This retrospective cohort study is an important contribution to the literature, but also serves to highlight the perils of retrospective studies.
Source: Armstrong GT, Chen Y, Yasui Y, et al. Reduction in late mortality among 5-year survivors of childhood cancer. N Engl J Med. 2016;374(9):833-842; doi:10.1056/NEJMoa1510795. See AAP Grand Rounds commentary by Dr. Mary-Jane Staba Hogan (subscription required).
PICO Question: Among children surviving ≥5 years after childhood cancer for which treatment was initiated from 1970-1999, what is the cumulative incidence of, and factors associated with, late mortality?
Question type: Descriptive
Study design: Retrospective cohort
The authors presented a new analysis from the Childhood Cancer Survivor Study of late (> 5 years post treatment) mortality in over 34,000 patients treated for cancer from 1970 through 1999. They found, not too surprisingly, a decrease in deaths from any cause as well as from health-related causes. Via a regression-based analysis, they found an association of decreased late mortality with reduction in 1) treatment exposure with cranial radiotherapy for acute lymphoblastic leukemia, 2) abdominal radiotherapy for Wilms' tumor, 3) chest radiotherapy for Hodgkin's lymphoma, and 4) anthracycline exposure.
The authors pointed out limitations of their study, all related to its retrospective nature and limitations of the data collected in the database. Thus, among other limitations, they could not directly attribute the improvements to lesser treatment exposure, nor could they factor in how much general improvement in healthcare contributed to improved outcomes. Therefore, I was a bit surprised at their conclusion in the study's abstract stating that lowered therapeutic exposure contributed to the decline in late mortality. We must remember that an association of events is not the same as cause and effect.
The AAP Grand Rounds editors called readers' attention to an 2001 posting in AAP Grand Rounds mentioning the Ecological Fallacy, meaning simply that summary group characteristics do not necessarily apply to every member of that group. The origins of the ecological fallacy likely are with Aristotle and, at least to me, refer more to how study results are applied than to how results are analyzed. The Survivor Study authors aren't really telling us that a child treated for cancer in 1997 will definitely have a longer post-treatment survival than an individual treated in 1972, which would be how the ecological fallacy would be applied here. The fallacy would be if we, as practitioners, explained the study in that fashion to one of our cancer-survivor patients.
So, we have good news for survivors of childhood cancers in that long-term survival is improving. We can't yet dissect out the contributions of various factors associated with that observation, but we can certainly celebrate the trend.
