Tissue microarray technology was introduced in 1998 by Kononen1, allowing hundreds of samples to be arrayed and studied in a single tissue block. The question arises as to whether the small cores present in the arrays are representative. The extent to which one, two, three or more cores will prove representative will vary with the heterogeneity of a particular tumour in its immunoreactivity for a particular antibody. The probability that any particular number of cores will be representative of the entire section has been studied:
Tumour type |
antibody |
one core |
two cores |
three cores |
four cores |
|
ovarian carcinoma |
91% |
97% |
98% |
99% |
||
ER2 |
92% |
97% |
98% |
99% |
||
p532 |
91% |
96% |
98% |
99% |
||
breast carcinoma |
ER3 |
92.5% |
96.2% |
98.1% |
99.1% |
|
PR3 |
94.5% |
97.3% |
98.6% |
99.3% |
||
Her23 |
90.4% |
95.2% |
97.6% |
98.8% |
||
gastric cancer biopsies |
p534 |
|
|
97% |
|
|
|
|
97% |
|
|||
VEGF4 |
|
|
97% |
|
||
fibroblastic tumours |
|
|
96% |
|
||
p537 |
|
|
97% |
|
||
pRB7 |
|
|
94% |
|||
Others have recommended the evaluation of 3 or 4 cores5,9.
Good correlation between TMAs and whole sections has been shown for a panel of lymphoid markers10.
Any limitation due to the sampling error in using cores to represent an entire section of a tumour is more than offset by the statistical advantage of being able to examine a very large number of cases of a particular tumour. For example, if 50% of 100 tumours are found to be immunoreactive, the 95% confidence interval for the true rate of reactivity for the population from which this sample is drawn is 40% to 60%. Most studies are of far smaller numbers of cases, and therefore the confidence intervals are correspondingly much wider.
References
This page last revised 30.8.2004.
©SMUHT/PW Bishop