Risk factors for clinically significant diffuse parenchymal lung abnormalities persisting after severe COVID-19 pneumonia

Several patients have persistent respiratory symptoms, reduced lung function and imaging abnormalities after coronavirus disease 2019 (COVID-19) pneumonia1-3. At six months, the reported prevalence of these abnormalities has variably ranged between 20 and 100 per cent4-9. Age, severity of hypoxemia and overall critical illness, co-morbidities and others have been found to be independent risk factors for prolonged mechanical ventilation in patients with COVID-1910. Age, peak C-reactive protein (CRP), the severity of hypoxemia at admission, mechanical ventilation and the length of hospital stay (LOS) have been associated with residual imaging abnormalities4,7,9,11,12. As these radiologic sequelae may be clinically insignificant, their relevance has been questioned13,14.

The true prevalence and predictors of ‘clinically significant’ lung abnormalities after COVID-19 pneumonia remain uncertain. In this study, we defined clinically significant diffuse parenchymal lung abnormalities (CS-DPLA) after severe COVID-19 pneumonia using a multidimensional model. We report the prevalence of CS-DPLA at two and six months after discharge and analyze the factors associated with CS-DPLA. Our primary objective was to assess whether the severity of acute COVID-19, defined by the neutrophil-lymphocyte ratio (NLR), a known surrogate marker of severity, was associated with CS-DPLA. Furthermore, it was explored whether the serum levels of biomarkers previously used in interstitial lung disease, namely surfactant protein D (SP-D), cancer antigen 15-3 (CA 15-3)/mucin 1 and transforming growth factor-β (TGF-β) would differ between participants with or without CS-DPLA.

Material & Methods

Study design & setting: This study was conducted at the department of Pulmonary Medicine, Postgraduate Institute of Medical Education & Research, Chandigarh, between January 2021 and February 2022. The Institutional Ethics Committee approved the study protocol (NK/6845/MD/785). A written informed consent was obtained from all the participants.

Participants: The participants were prospectively recruited from a follow up cohort of post-COVID-19 patients. The study included consecutive patients enrolled after recovery from the acute phase of the COVID-19 illness if they satisfied all the following criteria: (i) age ≥18 yr; (ii) diagnosis of COVID-19 by a real-time reverse transcriptase(RT)-PCR for severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) or a positive rapid antigen assay for SARS-CoV-2; (iii) severe COVID-19 pneumonia requiring hospitalization and supplemental oxygen or advanced respiratory support (ARS); and (iv) provided a written informed consent to participate in the study and the use of their data for research. Participants with pre-existing structural lung disease, pregnant or lactating women were excluded from the study.

The study participants were divided into two groups i.e. cases and controls. Both these groups were designated separately for the follow up visits at two and six months. Cases were defined as participants with CS-DPLA after COVID-19 pneumonia. Clinically significant DPLA was defined as radiologically diffuse involvement of ≥20 per cent of the lung parenchyma with any of the following: (i) dyspnoea severity on the modified Medical Research Council (mMRC) scale ≥2; (ii) resting oxygen saturation <95 per cent; (iii) ≥4 per cent desaturation on exercise testing; or (iv) abnormal spirometry. Controls were defined as participants without CS-DPLA. All study participants presenting at the two month visit and most participants presenting for the six month visit underwent chest computed tomography (CT), which was used to assess the radiological involvement. Participants unwilling for chest CT at six months underwent chest radiography with the subsequent assessment of radiologic severity.

For the biomarker study, the first 80 patients were selected who presented for a follow up visit. Twenty adult volunteers without any acute illness, chronic respiratory illness and without a history of severe COVID-19 were included as healthy controls.

Procedures: The following information was collected: age, sex, history of tobacco smoking, comorbid illness, the blood NLR at hospital admission, the ratio of oxygen saturation to fractional inspired oxygen (SFR) at admission, the worst value of serum lactate dehydrogenase (peak LDH), the worst value of serum C-reactive protein (CRP) during hospitalization (peak CRP), type of respiratory support [routine oxygen supplementation with nasal prongs or mask, oxygen delivery using highflow nasal cannula (HFNC), non-invasive ventilation (NIV) or invasive mechanical ventilation (IMV)], the peak requirement of fractional inspired oxygen (FiO₂) during the acute illness and the length of stay (LOS) in the hospital. The following assessments was performed at the baseline (at the time of discharge), two months and six months: clinical history and examination, dyspnoea severity on the mMRC scale (Supplementary Appendix), resting oxygen saturation on pulse oximetry, exercise testing using six-minute walk test (6MWT) or one-minute sit-to-stand test (in participants unable to perform 6MWT) and a thin-section CT of the chest (or chest radiography for unwilling subjects at six months). Spirometry was performed at the follow up visits only. The respiratory condition, dyspnoea severity and vital status of subjects who did not present for follow up were ascertained telephonically.

