Virtual Library

Start Your Search

You-lin 13910410711 Qiao



Author of

  • +

    JCSE01 - Perspectives for Lung Cancer Early Detection (ID 779)

    • Event: WCLC 2018
    • Type: Joint IASLC/CSCO/CAALC Session
    • Track: Screening and Early Detection
    • Presentations: 1
    • +

      JCSE01.04 - Risk Modeling for the Early Detection of Tin Miner Lung Cancer in China (ID 11397)

      08:40 - 09:00  |  Presenting Author(s): You-lin 13910410711 Qiao

      • Abstract

      Abstract

      Result of National Lung Cancer Screening Program has demonstrated a 20% reduced lung cancer mortality with low-dose computed tomography among current or former smokers with a smoking history of 30 or more pack-years[1]. Selecting high risk population for LDCT screening is a key issue for lung cancer screening. Many studies have suggested that lung cancer risk model which incorporating these factors can be more accurate to identify high risk individuals suitable for LDCT screening than the NLST criteria[2]. Thus, more precise evaluation of association between these factors are warranted to developing lung cancer risk models.

      In this study, we developed and internal validated a lung cancer risk model with data of a occupational screening cohort in Yunnan, China with the aim to exploring potential improvement of yield of lung cancer screening with Chest X-ray and Sputum cytology due to the improved risk stratification. This study was a prospective occupational cohort study among tin miners from Yunnan Tin Corporation (YTC) initiated in 1992[3]. Participants were tin miners aged 40 or over and had at least 10 years of underground work or smelting history. Participants also had received annual lung screening from 1992 to 1999, and were annually followed up through December 31, 2001. Previous studies suggested that age, education level, smoking, occupational radon and arsenic exposure, prior chronic bronchitis were risk factors of lung cancer risk in this cohort[4-6]. During the study period, a total of 443 lung cancer deaths were confirmed among 9295 participants. To reduce the potential information bias, 69 lung cancer death with 1 year since enrollment were not included into the analysis.

      To stratified those with higher lung cancer risk, we increased the age criteria from 40 to 50 years old(model 0), then further developed three risk prediction models with multivariate logistic regression respectively, and the predicted probability of lung cancer death for each participants were also calculated based on logistic regression model respectively(table1). The goodness of fit, discrimination and calibration ability of the model performance were evaluated with -2log likelihood, area under the receiver operator characteristic curve (AUC) (C-index) and Hosmer-Lemeshow test. We found that the model incorporated age, gender, smoking, educational prior chronic bronchitis, occupational radon and arsenic had the best discrimination performance with area under ROC as of 0.821(95%CI:0.805-0.836) (figure1a). The calibration performance of this model was also good(Hosmer–Lemeshow type χ2=5.413,p=0.773)(figure1b). The areas under ROC curve of model 2 and model 3 were significantly better than those of model1 and model 0(all p<0.001), however, no difference was found between model 2 and model 3. Besides, Bootstrapping techniques were used for internal validation of the model 3 to Correct for this overfit or optimism, and discrimination C-statistic from C-statics was the same to the original data.

      We stratified the participants into 4 quintiles for the predicted risk of death from lung cancer. The cumulative lung cancer death rate from quintile 1 with lowest risk to quintile 4 having the highest risk increased from 11.51, 47.66, 625.41 to 1732.37 per 105 person-years, while only 2.5% of all lung cancer deaths were in quintile 1 and 2. Similarly, in 210 screen-detected lung cancer deaths, the proportion in quintile 1 and 2 was only 2.4%.

      In conclusion, we have developed and internal validated a lung cancer risk model based on personal and occupation covariates in this occupational population, and this model showed good accuracy for the identification of lung cancer and might assist in identifying individuals at high risk of developing lung cancer in lung cancer screening in this occupational cohort.

      [1] Aberle DR, Adams AM, Berg CD, et al. Reduced lung-cancer mortality with low-dose computed tomographic screening. N Engl J Med. 2011. 365(5): 395-409.

      [2] Tammemägi MC. Application of risk prediction models to lung cancer screening: a review. J Thorac Imaging. 2015. 30(2): 88-100.

      [3] Qiao YL, Taylor PR, Yao SX, et al. Risk factors and early detection of lung cancer in a cohort of Chinese tin miners. Ann Epidemiol. 1997. 7(8): 533-41.

      [4] Lubin JH, Qiao YL, Taylor PR, et al. Quantitative evaluation of the radon and lung cancer association in a case control study of Chinese tin miners. Cancer Res. 1990. 50(1): 174-80.

      [5] Fan YG, Jiang Y, Chang RS, et al. Prior lung disease and lung cancer risk in an occupational-based cohort in Yunnan, China. Lung Cancer. 2011. 72(2): 258-63.

      [6] Yao SX, Lubin JH, Qiao YL, et al. Exposure to radon progeny, tobacco use and lung cancer in a case-control study in southern China. Radiat Res. 1994. 138(3): 326-36.