The aim of this study was to examine the effect of measurement bias in breast cancer and to create a more rational T-size categorization in tumor-node-metastasis staging in response to smaller, screen-detected cancers and measurement bias.

From 1987 to 2003, 10,853 invasive nonmetastatic breast cancers enlisted in the Rhode Island Cancer Registry with a known dimension were reviewed. Data analyzed by proposed classifications included the rate of lymph node metastases and the mortality rate from breast cancer.

The median diameter was 16 mm. Cancer measurements reflected the bias in pathologists’ dimension recording, which is centered strongly about whole- and half-centimeter sizes. A new T classification is proposed with the following sizes and frequencies in the Rhode Island Cancer Registry: 1 to 2 mm = T1 mic (3% of registered cases); 3 to 7 mm = T1a (11%); 8 to 12 mm = T1b (23%); 13 to 17 mm = T1c (18%); 18 to 22 mm = T2a (17%); 23 to 27 mm = T2b (8%); 28 to 32 mm = T2c (8%); 33 to 42 mm = T3a (6%); 43 to 52 mm = T3b (3%), and greater than 52 mm = T3c (4%). The unadjusted odds ratio for the probability of node metastases was 1.43 (confidence interval, 1.40–1.46; P < .001) with each increase in proposed grouping. The range in the lymph node metastatic rate was 5.5% for tumors 1 to 2 mm to 64% for cancers greater than 52 mm. By Cox proportional hazard, the unadjusted hazard ratio for death from breast cancer for each increase in proposed grouping was 1.33 (confidence interval, 1.29–1.37; P < .001). The 10-year survival rate ranged from 98.3% for tumors 1 to 2 mm to 70.3% for cancers greater than 52 mm.

A more rational T category for use in tumor-node-metastasis staging is presented to reflect the much smaller invasive breast cancers encountered by screening and to account for the dimension recording bias of pathologists. This new T category shows a clinically and statistically significant linear relationship for both incidence of lymph node metastases and hazard ratio of death.