Cumulative Incidences of Lung Cancer in Various Interstitial Lung Diseases

Takafumi Suzuki Tokyo Medical and Dental University Hiroyuki Sakashita Yokosuka Kyosai Hospital Masako Akiyama Tokyo Medical and Dental University Takayuki Honda Tokyo Medical and Dental University Masaru Ejima Tokyo Medical and Dental University Masahiro Ishizuka Tokyo Medical and Dental University Tsukasa Okamoto Tokyo Medical and Dental University Yasunari Miyazaki (  miyazaki.pilm@tmd.ac.jp ) Tokyo Medical and Dental University

Interstitial lung disease (ILD) patients often develop lung cancer. However, previous studies on the incidences of lung cancer in ILD patients focused on speci c aetiologies, such as idiopathic pulmonary brosis (IPF). The lung cancer incidences in these patients have not been investigated, and thus, we aimed to evaluate them here.
Methods ILD patients at our hospital were retrospectively reviewed. The cumulative incidences of lung cancer in patients with various ILDs were estimated with Kaplan-Meier curves and compared between ILD groups using log-rank tests. The association between several variables at initial diagnosis and lung cancer development was assessed with Cox proportional hazards regression analysis to identify predictors.

Conclusions
The lung cancer incidence is higher in IPF patients than in non-IPF idiopathic interstitial pneumonia patients and is equally high in patients with chronic hypersensitivity pneumonitis and IPF.

Background
Interstitial lung diseases (ILDs) are a group of diffuse parenchymal lung disorders with various forms.
ILDs are mainly divided into two categories: ILDs with and without any known causes. The former include hypersensitivity pneumonitis (HP), connective tissue disease-related ILDs (CTD-ILDs), ILDs with antineutrophil cytoplasmic antibody-associated vasculitis, and drug-induced pneumonia. The latter are called idiopathic interstitial pneumonias (IIPs), and the main subtype is idiopathic pulmonary brosis (IPF) [1].
In terms of the clinical course, ILDs can be acute or chronic [2]. The former include acute HP, acute eosinophilic pneumonia (EP), and acute or subacute IIPs, which consist of acute interstitial pneumonia, cryptogenic organizing pneumonia, nonspeci c interstitial pneumonia (NSIP), and acute exacerbation of IPF or NSIP [3]. The latter are IPF, chronic HP, CTD-ILDs, chronic EP, pulmonary sarcoidosis, asbestosis, radiation pneumonitis, pulmonary alveolar proteinosis.
There are a number of common comorbidities of IPF, a speci c form of chronic, progressive brosing ILD, including lung cancer (LC), pulmonary hypertension, chronic obstructive pulmonary disease (COPD)/emphysema, pulmonary embolism, and pulmonary infections [4][5][6][7]. Previous studies reported that patients with IPF are at high risk for developing LC [8,9]. IPF patients with LC had a worse prognosis than those without LC [10][11][12]. Choi et al. reported that the incidence of LC in IPF patients was higher than those in IIP patients and COPD patients [13]. High incidences of LC in patients with CTD-ILDs [14,15] and IIPs [13] have also been reported. The prevalence of LC in chronic HP patients has been reported, but the incidence has not yet [16].
The incidence of LC in patients with ILDs other than IPF, IIPs, and CTD-ILDs has not yet been revealed.
The present study was performed to evaluate the cumulative incidences of LC in patients with ILDs with various aetiologies.

Patient selection
This study was a retrospective single-centre cohort study involving patients with ILDs. In this study, 672 ILD patients who were diagnosed between January 2008 and December 2012 at Tokyo Medical and Dental University Hospital and who visited their physician regularly were recruited. The exclusion criteria were as follows: (i) patients who had LC at the time of the initial ILD diagnosis, (ii) patients with acute ILDs, and (iii) patients whose high-resolution computed tomography (HRCT) images were not available in the medical records (Fig. 1). Patients with acute ILDs were excluded from this study because the activity of acute ILDs is temporary and does not cause chronic in ammation and brosis. Patients enrolled in this study were observed for up to ve years because most chronic ILD patients, have a poor prognosis in general. This study was approved by the Institutional Review Board of Tokyo Medical and Dental University (approval number: M2019-087). Because of the retrospective nature of the study, written consent from the use of records was waived.

