The current meta-analysis accompanied the PRISMA statement (19). On February 29, 2024, PubMed, EMBASE, Web of science, and the Cochrane Library were retrieved for English language studies. The search term consists of a combination of the following medical subject heading (MeSH) and free text words: (lung carcinoma OR lung adenocarcinoma OR lung neoplasm OR lung cancer) AND (interstitial lung abnormality OR ILA OR subclinical ILD OR early ILD) AND (prognosis OR prognoses OR survival OR outcome) AND (computed tomography). Additionally, we manually retrieved the reference lists of the publications that qualified. Details of the literature search are presented in Supplementary Table 1 .

After the initial search and removal of the duplicates, screening of the title and abstract was done by 2 independent investigators (TXL and GXT). Full texts were retrieved for studies that satisfied the inclusion criteria. Any disagreements were resolved through discussion. Bibliography sections of the included studies were further searched for additional relevant papers.

Inclusion Criteria

Utilizing the Population-Intervention-Control-Outcome-Study (PICOS) framework, the inclusion criteria for the current study were developed as follows: patient population consisting of individuals diagnosed with lung carcinoma through pathological or histological means; Exposure-intervention involving a computed tomography (CT) scan of the lungs identifying ILA, with a positive diagnosis by a radiologist; Control group comprised of individuals whose lung CT scans showed no signs of ILA or interstitial lung disease; Desired outcomes consisted of English-language studies examining the relationship between ILA and prognosis in lung cancer patients, including OS, DFS, PFS, and CSS, as well as hazard ratio (HR) and 95% confidence interval (CI) related to patient survival; And study design encompassing both retrospective and prospective English-language publications. Classification of patients into ILA group and non-ILA group for the intervention and control groups was based on the definition of ILA identified in chest CT scans.

Exclusion Criteria

Conference abstracts, case reports, or comments were excluded. Studies that compared only the clinicopathologic features and not the prognostic value of ILA were excluded.

The main focus of data extraction was on the author, year of publication, region of study, design of the study, period of study, size of the sample, ILA incidence, types of cancer, treatment methods, duration of follow-up, and study outcomes. The quality of observational studies was assessed using the Newcastle-Ottawa Scale (NOS) score (20). Literature was considered high-quality if it scored above six. Two authors (GXT and TXL) independently double-checked all of the aforementioned steps, and any discrepancies were resolved by a third author (YSZ).

Stata 16.0 was utilized for the statistical analysis. OS is determined as the period from the initiation of therapy until decease, irrespective of the reason. A fatality linked to lung cancer or a respiratory ailment is counted as a cause-specific decease, and the period from the initiation of therapy until a cause-specific decease is designated as CSS. Within this meta-analysis, DFS and RFS in the included studies were jointly denoted as DFS, which is characterized as the duration from surgery until the reappearance of the disease or decease from any cause (event). The combined HR and 95%CI were utilized to assess the connection between ILA and prognosis in lung cancer patients. Multivariate HR and 95% CI data were directly obtained from the study. In cases where multivariate HR was not provided, univariate HR data was extracted. An HR greater than 1 (with a 95%CI not overlapping 1) suggested that ILA was linked to a poorer prognosis in patients, rather than the opposite being true. In terms of examining the relationship between ILA and clinical and pathological characteristics of lung cancer patients, the combined odds ratio (OR) and 95%CI were employed. An OR greater than or less than 1 (with a 95%CI not overlapping 1) indicated a correlation between ILA and clinicopathological features in lung cancer patients, and not the contrary. The chisquared test was employed to calculate the statistical heterogeneity. A random effect model was applied when P<0.1 and I²>50% indicated high heterogeneity; otherwise, the fixed effect model was utilized. Egger’s tests were conducted to assess publication bias, and in cases of significant bias, the trim-and-fill method was applied to adjust the results. Sensitivity analysis was carried out to evaluate result stability, with each study being excluded independently.

After the initial search, 57 duplicate studies were removed. Then there were 285 articles deleted after carefully reading the titles and abstracts. Later, the full texts of the remaining 45 articles were further assessed.