Volumetric acquisition for the chest CT and reconstructed thin-collimation (1 mm) sections (Supplementary Appendix) were also performed. Two experienced radiologists blinded to the study participants’ clinical and laboratory details independently categorized the extent of parenchymal involvement due to COVID-19 into ≥20 or <20 per cent by visually inspecting both the lung fields on the chest CT images15. Disagreements were resolved by discussion to reach a consensus. They also recorded the presence of ground-glass opacities, consolidation, parenchymal bands, reticulation and traction bronchiectasis (or honeycombing) using standard definitions16 (Fig. 1).

Close

Fig. 1

Representative images of thin-section computed tomography (CT) scans of the chest showing various radiological abnormalities in study participants. (A) Subtle ground glass opacities (asterisk) affecting <20 per cent of the lung parenchyma in a subject without clinically significant diffuse parenchymal lung abnormalities; (B) Ground glass opacities (asterisk) and parenchymal bands affecting ≥20 per cent of lung parenchyma; (C) Chest CT showing predominantly consolidation (arrow) and ground glass opacities; (D) Chest CT showing areas with parenchymal bands (small solid white arrows), reticulation (open arrow) and traction bronchiectasis (black arrow).

Export to PPT

The forced vital capacity (FVC), forced expiratory volume in the 1^st sec (FEV₁) and the FEV₁ to FVC ratio was recorded during spirometry17 (Supplementary Appendix). Spirometry was considered abnormal if either the FEV₁/FVC ratio (obstructive defect) or FVC [with a normal FEV₁/FVC (restrictive defect)] was below the corresponding lower limit of normal for our population18. Further, the 6MWT has performed and the distance walked during the test (6MWD) was recorded (Supplementary Appendix). The nadir oxygen saturation during the test was recorded using finger pulse oximetry. Desaturation was defined as ≥4 per cent fall in oxygen saturation anytime during the test. In subjects unable to perform the 6MWT, exertional oxygen desaturation was assessed during a one-minute sit-to-stand test (Supplementary Appendix). Percentage of the predicted (%pred) values for FVC and 6MWD18-20 were calculated.

Blood samples (5 ml) were collected from the study participants at baseline (at hospital discharge) and the isolated serum was stored at −80°C. The levels of surfactant proteins-D (SP-D), cancinoembryonic antigen (CA) 15-3 and transforming growth factor-beta (TGF)-β using the enzyme-linked immunoassay kits (NOVOS Biological, Centennial, CO, USA) were measured as per the manufacturer’s instructions (Supplementary Appendix).

Variables and outcomes: The primary exposure was the NLR at admission. Other markers associated with the severity of COVID-19 illness were also tested in univariate analysis, which included peak LDH, peak CRP, the SFR at admission and the peak FiO₂ required during acute illness. The potential confounders were age; sex; smoking; comorbidities such as diabetes mellitus, hypertension, coronary artery disease, obesity, chronic liver disease, chronic kidney disease, or obstructive airway disease; the requirement of ARS (mechanical ventilation or HFNC) and the LOS. The strength of association between the primary exposure and the confounders with the outcome of interest: CS-DPLA at the two and six month visits was assessed using odds ratios (ORs). The secondary outcome was the comparison of the levels of SP-D, CA 15-3 and TGF-β among cases, controls and healthy volunteers.

Statistical analysis: A formal sample size calculation was not performed, as the prevalence of CS-DPLA as defined in this study was not known at study initiation. The prospective cohort used as the source of the cases and controls in this study was time bound with the last inclusion in August 2021 and follow up till February 2022. All analyses were performed using the Statistical Package for the Social Sciences (SPSS) software for MS Windows (version 22; SPSS Inc.; Chicago, IL, USA). Descriptive data was presented as mean±standard deviation (SD), median (interquartile range) or number (percentage). The categorical variables were compared using the Chi-squared test (or Fischer’s exact test) and continuous variables using Student’s t test (or Mann-Whitney U test). Cohen’s Kappa statistics was used to assess the agreement between radiologists for scoring disease extent on chest CT. The association of the primary exposure and confounding factors with CS-DPLA was assessed by the univariate and multivariate logistic regression analyses. Multicollinearity was assessed amongst the markers of acute COVID-19 severity by calculating Spearman’s correlation coefficients and the variation inflation factor. For multivariate analysis, the primary exposure, the demographic variables known to be risk factors for diffuse parenchymal lung abnormalities (age, sex and smoking) and any variables with a P<0.1 in univariate analyses were entered as covariates. The OR and aOR was calculated for the tested covariates, along with the 95 per cent confidence intervals (CI).

