Background: Present clinical algorithms assign adjuvant chemotherapy according to prognosis, but clinical decision-making would be greatly improved if reliable predictive markers were available to identify which subsets of patients benefit most from treatment. We examined molecular markers in preserved tissue from patients with Dukes' B or C colon cancer randomised to receive, or not, adjuvant fluorouracil, and assessed each marker's prognostic and predictive value.
Methods: Formalin-fixed paraffin-embedded paired normal and tumour samples were obtained from 393 patients with colon cancer from the UK AXIS trial of postoperative portal vein infusion fluorouracil versus control. We measured loss of heterozygosity (LOH) and microsatellite instability at four loci: P53 (17p13), D18S61 (18q22.3), D18S851 (18q21.1), and DP1 (5q21). The prognostic value of each marker was assessed with the log-rank test, and the predictive value by comparison of treatment hazard ratios with the chi(2) test for heterogeneity (CSH).
Findings: In 228 (58%) patients informative for LOH at D18S61, this marker was significantly predictive: benefit from fluorouracil was significantly greater in patients retaining heterozygosity than in those with LOH (CSH p=0.02). Conversely, LOH at D18S61 was a significant prognostic marker of improved outcome in untreated patients. 314 (80%) patients were informative for LOH at at least one of the three 17p and 18q sites, of whom half retained heterozygosity at one or more site. The effect of chemotherapy in these patients was striking (hazard ratio 0.45, 95% CI 0.28-0.73), whereas chemotherapy had no effect in patients with no retained heterozygosity (0.91; 0.56-1.48), CSH p=0.039.
Interpretation: Retention of heterozygosity at one or more 17p or 18q sites was associated with the ability to benefit from adjuvant fluorouracil. These results support the principle of developing molecular markers as predictive factors in treatment decisions.