Based on the factors influencing missed diagnoses of colorectal polyps, we established and validated a nomogram to predict the risk of missed diagnoses of colorectal polyps after a single colonoscopy. Previous studies [12] have shown that most polyps newly detected during re-examination via endoscopy within 1 year were missed in the last colonoscopy, and only 11% were newly developed or recurrent cases. Therefore, we selected patients who had undergone two colonoscopies within 3 months as participants in this study, wherein it was feasible to define new polyps found during the second colonoscopy as missed polyps.
We observed that age, endoscopist experience, bowel preparation, retroflected view, withdrawal time, number of polyps in the right colon, and number of polyps ≥ 6 mm were independent influencing factors for missed diagnoses of colorectal polyps. Shin et al. [9] reported that the risk of missed diagnoses of adenomas in patients aged ≥ 60 years was twice that in patients aged < 60 years. As people grow older, the bends and folds in their colon gradually increase, and the incidence of colonic diverticula also increases [13]. These conditions are associated with inadequate bowel preparation and result in blind areas in the endoscopic field of vision, thus increasing the risk of missed diagnoses of polyps. In this study, the risk of missed diagnoses of polyps in patients with poor bowel preparation was 2.29 times that in those with good bowel preparation. Hassan et al. [14] proposed that high-quality bowel cleansing helped the detection of adenomas in each segment of the intestinal tract, thus improving the detection rate of single adenomas and reducing missed diagnoses of adenomas. Kluge et al. [15] showed that compared with well-prepared bowel segments, underprepared bowel segments had a higher polyp detection rate during re-examination via endoscopy after a short time, suggesting that inadequate bowel preparation could increase the risk of missed diagnosis of polyps.
A study [16] on 104,618 colonoscopies conducted by 201 endoscopists showed that the detection rate of adenomas varied greatly between endoscopists, ranging from 6.3–58.7%. This indicates that different endoscopists have different degrees of missed diagnoses of adenomas during colonoscopy. Yao et al. [17] also showed that there was a significant correlation between the experience of endoscopists and missed diagnoses of adenomas. The rate of missed diagnoses is higher among inexperienced endoscopists, possibly because they are not skilled enough in endoscopic procedures, have poor exposure to the endoscopic field, or have a poor ability to identify lesions. The withdrawal time of the colonoscope is an important indicator of the quality of colonoscopy. Sufficient withdrawal time guarantees careful observation of the colonic mucosa. Previous studies [18, 19] have reported a correlation between the withdrawal time of the colonoscope and the detection rate of colorectal adenomas. The longer the withdrawal time, the higher the detection rate of adenomas. Kumar et al. [20], after controlling for factors such as the patient’s sex, age, bowel preparation, adenoma-related characteristics, and experience of the physician, compared the rate of missed diagnoses of adenomas between a withdrawal time of 6 min and a withdrawal time of 3 min. They observed that the rate of missed diagnoses of adenomas was significantly lower when the withdrawal time was 6 min than when it was 3 min. In our study, the risk of missed diagnoses of polyps in patients with a withdrawal time < 6 min was 1.54 times that in patients with a withdrawal time ≥ 6 min.
Several studies [21–23] have reported that multiple adenomas, flat adenomas, and sessile serrated adenomas/polyps are mainly located in the right colon, are small in size, covered with mucus, and not easily detected during colonoscopy. Anatomically, the folds of the right colon are deep and difficult to expand, which leads to a poor field of vision in endoscopic examination and is also a reason why adenomas of the right colon are easily missed. Additionally, the right colon is connected to the small intestine, and it is easily covered by mucus and chyme released from the small intestine, which reduces the cleanliness of the right colon and visibility of the mucosa, resulting in a missed diagnosis of adenomas. Wang et al. [24] showed that compared with the left colon, the rate of missed diagnoses of adenomas in the right colon was significantly higher, and the rate of missed diagnoses of adenomas in the caecum and ascending colon was 35.6%. The retroflected view was originally used to examine and treat rectal lesions. Later, to improve the detection rate of adenomas in the right colon, endoscopists used this technique for the examination of the right colon. According to the literature [25], using a retroflected view in colonoscopy could significantly increase the detection rate of adenomas, especially in the right colon. Miyamoto et al. [26] studied the impact of retroflected view on the detection rate of adenomas in the right colon and found that after two standard forward views, newly missed adenomas could still be found using the retroflected view. This literature suggests that there might be unavoidable blind spots in the visual field during the standard forward view. As reported by Pickhardt et al. [27] in a study on the location of missed adenomas, 14 of 15 non-rectal missed adenomas (93.3%) were located on the folds, and 10 (71.4%) were located near the plica opening. In conventional colonoscopy, it is difficult to detect adenomas in these areas by front-view observation, while a retroflected view can help effectively observe these blind areas in the field of vision [28], improve the detection rate of adenomas, and reduce the rate of missed diagnoses.