Results

A total of 285 participants were screened, of which 236 were included in the prospective cohort (Fig. 2), with a mean age of 56.2 yr (34.3% women). Of these, 180 (76.3%) participants made a follow up visit at two and/or six months. Participants lost to follow up (n=56; 23.7%) had lower dyspnoea severity, lesser radiologic lung involvement and a lower prevalence of resting hypoxemia at baseline (Supplementary Table I). Three participants died during the follow up, all of them within the first two months. The probable causes of death were myocardial infarction, advanced renal failure and pulmonary embolism. One hundred and sixty and 110 participants presented for the two-month and six month visits, respectively. There was moderate agreement among the radiologists for categorizing the extent of lung parenchymal involvement with post COVID-19 DPLA in the two months (weighted Cohen’s kappa, 0·75) and six months (weighted Cohen’s kappa, 0·76) chest CT scans. About 176/180 (97.8%) of our participants received glucocorticoids during hospitalization, while 141/180 (78.3%) of the participants received low-medium dose glucocorticoids after discharge. About 14/180 (7.8%) of our participants developed infections during the follow up (4 urinary tract infection, 2 tuberculosis, 2 wound infection and pulmonary aspergillosis, tracheostomy site infection, infective diarrhoea, recurrent COVID-19, chickenpox and cellulitis in 1 participant each).

Close

Fig. 2

Participant flow chart.

Export to PPT

Prevalence of clinically significant diffuse parenchymal lung abnormalities (CS-DPLA) at two and six months: Of the 160 participants, who made a follow up visit at two months, 91 (56.9%) had CS-DPLA (cases), whereas 69 (43.1%) subjects qualified as controls. Thirteen participants (controls) had dyspnoea severity of mMRC grade 2 or more at two months but had <20 per cent lung involvement on the chest CT; all improved at six months. Of the 91 subjects with CS-DPLA (cases) at two months, 36 continued to have CS-DPLA at six months (counted as cases at six months), 19 did not have CS-DPLA at six months (counted as controls at six months), while 36 were lost to follow up. Among the 36 who were lost to follow up, 26 did not have significant dyspnoea and nine had significant dyspnoea on telephonic enquiry at six months. One participant could not be contacted.

At six months, 110 partcipants presented for follow up. Thirty four participants, who did not have CS-DPLA at two months, did not present for further follow up. On telephonic enquiry, none complained of respiratory symptoms. Therefore, we assumed these 34 participants as not having CS-DPLA at six months either. We made this assumption because all the controls from the two month visit, who presented for the six month visit (n=35) remained free of CS-DPLA at the latter visit also (Fig. 2). Thus, 144 participants were included in the six month analysis. Of these, 42 (29.2%) had CS-DPLA (cases), whereas 102 (70.8%) participants qualified as controls. The radiologic assessment was performed on chest radiography for 32 of the 110 subjects at six months as they refused to undergo chest CT scan (five qualified as cases, while 27 were designated as controls). Six participants did not present at two months but had CS-DPLA at six months. The dyspnoea severity was higher while the %pred FVC and the 6MWD were lower in the cases compared to controls at two and six months (Supplementary Table II).

About 13 and 60 per cent of our participants were free of dyspnoea (mMRC grade 0) at two months and six months, respectively (Fig. 3). Among participants with CS-DPLA at two months, 85.7 per cent had a restrictive defect, 2.8 per cent had obstructive defect, whereas 8.3 per cent had normal spirometry. Among participants with CS-DPLA at six months, 88.9 per cent had restrictive defect, 4.8 per cent had obstructive defect and 9.5 per cent had normal spirometry. Participants with CS-DPLA at two months had significant improvement in FVC at the six month follow up [mean (SD), 332 (287) ml] as compared to those without [mean (SD), 176 (237) ml; P=007]. About 74.4 per cent of the participants undergoing a chest CT at six months had persistent ground glass opacities, whereas 33.3 per cent had traction bronchiectasis (Supplementary Table III).

