Evaluation of Follow-Up CT Scans in Patients with Severe Initial Pulmonary Involvement by COVID-19

Objective To investigate the predictive factors of residual pulmonary opacity on midterm follow-up CT scans in patients hospitalized with COVID-19 pneumonia. Materials and Methods This prospective study was conducted in a tertiary referral university hospital in Iran, from March 2020 to December 2020. Patients hospitalized due to novel coronavirus pneumonia with bilateral pulmonary involvement in the first CT scan were included and underwent an 8-week follow-up CT scan. Pulmonary involvement (PI) severity was assessed using a 25-scale semiquantitative scoring system. Density of opacities was recorded using the Hounsfield unit (HU). Results The chest CT scans of 50 participants (mean age = 54.4 ± 14.2 years, 72% male) were reviewed, among whom 8 (16%) had residual findings on follow-up CT scans. The most common residual findings were faint ground-glass opacities (GGOs) (14%); fibrotic-like changes were observed in 2 (4%) patients. Demographic findings, underlying disease, and laboratory findings did not show significant association with remaining pulmonary opacities. The total PI score was significantly higher in participants with remaining parenchymal involvement (14.5 ± 6.5 versus 10.2 ± 3.7; P=0.02). On admission, the HU of patients with remaining opacities was significantly higher (−239.8 ± 107.6 versus −344.0 ± 157.4; P=0.01). Remaining pulmonary findings were more frequently detected in patients who had received antivirals, steroid pulse, or IVIG treatments (P=0.02, 0.02, and 0.001, respectively). Only the PI score remained statistically significant in multivariate logistic regression with 88.1% accuracy (OR = 1.2 [1.01–1.53]; P=0.03). Conclusion Pulmonary opacities are more likely to persist in midterm follow-up CT scans in patients with severe initial pulmonary involvement.


Introduction
In January 2020, the World Health Organization declared COVID-19 a global pandemic. COVID-19 is highly contagious, and in severe cases, it can lead to lung involvement and acute respiratory compromise or organ failure [1]. With gradual recognition of COVID-19 pneumonia, guidelines and criteria were steadily established so as to prevent its transmission and facilitate its diagnosis and treatment [2].
Using a chest-computed tomography (CT) scan is very important for rapid diagnosis and clinical decisionmaking [3]. According to the World Health Organization and the Centers for Disease Control and Prevention guidelines, chest radiography and CT were major diagnostic components when COVID-19 was probable [4]. Dominant CT fndings in COVID-19 pneumonia mostly include bilateral ground-glass opacities (GGOs), multifocal patchy consolidations, and peripherally distributed interstitial changes. Chest CT manifestations may difer in each patient during each stage of the disease, which can be used to diferentiate a diagnosis along with the severity of pulmonary involvement (PI), follow the changes, adjust the treatment, and predict the prognosis [5].
When it comes to COVID-19 pneumonia, the recent radiology literature focuses primarily on CT fndings as they are more sensitive than chest radiography [6][7][8]. A chest CT scan can detect small areas of GGO [9] and, therefore, is a promising imaging modality for monitoring the disease, if the radiation dose is balanced with radiologic principles. However, factors predicting long-term pulmonary sequelae have not yet been fully discussed [10]. Terefore, we designed this study to detect predisposing factors for residual pulmonary opacities on an 8-week follow-up CT scan. . Te current study was conducted in a tertiary referral university hospital from March 2020 to December 2020. Inclusion criteria were as follows: (a) being hospitalized with COVID-19 pneumonia confrmed by the positive rRT-PCR assay; (b) bilateral lung involvement in the initial chest CT scan; and (c) agreement to undergo an 8week interval follow-up CT scan. Of note, admission, discharge criteria, and treatment of all patients were based on the national protocol of COVID-19. Patients were categorized into the two following groups based on the follow-up CT scan: complete resolution and residual fndings. All demographic, clinical, and paraclinical data were also compared between the two groups.

Image Acquisition.
All chest CT images were obtained at the time of admission and eight weeks later using the Siemens SOMATOM Emotion (16 slices, Erlangen, Germany) MDCT scanner. All CT scans were performed in the supine position with deep inspiration, and the imaging parameters were as follows: 5 mm slice thickness; beam collimation of 1.2 mm; tube voltage, 130 kVp; and tube current, 70 mAs.

