Association between visual emphysema and lung nodules on low-dose CT scan in a Chinese Lung Cancer Screening Program (Nelcin-B3)

The here presented study used data from participants that were included in the Nelcin-B3 study. In the Nelcin-B3 study, a general Chinese population was recruited without inclusion criteria on smoking and pack-years to identify risk factors for the “Big 3” diseases (lung cancer, cardiovascular disease, and COPD), and reference values based on LDCT [16]. The study was approved by the Ethics Committee of Biomedicine Research of Second Military Medical University (registration number: NCT03992833). At the Radiology Department of Tianjin Medical University Cancer Institute and Hospital (TJMUCIH), 4000 participants from the general population, being Tianjin residents for 3 years or longer 3 years, aged between 40 and 74 years, without any history of cancer, were invited for LDCT lung cancer screening. The current analysis focused on a consecutive series of the participants included in the Nelcin-B3 study that underwent LDCT between May and October 2017 (see Fig. 1). Participants were excluded if they had incomplete data or pneumothorax. All participants signed the informed consent, and a structured face-to-face interview was conducted by trained interviewers to gather demographic information (age, sex, ethnicity, smoking status, pack-years, passive smoking, BMI). Passive smoking was defined as an indoor environment where participants inhale smoke produced by others ≥ 1 day a week for ≥ 15 min. About the pack-years, zero was scored for never smoker.

The CT chest examinations were performed with a 64-detector row CT system (Somaton Definition AS 64, Siemens) using a low-dose technique without the use of a contrast agent. Details of the technical parameters used in the CT scan protocol were as follows: 120 kVp, 35 mAs, pitch of 1.0, reconstruction kernels D45F (emphysema assessment), and B80F (lung nodules assessment) were applied to reconstruct images at 1.0/0.7 mm thickness and increment. All the participants were scanned in the supine position with head forward. The CT scans were obtained at deep inspiration with the breath-holding of subjects under the above technical setting.

After prospective collection of LDCT scans, the detection and features of suspected lung nodules were retrospectively assessed by one junior chest radiologist (Y.F.M. with 4 years of experience) and checked by another senior chest radiologist (D.M. with 10 years of experience). The maximum intensity projection (MIP) technique was used to detect lung nodules under the D45F kernel with 10-mm slice thickness. Three-dimensional segmentation software (MM Oncology Syngo.via., version VB30A, Siemens) was used to automatically measure the volume of each detected nodule under the B80F kernel. All CT scans used for lung nodule detection were read at both lung window (window center: −500 HU, window width: 1200 HU) and mediastinal window (window center: 35 HU, window width: 320 HU). Nodules were scored and classified as solid, part-solid, and non-solid nodules (pure ground-glass nodules). Nodule location was classified as upper lobe (right middle, left, or right upper lobe) or lower lobe (left or right lower lobe). Included in the here presented analysis were all non-calcified nodules with volume ≥ 30 mm³ [16‐18]. LDCT scan findings were assessed using Lung-RADS category (1, 2, 3, and 4) [19].

The visual emphysema assessment was performed using the Minimum Intensity Projection (Minip) technique with 10-mm thickness in the lung window setting (window center: −850 HU, window width: 400 HU) [20] and Multiplanar Reconstruction (MPR) technique with 1-mm thickness in window setting (window center: −750 HU, window width: 700 HU) [21] all based on the D45F reconstruction kernel CT images with the same software used for nodule assessment. All images were visually assessed by one radiologist (X.F.Y. with 5 years of experience) using a standard protocol based on validated criteria created by the Fleischner Society [22]. Interobserver agreement was determined based on 100 randomly selected cases assessed by a second radiologist (Z.H.Y. with 2 years of experience).

Emphysema was scored as well-defined or ill-defined low attenuation or lucencies following the Fleischner criteria [22]. If present (at least trace), emphysema was further categorized as one of the three predominant subtypes of emphysema (centrilobular[CLE], paraseptal [PSE], and panlobular). The severity of CLE was categorized into trace (< 0.5%), mild (0.5–5%), moderate (> 5%), confluent and advanced destructive (confluent-ADE) according to the morphology and most severe percentage of lucency in a lung zone.