Close

Fig. 3

Dyspnoea severity graded on the modified Medical Research Council (mMRC) scale in study participants at the two and six month visits.

Export to PPT

Predictors of CS-DPLA at two and six months: Participants with CS-DPLA at two months had higher NLR at admission and peak LDH, longer LOS and required ARS more commonly than the controls (Table I). Participants with CS-DPLA at six months had higher NLR at admission and a longer hospital LOS. Multicollinearity was absent amongst the measures of acute COVID-19 severity (NLR, peak LDH, peak CRP, SFR and peak FiO2). Multivariate analysis showed that hospital LOS was the only factor that independently predicted CS-DPLA (Table II) at two months [aOR (95% CI, 1.16 (1.07-1.25); P<0.001] and six months [aOR (95% CI, 1.07 [1.01-1.12]; P=0.01).

Biomarkers: The cases at two months had higher baseline serum levels of SP-D and CA 15-3 than the healthy volunteers and higher SP-D than the controls (Table III). Among those without CS-DPLA, the CA 15-3 was significantly elevated than the healthy volunteers. Participants with CS-DPLA at six months had higher baseline serum levels of TGF-β than the healthy volunteers.

Discussion

We did not find an independent association of the NLR at the time of hospital admission for severe COVID-19 with CS-DPLA in our study. The length of hospital stay was found to be the only factor independently associated with CS-DPLA at two and six months after severe COVID-19 pneumonia. We also found TGF-β as a potential biomarker of CS-DPLA at six months.

The prevalence of residual lung imaging abnormalities at six months has been reported to range between 20 and 100 per cent4-7. The wide variation is due to the reporting of even minor residual CT findings in several studies, which might be clinically inconsequential. About 10 per cent of patients with post-COVID-19 DPLA may have abnormal spirometry, while 20-60 per cent may have diffusion impairment, depending on the initial COVID-19 severity5,6. Viewing radiologic, lung function and oxygenation abnormalities in isolation fails to provide a precise understanding of their clinical significance. Our composite assessment approach based on breathlessness, radiologic abnormalities, spirometry and measures of gas exchange is thus clinically more relevant. Our choice of 20 per cent involvement on a visual assessment of the chest CT as the threshold for defining significant radiologic involvement was arbitrary. However, it was based on the conventional practice for defining clinically significant disease in systemic sclerosis-related ILD15. Dyspnoea severity of ≥2 on the mMRC scale was considered as clinically significant based on previous evidence for the clinical relevance of this threshold in patients with idiopathic pulmonary fibrosis21.

The prevalence of CS-DPLA at two and six months was about 60 and 30 per cent, respectively in hypoxemic patients discharged after hospitalization for severe COVID-19. The true prevalence is likely to be lower as nearly one fourth of our participants were lost to follow up. These participants had lower dyspnoea severity, higher oxygen saturation and lesser radiologic involvement at discharge and would be expected to have a lower prevalence of CS-DPLA than those who presented for follow up. The prevalence of post-COVID-19 sequelae in the longer term is generally considered to be 10 per cent in unselected patient populations11,22. About 60 per cent of our study participants were free of dyspnoea at six months. In three previously reported large follow up studies, 70-74 per cent of patients had no dyspnoea at six months of follow up5,6,8. This is attributable to the baseline difference in the study populations. All our study participants required supplemental oxygen and several required mechanical ventilation. Contrastingly, in two of the previous studies5,6, hospitalized participants with milder illnesses were also included; while, in the study by Wu et al8, patients who had been mechanically ventilated were excluded5,6,8.

The NLR was selected as a surrogate for COVID-19 severity as it has been found to correlate with the severity and outcomes of acute illness in several studies23-25. Confounders were selected based on the results of previous studies on COVID-19 sequelae4,7,11,12,26. Age, severity of the acute illness (measured using a variety of parameters in different studies including the severity of hypoxemia, tachycardia, radiologic score, LDH, CRP and others), mechanical ventilation and hospital LOS might predict residual radiologic and physiologic abnormalities4,7,11,12,26. The parameters of acute COVID-19 severity (NLR, peak LDH, peak CRP, SFR and peak FiO2) did not predict CS-DPLA at six months in our study, although NLR had a significant association in the univariate model. We found only the hospital LOS to be independently associated with CS-DPLA at six months. The length of hospital stay has been the most consistent predictor of pulmonary sequelae across studies4,7,9,26. This indicates that lung parenchymal injury during severe COVID-19 pneumonia is multifactorial11. The initial severity of the SARS-CoV-2 infection (reflected by NLR, LDH, CRP, hypoxemia and others) causing extensive diffuse alveolar damage is an important factor. However, it is possibely that the duration for which the lung injury persists, both because of the infection and the effects of the medical management including ARS, best predicts clinically significant lung sequelae in the longer term27. The duration of intensive care unit (ICU) stay and mechanical ventilation could be more precise measures of the duration of lung injury. However, due to the blurred boundaries between ICU and non-ICU units, extensive use of HFNC, frequently alternating usage of NIV and HFNC and changing triggers for IMV during the deluge of the COVID-19 pandemic, the two parameters could not be measured accurately28.

