Association Rule Mining Based HotSpot Analysis on SEER Lung Cancer Data

Introduction

Lung cancer ranks second in the list of most common cancers (National Cancer Institute, n. d.), and first in the list of most deadly cancers (Centers for Disease Control and Prevention, 2010), with the survival rate being about 15% after 5 years of diagnosis (Ries & Eisner, 2007).

The Surveillance, Epidemiology, and End Results (SEER) Program (National Cancer Institute, 2008) of the National Cancer Institute is an authoritative repository of cancer statistics in the United States (National Cancer Institute, 2010). It is a population-based cancer registry which covers about 26% of the US population across several geographic regions and is the largest publicly available domestic cancer dataset. The data includes patient demographics, cancer type and site, stage, first course of treatment, and follow-up vital status. The SEER program collects cancer data for all invasive and in situ cancers, except basal and squamous cell carcinomas of the skin and in situ carcinomas of the uterine cervix (Ries & Eisner, 2007). The SEER limited-use data is available from the SEER website on submitting a SEER limited-use data agreement form. Gloeckler Ries, Reichman, Lewis, Hankey, and Edwards (2003) present an overview study of the cancer data at all sites combined and on selected, frequently occurring cancers from the SEER data. The SEER data attributes can be broadly classified as demographic attributes (e.g., age, gender, location), diagnosis attributes (e.g., primary site, histology, grade, tumor size), treatment attributes (e.g., surgical procedure, radiation therapy), and outcome attributes (e.g., survival time, cause of death), which makes the SEER data ideal for performing outcome analysis studies.

With SEER data being available in the public domain, there is a mature literature on the statistics of SEER data (Yao et al., 2008; Rusthoven, Flaig, Raben, & others, 2008; Ries & Eisner, 2007; Coburn et al., 2008; Wang, Emery, et al., 2007; Wang, Fuller, Emery, & Thomas, 2007; Choi, Fuller, Thomas, & Wang, 2008), many of them using the the SEERStat software provided by SEER itself. Statistical studies using the SEER data include demographic and epidemiological studies of rare cancers (Yao et al., 2008), assessing susceptibility to secondary cancers that emerge after a primary diagnosis (Rusthoven et al., 2008), performing survival analysis (Ries & Eisner, 2007), studying the impact of a certain type of treatment on overall survival (Coburn et al., 2008), studying conditional survival (measuring prognosis of patients who have already survived a period of time after diagnosis) (Wang, Emery, et al., 2007; Wang, Fuller, et al., 2007; Choi et al., 2008), amongst many others. There also have been scattered applications of data mining using SEER data for breast cancer survival prediction (Lundin et al., 1999; Delen, Walker, & Kadam, 2005; Bellaachia & Guven, 2006; Endo, Shibata, & Tanaka, 2008) and a few studying lung cancer survival (Chen et al., 2009; Fradkin, 2006; Agrawal, Misra, Narayanan, Polepeddi, & Choudhary, 2011), but to the best of our knowledge there is no association rule mining analysis on lung cancer data.