Kappa statistics for the presence of emphysema and weighted kappa coefficients for CLE and PSE severity were calculated to assess interobserver agreement. Participants were classified as having no nodule versus having at least one nodule, and negative Lung-RADS (1 or 2) versus positive (3 or 4). Baseline characteristics of participants were described, overall and stratified by the presence or absence of lung nodules. To estimate the association between the baseline characteristics and the presence of lung nodules, univariate logistic regression analysis was performed to estimate odds ratios (ORs) and related 95% confidence intervals (95% CIs). To adjust for covariates, the following characteristics were selected: age, sex, smoking status, pack-years, passive smoking, and BMI. Those were based on previous literature [23, 24]. Multivariable logistic regression analysis was performed with “enter approach” to estimate aORs and related 95% CIs. In addition, an interaction term for smoking status and emphysema was included in the model. Given that analysis was significant, a stratified analysis was performed for smokers (current smokers) and non-smokers (never or former smokers). The predictors of interest were as follows: the presence of emphysema, the subtypes of emphysema, and the severity of CLE. The severity of emphysema was entered as an ordinal variable to investigate the effect in each category and as a numerical variable to test for linear trend. Mann-Whitney U testing and chi-square testing were conducted to analyze the association between emphysema and size, number, classification, and location of lung nodules. The nodule size, classification, and location were determined according to the largest nodule, and the highest Lung-RADS category was taken into account when more than one nodule was present. The number of nodules was categorized into three categories: 1 nodule, 2–3 nodules, 4 and more nodules. Statistical analysis was conducted using the SPSS 23.0 (IBM Corporation). p < 0.05 was considered a statistically significant difference.

In total, 1162 participants were included in this analysis (see Fig. 1). The mean age of the participants was 61.2 ± 6.9 years, 517 (44.5%) were male, and 781 (67.2%) were never smokers (see Table 1). Of the 1162 participants, 424 (36.5%) had lung nodules and 674 (58.0%) participants had emphysema. Concerning the predominant subtypes of emphysema, the most frequent emphysema subtype was centrilobular (93%) including 58 (9.3% ) moderate, confluent, or ADE severity, followed by PSE (7.0%). There were no participants with panlobular emphysema.

Agreement between radiologist for assessment of presence of emphysema was good (κ = 0.76 (95% CI: 0.63–0.89). Similarly, agreement was good for severity of CLE (κ_weighted = 0.77, 95% CI: 0.67–0.88) and PSE (κ_weighted=0.77, 95% CI: 0.58–0.96).

Participants with lung nodules were slightly older (mean 61.8 vs 60.8 years, OR: 1.02, 95% CI: 1.00–1.04), and were more frequently male (50.5% vs 49.5%, OR: 1.46, 95% CI: 1.15–1.86), compared to participants without lung nodules (see Table 1). Furthermore, participants with lung nodules had a higher prevalence of visual emphysema (66.7% vs 53.0%, OR: 1.78, 95% CI: 1.39–2.28). There were no significant differences regarding ethnicity, smoking status, pack-years, BMI, and passive smoking between participants with and without lung nodules.

Multivariable analysis showed that participants with emphysema based on visual assessment increased the risk of lung nodules by 71% (aOR: 1.71, 95% CI: 1.26–2.31; see Table 2) compared to participants without emphysema. Stratified results showed that the higher risk in participants with emphysema was more pronounced in smokers (aOR: 2.28, 95% CI: 1.17–4.43), but was also seen in non-smokers (aOR: 1.55, 95% CI: 1.16–2.06). PSE seemed to be associated with a somewhat higher risk of lung nodules than CLE (aOR: 2.43, 95% CI: 1.32–4.50, and aOR: 1.60, 95% CI: 1.23–2.09, respectively), compared with participants without any emphysema (see Table 3). With respect to emphysema severity, the aORs for lung nodules gradually increased (aOR range: 1.44–2.61, overall p < 0.01, see Table 4) with the severity of CLE. The aOR for lung nodule was still significantly different when emphysema severity was considered a continuous variable (aOR: 1.40, 95% CI: 1.19–1.64, p for trend < 0.001).

The multivariable analysis also showed that presence of emphysema in a participant increased the risk of positive Lung-RADS category by 70% (aOR: 1.70, 95% CI: 1.09–2.66; see Table S1) compared to a participant without emphysema. When stratified by subtype of emphysema, CLE was associated with positive Lung-RADS category (aOR: 1.69, 95% CI: 1.08–2.66), whereas this was not shown for PSE (aOR: 1.83, 95% CI: 0.71–4.72).

Of the 424 participants with lung nodules, 284 (67.0%, 284 of 424) had a single nodule. There was a difference in the total number of nodules between those with and without emphysema (overall p < 0.01, see Table 5). Participants with emphysema were more likely to have 2–3 nodules versus those without (31.8% vs 17.0%, respectively) (p < 0.01). Participants with emphysema had larger nodules (median 74 vs 62 mm³, p = 0.03) when compared to participants without. Most participants with nodules had solid nodules (90.8%, 385 of 424), followed by non-solid nodules (6.6%, 28 of 424) and part-solid nodules (2.6%, 11 of 424). This remained true when comparing participants with and without emphysema. In 223 (52.6%, 223 of 424) participants, nodules were located in the upper lobe. No significant difference was found when comparing participants with and without emphysema.