Substantial differences found between COVID-19 related and classical acute respiratory distress syndrome, could be due to the facts that the biomarkers indicative of endothelial injury, were found to be higher in COVID-19-related ARDS, supporting different pathophysiological pathways for these two syndromes29. Further, COVID-19 leaves behind substantial sequelae that mimic an interstitial lung disease process30. SP-D, a marker of lung injury and TGF-β, a key cytokine involved in lung fibrosis has been previously used as prognostic and severity markers for chronic ILDs31-33. SP-D was found elevated and correlated with disease severity during acute COVID-19 illness in a previous study34. Another study reported higher levels of TGF-β in patients with radiologic changes suggesting lung fibrosis after COVID-1935. In our study, patients with CS-DPLA at two months had elevated levels of SP-D and CA 15-3. CA 15-3 was also elevated in the control group as compared to healthy volunteers and thus seems to offer a lesser discriminant potential. Participants with CS-DPLA at six months had elevated TGF-β at the baseline visit possibly reflecting the initiation of a profibrotic pathway in such patients right at the time of hospital discharge (or even earlier). These serum cytokines, especially SP-D and TGF-β should be explored further as biomarkers for predicting significant residual lung sequelae after COVID-19. Besides, elevated TGF-β levels provide a pathophysiological basis for ongoing studies of TGF-β inhibitors such as pirfenidone and other antifibrotic agents in patients with persistent residual abnormalities after severe COVID-19 (Clinicaltrials.gov identifiers; NCT04653831, NCT04282902, NCT04607928, NCT04856111, NCT04541680, NCT04338802 and NCT04619680)30,36.

The findings of the present study indicate that a considerable proportion of patients may continue to have CS-DPLA at six months, especially those with severe acute disease and prolonged hospitalization. Elevated levels of TGF-β at the time of discharge may indicate a higher risk of having CS-DPLA at six months. Such participants must be followed up closely after discharge from the hospital. Besides, our study proposes a framework for a multidimensional assessment of patients after recovery from acute COVID-19.

The present investigation is limited by its single centre design and significant attrition. We tried mitigating the bias due to attrition by comparing the characteristics of included participants with those lost to follow up. We could not measure the biomarkers in all included participants due to logistic constraints. Furthermore, the diffusion capacity of the lungs and baseline spirometric parameters could not be measured due to COVID-19 restrictions. Several participants underwent only chest radiography during the six-month visit as they refused a repeat CT evaluation. Chest CT has its limitations as it involves radiation exposure, and thus, repeat studies cannot be obtained frequently. Electrical impedance tomography is an imaging technique that reconstructs the images of a specific body part based on the electrical conductivity of biological tissue. It does not involve radiation and can also provide information on the dynamic changes in the lung. Its successful use in COVID-19 survivors has been reported recently37. We assumed participants not having CS-DPLA at two months as not having CS-DPLA at six months either, as they did not have any respiratory complaints. Although a limitation, it is unlikely that participants who have disease resolution at two months would have any recurrence. Furthermore, most of our cases at six months were derived from those who had CS-DPLA at two months as well.

The findings of the present study suggest that among hypoxemic patients hospitalized for severe COVID-19, about 40 per cent participants continued to have dyspnoea severity of at least mMRC grade 1 at six months after discharge. CS-DPLA was present in about 60 and 30 per cent at the two and six month follow up, respectively. The length of hospital stay was found to be the only independent predictor of CS-DPLA at six months. SP-D and TGF-β should be further explored as biomarkers for assessing the risk of CS-DPLA after severe COVID-19 pneumonia.

Data statement: Data collected for the study, including de-identified participant data and related documents, including the protocol, statistical analysis plan and informed consent form, will be made available to qualified researchers after the publication of the manuscript upon reasonable request via application to the corresponding author.