Image Interpretation.
A board-certifed radiologist with 11 years of experience in thoracic radiology was blinded to patients' outcome and interpreted chest CT images, reviewing both lung and mediastinal window settings. Chest CT scan fndings were recorded according to the Fleischner Society glossary and published literature on viral pneumonia (2). Te following features were considered signifcant remaining pulmonary fndings: (a) fbrotic-like changes: traction bronchiectasis, parenchymal bands, or coarse reticular patterns; (b) faint GGO: subsegmental atelectasis, mosaic attenuation, and very faint GGO (HU <-500) were not considered signifcant remaining pulmonary fndings. Chest-CT-scan features included (a) predominant pattern: GGO or consolidation; (b) dominant distribution pattern: peripheral (peripheral one-third of the lung), axial, or diffuse; (c) number of involved lobes; (d) additional fndings: cardiomegaly, pleural efusion, subsegmental atelectasis, parenchymal band, crazy paving, and reverse halo sign; and (f ) density of opacities: using the Hounsfeld unit (HU) by placing a 10 mm region of interest.

Pulmonary Involvement (PI) Scoring System.
To assess PI, a semiquantitative scoring system was proposed. All fve lung lobes were visually reviewed for GGO and consolidation, and the total involvement of each lobe was scored from 0 to 5 according to the volume percentage of involvement (0: no involvement; 1: ≤5%; 2: 6-25%; 3: 26-50%; 4: 51-75%; and 5: ≥76%). Te total PI score was calculated as the sum of all the fve lobes' scores. Te PI score ranged from 0 (no involvement) to 25 (maximum involvement). Finally, the PI density index was calculated by dividing the total PI score by the number of involved lobes.

Statistical Analysis.
We performed analyses using SPSS for Windows ver. 18 (Chicago, IL, USA). All p values less than 0.05 were considered statistically signifcant. Nominal and continuous variables were reported as the frequency (%) and mean ± standard deviation, respectively. Te normality of data was evaluated by the Kolmogorov-Smirnov test. We conducted the comparisons by the independent sample t-test for normally distributed continuous variables; the Mann-Whitney U test was used for non normally distributed continuous variables, and (c) the Chi-squared test was used for nominal variables. An association between the HU and residual fndings was investigated for all 250 lobes of 50 patients.

Results
Data on 50 participants (male, 36 (72%); mean age, 54.4 ± 14.2 years) were analyzed. Among those, 8 (16%) had residual fndings on the follow-up CT scan. Te most common residual fndings were faint GGOs (14%) and fbrotic changes (4%). Tirty-two (64%) patients had at least one underlying disease. Te patients' demographic, clinical, and paraclinical data in the two groups are presented in Tables 1 and 2. Demographic data, age and sex, did not show a signifcant association with residual fndings in CT scans. Besides, onadmission SpO 2 in the complete resolution and residual groups was 81.5 ± 8.0 and 79.0 ± 9.7, respectively (P � 0.46). Underlying disease was detected in 66.7% of patients in the complete resolution group and 50% in the residual group (P � 0.25). On-admission laboratory tests also had no correlation with remaining parenchymal fndings (Table 1).
Further analyses revealed that treatment regimens had signifcant associations with midterm chest CT fndings. Remaining pulmonary opacities were more common in patients who had received antiviral, steroid pulse, or IVIG treatments (P � 0.02, 0.02, and 0.001, respectively).
Te total PI score was signifcantly higher in participants with remaining parenchymal involvement (14.5 ± 6.5 versus 10.2 ± 3.7; P � 0.02). However, the total GGO, consolidation, and PI density scores showed no diference between the two groups (P � 0.48, 0.41, and 0.14, respectively). Predominant involvement and the distribution pattern of initial compromised lung parenchyma were the same between the two groups (P � 0.56 and 0.61, respectively). Te number of involved lobes showed no association with remaining parenchymal fndings (P � 0.22). Besides, no signifcant diference was detected in additional fndings between the two groups (Table 2).
Additionally, of note, the HU of opacities in the onadmission CT scan was signifcantly higher in patients with remaining opacities than patients in resolved ones (−239.8 ± 107.6 in the group with residual opacities versus −344.0 ± 157.4 in resolved ones; P � 0.01) (Figures 1 and 2).
Backward multivariate logistic regression was applied. We used residual fndings as the outcome of interest and all variables with P value < 0.1 as independent variables. Only the total PI score remained statistically signifcant in multivariate logistic regression with 88.1% accuracy (OR � 1.2 [1.01-1.53]; P � 0.03).