Previous studies have reported [7, 8] that smaller polyps (< 10 mm, especially ≤ 5 mm) are more likely to be missed than larger polyps (≥ 10 mm). In a study on 2,158 colorectal polyps, Lee et al. [7] showed that polyps with a diameter of < 10 mm accounted for 98.4% of all missed polyps, which is inconsistent with our study results. One possible reason for this is that their study used polyps as the unit of observation. The larger the diameter of a single polyp, the easier it is to locate. However, our study used patients as the unit of observation. Patients may have multiple adenomas. When an endoscopist performs a colonoscopy and finds polyps ≥ 6 mm, their eyes may focus on these polyps (≥ 6 mm) and miss other types of polyps (such as polyps < 5 mm, flat polyps, etc.). A higher number of polyps sized ≥ 6 mm is associated with a higher risk of a missed diagnosis of polyps. It has been reported [11] that patients with small adenomas (6−9 mm) are more likely to develop multiple adenomas than those with small adenomas (< 5 mm). Kim et al.[8] reported that patients with multiple adenomas had a higher risk of missed diagnoses of adenomas. A possible reason was that after the endoscopist detected a certain number of adenomas, they were not as focused as they had been before. They might not have accurately recorded the information on the detected adenomas because they knew that these patients would receive follow-up treatment within a short period.
The advantages of this study are as follows. First, most previous prediction models of colorectal polyps have been related to the occurrence or detection of polyps [29–31]. Most of these were scoring systems, and some models lacked validation. Compared with traditional scoring systems, the nomogram established in this study could integrate more predictive factors, graph and visualize the results of the logistic regression analysis, and intuitively evaluate the risk of missed polyp diagnosis in each patient after colonoscopy. Second, in some previous literature on the study of factors for missed diagnoses of adenomas [7, 9, 24], when analysing the influence of clinical characteristics of polyps (such as location, shape, size, etc.) on missed diagnoses of adenomas, it was a common practice to record the clinical characteristics of only one polyp in one patient. However, clinically, a patient may have multiple polyps; therefore, some polyps were not included in the statistical analysis, which might have adversely affected the results of their study. In our study, we independently analysed the influence of the number of polyps of various types, such as the number of polyps at various locations, polyps of various shapes, and polyps of different diameters, on missed diagnoses of polyps in patients. This method can simultaneously include multiple polyps in a patient for statistical analysis; thus, our results may be more reliable. Third, after completion of colonoscopy, the nomogram established in this study could quickly assess the risk of missed diagnoses of colorectal polyps in the patient and provide possible risk factors, namely, the main factors leading to the risk of missed diagnoses. Endoscopists could formulate appropriate and timely follow-up strategies for patients based on the prediction results. In the subsequent endoscopic re-examination of the patient, the endoscopist needs to pay attention to these possible risk factors and try to reduce the risk of missed polyps during colonoscopy. Fourth, the bootstrap method was used for internal validation of the established nomogram and the validation dataset was used for external validation of the nomogram. The nomogram showed good discrimination in both the training and validation sets, with C-indices of 0.765 and 0.726, respectively. A good calibration curve also proved the consistency between the predicted and actual observed values. Through calculation of the Youden index, 152.2 points were taken as the optimal cut-off value of the nomogram. Under the cut-off value, the sensitivity, specificity, positive predictive value, and negative predictive value performed well.
The limitations of this study are as follows. As this was a retrospective study, we were unable to include the risk factors affecting the occurrence and development of polyps, such as smoking, alcohol consumption, body mass index, dietary factors, and metabolic factors. Our study could not confirm whether these factors were also risk factors associated with missed diagnoses of polyps. The prediction performance of the nomogram established in this study was affected by the quality of the second colonoscopy, since the presence of new polyps detected during the second colonoscopy determined whether the patient was included in the missed diagnosis group, thus affecting the results of our study. Some new types of endoscopes that can improve the detection rate of adenomas, such as high-definition colonoscopes, wide-angle colonoscopes, panoramic endoscopes, and virtual staining endoscopes, were not used in this study, which might have affected the results of this study. Moreover, this study had a small sample size and used a single source of cases. Although there was external validation, the validation data were from the same hospital in different time periods. The transferability and generalizability of the test model were poorer than those of the external space validation. Therefore, a multi-centre and forward-looking research with a larger sample size is required for external space validation to confirm the reliability of the nomogram.