Thus, one could have expected that inactivation of Smad4 might result in a TGF�� resistance that would favour tumour expansion. Interestingly, the patients with deletion of Smad4 did not show a significantly worse thereby prognosis than those without a deletion (Boulay et al, 2002). In contrast, in the same population, Smad4 seemed to be a predictive marker for 5FU/mitomycin adjuvant chemotherapy. However, whether Smad4 plays a key role in tumorigenesis of colorectal cancer is still unclear. To date, a significant number of Smad4 point mutations have been found only in pancreatic carcinomas (50�C60%), biliary tract carcinomas (15%) or colorectal carcinomas (5�C20%) (Hahn et al, 1996; Schutte et al, 1996; MacGrogan et al, 1997).

Although the existence of additional unknown target tumour-suppressor genes in the region of 18q21 cannot be ruled out, recently published results strongly suggest a significant contribution of Smad4 gene inactivation in advanced tumour stages. Metastatic colorectal carcinomas including carcinomas metastasised to the liver showed a considerably higher frequency (31�C35%) than invasive carcinomas without distant metastasis (7%) (Miyaki et al, 1999; Ohtaki et al, 2001). Our findings of less than 5% point mutations are at the lower end of the spectrum and confirm the low frequency of point mutations of Smad4 in early-stage colorectal cancer without distant metastasis. The limitation to patients with loss of one Smad4 allele�Cinitially used to select a population with a presumably high mutation frequency�Cis one possible theoretical explanation for our findings.

However, in pancreatic and biliary tract carcinomas, patients with LOH represent a group with an especially high point mutation frequency in the remaining gene, making this explanation highly unlikely (Hahn et al, 1998; Barbera et al, 2000). Other possible explanations for the absence of Smad4 point mutations in colorectal cancer at this stage include methylation changes at the promoter and alternative splicing or changes in mRNA stability (Roth et al, 2000). The importance of genes that undergo alterations at low prevalence, however, may as yet be underestimated. Such events may contribute significantly to the genetic variety within a tumour type and, thus, to the complexity of human tumorigenesis. Since it is likely that many alterations of low prevalence exist in human cancers, an individual tumour might still acquire several of these different alterations with a high probability, making low prevalence alterations a powerful driving force of the carcinogenic process. In conclusion, our findings indicate that Smad4 point mutations are infrequent in early Brefeldin_A stages of colorectal cancer.