Discussion
We evaluated the potential predictive factors for remaining parenchymal fndings in chest CT scans obtained 8 weeks after discharge. Follow-up imaging fndings mainly included faint GGOs and fbrotic-like changes. We found that remaining pulmonary involvement was signifcantly more common in patients with a higher on-admission PI score or who received antiviral, steroid pulse, or IVIG treatments. Besides, denser opacities were more likely to persist in midterm follow-up CT scans.
Factors predicting the midterm chest CT fndings were investigated through limited previous studies. Te largest study investigated the predictive factors of parenchymal fbrotic changes in a six-month CT scan of 114 patients. Fibrotic changes were detected in 35% of patients. A higher initial PI score was found in patients with remaining parenchymal involvements, which aligned with our fndings. [14]. Similar results were also reported in another study of 41 patients in a seven-month follow-up investigation [15]. Another study evaluating the three-month follow-up CT scans of 52 COVID-19 patients reported GGOs (54.5%) and subpleural parenchymal bands (31.8%) as the most common fndings on midterm images. Also, a higher on-admission PI score was found in patients with residual fndings [12]. Another retrospective study evaluated 14 patients with fbrosis and 18 patients without fbrosis on the follow-up CT scan. It was found that patients with older age and higher levels of infammatory indicators (CRP and interleukin-6) end up more frequently in the fbrotic group. Longer hospital stay, pulsed steroid therapy, and antiviral therapy were also associated with fbrosis [11]. A prospective study on 80 participants showed 48% GGOs (48%), bands (37%), and fbrosis (12%) in COVID-19 patients after about 3 months. Te predictors of midterm sequelae in this study were levels of CRP, fbrinogen, urea, and creatinine, age, hospital stay, and mechanical ventilation [16].
Te PI score has been confrmed to predict the shortterm outcome and prognosis in COVID-19 patients [8]. In this study, we found that patients with extended parenchymal involvement are more likely to show residual fndings in follow-up images. Terefore, they may beneft from follow-up CT examinations due to some extent of pulmonary sequelae. However, studies with long-term follow-up are required to confrm our results.
Our further analyses showed higher rates of residual fndings in patients who had received antiviral, steroid pulse, or IVIG treatments during hospitalization. Antibiotic therapy and mechanical ventilation were not signifcantly associated with residual fndings. Yu et al. also found a signifcant association between pulmonary sequelae and steroid pulse and antiviral treatments in 32 COVID-19 patients [11]. Another study also found a higher probability of pulmonary fbrotic changes in patients who underwent steroid therapy or mechanical ventilation in a seven-month CT scan [15]. However, only steroid therapy remained statistically signifcant in multivariate logistic regression, and the PI score was the only independent predictive variable in our multivariate logistic regression. We assume that since patients with more severe initial pulmonary involvement undergo more intensive treatments, the relationship between these treatments and residual fndings might be biased. Altogether, we believe that extent of initial PI is the best predictor of midterm pulmonary fndings.
Our study has several limitations. We only included hospitalized patients that may not properly refect the midterm sequelae of COVID-19 in the community. Te subjective nature of qualitative and semiquantitative chest Canadian Respiratory Journal 3 CT scan evaluations can result in diferent imaging fndings and should be replaced with more accurate quantitative measurements in later studies. Future studies on larger populations and long-term imaging follow-ups of COVID-19 patients could reveal the potential burden of this disease on the involved organs including the lungs, which changes further approach to therapeutic strategies of COVID-19. We also recommend spirometry measurements to evaluate the long-term consequences of novel coronavirus pneumonia in the future. However, sufcient evidence of chronic complications and remaining pulmonary sequelae in COVID-19 patients is still lacking due to the novelty of this virus and diversity in its fndings.  In conclusion, we found signifcantly higher residual pulmonary parenchymal involvement in patients with higher initial PI scores and those receiving antiviral therapy, steroid pulse, or IVIG treatments. Besides, opacities with higher HUs were more likely to persist in midterm imaging. We recommend that patients with severe initial pulmonary involvement should receive intensive care and be followed to minimize further sequelae.

Data Availability
Te data used to support the fndings of this study are available from the corresponding author upon request.

Conflicts of Interest
Te authors declare that there are no conficts of interest regarding the publication of this paper. Canadian Respiratory Journal 5