De nition of each ILD
The diagnosis of IPF was based on the clinical practice guidelines from the American Thoracic Society (ATS), European Respiratory Society (ERS), Japanese Respiratory Society (JRS), and Latin American Thoracic Association (ALAT) [17]. The diagnosis of IIPs was based on the ATS/ERS statement about the classi cation of IIPs [5]. IIPs other than IPF were de ned as non-IPF IIPs in this study. The diagnosis of chronic HP in this study was based on the criteria as previously described [18]. Connective tissue disease (CTD) comprises a group of chronic and systemic autoimmune disorders, such as rheumatoid arthritis, systemic sclerosis, polymyositis or dermatomyositis, Sjögren syndrome, and systemic lupus erythematosus [14]. Patients with CTD were diagnosed by rheumatologists and/or dermatologists in our hospital or ful lled the criteria for a speci c CTD [19][20][21][22][23][24]. The diagnosis of sarcoidosis was established based on the guidelines of the ATS/ERS/World Association for sarcoidosis and other granulomatous disorders [25]. Pulmonary sarcoidosis patients with radiographic stage II or higher were included [26].

Study design
The diagnosis of LC was based on histological and/or cytological ndings of carcinoma. Metastatic disease of the lung was excluded. The eighth edition of the TNM classi cation system for LC was used in this study [27]. Sex, age, pack-years, and radiological usual interstitial pneumonia (UIP) pattern at the time of the initial ILD diagnosis were recorded. The forced vital capacity (FVC), forced expiratory volume in one second (FEV 1 )/FVC values, and serum levels of Krebs von den Lungen-6 (KL-6) and surfactant protein D (SP-D) were recorded for the ILD patients within six months of the initial ILD diagnosis. The radiological UIP pattern was based on HRCT scanning patterns indicated in the clinical practice guidelines published by the ATS, ERS, JRS, and ALAT [17]. The cumulative incidences of LC in ILD patients, and the predictive factors were estimated. ILD patients who developed LC were reviewed for radiological characteristics as well as for LC stage and treatment at the time of LC diagnosis. The diagnosis criteria for emphysema was obtained from previous reports [9]. HRCT at the time of LC diagnosis was retrospectively interpreted by two pulmonary specialists (T.S., M.E.).

Statistical analysis
Data are represented as numbers or medians. Statistical analyses were performed using GraphPad Prism 8 software (GraphPad, San Diego, CA, USA) and EZR (Saitama Medical Centre, Jichi Medical University, Saitama, Japan) [28]. Data are expressed as medians and ranges. One-way analysis of variance or the Kruskal-Wallis test was used as appropriate to compare the values between different groups. When signi cant differences between groups were observed, intergroup comparisons were assessed using Tukey's multiple comparisons test or Dunn's multiple comparisons test, as appropriate. The cumulative incidences of LC in ILD patients were evaluated with Kaplan-Meier curves and compared between groups with log-rank tests. Multiple adjustment was not performed. Cox proportional hazards regression analysis was used to identify the signi cant variables predicting the development of LC. Nine variables were examined for their association with the development of LC in patients with ILDs, namely, sex, age, packyears, radiological UIP pattern on HRCT, FVC value, FEV 1 /FVC value, serum KL-6 level, serum SP-D level, and aetiology. In univariate Cox proportional regression analysis, all available data were used for every variable. All reported p values were two-sided, and a value < 0.05 was considered to be statistically signi cant.

