Assessment of selection bias in clinic-based populations of childhood cancer survivors: a report from the childhood cancer survivor study

Pediatr Blood Cancer. 2009 Mar;52(3):379-86. doi: 10.1002/pbc.21829.


Background: It is not known to what extent prevalence estimates of late effects among childhood cancer survivors derived from clinic based samples represent the actual estimates that would be derived if the entire population of childhood cancer survivors was recruited and evaluated for a particular outcome.

Procedure: In a large retrospective cohort study of childhood cancer survivors, the Childhood Cancer Survivor Study (CCSS), the prevalence of chronic health conditions among participants who reported being seen in a cancer center or long-term follow-up clinic was compared to the prevalence of chronic conditions in the entire cohort.

Results: When compared to survivors who had no medical care in the previous 2 years, survivors accessing medical follow-up were significantly more likely to have chronic health conditions. Relative risks of reporting a chronic health condition were 1.4 (95% CI: 1.3-1.5) if seen in a cancer center or long-term follow-up clinic and 1.2 (95% CI: 1.1-1.3) if seen in a general medical care setting. Estimates derived from only those childhood cancer survivors who were seen in a cancer center or long-term follow-up clinic overestimate the prevalence of any chronic disease by 9.3% (95% CI: 7.0-11.6).

Conclusions: Applying chronic condition prevalence estimates from a clinical population to the general population of childhood cancer survivors must be undertaken with caution. Survivorship research must maintain a high level of scientific rigor to ensure that results reported in the literature are interpreted within the appropriate context.

Publication types

  • Research Support, N.I.H., Extramural
  • Research Support, Non-U.S. Gov't

MeSH terms

  • Adolescent
  • Adult
  • Humans
  • Neoplasms / diagnosis
  • Neoplasms / epidemiology*
  • Risk Factors
  • Survival Rate
  • Survivors*