OBJECTIVE: The goal of this study is the improvement of the prognostic
information associated with conventional diagnoses. Our previous fact
or analysis of 26 reliably identified histological features in infrate
ntorial childhood neuroglial tumors yielded five interpretable, uncorr
elated, quantitative histological factors that we named spongy, fibril
lary, proliferative, nuclear, and ring. Five quantitative scores, one
for each of the five factors, provide an objective method for quantify
ing the histological heterogeneity of a tumor. The scores, alone or in
conjunction with conventional diagnoses, identify groups of histologi
cally homogeneous tumors. METHODS: Multivariate Cox proportional hazar
ds models were developed to assess the contribution of each factor to
survival prognosis, after allowing patient-specific demographic and cl
inical data in the models as covariates. Hazard ratios, estimated for
each statistically significant factor and covariate in the multivariat
e model, provide the basis for the determination of the prognosis. The
hazard ratio is the ratio of the hazard function for subjects with an
attribute, e.g., an age of 10 years, to the hazard function for subje
cts who have some chosen baseline attribute, e.g., an age of 1 year. T
he important criterion of this ratio is beta, a statistic estimated fr
om the survival data in the Childhood Brain Tumor Consortium database
of infratentorial neuroglial tumors. Kaplan-Meier survival curves were
used to investigate differences in the survival of factor-determined
subgroups of patients with various diagnoses. RESULTS: An increased li
kelihood of survival is associated with older age, more tumor removal,
more recent decade of surgical intervention, and high spongy and fibr
illary factor scores. A decreased likelihood of survival is associated
with high nuclear, proliferative, and ring factor scores. Gender, loc
ation within the infratentorial compartment, and subsequent treatment
did not add prognostic information. For certain subgroups of astrocyto
ma and for ependymoma and medulloblastoma, factors are important in pr
edicting survival with greater accuracy. CONCLUSION: Factor scores pro
vide clinically useful quantitative estimates of survival probability
that are more specific and accurate than the general estimates based o
n the conventional diagnosis alone.