Study design and participants

This is a prospective observational study involving four University-affiliated hospitals in Shanghai, China. A total of 2990 patients with chest pain were consecutively recruited in chest pain outpatient during Monday to Friday between August 2016 and October 2017, and 2773 patients were finally admitted to wards of cardiology after prescreening by the exclusion criteria. Sample size calculation is shown in Supplementary File. Patients were eligible if they were aged ≥18 and ≤80 years, presented with signs and symptoms of chest pain and agreed to participate in this study. The exclusion criteria were pregnant women, organ transplant patients, patients suffering from bleeding disorders, patients with neoplasms with a life expectancy <1 year, chronic kidney disease (estimated glomerular filtration rate <60 mL/min/1.73 m²) and patients with a clear non-cardiac chest pain. A flowchart describing the inclusion and exclusion process is shown in figure 1. Full study protocol is provided on request to the corresponding author.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Flowchart. ACS, acute coronary syndrome; eGFR, estimated glomerular filtration rate.

The study complies with the Declaration of Helsinki and was approved by the Ethics Committee of Zhongshan Hospital, the Ethics Committee of Tongji Hospital, the Ethics Committee of Shanghai Jiao Tong University Affiliated Sixth People’s Hospital and the Ethics Committee of Minhang Hospital. Written informed consent was obtained from all participants.

Clinical and laboratory data

Clinical history, physical examination findings and blood tests including renal and liver functions, C reactive protein (CRP), D-dimer and high-sensitive cardiac troponin T (hs-cTnT; Roche) levels, ECG, coronary angiogram and echocardiogram findings were recorded in all patients. Troponin levels were measured by electrochemiluminescence method using hs-cTnT assay (Roche Diagnostics) on Roche Cobas e601. The coefficient of variation in the hs-cTnT assay is ≤10% at the cut-off value of 13 ng/L. The 99 percentile upper reference limit of hs-cTnT assay is 14 ng/L. Besides, the assay also has a limit of blank of 3 ng/L and a limit of detection of 5 ng/L, and the analytical range is 3–10 000 ng/L. Regular blood lipid tests were conducted using standard methods. Specifically, total cholesterol (TC) was measured by enzymatic cholesterol method using cholesterol oxidase/peroxidase aminophenazone reagent, while total triglyceride (TG) was measured by glycerol-3-phosphate oxidase/peroxidase anti-peroxidase method. LDL cholesterol (LDL-C) and high-density lipoprotein cholesterol (HDL-C) were measured by homogenous polyethylene glycol modification enzyme method. Other blood biomedical assay of apolipoprotein (AI, AII, B and E), lipoprotein (a) and lipoprotein-associated phospholipase were performed by immune transmission turbidity method. Performers of the above tests were blinded to the diagnosis of diseases. The detection limit, analytical range and reference interval or decision limit are summarised in online supplementary table 1.

Diagnosis of ACS

The diagnosis of ACS including UA, non-ST segment elevation MI (NSTEMI) and ST-segment elevation MI (STEMI) was made by two independent cardiologists by reviewing all patients’ notes, including symptoms, 12-lead ECG and blood tests results according to the 2014 American Heart Association/American College of Cardiology guidelines and international definition of MI version 3. Type 1 and type 2 MI were both included.20 21 The agreement level and kappa value between the two cardiologists were 0.99 and 0.98. The cases where the two cardiologists disagreed were reviewed by a senior cardiologist. All the cardiologists were blinded to ceramide levels. Cardiac chest pain with persistent ECG findings of elevation in the ST segment indicate STEMI. Cardiac chest pain and elevations in troponins levels without ST-elevation indicate NSTEMI. Those with presence of one or more of three principal ischaemic symptoms ((1) rest angina (lasting >20 min), (2) new-onset (<2 months previously) severe angina and (3) a crescendo pattern of occurrence (increasing in intensity, duration, frequency or any combination of these factors)) without elevations in cardiac troponins are defined as UA.5 Participants were diagnosed with ACS when there is evidence of UA, NSTEMI or STEMI.

Quantification of plasma ceramides

Potential ceramides were generated from previous untargeted and targeted ceramide profile studies.22 23 Analyses of 12 plasma ceramides were performed using a Waters Xevo TQ-S mass spectrometer (Manchester, UK) equipped with Waters Acuity UPLC I-Class (Milford, Massachusetts, USA) in a hospital and Qlife lab collaborative laboratory based in Department of Cardiology. The mass spectrometry was operated in multiple-reaction monitoring mode with  electro spray ionisation interface (ESI)-positive ionisation. The capillary voltage was set at 3.0 kV, and the source temperature was 120℃. The desolvation temperature and gas flow were 400℃ and 800 L/h, respectively. The source offset was maintained at 60 V.

Blood samples for ceramides test were collected and centrifuged using EDTA anticoagulation tube at admission. Plasma were immediately stored in −80℃ for future analysis. Ceramides tests were taken immediately after samples of all patients were collected. Before analysis, the samples were thawed at room temperature, then a volume of 800 µL of protein precipitation solution (isopropanol) that containing D7-Cer d18:1/16:0 (0.01 pmol/µL), D7-Cer d18:1/18:0 (0.005 pmol/µL), D7-Cer d18:1/24:0 (0.015 pmol/µL) and D7-Cer d18:1/24:1 (0.015 pmol/µL) was pipetted into 1.5 mL Eppendorf tube after addition of 50 µL of plasma sample. The mixture was thoroughly vortexed for 10 min followed by 5-min centrifugation at 14 000 rpm. Supernatant (70 µL) was collected for liquid chromatography–mass spectrometry (LC/MS/MS) analysis. Ceramides ratios were calculated using Cer(d18:1/16:0), Cer(d18:1/18:0), Cer(d18:1/24:1(15Z)) and Cer(d18:1/24:0).23 Lab technicians who were in charge of ceramides test were blinded to disease diagnosis.

