Development and validation of an online dynamic nomogram based on the atherogenic index of plasma to screen nonalcoholic fatty liver disease

Baseline characteristics of participants

As shown in Fig. 1, a total of 2,318 individuals were included in the study and randomly divided into a "training set" (n = 1,200) and a "validation set" (n = 1,118). The demographic and clinical characteristics of the training and validation sets are summarized in Table 1. Participants in the two sets have similar characteristics. Of the 1,200 individuals comprising the training set, 46.2% were male, with a median (interquartile range) age of 43 (31–51). In the validation set, the median age was 43 years, and 507 (45.3) were male. The prevalence of NAFLD in the training set and validation set was 22.7% and 23.0%, respectively.

Table 1 Baseline characteristics of the participants in the training set and validation setDevelopment of the nomogram

In the training set, shown in Table 2, age (OR = 1.02, 95% CI = 1.01–1.04), WC (OR = 1.09, 95% CI = 1.06–1.12), BMI (overweight vs. nonoverweight, OR = 2.26, 95% CI = 1.49–3.41), serum ALT (abnormal vs. normal, OR = 7.66, 95% CI = 3.81–15.42), and AIP (median vs. low, OR = 1.55, 95% CI = 0.88–2.71; high vs. low, OR = 2.98, 95% CI = 2.04–4.35) were positively associated with NAFLD risk. In contrast, serum AST (Abnormal vs. Normal, OR = 0.27, 95% CI = 0.09–0.83) was inversely related to the risk of NAFLD. Similar relationships were observed in the validation and NHANES sets, except for AST (OR = 0.57, 95% CI = 0.20–1.59 and OR = 1.28, 95% CI = 0.81–2.01, respectively).

Table 2 Multivariate logistic regression models in different populations

As shown in Fig. 2, the final nomogram was developed based on the six variables, including age, BMI (overweight vs. nonoverweight), WC, serum ALT (abnormal vs. normal), AST (abnormal vs. normal), and AIP (low vs. median vs. high), and was available online (https://fmumodel.shinyapps.io/NAFLD_screen_DN/). Each predictor corresponds to a specific score by finding its position on its scale and plotting a straight line to the scale above. The cumulative sum of each "point" is the "total point", which is further converted to the probability of NAFLD. For instance, a 67-year-old participant with a BMI = 21.3 kg/m2, WC = 86 cm, ALT = 12 IU/L, AST = 20 IU/L, and AIP = -0.07 had a significant probability of NAFLD of approximately 19.0% (95% CI = 12.8%-27.4%).

Fig. 2figure 2

A Nomogram developed in the training set for predicting the risk of NAFLD. B Online dynamic nomogram accessible at https://fmumodel.shinyapps.io/NAFLD_screen_DN/, depicting an example for predicting the probability of NAFLD for a 67-year-old participant, with BMI = 21.3 kg/m2, WC = 86 cm, ALT = 12 IU/L, AST = 20 IU/L, and AIP = -0.07. BMI, body mass index; WC, waist circumference; ALT, alanine transferase; AST, aspartate aminotransferase; AIP, atherogenic index of plasma; NAFLD, nonalcoholic fatty liver disease

Diagnostic performance of the nomogram

The ROC curves for the nomogram, FLI, HSI, and AIP are shown in Additional file 2. The performance of these models is detailed in Table 3 and Additional file 3. The AUROC of the nomogram in the training set (0.863, 95% CI = 0.840–0.886) was similar to that of the FLI (0.862, 95% CI = 0.838–0.886, P = 0.850) and higher than that of the HSI (0.835, 95% CI = 0.808–0.862, P = 0.019) and AIP (0.782, 95% CI = 0.752–0.811, P < 0.001). Similar significant results were observed in the validation and NHANES sets.

Table 3 Diagnostic performance of the nomogram, FLI, HSI, and AIP for predicting NAFLD in the training and validation sets

Calibration curves indicated great agreement between the probabilities predicted by the nomogram and the actual prevalence of NAFLD in the training set, showing that the nomogram provided good calibration. Good calibration of the model was also confirmed in the validation and NHANES sets (see Additional file 4).

DCA and CIC for clinical utility of the nomogram

As shown in Fig. 3A, B, and C, DCA was performed to evaluate the clinical relevance of the nomogram in the training, validation, and NHANES sets. In the training set, the nomogram, FLI, HSI, and AIP showed better net benefit than treating all and treating none from a threshold probability of < 100%, < 78%, < 79%, and < 60%, respectively. The nomogram and FLI exhibited the best performance from threshold probabilities of < 33% and > 33%, respectively. In the validation set, from a threshold probability of < 33%, we could obtain more net benefit guided by the nomogram than the referenced strategies (FLI, HSI, and AIP). For example, in the training set, at a threshold of 30%, the nomogram provided a net benefit of 12% (95% CI = 11–14), with a sensitivity of 73% (95% CI = 69–78) and specificity of 82% (95% CI = 80–84), implying that an additional 50% of NAFLD cases could be prevented (standardized net benefit) (see Additional file 5).

Fig. 3figure 3

DCA for the nomogram, FLI, HSI, and AIP for prediction of NAFLD, and CIC of the nomogram for prediction of NAFLD. A) DCA in the training set. B) DCA in the validation set. C) DCA in the NHANES set. D) CIC in the training set. E) CIC in the validation set. F) CIC in the NHANES set. DCA, decision curve analysis; CIC, clinical impact curve; FLI, fatty liver index; HSI, hepatic steatosis index; AIP, atherogenic index of plasma; NAFLD, nonalcoholic fatty liver disease

The CIC of the nomogram in the training, validation, and NHANES sets (Fig. 3D, E and F) illustrated that the nomogram possesses significant predictive value: the predicted number of high-risk patients was always greater than the number of low-risk patients within the wide and practical ranges of threshold probabilities, and the cost‒benefit ratio would be acceptable in the same range.

Subgroup analysis

Present model still shows good applicability across strata of age, sex, and presence of diabetes and hypertension (see Additional files 6, 7, and 8). Good calibration and clinical utility of the model were confirmed in participants of different ages (age < 40, 40–60 years, and > 60 years), male, female, diabetes, non-diabetes, hypertension, and non-hypertension groups.

留言 (0)

沒有登入
gif