11 (0.01-0.93), P for view more linear trend = 0.03) (Table (Table22). Table 2 Odds ratios and 95% confidence intervals for quartiles of each biomarker level and colorectal cancer risk from multivariate conditional logistic regression models1 Quartiles of plasma sVCAM-1 level were positively associated with increased colorectal cancer risk (P for linear trend = 0.02) (Table (Table2).2). This association was borderline non-significant when sVCAM-1 was coded as a continuous variable (P = 0.07). Unadjusted models (matching factors only) showed similar results (data not shown). A sensitivity analysis excluding cases that were diagnosed during the first two years of follow-up (7 cases) did not modify the findings, nor did sensitivity analyses excluding subjects with high hs-CRP values (> 15.5 ng/mL, i.e.

, mean + 3SD, n = 3 subjects; data not shown). Indicators of the predictive potential of colorectal cancer risk models (Table (Table3)3) showed improvement when adiponectin alone was included in the multivariate model (P for AUC improvement = 0.009). The RIDI statistic indicated a 12.2% (10.9-13.6) improvement. Improvement in the prediction of colorectal cancer risk was limited when sVCAM-1 only was introduced into the multivariate model (P for AUC improvement = 0.09), with 9.9% (8.7-11.0) improvement, as indicated by the RIDI statistic. Prediction was substantially improved when adiponectin and sVCAM-1 were simultaneously included in the multivariate model: P for AUC improvement was equal to 0.01, and the RIDI reached 26.5% (24.4-28.7).

Table 3 Predictive potential of adiponectin and soluble vascular cell adhesion molecule-1 regarding colorectal cancer risk: Relative integrated discrimination improvement and improvement of area under the curve DISCUSSION In this prospective study, pre-diagnostic plasma adiponectin level was associated with decreased colorectal cancer risk, independently of other known risk factors. On the contrary, plasma sVCAM-1 level was associated with increased colorectal cancer risk. Models including these two biomarkers showed significantly improved discriminatory capabilities compared to models including only established risk factors. Lower levels of circulating adiponectin have been observed in prevalent colorectal cancer cases compared to controls[10,13,24-26].

Single nucleotide polymorphism analyses have found that some variants of the adiponectin genes are related to either increased (rs822395, rs1342387) Entinostat or decreased (rs266729) colorectal cancer risk[27], although no association was detected in a recent study in the United Kingdom[28]. Another study suggested that variants of the adipokine genes may affect colorectal cancer risk in combination with variants in diabetes-related genes[29]. Studies with colorectal cancer patients showed that higher adiponectin levels were associated with a better prognosis[10,13,30].