Clinical characteristics of patients with ILDs
Of the 672 ILD patients, 66 were excluded: 55 patients had LC at the time of the initial diagnosis of an ILD, 8 patients had acute ILDs, and 3 patients did not have HRCT images available (Fig. 1). The 606 patients with ILDs were divided into 6 groups based on aetiology: 161 with IPF, 133 with non-IPF IIPs, 160 with chronic HP, 87 with CTD-ILDs, 19 with pulmonary sarcoidosis, and 46 with other ILDs. Among the 46 patients with other ILDs, 15 had ILDs with anti-neutrophil cytoplasmic antibody-associated vasculitis, 9 had suspected or diagnosed drug-induced ILDs, 7 had suspected or diagnosed asbestosis, 4 had suspected or diagnosed radiation pneumonitis, 3 had chronic EP, 3 had suspected but undiagnosed pulmonary sarcoidosis, 2 had pulmonary alveolar proteinosis, 2 had suspected but undiagnosed CTD-ILDs, and 1 had Castleman's disease.
The characteristics of the ILD patients are summarized in Table 1. The group of patients with CTD-ILDs had a higher proportion of female patients than IPF, non-IPF IIPs, chronic HP, and other ILDs. The group with non-IPF IIPs had a higher proportion of female patients than IPF. No differences were seen in age at the time of initial ILD diagnosis among the groups. The number of pack-years at the time of initial ILD diagnosis was higher in the group with IPF than non-IPF IIPs, chronic HP, CTD-ILDs, and pulmonary sarcoidosis. The IPF group had a higher proportion of patients with radiological UIP pattern at the time of initial diagnosis than any other groups, and the chronic HP group had the second highest proportion among the groups. The IPF, chronic HP, and CTD-ILD groups had lower FVC values than the non-IPF IIP and pulmonary sarcoidosis groups. The groups of chronic HP and CTD-ILD groups had lower FVC values than the group of other ILDs. The IPF and non-IPF IIP groups had lower FVC values than chronic HP groups. The other ILD group had a lower FEV 1 /FVC value than the IPF, non-IPF IIP, and CTD-ILD group.
The chronic HP group had a higher serum KL-6 and SP-D levels than any other groups.

Characteristics of ILD patients who developed LC
The characteristics of the ILD patients who developed LC are summarized in Table 2. The median observation period of ILD patients was 45 (range, 4-60) months. Of the 606 ILD patients, 28 developed LC during the observation period. The median interval between diagnosis and the development of LC was 32 (range, 4-60) months. Among the 28 ILD patients who developed LC, 12 had IPF, 10 had chronic HP, 3 had non-IPF IIPs, 2 had CTD-ILDs, and 1 had pulmonary sarcoidosis. Of the 28 ILD patients with LC development, 25 (89%) were male smokers. 24 patients (86%) had radiological UIP pattern at the diagnosis of ILDs, and 17 (71%) had LC lesion adjacent to honeycombing. Squamous cell carcinoma was the most frequent cell histological type (n = 13 [46%]), followed by adenocarcinoma (n = 17 [25%]). Of the 14 patients performed surgery on, all 11 excluding 3, whose pathological ndings were not available, had pathological UIP pattern.

Cumulative incidences of LC in ILD patients
The cumulative incidences of LC at 1, 3, and 5 years were 1.9, 5.7, and 12.3% in the IPF group; 0.8, 0.8, and 4.0% in the non-IPF IIPs group; 2.0, 4.6, and 11.0% in the chronic HP group; and 1.1, 1.1, and 2.9% in the CTD-ILD group (Fig. 2). As a result of comparing the four groups together, no signi cant difference was found in the cumulative incidence of LC among the IPF, non-IPF IIP, chronic HP, and CTD-ILD groups (p = 0.074). The incidence densities of LC in the IPF, non-IPF IIPs, chronic HP, and CTD-ILD groups were 2.36, 0.70, 1.98, and 0.64 per 100 person-years, respectively. When each two groups were compared, IPF patients had a higher incidence of LC than non-IPF IIP patients (p = 0.036) (Fig. 2). Cox proportional regression hazards models showed that IPF was also signi cantly associated with the development of LC compared to non-IPF IIPs (hazard ratios 6.042, 95% con dence interval [CI] 1.284-28.440, p = 0.023).

Factors predictive of LC development in ILD patients
To determine the factors in ILD patients that were predictive of LC development at the time of the initial ILD diagnosis, nine factors were rst assessed with univariate Cox proportional regression hazards models. Male sex, pack-years, radiological UIP pattern, and FVC value were signi cantly associated with the development of LC (Table 3). Whereas, age, and serum KL-6 and SP-D levels were not associated with LC development. Multivariate Cox proportional regression hazards models, which included the factors that were signi cant in the univariate analysis, showed that a radiological UIP pattern, FVC value, and pack-years were independent predictive factors for the development of LC (Table 4). Focusing on the IPF and chronic HP groups, univariate Cox proportional regression hazards models showed that a FVC value and pack-years were signi cantly associated with the development of LC in IPF patients and in chronic HP patients, respectively (Tables 5 and 6).

