A new insight into fecal hemoglobin concentration-dependent predictor for colorectal neoplasia

We sought to assess how much of the variation in incidence of colorectal neoplasia is explained by baseline fecal hemoglobin concentration (FHbC) and also to assess the additional predictive value of conventional risk factors. We enrolled subjects aged 40 years and over who attended screening for colorectal cancer with the fecal immunochemical test (FIT) in Keelung community-based integrated screening program. The accelerated failure time model was used to train the clinical weights of covariates in the prediction model. Datasets from two external communities were used for external validation. The area under curve (AUC) for the model containing only FHbC was 83.0% (95% CI: 81.5-84.4%), which was considerably greater than the one containing only conventional risk factors (65.8%, 95% CI: 64.2-67.4%). Adding conventional risk factors did not make significant additional contribution (p = 0.62, AUC = 83.5%, 95% CI: 82.1-84.9%) to the predictive model with FHbC only. Males showed a stronger linear dose-response relationship than females, yielding gender-specific FHbC predictive models. External validation confirms these results. The high predictive ability supported by a dose-dependent relationship between baseline FHbC and the risk of developing colorectal neoplasia suggests that FHbC may be useful for identifying cases requiring closer postdiagnosis clinical surveillance as well as being an early indicator of colorectal neoplasia risk in the general population. Our findings may also make contribution to the development of the FHbC-guided screening policy but its pros and cons in connection with cost and effectiveness of screening should be evaluated before it can be applied to population-based screening for colorectal cancer.