American College of Surgeons. Cancer programs. National Cancer Data Base. www.facs.org/cancer/ncdb/. One of the main advantages of large datasets is the study of practical models at a disease site. This may be particularly useful in the field of lung radiation therapy, as there are differences in practice between providers based on available knowledge. The introduction of new techniques or technologies, including SBRT for early-stage non-small cell lung cancer (NSCLC) and metastatic pulmonary diseases, is a frequent application of this issue. Because of the high rates of local control observed in prospective studies evaluating the effectiveness of SBRT (26.27) in the NSCLC, several groups have used NCDB (28-30) to study how this has changed over the past 10 to 15 years. Another example was the emergence of IMRT in the treatment of locally advanced CBNPC. An analysis of SEER-Medicare revealed that the rate of IMRT use increased from 0.5% in 2001 to 14.7% in 2007 in patients with stage III NSCLC (31). Well-conducted observational studies with large databases promise to improve our understanding of lung cancer management. Given the need to provide quality care to patients from all segments of the population, observational analyses can help clarify ideal management strategies for certain categories of patients and determine which patients may not be optimally treated. In addition, the emergence of machine learning can help eliminate unknown disruptive factors and improve the accuracy and prediction of models based on large datasets (84).
While RCTs remain the gold standard on which we rely on our treatment choices, observational analyses with large registries can provide important hypothesis data from which future prospective studies that will change practice can be constructed. Important lessons have been learned from the NCDB alpha-PUF (Participant User Files). As an alpha-PUF test site, we have had the opportunity to experience many of these lessons up close. The complexity and staffing required for an effective results research program was evident early in the Alpha PUF test program. Eighteen months after the program, only a small fraction of the first six trial sites for alpha-PUF cancer received published results. This was due, at least in part, to an underestimation on our part (and I would start from a similar point of view from the other five Alpha PUF sites) the significant resources needed to make a program successful. These include the resources needed to conduct quantitative analyses, but also a lack of familiarity with the available data elements, making it difficult to ask clinically important questions that have a reasonable chance of being accurately addressed by the available data.