Comparison of GPs patients in young, middle-aged and elderly groups
As age increased, the proportion of neoplastic polyps increased, reaching 7.3% in young adults, 7.7% in middle-aged and 11.9% in elderly groups, of which the proportion of malignant polyps was 0.2%, 0.9% and 5.1%, respectively (Fig 1-A, P<0.05). Interestingly, there were no malignant polyps in GPs patients with long diameter of 6-9 mm, and all the malignant polyps were in long diameter of 10-20 mm. Likewise, the proportion of neoplastic polyps were significantly higher in the elderly group than in the young adults and middle-aged groups (Fig 1-B, P>0.05 and Fig 1-C, P<0.05). Besides, the mean values of long diameter (Fig 1-D, P<0.05) and short diameter (Fig 1-E, P>0.05) in the elderly group were higher than those in the young adults and middle-aged groups. The mean long diameter of neoplastic polyps in the elderly group was also higher than that in the young adults and middle-aged groups (Fig 1-F, P<0.05).
By comparing the characteristics of GPs in different age groups, Sex, CEA, CA-125, number of polyps, polyp size (long diameter), polyp size (short diameter), fundus, thickness of gallbladder wall, and echogenicity were statistically different (P<0.05). Therefore, there were differences among different age groups in clinical and preoperative ultrasound features of GPs. Details are shown in Table 1.
Univariate and multivariate analysis of neoplastic polyps
According to the ROC curve analysis (Fig 2 A-C), we determined the best cut-off values of long- and short diameter of GPs were 10.5 mm and 8.0 mm in different age groups, respectively. Comparison of non-neoplastic polyps and neoplastic polyps in different age groups is shown in Table 2. Univariate analysis showed that number of polyps, polyp size (long diameter), polyp size (short diameter) and funds were associated with neoplastic polyps in different age groups (P<0.05). Multivariate analysis showed that number of polyps (single), polyp size (long diameter≥10.5 mm), and fundus (broad base) were the independent risk factors of neoplastic polyps in young adults and elderly groups, number of polyps (single), polyp size (long diameter≥10.5 mm), and polyp size (short diameter≥8 mm) were the independent risk factors of neoplastic polyps in middle-aged groups.
Linear scoring models development for different age groups
Based on the independent risk factors for neoplastic polyps identified by the logistic regression model for different age groups, the linear scoring models were developed (Table 3). For young adults group, the total score of its linear model = number of polyps (single assigned 4 points and multiple assigned 0 point) + polyp size [(long diameter), >10.5 mm assigned 4 points and ≤10.5 mm assigned 0 point] + fundus (broad base assigned 3 points and pedicle assigned 0 point). Similarly, for middle-aged group, the total score of its linear model = number of polyps (single assigned 4 points and multiple assigned 0 point) + polyp size [(long diameter), >10.5 mm assigned 6 points and ≤10.5 mm assigned 0 point] + polyp size [(short diameter), >8 mm assigned 4 points and ≤8 mm assigned 0 point]; for elderly group, the total score of its linear model = number of polyps (single assigned 8 points and multiple assigned 0 point) + polyp size [(long diameter), >10.5 mm assigned 6 points and ≤10.5 mm assigned 0 point] + fundus (broad base assigned 9 points and pedicle assigned 0 point).
According to the ROC curve analysis (Fig 3 A-C), we determined a total score of 5.5, 5.0, and 14.5 points as the best cut-off values for differentiating neoplastic and non-neoplastic polyps, and the total score ≤ 5.5 was defined as low-risk group and >5.5 as high risk group for neoplastic polyps in young adults group; the total score ≤5 was defined as low-risk group and >5 as high risk group for neoplastic polyps in middle-aged group; the total score ≤ 14.5 was defined as low-risk group and >14.5 as high risk group for neoplastic polyps in elderly group.
Compared with long diameter ≤10.5 mm and short diameter ≤8.0 mm, the proportion of neoplastic polyps in the low-risk group of the linear models decreased from 3.7% (17/451), 6.0% (34/564) to 2.6% (13/500) in young adults group; 2.9% (15/508), 4.8% (26/542), to 2.0% (10/504) in middle-aged group; 6.0% (10/167); 7.6% (15/198) to 4.1% (8/197) in elderly group. Conversely, compared with long diameter > 10.5 mm and short diameter >8.0 mm and the proportion of neoplastic polyps in the high-risk group of the linear models increased from 16.4% (28/171), 18.9% (11/58) to 26.2% (32/122) in young adults group; 22.9% (36/157), 20.3% (25/123) to 25.5% (41/161) in middle-aged group; 26.1% (18/69), 34.2% (13/38) to 51.3% (20/39) in elderly group (Fig 4 A-C). Thus, cholecystectomy should be recommended for GP patients at high risk of neoplastic polyps.
Analysis of predictive ability of linear scoring models and long- and short diameter GPs
In young adults group, the AUC of its linear scoring model was 0.794 (95%CI: 0.734~0.853), which was higher than the AUCs of the long diameter (AUC: 0.712, 95%CI: 0.624~0.800) and short diameter (AUC: 0.654, 95%CI: 0.562~0.746) of GPs for differentiating neoplastic and non-neoplastic polyps (all P<0.05). Similarly, in the middle-aged group, the AUC of its linear scoring model was 0.857 (95%CI: 0.805~0.908), which was higher than the AUCs of the long diameter (AUC: 0.802, 95%CI: 0.738~0.867) and short diameter (AUC: 0.740, 95%CI: 0.657~0.823) of GPs (all P<0.05); in elderly group, the AUC of its linear scoring model was 0.849 (95%CI: 0.784~0.914), which was higher than the AUCs of the long diameter (AUC: 0.714, 95%CI: 0.593~0.835) and short diameter (AUC: 0.693, 95%CI: 0.564~0.822) of GPs (all P<0.05).
Hosmer-Lemeshow goodness of fit test indicated that it was more acceptable for the predictive probabilities of the linear scoring models than the long- and short diameter of GPs to fit the actual probabilities (all P>0.05), which showed that the prediction accuracy of the linear scoring models in different age groups was higher than the long- and short diameter of GPs for predicting the neoplastic polyps. Details are shown in Table 4.