Is Computed Tomography Simulation Associated With Improved Survival, or Is This Finding a Result of Confounding Variables?

Abstract

TO THEEDITOR: In a recent article, Chen et al 1 analyzed the use of computed tomography (CT) simulation in the treatment of stage III non–small-cell lung cancer (NSCLC). They found that CT simulation was associated with a lower risk of death with an adjusted hazard ratio of 0.77. For this large improvement, on par with the reduction seen with the addition of concurrent chemotherapy, to be associated with CT simulation alone is simply not plausible. The authors imply that the use of CT simulation in lung cancer is associated with higher thoracic radiotherapy doses and better targeting and that these may have led to the improved survival. These conclusions are mitigated by the fact that radiation oncologists have been using diagnostic CT scans to shape their fields since the early 1980s, as stated in the editorial 2 that accompanied the article by Chen at al. 1 What then can explain these results? It is likely that there is a confounding variable such as the increasing use of positron emission tomography (PET) scans during this same time period. The authors do not mention whether they accounted for PET scans as an instrumental variable when they assessed for confounding by unobserved factors. PET began to be increasingly used after the year 1998 in the staging of NSCLC after the approval of PET for reimbursement by Medicare. 3 The editorial 2 that accompanied the article by Chen et al 1 states that the receipt of CT simulation may have been a marker for overall quality of care. However, CT simulation could also well be a marker for patients who underwent advanced imaging studies. Centers with access to a CT simulator would have been more likely to offer their patients access to PET scans as well. PET scanning has been shown in prospective studies to detect metastatic disease 15% to 20% of the time when other staging studies are negative. 4 The use of PET scans in patients with NSCLC during the time period of the study by Chen et al 1 would have caused a stage migration. Patients with stage III NSCLC who received a PET scan would have had an improved survival compared with those who did not because of the exclusion of patients with PET-detected metastatic disease. This would have been independent of the CT simulation procedure. However, by not accounting for stage migration and the increasing use of PET scanning during the time period studied, the authors may have erroneously associated CT simulation with an improved survival. In fact, a large population-based study of patients with stage III cancer in a pre-PET era (1994-1998) and a PET era (1998-2004) found that PET use was significantly associated with a reduced hazard ratio of 0.77 (95% CI, 0.69 to 0.85). 5 This reduction in hazard ratio is similar to the reduction in hazard ratio of 0.77 that Chen et al 1 are associating with CT simulation. No one in the oncologic community questions the real benefits of CT simulation with respect to target delineation, dose calculations, and precise radiation therapy planning. However, before associating the improved hazard ratio with CT simulation, we should take into account the well-documented Will Rogers effect and stage migration as a result of PET scanning that occurred in stage III lung cancer during the time period studied. 6