Purpose: To determine the frequency and patterns of lore mortality amo
ng long-term survivors of childhood cancer. Materials and Methods: Med
ical records of patients who survived at least 5 years after the diagn
osis of childhood cancer were reviewed to determine the causes of subs
equent deaths. Estimated 15-year survival and standardized mortality r
atios for deaths from nonneoplastic treatment complications were compa
red with adjusted United States population estimates, The study includ
ed 2,053 patients who had survived greater than or equal to 5 years, g
rouped by treatment eras that reflected increased intensity of therapy
and significantly improved survival (early era, 1962 to 1970; recent
era, 1971 to 1983). Results: There have been 258 subsequent deaths in
the 2,053 childhood cancer survivors; 169 occurred 5 to 10 years postd
iagnosis and 89 greater than or equal to 10 years post diagnosis. For
the study period as a whole, deaths were attributed to recurrent prima
ry malignancy in 61% of cases, second malignancy in 20%, nonneoplastic
treatment complication in 10%, and unintentional injury/suicide in 8%
. Late death from recurrent disease decreased significantly for surviv
ors treated in the recent era (P < .0001), while the risk of death fro
m second malignancies increased, although not statistically significan
tly (P = .10). Projected 15-year survival estimates for all greater th
an or equal to 5-year survivors in both treatment ems was greater than
90%, but differed from expected rates, Conclusion: Late mortality fro
m recurrence after treatment for childhood cancer decreases with more
effective initial therapy. Prolonged disease-free status is associated
with an expected survival that approaches that of the general populat
ion for patients treated from 1971 through 1983. The impact of more re
cent intensified and novel therapies for high-risk patients remains to
be determined. (C) 1997 by American Society of Clinical Oncology.