IDDF2018-ABS-0224 Evaluating stratification performance of high-risk factors questionnaire (HRFQ) for colorectal cancer and advanced colorectal neoplasia in chinese asymptomatic subjects

Background High-Risk Factors Questionnaire (HRFQ) has been widely applied in Chinese organised colorectal cancer (CRC) screening programmes since the 1990s. It served as a screening test to identify the possibly missing high-risk population whose faecal occult blood test result was negative. Nevertheless, its performance to stratify CRC and advanced colorectal neoplasia (ACN) have not been adequately examined. There is no comparison between its discriminatory capability and that of other risk prediction models commonly adopted in China, such as the Asia Pacific Colorectal Screening (APCS). Methods A prospective, multi-centre colonoscopy study was held in 7 endoscopic units across North Shanghai, China from 2016 to 2017. Average-risk subjects were recruited when they received screening colonoscopy. Information required by HRFQ, including family history of CRC in first degree relatives, personal history of any malignancy, past history of colorectal polyps, and other required variables were collected by self-reported questionnaires. Experienced colonoscopists performed all the colonoscopy, while pathologists who were blinded to the survey findings, reported the histology based on standard diagnostic criteria. The sensitivity, specificity, positive/negative predictive value(PPV/NPV), and the C-statistics were calculated to evaluate the discriminatory value of HRFQ for CRC and ACN. We also examined the above parameters when APCS scoring system was used, using a score of 4–7 points as ‘high-risk’. Results Among all 2813 enrolled individuals (mean age 58.7±8.4 years, male 52.0%), 142 cases (5.0%) had ACN and 29 patients (1.0%) had invasive cancers. The proportion of high-risk population identified by HRFQ and APCS were 46.9% (n=1,318) and 31.8% (n=894), respectively. For CRC the difference between HRFQ and APCS was not significant in C-statistics. However, for ACN the performance of APCS with respect to sensitivity, specificity, PPV, NPV and C-statistics, were significantly superior to that of HRFQ (table 1). Conclusions These findings implied that the use of HRFQ might need to be re-examined to attain a satisfactory predictive capability. Due to the current limited CRC case number, we recommend future studies in other population groups to enhance its generalizability.