Colonoscopy is the gold standard for the screening and prevention of colorectal cancer, significantly reducing colorectal cancer-related morbidity and mortality [12]. Consequently, there have been substantial efforts to enhance colonoscopy training. Various quality assessment metrics have been proposed to evaluate the effectiveness of training programs, including the most widely used adenoma detection rate (ADR), and the polyp detection rate (PDR) [3456]. However, these indicators focus more on lesion detection rates, than on the ability to differentiate between lesions that require resection and those that do not. As a result, they fail to assess cognitive and decision-making skills that are essential for high-quality colonoscopies [78]. Accurate lesion discrimination is critical for preventing unnecessary resections while ensuring the detection of clinically significant adenomas; however, no objective metric has been established to measure this skill, highlighting a gap in colonoscopy training evaluation.

Adenoma per polypectomy (APP), defined as the number of histologically confirmed adenomas per total polypectomy, has been proposed as an indicator of visual lesion discrimination ability [9]. This study is the first to introduce APP as a structured measure of lesion discrimination ability among trainees. By analyzing APP progression across 3 training phases, this study evaluated its potential as a novel indicator of lesion discrimination ability in colonoscopy training.

From March 2020 to February 2023, 8,072 colonoscopies were performed by 9 trainees; and 11,687 colonoscopies were performed by 5 experts at the National Cancer Center. The patient demographics and procedural characteristics are summarized in Table 1. Except for age, the patient characteristics of the trainee and expert groups showed significant differences.

Table 2 presents the colonoscopy results for both groups. The cecal intubation time was significantly longer in the trainee group than in the expert group (7 minutes vs. 5 minutes, P < 0.001). Despite the longer observation time in the trainee group (13 minutes vs. 11 minutes, P < 0.001), both the PDR and ADR were significantly lower than those in the expert group (PDR: 57.3% vs. 63.3%, P < 0.001; ADR: 45.4% vs. 53.3%, P < 0.001). Furthermore, the APP in the trainee group was significantly lower than that in the expert group (67.0% vs. 73.9%, P < 0.001).

Fig. 1 illustrates the changes in the number of resected polyps and APP for every 100 colonoscopies performed by the 9 trainees. From the beginning of training, at up to 200 colonoscopies, both the number of resected polyps and APP increased simultaneously. Starting from 300 colonoscopies, the number of resected polyps increased sharply compared to the previous phase, whereas the APP declined. This trend continued until 500 colonoscopies were performed, at which point the number of resected polyps remained elevated. However, upon reaching 600 colonoscopies, the number of resected polyps markedly decreased, whereas the APP showed a slight increase compared to the previous phase.

Based on Fig. 1, the training phases were categorized into 3 stages: phase 1 (from the beginning of training to 200 colonoscopies), phase 2 (200–500 colonoscopies), and phase 3 (>500 colonoscopies). Table 3 presents a comparison of the colonoscopy outcomes between trainees and expert endoscopists in each phase. The PDR gradually increased with training and, in phase 3, showed no statistically significant difference compared to the expert group (59.8% vs. 63.3%, P = 0.062). However, the ADR also increased progressively throughout the training but remained significantly lower than that of the experts, even in phase 3 (48.3% vs. 53.3%, P = 0.003). The APP showed a decline in phase 2 and then increased again in phase 3, reaching a higher value than that in phase 1; however, it still showed a statistically significant difference compared with the expert group (69.9% vs. 73.9%, P = 0.027).

Table 4 presents the individual colonoscopy results of the 9 trainees. The PDR, ADR, and APP showed statistically significant differences among the trainees. In addition, none of the 9 trainees achieved an expert APP value of 73.9%.

This study introduced the concept of APP to assess changes in the visual discrimination ability of lesions during colonoscopy. An assessment of the APP progression over time revealed a pattern that could be categorized into 3 distinct phases (Fig. 1, Table 3). Phase 1 represents the familiarization stage with colonoscopy, during which the APP was relatively low, and the number of resected polyps was small. This phase suggests limited polyp detection, leading to a lower PDR and ADR. This pattern was observed from the beginning of training to approximately 200 colonoscopies. Phase 2 marks the stage in which trainees become accustomed to colonoscopy, leading to an increase in polyp detection. As a result, the number of resected polyps increased, whereas the APP declined. The PDR and ADR improved during this phase, which persisted until approximately 500 colonoscopies were performed. In phase 3, the number of resected polyps decreased, compared with that in the previous phase, whereas the APP increased. Furthermore, the PDR and ADR continued to increase during this stage. These findings suggest that, while trainees had improved in detecting polyps, they also developed the ability to assess polyp morphology and decide whether resection was necessary, leading to a reduction in the total number of resected polyps and an increase in APP. However, despite the improvement in the APP, the trainees did not reach the expert level, indicating that visual discrimination skills for polyp assessment should continue to develop even after the training period.

