Chest CT Patterns from Diagnosis to 1 Year of Follow-up in Patients with COVID-19

Study Sample

We prospectively evaluated participants with COVID-19 consecutively discharged from two isolation centers (ie, Western Campus and Zhuankou Fangcang Shelter Hospitals, affiliated with Union Hospital of Tongji Medical College at Huazhong University of Science and Technology) between January 27, 2020, and March 31, 2020. This cohort included participants reported in a previous study that involved only data at admission and discharge for the longitudinal analysis in our study (7). The criteria for diagnosis, therapy, and discharge followed the nationally standardized protocols (19). Patients with a respiratory rate greater than 30 breaths per minute or an oxygen saturation of 93% or less while breathing room air were classified as having severe COVID-19, whereas ARDS was diagnosed when the ratio of arterial partial pressure of oxygen to fraction of inspired oxygen was 300 mm Hg or less (19–21). The inclusion criteria were as follows: (a) age 18 years or older; (b) no medical history of pulmonary, autoimmune, or malignant disease; (c) COVID-19 survivors who underwent hospitalized treatment; (d) chest CT performed at admission and discharge; (e) chest CT at discharge showing residual lung lesions related to COVID-19; and (f) voluntary, written informed consent for chest CT follow-up. We excluded participants who dropped out of the study, had a subsequent positive result after a COVID-19 nucleic acid test, or acquired an additional lung infection after discharge.

Chest CT Protocol

Unenhanced chest CT examinations were performed using the same commercial CT scanner (Ingenuity Core 128, Philips Medical Systems). Images were obtained during breath holding at full inspiration. The fixed tube voltage was set to 120 kVp with adaptive current modulation, resulting in a mean volume CT dose index of 7.1 mGy ± 2.2 (standard deviation) (range, 3.1–12.2 mGy) and a dose-length product of 286.0 mGy · cm ± 97.2 (range, 99.2–506.3 mGy · cm). Other acquisition parameters were as follows: pitch, 0.999; collimation, 64 × 0.625 mm; gantry rotation time, 0.75 second; and DoseRight (Philips) index, 18. From the raw data, axial CT scans were reconstructed with a matrix size of 512 × 512 (thickness of 1.5 mm and interval of 1.5 mm) using iterative reconstruction (iDose level 5, Philips Healthcare).

Serial Chest CT Assessments

Chest CT data at admission and discharge were retrospectively collected from the institutional picture archiving and communication system (version 11.3.5.8902, Vue PACS; Carestream Health). In addition, three serial chest CT examinations were performed at 3, 7, and 12 months since initial symptom onset. Visual assessments were independently performed by three senior radiologists (B.L., L.Y., and C.Z., with 26, 23, and 27 years of clinical experience in thoracic radiology, respectively). Afterward, any disagreement was resolved by discussion and consensus. Three radiologists were blinded to the clinical progress of all participants. Abnormalities were described using the standard terms defined in the Fleischner Society glossary and peer-reviewed literature (Fig E1 [online]) (7,8,11,17,22–24). Complete resolution was defined as the disappearance of any COVID-19–related lung abnormalities at chest CT (7). The distribution of lung lesions was categorized as a subpleural, diffuse, or random distribution (8). Following the methods in previous studies of COVID-19, we used the same semiquantitative CT scoring system to estimate the involvement of the lung lesions in each lobe from 0 to 5 points (ie, 0, no lesion; 1, <5%; 2, 5%–25%; 3, 26%–49%; 4, 50%–75%; 5, >75%) (8,17,25). The sum of the CT score in each lobe was calculated as the total CT score, ranging from 0 to 25. At 12 months after symptom onset, participants with or without residual CT abnormalities were compared.

Statistical Analysis

All statistical analyses were performed using SPSS statistics software (version 26, IBM). Quantitative and counting data were presented as means ± standard deviations, with ranges and the percentages of the total, respectively. Mann-Whitney tests were performed to estimate continuous variables between different groups according to the nonnormal distribution assessed with the Shapiro-Wilk tests. We performed χ² tests to evaluate categoric variables between different groups. The Fisher exact test was performed instead of a χ² test if the expected count was less than five. Univariable and multivariable logistic regression analyses (ie, forward conditional method) were used to investigate the independent risk factors for chronic CT changes after 1 year, and odds ratios with 95% CIs were calculated. Linear mixed models of repeated measures were established to estimate the fixed effects of the time and participant grouping on the number of involved lung lobes and total CT score using type III tests. Further intertime point comparisons were performed using Bonferroni adjustments. Statistical significance was defined as two-tailed P < .001.

Characteristics of Study Sample

A total of 2171 patients discharged from the hospital were screened for our study. Of these, 209 participants were included in the final analysis, 1063 patients were excluded owing to the lack of chest CT at admission, and 285 patients were excluded because of complete resolution of CT abnormalities at discharge (Fig 1). During follow-up, 15 patients were excluded because of study dropout, and three were excluded because of secondary lung infection (Fig 1). The mean age of participants was 49 years ± 13 (range, 20–82 years), with a 1:1.25 male-to-female ratio. Twenty percent of participants (41 of 209) had comorbidities, including hypertension, type 2 diabetes mellitus, and coronary heart disease. Four of 209 participants (2%) had a history of smoking. The mean interval between symptom onset and admission was 6.9 days ± 4.4 (range, 1.0–18.0 days). Pneumonia, severe pneumonia, and ARDS were diagnosed in 107 (51%), 80 (38%), and 22 (11%) participants, respectively. All participants who developed ARDS underwent mechanical ventilation in an intensive care unit. The mean hospitalized period was 24.8 days ± 13.3 (range, 5.0–80.0 days). The chest CT examinations at admission and discharge were performed at 6.3 days ± 4.5 and 28.3 days ± 13.8 after symptom onset, respectively. The detailed information is summarized in Table 1.