12 articles (7–18) involving 4416 patients were ultimately included. The PRISMA flow diagram is provided in Figure 1 . The main characteristics of the studies included are shown in Table 1 . All the 12 studies were retrospective. Four studies were from Korea and Japan, respectively, three studies were carried out in USA, the remaining one in China. Ten studies involved the treatment of NSCLC, one studies included small-cell lung cancer (SCLC), and one study had both NSCLC and SCLC arms. The sample sizes varied from 73 to 1,524. The incidence rate of ILA ranged from 3.93% to 37.87%, with an average incidence of 14.77%. The NOS scores for the 12 articles ranged from 6 to 8, which represented a low risk of bias ( Table 1 ).

In total, 11 articles (7–11, 13–18) involving 4188 patients explored the association between ILA and OS in lung cancer patients. The pooled HR was (n=11; HR=2.22; 95% CI=1.68-2.95; P<0.001; I² = 72.0%; Ph<0.001), implying that the lung cancer patients with ILA raised death risk by 122% compared to those without ( Figure 2 ).

Since there was significant heterogeneity, a random effects model was used (I² = 73.9%, P<0.001). We then conducted subgroup analyses based on study region, sample size, cancer types, treatment and analysis. The results were consistent with the above findings ( Table 2 , Supplementary Figures S1–5 ).

The relationship between ILA and PFS was examined using prognostic data from 3 studies (16–18) involving 355 participants.A combined analysis demonstrated that patients with ILA was significantly correlated with shortened PFS (HR=1.59; 95% CI=1.08-2.32; P=0.017) than those without, with no significant heterogeneity identified between studies (I² = 0%; P=0.772) ( Figure 3 ).

There were 2 studies (7, 9) with a total of 1906 patients investigating the correlation between ILA and DFS. Significant heterogeneity was observed in the included studies (I² = 90.4%, P = 0.001), so a random effects model was used. Interestingly, the results indicated that the combined ILA patients had no significant association with DFS (HR: 2.07, 95% CI: 0.94–7.02, P = 0.066) ( Figure 4 ).

The association between ILA and CSS in cancer patients was explored in two articles (12, 17) with 341 individuals. We found significant correlation between ILA and CSS in lung cancer patients (HR=4.00, 95% CI: 1.94-8.25, P<0.001) using a fixed effect model (I² = 0%, P= 0.594) ( Figure 5 ).

Of the articles included for analysis in the present study, a total of 8 studies, involving 2330 patients with lung cancer, reported a relationship between the ILA and clinicopathological factors of lung cancer. As shown by the combined results in Table 3 , patients with ILA were significantly associated with male (OR= 2.43; 95% CI= 1.48-3.98; P<0.001), smoking history (OR= 2.11; 95% CI=1.37-3.25; P<0.001), advanced age (OR=2.50; 95% CI= 1.56-4.03; P<0.001), squamous carcinoma (OR=0.42; 95% CI=0.24-0.71; P=0.01), and EGFR mutation (OR=0.50; 95% CI=0.32-0.78; P=0.002). There was no significant correlation, however, between the ILA patients and Race (OR=1.92; 95% CI= 0.51-7.18; P= 0.334), Stage (OR= 0.93; 95% CI= 0.65-1.34; P= 0.708), ALK (OR= 0.30; 95% CI= 0.11-1.38; P= 0.144) ( Supplementary Figures S6–13 ).

In the present study, we performed a sensitivity analysis on the relationship between the ILA and OS ( Figure 6 ) and PFS ( Figure 7 ), through which we determined that the significance of ILA in predicting OS and PFS in patients with lung cancer did not change after eliminating any single article.

Funnel plot and Egger’s test evaluated the inclusion for publication bias. For the Meta-analysis of the relationship between ILA and OS, the meta-analysis was biased (P = 0.04) ( Figure 8 ); Next, the trim and fill method was utilized to calculate the number of missing studies in OS. By factoring in the missing hypothesis studies, the combined HR of OS was recalculated but was not substantially different. As a result, the publication bias had little impact, and the outcome was stable ( Figure 9 ).