A total of 8382 patients were analyzed (as shown in Fig. 1). The mean age of the patients was 61 (51 to 68) years old; 5348 (63.8%) were male; 6504 (77.6%) were local residents; 3060 (36.5%) had a smoking habit. And there were 4338 (51.8%) patients with coronary heart diseases (CHDs); 3048 (36.4%) patients with acute coronary syndrome (ACS); 5117 (61.0%) patients with hypertension; 976 (11.6%) patients with CKD, and 1117 (13.3%) patients with stroke.

A flow diagram of the study

As is shown in Table 1, the baseline LDL-c concentration was 3.16 (2.53 to 3.82) mmol/L, HDL-c concentration was 1.09 (0.93 to 1.30) mmol/L, TC concentration was 4.82 (4.02 to 5.70) mmol/L and TG concentration was 1.43 (1.01 to 2.04) mmol/L. After discharge, 6740 (84.4%) patients were taking statins, 4531 (56.7%) taking aspirin, 4132 (51.7%) taking clopidogrel, 4905 (61.4%) taking ACEIs/ ARBs, 4611 (57.7%) taking β-blockers and 2643 (33.1%) taking CCBs. Patients with higher Lp(a) levels were female, older, and had more hypertension, CHDs, ACS, stroke, CKD (P < 0.05), and higher CCIs, LDL-c and TC (Table 1).

During an average follow-up of 619 (320 to 1061) days, the rate of loss to follow-up was 23.7%. And 1361 (16.2%) MACEs, 124 (1.5%) myocardial infarction (MI), 418 (5.0%) unstable angina (UA), 585 (7.0%) cases of heart failure (HF), 291 (3.5%) stoke and 125 (1.5%) all-cause death were collected. The clinical outcomes are shown in Table 2.

Kaplan–Meier analysis (Fig. 2) exhibited that increased serum Lp(a) concentrations was significantly associated with the occurrence of MACEs (log-rank X2 = 25.767, P < 0.001) and all-cause deaths (log-rank X2 = 15.016, P = 0.002) in patients with cardiovascular diseases.

Kaplan–Meier curves according to quartiles of serum Lp(a) levels. A Major adverse cardiovascular events (MACEs); and B all-cause death. Lp(a) quartiles: Q1 ≤ 80.00 mg/L; Q2: 80.01 ~ 160.90 mg/L; Q3: 160.91 ~ 336.41 mg/L; and Q4: > 336.41 mg/L

As shown in Table 3, the incidence of MACEs was 7.65, 8.24, 9.73 and 10.75 per 100 person-years in each Lp(a) quartile, respectively; the all-cause mortality was 0.48, 0.69, 0.64 and 1.18 per 100 person-years in each Lp(a) quartile, respectively. The incidences of both two events were found to correlate with the increase in Lp(a).

When Lp(a) was treated as a continuous variable, the incidence of MACEs would increased by 31.6% (HR: 1.316, [95% CI: 1.162 to 1.491], P < 0.001) for every 1 mg/L increase in Lp(a). For all-cause deaths, mortality was increased by 91.3% for every 1 mg/L increase in Lp(a) (HR: 1.913, [95% CI: 1.255 to 2.916], P = 0.003). After adjusting the age, sex, hypertension, history of CHDs, CKD, stroke, CCI, statins, HDL-c, TC, and LDL-c, the trends had not changed.

According to Table 3, when Lp(a) was used as a categorical variable, compared with group Q1, the unadjusted HR of group Q2, Q3 and Q4 in MACEs was 1.086 (95% CI: 0.921 to1.280, P = 0.325) and 1.287 (95% CI:1.099 to1.507, P = 0.002) and 1.424 (95% CI: 1.221 to 1.662, P < 0.001), respectively. Similarly, compared with group Q1, the unadjusted HR for groups Q2, Q3 and Q4 in all-cause death was 1.446 (95% CI: 0.800 to 1.365, P = 0.223) and 1.335 (95% CI:0.735 to 2.355, P = 0.343) and 2.464 (95% CI: 1.439 to 4.219, P = 0.001), respectively. After adjusting the age, sex, hypertension, history of CHDs, CKD, stroke, CCI, statins, HDL-c, TC, and LDL-c, the trends were similar to before, indicated that the risk of MACEs was increased when Lp(a) > 160.90 mg/L, and the risk of all-cause death was increased when Lp(a) > 336.41 mg/L.

In addition, after adjusting the age, sex, hypertension, history of CHDs, CKD, stroke, CCI, statins, HDL-c, TC, and LDL-c, multivariate COX regression analysis suggested that male, CKD, stroke, CCI and TC were independent risk factors for MACEs. And there was no significant correlations between MACEs and age, smoking, place of residence, hypertension and CHD. For all-cause death, hypertension, CKD, stroke, CCI were independent risk factors. And statins were effective protective factors in both two events. After adding an interaction term between Lp(a) and statin into the Cox model for the overall population, the result suggested that there was no interaction term between Lp(a) and statins (HR: 0.824, [95% CI: 0.590 to 1.152], P = 0.258) (Tables 4, 5).

The ROC curves are shown in Fig. 2. For MACEs, the AUC of serum Lp(a) concentrations [0.559, (95% CI: 0.542 to 0.575), P < 0.001] was greater than LDL-c [0.485, (95% CI: 0.467 to 0.502), P = 0.073]. When combined the Lp(a) with LDL-c, the combination AUC of Lp(a) and LDL-c [0.556, (95% CI: 0.539 to 0.574), P < 0.001] did not show any advantages than Lp(a) (Fig. 3A). For all-cause death in Fig. 3B, the AUC of serum Lp(a) concentrations 0.600 [0.600, (95% CI: 0.549 to 0.651), P < 0.001] was greater than LDL-c [0.462, (95% CI: 0.142 to 0.406), P = 0.142]. The combination AUC of Lp(a) and LDL-c [0.611, (95% CI: 0.563 to 0.658), P < 0.001] had shown remarkable advantages than Lp(a) and LDL-c alone.

The receiver operating characteristic (ROC) curve. A MACEs, B all-cause death