Roel Bogie

Chapter 12

Implications for clinical practice and future perspectives Sub-classification of LSTs is a useful first step in assessing risk of submucosal invasion ( Chapter 2 ) that can be easily taught to novices using an e-learning ( Chapter 4 ). Additional classifications using pit patterns and vascularization in combination with modern visualization tools like (virtual) chromoendoscopy and magnification endoscopy will be used more frequently. 75, 76 These developments may help to improve clinical decision making with early differentiation between endoscopic and surgical treatment strategies. Artificial intelligence (AI) may become the new standard, either as a detection aid pointing to the presence of neoplasia and/or as a determination aid by predicting histology based on endoscopic appearance. 77 Currently, many difficulties and concerns exist with respect to detection and determination of LSTs and sessile serrated lesions by artificial intelligence. 78 Up to now, artificial intelligence used for assessing whether a detected lesion is premalignant or not has a high positive predictive value, but rather low negative predictive value, especially in large and sessile serrated lesions. 79 As outlined in this thesis, these neoplasms are believed to have the highest risk of developing into a PCCRC. Without improvement of the AI algorithms, by training them more specifically for LSTs and sessile serrated lesions, 80 the aim of further reducing PCCRCs will not be achieved. As shown in Chapter 3 and Chapter 5 , the efficacy of endoscopic resection should further improve. A prospective study with systematic training of endoscopists in the diagnosis and endoscopic treatment of large non-pedunculated neoplasms is on its way. The ultimate goal of all colonoscopic interventions is to prevent development of CRCs. Monitoring of PCCRC incidence is an important quality indicator of CRC screening programs. With the introduction of the national CRC screening program in the Netherlands, systematic registration of the occurrence of interval CRC was started. 81 These data will be used to monitor quality at the level of endoscopists, but these data could also been used as prospective database for studies on etiology. In the meantime, the PCCRC definition has changed. Recently a consensus statement of the World Endoscopy Organization resulted in changes in the upper time limit for PCCRCs to 10 years after a negative colonoscopy. 82 Therefore, we used an older definition as used by Pabby and all, 83 in Chapter 11 , while in Chapter 10 we applied the WEO definition. The definitions for most likely etiology have also changed. Since the sojourn time (time between occurrence of preclinical cancer and detected cancer) has been estimated between 4.5 to 5.8 years, 84 cutoff time for newly developed CRC became higher (4 instead of 3 years). This new international classification is a prerequisite for further studies and international comparison of data. As shown in Chapter 9 , commonly occurring colorectal neoplasms are the presumed precursor lesions of PCCRC. Prevention should therefore focus on detection, determination and resection of colorectal neoplasms. More detailed knowledge about specific molecular profiles of the potentially most malignant precursor lesions is welcomed. This knowledge can not only provide insight in neoplastic growth patterns which could lead to pathway specific treatment strategies, but could also assist in the diagnostic process. Because of capacity problems, costs, discomfort and potential risks of colonoscopies, alternative methods for screening and surveillance are currently being investigated. Molecular stool and blood tests are being developed that may help to detect specific molecular features of CRCs. 85, 86 Such tests could help to reduce the occurrence of interval carcinomas within the screening program since the FIT test has less sensitivity for right sided lesions and for non-polypoid lesions. 87 Within the Dutch CRC screening program fewer proximally located colorectal neoplasms are detected compared to symptomatic patients with a CRC diagnosed outside the screening program, in regular care. 88 It is important that these newmolecular stool tests

232