OBJECTIVE: The largest tumor diameter, D(T), is a variable of great cl
inical value in breast cancer but a biased and imprecise estimator of
real tumor size. Three-dimensional, shape-independent estimates would
more realistically reflect the tumor bulk and provide more accurate cl
inical staging. For experimental oncology, the measurements may be use
ful for precise assessment of tumor burden. STUDY DESIGN: In 64 prospe
ctively collected breast cancers, unbiased stereology was used for est
imating the gross tumor volume, V(T), cutting specimens into parallel,
equally thick sections with subsequent determination of total tumor s
ectional area. For comparison, the volume of invasive tumor epithelium
, ''V'' (epi), tons obtained by microscopic examination of systematica
lly sampled tissue fractions of the same tumors, embedded in both meth
acrylate and paraffin.RESULTS: The median D(T) was 2.2 cm, and the med
ian V(T) tons 6.72 cm(3). The correlation between these variables was
not very close (r = .77 ), and the slope of the regression line was st
eeper than expected, presumably reflecting a change in tumor shape wit
h growing size. The sampling scheme used for estimation of V(T) proved
highly efficient, yielding a mean error coefficient of 9%. The median
''V''(epi) in methacrylate was 1.19 cm(3), 21% larger than in paraffi
n. Estimates of ''V'' (epi) were highly reproducible (r = .97) and cor
related highly with point counting-based estimates of the feature (r =
.96). ''V'' (epi) correlated with V(T) (r greater than or equal to .7
5), but the slopes of the regression lines were steeper than expected,
corresponding with the correlation found between epithelial volume fr
action and tumor size (r = .26). On average, about 25% of the gross tu
mor was composed of invasive epithelium, but with a wide range. CONCLU
SION: In breast cancer, realistic estimates of tumor volume, volume of
invasive epithelium and epithelial volume fraction can be obtained by
efficient stereologic techniques, which seem useful for clinical and
experimental oncology. In the present methodologic study, baseline dat
a were generated. Further studies are needed to assess the clinical va
lue of stereologic tumor size estimates as compared with traditional s
taging parameters.