In this study we have demonstrated that positive MSI screening status is associated with a more favourable prognosis and certain clinicopathological characteristics of CRC, i.e. older age, female sex, proximal tumour location and low differentiation grade. This is accordant to previous studies [16, 17], thus consolidating our data. The correlation between MSI and a prolonged CSS was significant in both uni- and multivariable analysis for the whole cohort and for stage III-IV disease, but not for stage I-II disease. A possible interpretation of this observation could be that MSI reflects a less agressive phenotype, with an improved CSS even in the presence of metastatic disease and despite adverse pathological characteristics such as low differentiation grade.
Here, we could not demonstrate any association between MSI status and response to adjuvant treatment in curatively treated patients with stage III disease. On one hand, the subgroups available for analysis of the effects of adjuvant treatment were rather small, thus limiting statistical power. On the other hand, since this study started already in the mid 90s, when adjuvant chemotherapy was not yet standard of care in Sweden, nearly half of the curatively treated patients with stage III disease did not receive adjuvant treatment. Although treatment predictive markers are best evaluated in a controlled, prospective setting, the value of retrospective analyses on tumours from less extensively treated patients should not be underestimated.
Moreover, the feasibility of MSI as a predictor of therapeutic response has been debated, and a recent meta-analysis showed results concordant with ours, i.e. no treatment predictive value of MSI status .
Beta-catenin overexpression was also associated with a reduced risk of death in the full cohort and in patients with stage III-IV disease, although these associations were weaker than for MSI status. Previous studies on beta-catenin as a prognostic marker have been conflicting, with some studies reporting a favourable or no prognostic value [14, 35–38] and some an association with poor clinical outcome [12, 13, 39, 40]. Many factors could explain the differing results regarding the prognostic value of beta-catenin overexpression; e.g. intrinsic tumour heterogeneity , different immunohistochemical staining and visualization methods with varying degrees of sensitivity, and lack of standardization of what constitutes a “positive” or “negative” result. In many cases, nuclear beta-catenin expression is predominantly seen in the margin and not in the center of the tumour . Horst et al. postulated that lost beta-catenin regulation is presented rather as an altered intratumoral distribution of nuclear beta-catenin expression  and, in light of this observation, the TMA-technique may not be an ideal tool for its assessment. Moreover, the degree and prognostic implications of beta-catenin activation may differ in primary and metastatic CRC .
The herein observed inverse association between MSI and beta-catenin overexpression is also in accordance with previous studies [45, 46]. Assuming that MSI status and beta-catenin overexpression represent two different pathways to colorectal tumorigenesis, it might be perceived as contradictory that both reflect an improved CSS. However, as these pathways are not mutually exclusive and likely overlap, their complex interactions need further study.
The associations between MSI and cell cycle regulators cyclin D1 and p21 have been described previously [47, 48] but the underlying mechanisms still remain unclear. Similarly, the inverse associations of MSI tumours with p53 mutation status and p27 expression are well known [30, 48]. The positive correlation between beta-catenin overexpression and p53 status is also in line with previous studies [14, 39, 45].
The observation of a positive association of cyclin D1 expression with MSI screening status and beta-catenin overexpression, respectively, is in line with the previously demonstrated findings of cyclin D1 expression being associated with a prolonged CSS . Notably, the beneficial prognostic value of high cyclin D1 expression was only evident in male and not in female CRC patients . In this study, the prognostic and treatment predictive value of MSI and beta-catenin overexpression did not differ according to sex.
Since the MDCS is a population-based cohort study, a potential selection bias compared with the general population must be taken into consideration, but the distribution of TNM-stages at diagnosis is in line with the expected, with no predilection for less advanced stages [26, 28]. Notably, while the distribution of clinicopathological factors and survival from the disease did not differ between women and men , a recent study demonstrated sex-related differences in the associations of anthropometric factors with CRC risk by TNM stage and tumour location in the present cohort . These findings provide further evidence for colorectal cancer being a disease in which the influence of sex hormones should not be underestimated. Therefore, the prognostic and treatment predictive value of investigative CRC biomarkers should always be considered in strata according to gender. Moreover, it will be of interest to investigate the influence of anthropometric, hormonal and life-style factors on risk of CRC defined by MSI status and beta-catenin alterations in the present cohort.
Since the number of events was identical for CRC-specific and overall survival in patients with metastatic disease, as previously demonstrated , the use of CRC-specific survival as primary endpoint in the survival analyses in this study should be a reasonable surrogate for cancer-specific outcome. Moreover, similar survival trends were observed for 5-year overall survival.