Morphometry in predicting survival in breast cancer

Abstract

To assess the value of morphometry in predicting prognosis of patients with invasive breast cancer (BC) 102 patients with infiltrating ductal and infiltrating lobular cancers were selected for study from the 5 yrs up to 1979. Of these 46 died, 39 from BC and 7 from other causes. Among the cellular and nuclear features measured were mitotic activity index (MAI); cellularity index; N/C ratio; nuclear longest axis, shortest axis, area, perimeter, size, shape factor; and nucleolar area, perimeter and area to perimeter ratio. Two systems for histological grading of invasive BC were compared, the NSABP method and our modification of the WHO system. The number of axillary lymph nodes involved was also noted. The statistically significant χ 2 values for the 3 Cox Regression analyses (CRA) are shown below. VARIABLES CRA 1 CRA 2 CRA 3 Positive nodes >3 8.82 10.20 10.24 MAI (mitoses/10 HPF) 5.61 7.75 4.94 Mitoses/1000 nuclei 6.07 10.14 6.86 Modified WHO histo-grade (3.40) 7.66 6.22 NSABP histo-grade (2.84) 4.93 (0.89) Age >75yrs 4.79 (1.40) (0.46) CRA 1; time to all deaths. CRA 2: time to BC-specific deaths. CRA 3: disease-free interval. ( ) not significant χ 1 2 = 3.84 Of all the morphometric data mitoses/1000 cells achieved the highest statistical significance in separating survivors from non-survivors and was also the single most important feature in both the NSABP grading method and our modified WHO histological grading scores. Of the 3 regressions, the BC-specific death regression (CRA-2) was considered the best χ 2 2 =29.4, p Conclusions : A PI based on mitoses/1000 cells and positive nodes > 3 or ≤ 3 will be useful in identifying and classifying more accurately groups of patients of differing prognosis after the present learning set has been enlarged and tested against a new group of patients (test set) and subjected to statistical evaluation of the accuracy and sensitivity of the classification rule.