Jump to comment:

• Longer career, longer follow up, immortality bias
  Juliana Antero, Aurélien Latouche and Jean-François Toussaint
  Published on: 29 January 2020

• Published on: 29 January 2020

  Longer career, longer follow up, immortality bias

  • Juliana Antero, Researcher The French National Sports Institute (INSEP)
  • Other Contributors:
    • Aurélien Latouche, Professor
    • Jean-François Toussaint, Professor

  The article Mortality of Japanese Olympic athletes: 1952 – 2017 cohort study currently fails to account for a probable source of non-negligible bias. Whilst the main finding, stating that there is a decreased mortality rate among Japanese Olympians appears to be methodologically reliable, the authors also state that “higher mortality was observed among those who participated in the Olympics twice and three times or more compared with those who participated just once,” which we believe to be a potentially inaccurate finding.
  A delayed entry exists at baseline, as those with longer careers tend to be older than those with shorter careers. For instance, someone who has participated in 3 Olympic Games started her/his Olympic career 12 years before those who have participated only once. Thus, there might be an important period effect underlying these findings. Even if the authors have adjusted the analysis by age, this is not equivalent to adjusting for period, as earlier periods in time are related with higher mortality rates, which could explain the findings.
  Additionally, another problem often seen in survival studies is ignoring when a death event occurs. This is problematic because deaths will be observed more frequently in subjects with longer follow-up times. In the presence of time-dependent bias, the hazard ratio is artificially underestimated and the length bias leads to an artificial underestimation of the overall hazard [2]. Therefore, faulty interpretat...

  Show More

  The article Mortality of Japanese Olympic athletes: 1952 – 2017 cohort study currently fails to account for a probable source of non-negligible bias. Whilst the main finding, stating that there is a decreased mortality rate among Japanese Olympians appears to be methodologically reliable, the authors also state that “higher mortality was observed among those who participated in the Olympics twice and three times or more compared with those who participated just once,” which we believe to be a potentially inaccurate finding.
  A delayed entry exists at baseline, as those with longer careers tend to be older than those with shorter careers. For instance, someone who has participated in 3 Olympic Games started her/his Olympic career 12 years before those who have participated only once. Thus, there might be an important period effect underlying these findings. Even if the authors have adjusted the analysis by age, this is not equivalent to adjusting for period, as earlier periods in time are related with higher mortality rates, which could explain the findings.
  Additionally, another problem often seen in survival studies is ignoring when a death event occurs. This is problematic because deaths will be observed more frequently in subjects with longer follow-up times. In the presence of time-dependent bias, the hazard ratio is artificially underestimated and the length bias leads to an artificial underestimation of the overall hazard [2]. Therefore, faulty interpretation may result from the underlying assumptions, which is potentially the case in this study.
  The authors argue that the higher mortality rate found among those who participated in multiple Olympics could be related to longer periods of intense training. Yet, this is contradictory to their main finding which demonstrates far lower mortality rates among Olympians in comparison to the general population.
  Finally, the higher mortality found among those engaged in multiple Games could also be explained by the missing life status confirmations. There were 1324 Olympians who were assumed to be alive when missing date of death. Previous studies that dealt with the same data source and the submission of life status to national data certification for confirmation found that there was more reliable data for well-known athletes [3]. Thus, date of death is probably more widespread for athletes engaged in multiple games and therefore less well-known athletes, participating in only one Games, may have been mislabeled alive when death information was missing.
  The main findings that demonstrate lower mortality rates among Japanese Olympians compared to the general population rely on standard methods and seem to be reliable. The secondary findings that compare mortality among Olympians engaged in multiple games should be revised accounting for the biases described in this letter.

  1 Zwiener I, Blettner M, Hommel G. Survival analysis: part 15 of a series on evaluation of scientific publications. Dtsch Ärztebl Int 2011;108:163–9. doi:10.3238/arztebl.2010.0163
  2 Wolkewitz M, Allignol A, Harbarth S, et al. Time-dependent study entries and exposures in cohort studies can easily be sources of different and avoidable types of bias. J Clin Epidemiol 2012;65:1171–80. doi:10.1016/j.jclinepi.2012.04.008
  3 Antero-Jacquemin J, Pohar-Perme M, Rey G, et al. The heart of the matter: years-saved from cardiovascular and cancer deaths in an elite athlete cohort with over a century of follow-up. Eur J Epidemiol Published Online First: 5 May 2018. doi:10.1007/s10654-018-0401-0

  Show Less

  Conflict of Interest:
  None declared.

  • Back to top