Various lesions can develop in the colon; however, not all detected lesions are pathological. Therefore, efforts have been made to effectively differentiate these lesions. The most well-known classification systems include the Kudo classification, which utilizes pit patterns; the Japan NBI Expert Team (JNET) classification, based on capillary patterns; and the Paris classification [78]. These classification systems help endoscopists to distinguish between lesions and determine whether resection is necessary. Although it is possible to resect all detected lesions, doing so presents several challenges, including increased patient discomfort due to a prolonged procedure time, a higher risk of bleeding and perforation from excessive resection, increased physician fatigue, and elevated procedural costs [111213]. Therefore, endoscopists are required to resect lesions that are deemed necessary, accurately and selectively. Lesion discrimination is an essential skill for endoscopists, and endoscopic training programs should incorporate methods to teach and refine this ability. However, to date, there is no direct metric for objectively assessing lesion discrimination skills. Previous studies have primarily focused on the ADR and PDR as key training metrics, with the assumption that higher detection rates reflect improved endoscopic skills [3456]. However, these metrics fail to account for the decision-making process involved in lesion resection, a crucial aspect of colonoscopic proficiency. Our findings suggest that the APP may provide additional insights into how trainees develop lesion discrimination skills over time. Unlike the ADR, which primarily measures detection ability, the APP reflects a trainee’s ability to determine whether a detected lesion warrants resection. This distinction highlights the APP as a complementary training metric that may enhance the assessment of colonoscopy training along with the ADR.

This study included 9 trainees, with 3 enrolled each year over a 3-year period. They underwent the same training using a standardized 1-year colonoscopy training system, which was structured based on previous research [10]. In the early stages of training, experts provided supervision and guidance throughout the colonoscopy procedure; however, as the training progressed, greater autonomy was granted to the trainees, with supervisors intervening only when necessary. To enhance endoscopy knowledge, structured educational sessions were conducted weekly, including lectures from experts, trainee-led topic presentations, and journal reviews. This training approach ensured adequate quality control in colonoscopy, as evidenced by an ADR of 41.9% in phase 1, which exceeded the recommended minimum ADR threshold of 25%, even in the early stages of training [1415]. Additionally, in phase 3, when supervisors were less involved, the ADR further increased to 48.3%, demonstrating the effectiveness of the training program. Similarly, the APP was the highest in phase 3, where expert intervention was relatively minimal, suggesting that trainees developed independent lesion discrimination skills through structured education. These findings highlight the importance of incorporating hands-on colonoscopy practice and structured learning from experts and self-directed studies through textbooks and journal reviews in colonoscopy training programs.

The study findings have several implications. First, acquiring the ability to discriminate between lesions requires extensive experience. Considering that this ability is essential for avoiding unnecessary resections, phase 3, in which trainees maintained an adequate level of ADR while performing fewer polypectomies and achieving a higher APP, can be interpreted as the stage at which lesion discrimination skills were acquired. This phase was reached after >500 colonoscopies, which exceeds the relatively lower number of colonoscopies traditionally considered sufficient for colonoscopy training [161718]. Second, the APP varied among endoscopists. Some ambiguous lesions may require polypectomy for biopsy purposes, and such lesions cannot be incorrectly resected solely because they are not pathological. As a result, differences in interpretation exist among endoscopists, leading to variations in the APP. Notably, such discrepancies can also be observed among experts with well-established lesion discrimination skills, suggesting that the APP alone is insufficient as an absolute metric for evaluating endoscopists. Instead, it should be recognized as a complementary training assessment tool that can be used alongside other quality indicators, such as the ADR. Third, trainees still exhibited differences from experts even after completing colonoscopy training. In phase 3, which corresponded to the later stage of training, the ADR and APP among trainees remained lower than those of the experts (Table 3). This suggests that there is room for further technical improvement, even after completing training. Therefore, trainees should continue learning through self-directed studies and seek ongoing guidance from experts to refine their skills further. As a reference, the pathologic results of the 2 groups are summarized in Supplementary Table 1, which shows that nonspecific changes were the most common nonpathologic findings in both groups, followed by hyperplastic and inflammatory polyps. The higher proportion of nonspecific and hyperplastic polyps in trainees likely contributed to their lower APP.

In addition, this study has some limitations. First, it was a retrospective study. Given the large number of colonoscopy results, inaccurate data entry is possible. Second, this study does not establish an optimal threshold for APP. As previously discussed, the APP is influenced more by individual endoscopist tendencies than by other colonoscopy quality indicators. Furthermore, while APP may be useful in reflecting lesion discrimination skills, overemphasizing this metric could unintentionally discourage polypectomy for borderline or ambiguous lesions. This could increase the risk of missing adenoma, especially if polypectomy decisions are influenced primarily by the efforts to improve APP. Therefore, a higher APP cannot be universally regarded as favorable. To address this limitation, we provided the APP values of the experts as a reference; however, this cannot be definitively considered the optimal APP threshold. Further studies are needed to determine the appropriate APP range by analyzing data from a larger number of endoscopists. Third, selection bias should be considered as the study was conducted at a single institution with a structured training program. Future studies should validate these findings across multiple centers to enhance generalizability. Fourth, the results were derived from the average APP of trainees, which means that individual trainees may not fully reflect the overall findings of this study. Additionally, there was considerable variation in the number of colonoscopies performed by trainees over the course of 1 year. This discrepancy raises the potential concern that not all trainees undergo the same training process.

Despite these limitations, this study has several strengths. Compared to previous studies on colonoscopy training, this study included a larger number of colonoscopies performed by trainees, allowing for a more comprehensive evaluation of training outcomes, particularly in the later stages. Furthermore, this study provides a direct comparison by including colonoscopy results obtained by experts within the same time frame. This allows for an immediate evaluation of trainee performance compared with that of experts. The current 1-year colonoscopy training course has been established for 20 years, ensuring a standardized curriculum. This allowed us to assume that all trainees underwent the same educational process, which was another strength of this study.

In conclusion, the APP is a valuable metric for evaluating lesion discrimination ability in colonoscopy training. It underwent a 3-stage progression demonstrating how trainees developed their lesion assessment skills over time. This pattern of change suggests that APP could serve as an additional indicator for assessing the effectiveness of colonoscopy training.