5.6 s), spirometry, and TLC, with z-scores from reference equations. COVID-19 survivors were tested 32–575 days post-infection (median 130 days). Missing TLC data was imputed. Mixed model binary logistic regression was used with “Study Site” as a random intercept, COVID-19 (1) vs. control (0) as the dependent variable, and z-scores for spirometry, TLC, and diffusion capacity as independent variables. Matthews correlation coefficient (MCC) assessed model classification using a threshold that maximized Youden's J. Notable Findings: Among 572 post-COVID subjects aged 20–90 (median 60), 61 (11%) had confirmed restriction (TLC < LLN, z-score < -1.645). Additionally, 53% showed some type of impairment (obstruction, restriction, mixed disorder, or DLNO/DLCO < LLN). Of 18 models evaluated, the one with the lowest Bayesian Information Criterion (BIC) was a combined DLNO+DLCO z-score model (DLNO and DLCO z-scores summed), showing superior COVID-19 detection. Four of the top five models were combined DLNO+DLCO or DLNO-only, while one was DLCO-only. MCC values showed all top five models were either combined or DLNO-only, with MCCs around 0.48. Dyspnea severity also correlated with combined, DLNO-only, and DLCO-only z-scores (p < 0.001). Data Interpretation and Implications: This study highlights that DLNO, when combined with DLCO, provides insights into COVID-19-altered lung diffusion, especially across the alveolar-capillary membrane. Combined z-scores improve post-COVID abnormality classification, supporting targeted follow-up for symptomatic survivors. Wider adoption of this approach would require regulatory approval and standardization. The combined technique shows promise in detecting subclinical COVID-19-related lung impairments, enhancing sensitivity in long-term monitoring.
Files
Steps to reproduce
This study is an individual participant data meta-analysis that compiled data from 11 previous studies (filtered data) [1–11]. White individuals with and without previous COVID-19 were included. Data from six centers (Italy = Genova, Bergamo, Milan; Spain = Vigo; France = Bobigny; Denmark = Copenhagen) were collected from Jan 2020 to Dec 2022 [1–10]; the 11th study was solely controls, collected in 2011 [11]. Local ethical standards allowed anonymized data use, eliminating the need for IRB approval for this analysis. The pooled dataset underwent rigorous screening and cleaning, with eligibility restricted to White subjects aged 18+. Data cleaning excluded entries with specific quality control issues, including: (A) alveolar volume (VA) exceeded total lung capacity (TLC) by ≥0.20 liters; (B) forced expiratory volume in 1 second (FEV1) was greater than forced vital capacity; (C) forced vital capacity (FVC) exceeded TLC; (D) breath-hold time >12 seconds; (E) control subjects had z-scores < -2.500 or > +2.500. This ensured accurate, reliable data and strengthened result validity. After cleaning, the dataset used was "Covid19_&_Controls_Filtered.sav". Z-scores were generated for pulmonary diffusing capacity for carbon monoxide (DLCO) and nitric oxide (DLNO), VA, TLC, FEV1, FVC, and FEV1/FVC ratio. Z-scores were derived using GLI reference equations for spirometry, TLC, and DLCO for White individuals [12–14], plus double-diffusion reference equations for White individuals [15–16]. To predict those with previous COVID-19, a mixed-effects logistic regression with “Study Site” as a random effect was used; DLNO and other variables were treated as fixed. MCC classified disease. Files are saved in SPSS (.sav) and .csv formats. "Variable_Labels_&Values1.csv" goes with "Covid19_&_Controls.sav or .csv"; "Variable_Labels&Values2.csv" with "Covid19&Controls_Filtered.sav/.csv"; and "Variable_Labels&_Values3.csv" with "Dyspnea_vs_Diffusing_capacity.sav or .csv". References 1. Barisione G & Brusasco V. Physiol Rep 2021: 9(4): e14748 2. Barisione G. & Brusasco V. ERJ Open Res 2023: 9(2) 3. Dal Negro RW, et al., Multidiscip Respir Med 2022: 17: 875 4. Imeri G, et al. Multidiscip Respir Med 2024: J Clin Med 2021: 10(10) 5. Nunez-Fernandez M, et al. J Clin Med 2021; 10 6. Seccombe LM, et al. Physiol Rep 2023: 11(7): e15660 7. Lytzen et al. (2024). Exp Physiol 109(5):652-661 8. Agostoni P et al. (2024). Respir Res. Feb 8;25(1): 82 9. Sesé L, et al. European Respiratory Journal 2022: 60(suppl 66): 2662 10. Sesé L, et al. [FRENCH]. Revue des Maladies Respiratoires Actualités 2022: 14(1): 138-139 11. Magini et al. (2015) Eur J Prev Cardiol. 2015 Feb;22(2):206-12. (control subjects only) 12. Quanjer PH, et al. Eur Respir J 2012: 40(6): 1324-1343 13 Hall GL, et al. Eur Respir J 2021: 57(3) 14. Stanojevic S, et al. Eur Respir J 2017: 50(3) 15. Munkholm M, et al. Eur Respir J 2018: 52(1): 1500677 16. Zavorsky GS & Cao J. BMJ Open Respir Res 2022: 9(1)