Big Data and Clinical Research in Oncology: The Good, the Bad, the Challenges, and the Opportunities

Abstract

National cancer registries provide an ever-growing volume of data and increasing access to that data for the purpose of clinical research in oncology. In the United States, three national cancer-specific registries have been developed to collect data on cancer patients, their cancers, how they are treated, and their outcomes. Beginning with the passage of the National Cancer Act of 1971 and funded since 1973, the National Cancer Institute (NCI)’s Surveillance Epidemiology and End Results (SEER) program is a population-based registry from 20 U.S. geographic areas, covering *28 % of the U.S. population. In 1989, the American College of Surgeons (ACoS) Commission on Cancer (CoC) started a joint program with the American Cancer Society: the National Cancer Data Base (NCDB). Approximately 70 % of all new cancer cases diagnosed in the United States each year are currently captured by the NCDB, which contains the records of *29 million patients from *1,500 institutions (making the NCDB a hospitalbased, not population-based, registry). Finally, the National Program of Cancer Registries (NPCR) was established in 1992 and is administered by the U.S. Centers for Disease Control and Prevention (CDC). NPCR supports cancer registries in 45 states, representing 96 % of the U.S. population. The data entered into each of these three national cancer registries are not collected completely independently of the others. The processes, system, and rules that govern the data collection for all three registries significantly overlap, as do the professionals (cancer registrars/CTRs) who actually collect and enter the data. The North American Association of Central Cancer Registries Inc. (NAACR), established in 1987, is a collaborative umbrella organization that develops and promotes uniform data standards for the cancer registries. All central cancer registries in the United States (and Canada) are members of NAACR, although each registry may require a different subset of data elements to be reported. In this issue of Annals of Surgical Oncology, In et al. from the ACoS CoC and nearby Chicago hospitals, bring to light one of several important weaknesses of the data currently collected and reported in these national cancer registries: accurate information on local, regional, and distant recurrence rates and the timing of those recurrences after a first course of treatment is completed. The authors point out that the data reported by each of these registries have traditionally concentrated on the initial presentation and first course of treatment, with little follow-up information reliably collected, except death. The NCDB does attempt to collect data on the time of first recurrence as well as the type of recurrence (local, regional, distant), and the authors examined the completeness of these data points in the NCDB for more than 700,000 patients with five common tumor types diagnosed between 2002 and 2005. Disappointingly, they report that complete information to allow an accurate determination if and when a recurrence had occurred after completion of the first course of treatment, and the type of recurrence, was lacking at a majority of the more than 1,400 hospitals for more than half of their patients. On average, hospitals had incomplete recurrence information on 56.7–66.7 % of patients studied. Only 9.0 % of hospitals collected recurrence information well on all five of the cancer sites examined. The absence of reliable information on recurrence after treatment of primary cancer makes the determination of This is an editorial to the article available at doi:10.1245/s10434-0143516-x.