Table 1 Predictive models based on conventional clinical data in the last decade
Name | Published Time | Major Tool | Relevant Factors | Evaluation |
|---|---|---|---|---|
Yang’s Model(6) | 2018-03 | Ultrasound | ✓ Age ✓ Number (single/ multiple) ✓ Sessile/ pedunculated ✓ Polyp size | ✓ Predictive score (PS) = − 7.3633 + 0.0374 × [Age] + 0.6667 × [Number] + 1.5784 × [Sessile] + 0.2189 × [Size]. Probability of neoplastic polyp = ePS / (1 + ePS), where e = 2.7182. ✓ The AUC was 0.83 and 0.90 in the modeling group and validation group, respectively. When the cut-off value of the neoplastic probability was 7.4%, the sensitivity of the model was 78.5% and the specificity was 77.5%. In the entire cohort, only 2 individuals (6.7%) with malignant polyp were missed with this cut-off value. ✓ It proved the model could be useful in clinical practice to predict neoplastic potential of gallbladder polyp more accurately than only considering each risk factor of neoplastic gallbladder polyp. |
Wennmacker’s Model [9] | 2018-09 | Pathology report | ✓ Polyp size ✓ Number ✓ Wall thickening ✓ Protruding polyp ✓ Presence of gallstones | ✓ The decision tree using the surgical threshold data and clinicopathological characteristics of neoplastic and non-neoplastic polyps was established, which results in the prediction for each of the 16 possible combinations of clinicopathological characteristics. ✓ The AUC was 0.75. The curve showed that 1 cm is the most optimal size threshold for differentiating neoplastic and non-neoplastic polyps. Sensitivity of the surgical threshold for indicating neoplastic polyps was 68.1% and specificity was 70.2%. |
Chen’s Model [13] | 2019-10 | Contrast-enhanced computed tomography | ✓ Size ✓ Stone ✓ Mucosal smoothness (smooth/ irregular) ✓ Layered pattern of gallbladder wall on portal vein phase ✓ Gallbladder wall enhanced ✓ ∆CT value of mass (portal phase – delayed phase) ✓ Age ✓ CA199 | ✓ The nomogram was established by 6 radiological features and 2 clinical factors. ✓ The AUC in the internal and external validation cohorts were up to 0.91 and 0.89, respectively. it also demonstrated superior sensitivity (95.6%) and accuracy (95.2%) in the diagnosis of GBC in the training cohort. Most of the gallbladder polyps, which were misdiagnosed as benign lesions, were successfully identified using this nomogram. ✓ The nomogram added significant strength for early detection of malignancy in the gallbladder, especially for T1-2 tumors. |
Kim E’s Model [7] | 2020-08 | Bile | ✓ Age ✓ Bile viscosity ✓ Bile cholesterol | ✓ A predictive scoring model was developed for polypoid lesions of the gallbladder larger than 1 cm to distinguish adenomatous polyps from cholesterol ones. ✓ The AUC was 0.845. The model had a sensitivity of 90.9% and a specificity of 80.2% at a cutoff of ≥ 6 points. The performance of the model was superior to Wennmacker’s Model [7]. ✓ The process of obtaining bile acids is invasive. Moreover, this study only explored the differences between adenoma and cholesterol polyps, but did not explore the differences between other non-cholesterol benign polyps such as hyperplastic polyps and inflammatory polyps. |
Zhang D’s Model [10] | 2021-06 | Ultrasound | ✓ Number of polyps ✓ Fundus (pedicle/ broad base) ✓ Echogenicity ✓ Polyp size (long diameter) ✓ Polyp size (short diameter) | ✓ The nomogram prediction model for gallbladder polyps with malignant tendency with a long diameter of 10–15 mm was constructed, which is available at https://docliqi.shinyapps.io/dynnom/. ✓ The consistency index of the model was 0.778 and the internal validation was 0.768. |
Zhang X’s 2021 Model [43] | 2021-06 | Ultrasound | ✓ Age ✓ Cholelithiasis ✓ CEA ✓ Polyp size ✓ Sessile | ✓ The formula for neoplastic risk in patients with gallbladder polyps was: Y = 1.194 × [age] + 1.177 × [cholelithiasis] + 1.171 × [CEA] + 1.112 × [polyp size] + 1.066 × [sessile] − 3.944. ✓ The AUC was 0.846 and 0.835 in the training and validation cohorts, respectively. The nomogram achieved an overall accuracy rate of 84.1%, with a sensitivity of 68.1% and a specificity of 88.2%. ✓ Compared with Yang’s model(6) and three different management guidelines(JSHBPS, ESGAR and CCBS) at that time, the nomogram achieved significantly better diagnostic performance and provided more clinical benefit. |
Ma’s Model [8] | 2022-01 | Ultrasound | ✓ Cross-sectional area ✓ Positive blood flow ✓ Age ✓ ALT ✓ ALT/AST | ✓ The scoring model for predicting true polyps was established, and a new reference parameter, the cross-sectional area of a gallbladder polyp, was innovatively introduced. ✓ The AUC was 0.883. A total score of 6.5 was the optimal cut-off value for distinguishing between true polyps and pseudo-polyps, with a sensitivity of 72.7% and a specificity of 89.6% in the reference group. |
Zhang X’s 2022 Model [14] | 2022-03 | Ultrasound | ✓ Age ✓ Polyp size ✓ CEA ✓ Gallstone ✓ Sessile shape | ✓ The formula for the prediction model was: Y = 1.084 × [age] + 0.937 × [polyp size] + 1.465 × [CEA] + 0.927 × [gallstone] + 0.862 × [sessile] – 4.236. ✓ The nomogram achieved an overall accuracy rate of 86.3% with a sensitivity of 69.5%, a specificity of 90.7%. The model yielded the AUC of 0.845 in the validation cohort. ✓ The model showed better diagnostic performance than Yang’s model(6) and three guidelines(JSHBPS, ESGAR and CCBS) at that time. ✓ Limited by the number of ultrasound images, the minimum caliper diameter was not obtained in the model, despite the research proved that it was the important independent predictor for malignant gallbladder polypoid lesions. |
Liu’s Model [12] | 2022-06 | Ultrasound | ✓ Number of polyps ✓ Maximal diameter ✓ Shape (irregular/regular) | ✓ The regression equation was logit(P) = -3.828 + 1.083 × number of GPLs + 0.218 × diameter of GPLs + 1.714 × shape of GPLs. ✓ AUC was 0.828. When logit P > 0.204, the sensitivity of estimating adenomatous polyps was 79.5%, the specificity was 70.6% and the whole correct ratio was 73.3%. ✓ The model reduced confounding factors in diagnosing adenomas, and its prediction efficiency is better than Wennmacker’s Model(7). |
Li’s Model [11] | 2022-08 | Ultrasound | ✓ Age ✓ Number of polyps ✓ Polyp size (long diameter) ✓ Polyp size (short diameter) ✓ Fundus | ✓ A Bayesian network prediction model was available at https://simulator.bayesialab.com/#!simulator/204709691197. ✓ The AUC was 77.38% and 75.13%, and the model accuracy was 75.58% and 80.47% for the Bayesian network model in the training set and testing set, respectively. ✓ The model was accurate and practical for predicting gallbladder polyps with malignant potential patients in a long diameter of 8-15 mm. ✓ The model took not only the long diameter of polyp size, but also the short diameter into consideration. |