This cross-sectional study validated the performance of the APCS score and its modified versions in an asymptomatic population in China. The results show that all scoring systems have certain ability to predict the risk of CAN and reduce the resource load. However, in this external validation, there was no essential difference in the predictive power between the APCS score and its modified versions. The modified APCS score and the KCS score seemed the preferable systems to classify high risk subjects based on its highest RR and sensitivity in the selected population.
The most important finding of this study is that the APCS score and its modified versions have certain ability to predict CAN in asymptomatic population in Xuzhou. This result is consistent with previous validation of the APCS score in Beijing (13) and Ningxia (14), which may mean that risk scoring scores can be used as a preliminary screening for colorectal cancer screening in China. Even more, as the updated Asia Pacific Consensus Recommendations on colorectal cancer screening recommended (6), the risk scoring system can select high-risk patients for early colonoscopy and reduce the colonoscopy resource load required to detect one CAN. Since questionnaire survey is one of the basic methods of colorectal cancer screening in China, the use of the scoring system as a preliminary screening may improve the cost-effectiveness of colorectal cancer screening.
Although there were no statistically significant differences in the predictive ability of each scoring system for CAN, there were some differences in the performance characteristic. Compared with the APCS score, the CNP scores performed similarly. The modified APCS score and the KCS score improved the sensitivity and reduced the missed diagnosis of CAN, but the specificity decreased and the misdiagnosis increased. On the contrary, the 8-point risk score improved the specificity and reduced the misdiagnosis, but the sensitivity was decreased, which was easy to cause missed diagnosis. Since colorectal cancer screening is the process of detecting and intervening early-stage colorectal cancers and precancerous lesions in asymptomatic population (15–17), it is more important to reduce missed diagnoses with little difference in resource loads. The modified APCS score and the KCS score seemed to be preferable systems to classify high risk subjects based on its highest sensitivity in the selected population.
Using the APCS scoring system as a comparator, the modified APCS score, the KCS score and the 8-point risk score added BMI as a risk factor, while the CNP score added BMI and diabetes as risk factors. BMI is a typical value derived from the weight and height to define overweight (25 ≤ BMI < 30) and obesity (BMI ≥ 30) in adult. Obesity is regarded as one of the key risk actors for the pathogenesis of colorectal cancer with 11% of colorectal cancer cases in Europe linked to being overweight (18, 19). According to the 2018 WCRF/AICR report (20), each 5 kg/m2 increase in BMI was associated with a 5% increase in colorectal cancer risk (RR = 1.05, 95%C.I.: 1.03 to 1.07). Diabetes mellitus is also widely believed to be involved in the development of colorectal cancer. In a pooled analysis of 19 prospective population-based cohorts in East and South Asia, Chen et al. (21) found a 41% increased risk of colorectal cancer in patients with diabetes (HR = 1.41, 95%C.I.: 1.26–1.57). The China Kadoorie Biobank (CKB) study, which included a follow-up study of 500,000 participant, found a 44% increased risk of colorectal cancer among screen-detected diabetics (HR = 1.44, 95%C.I.: 1.18–1.77) (22). In addition, several factors considered to be associated with colorectal cancer were not added to the scoring system. Ulcerative colitis (23), red and processed meat intake (20, 24) and excessive alcohol consumption (20, 25) are considered risk factors for colorectal cancer. Regularly aspirin taking (26, 27), dietary fiber intake (20, 28) and reasonable physical activity (20, 29) have been shown to reduce the risk of colorectal cancer. Further improvements of the scoring system by continuing to add risk factors may improve the predictive power of CAN, and which factors need be added need to be explored in future research.
To identify high-risk individuals of colorectal cancer accurately, several recent studies had attempted to combine risk-scoring systems with laboratory test results (30–32). Since FIT is the most widely used stool-based test and has convenience, speed and economic advantages, it is of course the first choice for the combination of risk scoring system. Chen et al. (30) conducted a randomized controlled trial in Chinese population, and suggested that the individualized screening strategy combining the modified APCS score and FIT could ensure a higher screening participation rate, and the detection rate of CAN was higher than that of FIT alone. Sekiguchi et al. (31) combined the 8-point risk score with FIT and found that the sensitivity of CAN diagnosis was improved compared with that of FIT alone. Park et al. (32) used fecal hemoglobin (f-Hb) concentration as one of the risk factors to establish a scoring system for CAN, with a c-statistic of 0.75(0.73–0.78). The combined application of risk scoring system and laboratory test results may become the development trend of colorectal cancer screening.
This study has several strengths. First, to our knowledge, this is the first cross-sectional study that validated the performance of the APCS score and its modified versions in asymptomatic population in China. Second, this study was conducted under the framework of CanSPUC, which used rigorous standards to guarantee the integrity and accuracy of the collected data, including a review mechanism to ensure the quality of data and development of a data system to monitor all the processes of the study. Thirdly, we used several evaluation indices to evaluate the prediction ability of the scoring systems from various aspects.
This study also has several limitations. First, for practical reasons, only the colorectal cancer screening data of asymptomatic population in Xuzhou were used in this study. Second, although cluster sampling was used during the population recruitment, participation in questionnaire and colonoscopy was voluntary, which may lead to selection bias. In addition, only participants who met the high-risk conditions for colorectal cancer were recommended to undergo colonoscopy examination due to limited resources when CanSPUC was conducted. This leads to a higher high-risk rate in the risk stratification results of this study.
In summary, in this external validation, the APCS score and its modified versions have certain ability to predict the risk of advanced neoplasia and reduce the resource load. The modified APCS score and the KCS score seemed the preferable systems to classify high risk subjects based on its high RR and sensitivity in the selected population. Future research could focus on adding risk factors or combining with laboratory test results to improve the predictive power of the scoring system.