More information about text formats
We read with great interest the report from Hamer and colleagues that examined the hypothesis that associations between television (TV) viewing and mortality from heart disease (HD) are due to confounding (1). They employed a negative control approach (2) and report evidence of associations between TV viewing and HD mortality (HR=1.09 [1.06, 1.12] per 1 hr/day increase in TV) and accidental deaths (the negative control outcome; HR=1.06 [0.98, 1.15]) after adjusting for age, sex, smoking, education, and prevalent HD (1)
The positive association between TV and accidental deaths was interpreted as evidence that the TV-HD mortality association was due to confounding. Although key study limitations were noted including a small number of accidental deaths and limited adjustment for confounding, the authors concluded that “observed associations between TV and HD are likely to be driven by confounding”. Although we agree that confounding is a worrisome threat to the internal validity of epidemiologic studies, we believe that the conclusion in the Hamer report is overstated.
A critical additional strategy to understand bias due to confounding, one that was not employed in the current study, is to examine relevant results from published studies conducted in different study populations using different methods. (2) We previously reported results in two studies that examined associations for accidental deaths and HD mortality with TV viewing (3) and leisure-time sitti...
A critical additional strategy to understand bias due to confounding, one that was not employed in the current study, is to examine relevant results from published studies conducted in different study populations using different methods. (2) We previously reported results in two studies that examined associations for accidental deaths and HD mortality with TV viewing (3) and leisure-time sitting (predominantly TV viewing) (4), but these findings were not cited in the Hamer report. Our studies each examined 3-5 times more accidental deaths than the Hamer report, and adjusted for a much larger number of confounding factors. (3,4) We found significant positive associations with TV-HD mortality, but no evidence of association between accidental deaths and greater TV time (HR=1.01 [0.62, 1.64]; 7+ vs. < 1 hr/day) (3) or leisure-time sitting (HR=0.91 [0.76, 1.10]; 6+ vs. < 3 hr/d). (4) In context of the negative control outcome framework, our results provide no evidence that previously observed HD associations, or associations with several other causes of death, were due only to confounding. (3,4)
Many studies have reported positive associations between disease/mortality outcomes and TV, a prevalent leisure-time behavior that, as Hamer and colleagues note, is likely to displace time spent in more healthful physically active pursuits. We strongly support efforts to better understand these relationships, including careful consideration of bias and threats to validity. As we do so, it is critical that we consider the broad range of information available before drawing strong conclusions based on a single study.
1. Hamer M, Ding D, Chau J, Duncan MJ, Stamatakis E. Association between TV viewing and heart disease mortality: observational study using negative control outcome. Journal of Epidemiology and Community Health. 2020:jech-2019-212739.
2. Pearce N, Vandenbroucke JP, Lawlor DA. Causal Inference in Environmental Epidemiology: Old and New Approaches. Epidemiology. 2019;30(3):311-316.
3. Keadle SK, Moore SC, Sampson JN, Xiao Q, Albanes D, Matthews CE. Causes of Death Associated With Prolonged TV Viewing: NIH-AARP Diet and Health Study. American Journal of Preventive Medicine. 2015;49(6):811-821.
4. Patel AV, Maliniak ML, Rees-Punia E, Matthews CE, Gapstur SM. Prolonged Leisure Time Spent Sitting in Relation to Cause-Specific Mortality in a Large US Cohort. American Journal of Epidemiology. 2018;187(10):2151-2158.