Baseline characteristics

In our study, 601 patients were included in the full analysis set. During the follow-up, there were 8 missing data, 3 missing samples, and 11 missing patients, but they showed no influence on the statistical results. Finally, 579 T2DM patients were enrolled in this work, including 369 males (63.7%) and 210 females (36.3%). All subjects were diagnosed with T2DM and the longest follow-up period was 48 months[19, (10 ~ 32) ]. During the follow-up period, there were 15 cases with cardiovascular or cerebrovascular death, 76 cases with acute coronary syndrome, 69 cases with coronary stent implantation, and 17 cases with stroke. The patients were divided into a positive group and a negative group according to whether the predefined MACCE occurred during follow-up. Compared with MACCE negative group, the proportion of anti-hypertensive therapy (P = 0.029), lipid-lowering therapy (P = 0.028), the levels of NGAL (P = 0.000), sdlDL-C (P = 0.000), HbA1c (P = 0.001), LDL-C (P = 0.000), ApoB (P = 0.002) in the positive group were significantly higher; and ApoA–I (P = 0.002) levels were lower. No significant differences were found in age (P = 0.907), gender (P = 0.653), body mass index (BMI) (P = 0.988), smoking or not (P = 0.886), hypertension (P = 0.757), dyslipidemia (P = 0.892), SBP (P = 0.055), DBP (P = 0.572), fasting plasma glucose (FPG) (P = 0.304), insulin (P = 0.513), TG (P = 0.728), TC (P = 0.959), HDL-C (P = 0.782), Lp(a) (P = 0.183), hsCRP (P = 0.970), neutrophils (P = 0.206), Scr (P = 0.931), and UA (P = 0.938) between two groups (Table 1). In this study, oral hypoglycemic agents included insulin (P = 0.131), other AHAs (P = 0.7866), metformin (P = 0.477), sulfonylurea (P = 0.619), thiazolidinedione (P = 0.852), glinides (P = 0.481), glitazones (P = 0.729), α-glucosidase inhibitor (P = 0.934), DDP-4i (P = 0.757), we have made a corresponding analysis on the use of these drugs and their impact on NGAL and sdlLDL, and outcomes of interest. Unfortunately, these drugs have no impact on NGAL and sdlLDL and their corresponding results (P > 0.05 respectively). In general, the baseline characteristics were well-matched between the two groups.

Table 1 Clinical, biochemical, and angiographic characteristics of study subjects in different groups Correlations of NGAL and sdlDL-C

Correlations of NGAL and sdlDL-C with other variables were searched across the whole study by Spearman’s rank correlation analysis. It was found that the NGAL was significantly positively correlated with BMI (r = 0.391, p = 0.001), TG (r = 0.228, p = 0.032), hsCRP (r = 0.251, p = 0.007), and neutrophils (r = 0.454, p = 0.001). On the other hand, sdlDL–C level was positively correlated with LDL-C (r = 0.413, p = 0.001), TG (r = 0.432, p = 0.001), and ApoB (r = 0.232, p = 0.002), and negatively correlated with HDL-C (r = -0.362, p = 0.031) and ApoA–I (r = -0.402, p = 0.001). The specific results were given in Table 2.

Table 2 Relationship between serum NGAL and sdlDL-C levels and other variables

A total of 25 independent variables were included. As the correlation coefficient analysis among independent variables was conducted before, it was found that there was a certain correlation between different independent variables, so the dimension was reduced, the most representative high-risk predictors were screened, and LASSO regression analysis was conducted for all independent variables. With penalty coefficient λ, the coefficients of the independent variables initially included in the model are gradually compressed, and the last part of the independent variable coefficients are compressed to 0, avoiding over fitting of the model. Using 10 times cross validation of minimum criterion to identify the optimal penalty coefficient in LASSO regression model λ, When λ value continues to increase to 1 standard error, The λ is the optimal value of the model, and the final independent variables were ApoA-i, APOB, HbA1c, NGAL, sdLDL-c (Fig. S1A, B).

Next, MACCE and follow-up time were the dependent variables, while age, BMI, smoking, hypertension, dyslipidemia, SBP, DBP, FPG, HbA1c, insulin, TG, TC, HDL-C, LDL-C, sdIDL-C, ApoA-I, ApoB, Lp(a), hsCRP, neutrophils, Scr, UA, and NGAL were set as independent variables. The results showed that high levels of HbA1c (HR = 1.112, 95% CI: 1.006–1.228, P = 0.038), sdIDL-C (HR = 1.052, 95% CI: 1.037–1.066, P < 0.001), NGAL (HR = 1.006, 95% CI: 1.005–1.008, P < 0.001) were independent risk factors for MACCE in T2DM patients aged ≥ 65 years (Table 3).

Table 3 Multivariate logistic regression analysis results of factors independently associated with the occurrence of MACCE of elder T2DM patients (HR, 95% CI)

We re-validate the robustness of the model, that was the internal validation, the calibration curves of the Cox model were drawn. The results showed that the predicted probability curve of the model fitted well with the reference probability, suggesting that the model had high accuracy(Fig S2).

Predictive value of NGAL and sdlDL-C

ROC curve analysis showed that both NGAL (AUC = 0.79, 95% CI: 0.75–0.84, P < 0.001) and sdlDL-C (AUC = 0.76, 95% CI: 0.72–0.80, P < 0.001) could predict the occurrence of MACCE in T2DM patients aged ≥ 65 years. Moreover, the combination of NGAL and sdlDL-C improved the prediction capability (AUC = 0.87, 95% CI: 0.84–0.90, P < 0.001). The NGAL level of 131.9 ug/L and sdlDL-C level of 32.35 mg/dl were determined as the best cut-off points to predict the risk of MACCE of T2DM patients aged ≥ 65 years, with a sensitivity of 74.01% and 69.49% and a specificity of 78.86% and 73.38%, respectively (Fig. 1; Table 4).

Fig. 1

Receiver operating characteristic (ROC) analysis on the predictive capacity of the created multivariable Cox regression model for the occurrence of CAD in patients with T2DM aged ≥ 65 years. ROC analysis on the predictive capacity of the NGAL (AUC = 0.79, 95% CI: 0.75–0.84, P < 0.001) and sdlDL-C (AUC = 0.76, 95% CI: 0.72–0.80, P < 0.001) for the identification of the hazard for the primary composite outcome of major adverse cardiovascular or cerebrovascular events (MACCE: cardiovascular or cerebrovascular death, acute coronary syndrome, coronary stent implantation, and stroke), with a sensitivity of 74.01% and 69.49% and a specificity of 78.86% and 73.38%, respectively. Moreover, the combination of NGAL and sdlDL-C improved the prediction capability (AUC = 0.87, 95% CI: 0.84–0.90, P < 0.001)

Table 4 ROC analysis on the predictive capacity of NGAL and sdlDL-C for the occurrence of MACCE in T2DM patients aged ≥ 65 years