Figure 1:

Download as PowerPoint

Table 1: Basic Characteristics of Study Sample

Visual CT Assessments

The complete resolution rate gradually increased over time from 61% (128 of 209 participants) at 3 months to 75% (156 of 209 participants) at 12 months after symptom onset (Fig 2A and Table E1 [online]). At admission, GGOs (201 of 209 participants [96%]), consolidation (124 of 209 participants [59%]), and “crazy paving” pattern (53 of 209 participants [25%]) were the most common findings (Fig 2B and Table E1 [online]). Thereafter, consolidation and crazy paving gradually improved over time, whereas other pulmonary abnormalities, including GGOs, parenchymal bands, reticular lesions, bronchial dilation, and volume loss, tended to increase after admission but started to resolve after discharge (Fig 2B and Table E1 [online]). However, at 3 months after symptom onset, CT abnormalities had stabilized except for GGOs and parenchymal bands, which gradually resorbed (Fig 2B and Table E1 [online]). Nevertheless, GGOs were the predominant radiologic pattern throughout (Fig 2C and Table E1 [online]). The mean total CT score and the number of involved lung lobes gradually decreased from admission (9.7 ± 7.0 and 3.7 ± 1.5, respectively) to 12 months (1.8 ± 3.9 and 0.9 ± 1.7, respectively) after symptom onset (P < .001 for both) (Fig 2D, Table E1 [online]). After 3 months, the decrease in mean total CT score and mean number of involved lung lobes became nonsignificant (P ≥ 0.33 and P ≥ .06, respectively) (Fig 2D).

Figure 2:

Download as PowerPoint

Comparisons between Participants with or without Complete Resolution at 12 Months

Participants were divided into two groups based on the CT abnormalities at 12 months after symptom onset—participants with complete resolution (156 of 209 participants [75%]) (Fig 3) and participants with residual CT abnormalities (53 of 209 participants [25%]) (Figs 4, 5). In participants with complete resolution, no recurrent pulmonary abnormalities were observed at further follow-up after complete resolution was initially achieved (Fig 3). Age, previous comorbidities, COVID-19 severities, time between admission and symptom onset, hospitalized period, lymphocyte count, lactate dehydrogenase level, albumin level, number of involved lobes, and total CT score at admission all demonstrated differences between the two groups (P ≤ .001 for each comparison) (Tables 2, 3). Age 50 years or older, lymphopenia, and severe or aggravation of ARDS were independent risk factors for residual CT abnormalities at 1 year (odds ratio = 15.9, 18.9, and 43.9, respectively; P < .001 for each comparison) (Table 4).

Figure 3:

Download as PowerPoint

Figure 4:

Download as PowerPoint

Figure 5:

Download as PowerPoint

Table 2: Comparison of Characteristics between Two Groups

Table 3: Comparison of CT Findings between Two Groups at Admission and Discharge

Table 4: Relationship of Risk Factors with Residual CT Abnormalities after 1 Year

Diffuse consolidation was more common at chest CT at admission and discharge in participants with residual CT abnormalities than in those with complete resolution (Table 3). At discharge, complicated CT abnormalities, including reticular lesions, bronchial dilation, and volume loss, were more commonly observed in participants with residual CT abnormalities than in participants with complete resolution, whereas predominant GGOs were common to both groups (Table 3). Over time, the total CT score and number of involved lung lobes gradually decreased, with statistically significant improvements at 3 months (Fig 6) but were higher in participants with residual CT abnormalities than participants with complete resolution (Table E2 [online]). Noticeably in participants with complete resolution, reticular lesions (21 of 156 participants [14%]) and bronchial dilation (18 of 156 participants [12%]) were observed at discharge but had almost completely resorbed at 3 months (Fig 6 and Fig E2 [online]). In participants with residual CT abnormalities, reticular lesions (41 of 53 participants [77%]) and bronchial dilation (39 of 53 participants [74%]) observed at discharge were incompletely resorbed and persisted in 53% (28 of 53) and 45% (24 of 53) of participants at 12 months, respectively (Figs 4–6; Table E2 [online]).

Figure 6:

Download as PowerPoint

Different Residual Abnormalities at 12 Months

On the basis of the different residual CT abnormalities at 12 months, participants could be further divided into two subgroups (Table E3 [online]): participants with residual linear opacities (25 of 53 participants [47%]) and those with multifocal reticular or cystic lesions (28 of 53 participants [53%]). Participants with residual linear opacities demonstrated gradual resolution of pulmonary lesions since discharge, leaving residual parenchymal bands or thin linear opacities (Fig 4; Figs E3–E4 [online]; Table E4 [online]). In contrast, participants with multifocal reticular or cystic lesions presented persistent subpleural reticular or cystic lesions at follow-up (Fig 5; Figs E4–E5, and Table E4 [online]). At 12 months, the total CT score was 3.9 ± 2.6 versus 10.1 ± 4.3, and the number of involved lung lobes was 2.6 ± 1.0 versus 4.5 ± 1.2 in participants with residual linear opacities and multifocal reticular or cystic lesions, respectively (P < .001 for both comparisons) (Tables E3, E4 [online]). However, the improvements in total CT scores in both subgroups were observed at 3 months after symptom onset (P < .001) (Fig E4 [online]).