Risk Factors and Clinical Characteristics of Lung Cancer in Idiopathic Pulmonary Fibrosis

A Retrospective Cohort Study

Hongseok Yoo; Byeong-Ho Jeong; Myung Jin Chung; Kyung Soo Lee; O. Jung Kwon; Man Pyo Chung


BMC Pulm Med. 2019;19(149) 

In This Article


We conducted a cohort study based on a prospective registry of patients with IPF at Samsung Medical Center (a 1961-bed, university-affiliated, tertiary referral hospital in Seoul, Republic of Korea). In our hospital, all consecutive patients diagnosed with interstitial lung disease in the interstitial lung disease clinic are prospectively registered in an interstitial lung disease database since January, 1998. From the database, we identified 1360 patients who were diagnosed with IPF based on the diagnostic criteria of the American Thoracic Society and European Respiratory Society[1] between January 1998 and April 2013. Records of these patients were reviewed for possible inclusion in analysis. Excluded were 235 patients who were followed at our hospital for less than 6 months and 10 who did not have sufficient data for analysis. In addition, 172 patients diagnosed with lung cancer concurrently or within 6 months of IPF diagnosis and 5 patients transferred to our hospital with treated lung cancer were excluded. Finally, 938 patients without lung cancer at the time of IPF diagnosis and with fully available data were considered eligible for analysis. The Institutional Review Board of Samsung Medical Center approved the collection, analysis, and publication of the data and informed consent was waived due to the retrospective nature of the study.

The following clinical data were obtained from the medical records: age, gender, comorbidities, symptoms at the time of lung cancer diagnosis, use of corticosteroid or azathioprine for IPF treatment, results of pulmonary function tests at initial and 1 year after IPF diagnosis and at the time of lung cancer diagnosis, radiologic findings at the time of IPF and lung cancer diagnosis, histologic type and stage of lung cancer, and mortality. Chest radiography and chest computed tomography (CT) images were thoroughly reviewed by two authors (MJC and KSL). CT fibrosis score was defined as percentage of lung affected by fibrosis that included reticulation/honeycombing and CT emphysema score was defined as percentage of lung affected by emphysema.[17,18] Percentages were rounded to the nearest 5%. Rapid decline of forced vital capacity (FVC) and diffusing capacity for carbon monoxide (DLco) were defined as annual decline of FVC of 10% or more and DLco of 15% or more, respectively.[19,20]

Categorical variables are reported as numbers (percentages). Continuous variables with normal distribution are reported as mean with standard deviation while variables with nonnormal distribution are reported as median with interquartile ranges (IQR, 25th to 75th percentiles). Categorical variables were compared using chi-square test and continuous variables using either independent t-test or Mann Whitney U test according to normalness of distribution. The annual decline of FVC and DLco was calculated using the results of pulmonary function tests at the time of IPF diagnosis and at 1 year after the IPF diagnosis. Regarding the patients in whom lung cancer developed within 1 year of IPF diagnosis, the last pulmonary function test undertaken at the time without the evidence of lung cancer was adopted for estimation of annual decline. We estimated the cumulative incidence of lung cancer with the Kaplan-Meier method. Cox proportional hazard model using backward stepwise selection method was used to identify independent predictive factors for lung cancer development with careful selection of variables after univariate regression analysis. The cumulative incidence of lung cancer according to the predictive factors identified by Cox proportional hazard model was estimated by Kaplan-Meier method. The statistical significance was determined by log-rank test. Kaplan-Meier estimation was used to determine the survival curves for patients with and without lung cancer, which were then compared using the log-rank test.

All tests were two-sided and a P value < 0.05 was considered significant. Data were analyzed using IBM SPSS Statistics 20 (IBM, Chicago, IL, USA).