Discussion
There were three main ndings in the present study. First, the incidence of LC in IPF patients is higher than that in non-IPF IIP patients. The incidence densities of LC in IPF patients ranged from 0.81 to 4.71 per 100 person-years in previous reports [8 -13, 29-35]. The incidence rate in this study was comparable with those in previous reports on IPF patients. The incidence densities of LC in patients with CTD-ILDs were also reported in two previous studies as 0.98 and 1.66 per 100 person-years [14,15]. Thus, the incidence density of LC in CTD-ILD patients in this study was slightly lower than the values in previous reports [14,15]. Moreover, Choi et al. reported that the incidence densities of LC in IPF and IIP patients were 3.81 and 1.57 per 100 person-years, respectively, and that the incidence of LC in IPF patients was higher than that in IIP patients based on data from the Korean national database [13]. Although the IIP patient group included the IPF patient group in the previous report, our result was consistent with theirs. Our study revealed that the incidence of LC is higher in IPF patients than in patients with IIPs other than IPF when diagnosed according to the strict criteria for each disease. Our results revealed that the number of pack-years was higher in IPF patients than in non-IPF IIP, chronic HP, CTD-ILD, and pulmonary sarcoidosis patients. Thus, a hypothesis that a history of smoking simply affects LC development in IPF patients might be considered. However, the median incidence of LC in IPF patients from 11 cohort studies was 2.07 per 100 person-years (95% CI, 1.46-2.67), which is obviously higher than the incidence rates of 0.2-0.7 per 100 person-years reported in LC screening trials in heavy smokers without ILD [36]. This suggests that IPF increases the risk of LC development beyond what would be expected in older populations and in those with smoking history [36]. This suggest that IPF itself may participate in LC development more than non-IPF IIPs do. Second, we determined the incidence of LC in chronic HP patients. A previous study on the association between chronic HP and LC revealed that the prevalence of LC in chronic HP patients was 10.3% [16]. In our study, the incidence density of LC in the chronic HP group (1.98 per person-years) was the second highest among groups and was not signi cantly different from the incidence density in the IPF group (2.36 per person-years). Since previous studies reported that the incidence densities of LC in IPF patients were 0.81 to 4.71 per 100 person-years [8, 9,[11][12][13][29][30][31][32][33][34][35], the incidence density in chronic HP patients in our study was approximately equal to that in IPF patients in previous reports.
Third, in a multivariate proportional hazards model, a radiological UIP pattern, FVC value, and pack-years were predictive factors for the development of LC in ILD patients. K.J. Lee et al. revealed that most LCs were located in the brosis of IPF lesions, and suggested that brotic lesions of IPF may be involved in lung carcinogenesis [10]. In our study, all 11 patients who underwent surgery for LC except 3 lacking pathological data had pathological UIP pattern lesions. Therefore, that a radiological UIP pattern was one of predictive factors of LC development in ILD patients may indicate that radiological UIP lesions affect the development of LC. In the present study, a FVC value in ILD patients was also associated with LC . Therefore, in ILD patients, a higher FVC value, linked to emphysema, might induce LC development. This helps to explain the result of univariate Cox proportional regression analysis in IPF patients. In our study, pack-years was identi ed as a predictive factor for LC in ILD patients, which agreed with the ndings in previous reports revealing an association between IPF and LC development [9,10]. This might elucidate the relation between pack-years and LC development in chronic HP patients in univariate Cox regression analysis.
This study had some limitations. First, this was a retrospective, single-institution cohort study with a limited sample size. Second, we specialize in HP, and patients with HP visit our hospital from all over the country. The proportion of HP patients among those with ILDs seems to be higher than that at other institutions. Future prospective and multicentre studies with including a larger series of ILD patients are needed.

Conclusions
The incidence of LC is higher in IPF patients than in non-IPF IIP patients and is comparatively high in patients with chronic HP and those with IPF. Moreover, a radiological UIP pattern, FVC value, and pack-years might be associated with the development of LC in ILD patients.

Consent for publication
Not applicable.

Availability of data and materials
The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Competing interests
The authors declare that they have no competing interests.

Funding
This research did not receive any speci c grant from funding agencies in the public, commercial, or notfor-pro t sectors.
Authors' contribution YM takes full responsibility for the content of this manuscript, including data and analysis. TS, HS, and TO contributed to the study design, analysis and interpretation of the data, and the writing of this manuscript. MA, TH contributed to data analysis and interpretation. ME and MI contributed to data analysis. All authors critically revised the manuscript for intellectual content, approved the nal draft, and agreed to be accountable for all aspects of the work.  Data were presented as numbers or medians (percentages or ranges).