Statistical methods

Continuous variables are summarised as mean (SD) or median (IQR) and categorical data as count (percentage). Student’s t-test, Mann-Whitney U and Wilcoxon signed rank tests were used to test differences in continuous variables where appropriate, and χ² test was used for proportions. Univariate logistic regression was used to analyse the association of each variable and binary outcome. Multivariable logistic regression analyses were undertaken to identify the demographic, clinical and laboratory factors associated with ACS in patients with chest pain after adjusting for potential confounders, and the final models included variables significantly associated with ACS.

Training and validation cohorts were developed via a 10-fold cross-validation approach repeated 10 times.24 The estimate of prediction error is almost unbiased under 10-fold cross-validation. Repeated 10 times that minimise the cross-validated error is then used to build the final model.25 26 The performance and discrimination ability of the five models were assessed using the area under the receiver operating curve (AUROC) and net reclassification index (NRI). ROC curves and NRI values were compared among models in training and validation datasets.

All tests were two-tailed, and p values <0.05 were considered statistically significant. OR with 95% CIs for the chosen clinical and laboratory variables in the logistic regression models, and AUROC and NRI values with 95% CIs for different test datasets were calculated. Statistical analyses were done with statistical software R V.3.4.3.

Patient and public involvement

This study was conducted without patient and public involvement.

Demographics, lipid profiles and laboratory findings of patients

The demographics, clinical and laboratory findings of all participants are reported in table 1. Serial troponin levels were measured at 0–3 h after admission in all recruited patients, and all patients underwent echocardiography and coronary angiography. In 2773 patients with chest pain, 354 (12.8%) were diagnosed with ACS, among whom 116 (4.2%) were UA, 114 (4.1%) were NSTEMI and 124 (4.5%) were STEMI. All 116 patients with UA had a significant stenosis on coronary angiography. Compared with those without ACS, patients with ACS are more likely to be men (73.44% vs 67.57%), to be current smokers (31.35% vs 21.62%) and to have hypertension (68.92% vs 62.00%). Lipid profiles and lipoproteins including TG, TC, LDL-C, ApoB, ApoE and lipoprotein(a) (Lp(a)) were significantly elevated in patients diagnosed with ACS, while HDL-C and apolipoprotein A1 (ApoA1) were significantly lower in ACS patients (all p<0.001, except for Lp(a), p=0.023). CRP and cardiac damage markers including hs-cTnT and N-terminal pro-brain natriuretic peptide (NTproBNP) levels were significantly higher in ACS patients (all p<0.001). Demographics, lipid profiles and laboratory findings of patients diagnosed with STEMI, NSTEMI and UA were also reported in online supplementary table 2.

Plasma ceramides levels in ACS patients

Table 2 presents the levels of 12 ceramides molecules in patients with and without ACS. The levels of 12 ceramide molecules (Cer(d18:1/16:0), Cer(d18:1/18:0), Cer(d18:1/24:0), Cer(d18:1/24:1(15Z)), Cer(d18:1/14:0), Cer(d18:1/20:0), Cer(d18:1/22:0), Cer(d18:0/16:0), Cer(d18:0/18:0), Cer(d18:0/24:0), Cer(d18:0/24:1(15Z)) and Cer(d18:1/24:1)) were all significantly elevated in patients with ACS compared with those without (all p<0.001, except for Cer(d18:1/24:0), p=0.028; table 2). In addition, three ceramide ratios (Cer(d18:1/16:0)/Cer(d18:1/24:0), Cer(d18:1/18:0)/Cer(d18:1/24:0) and Cer(d18:1/24:1(15Z))/Cer(d18:1/24:0)) were also significantly increased in ACS patients with all p values <0.001. In patients diagnosed with UA and NSTEMI, levels of ceramides and ceramide ratios were also significantly higher compared with patients diagnosed with non-ACS (see online supplementary table 3).

Association between plasma ceramides and traditional and cardiac damage markers

The associations between plasma ceramides with traditional ACS risk factors and cardiac markers for myocardial injury are shown in online supplementary table 4. Most of the 12 ceramide molecules and ceramide ratios were moderately correlated with lipid profiles including TG, TC, LDL-C, ApoB and ApoE, while CRP was shown to be weakly correlated with ceramides. There was no significant correlation between ceramides and myocardial damage markers including hs-cTnT and NTproBNP.

Clinical predictors and multivariable models for ACS

Clinical predictors for ACS were identified in univariate logistic regression analysis and are shown in table 3. Being male, having hypertension and being current smoker were significantly associated with ACS (p<0.05). Besides, most of the lipid profile (TG, TC, HDL-C, LDL-C, ApoA1 and ApoB) were significantly associated with ACS (all p<0.001). Also, increased levels of hs-cTnT (OR and 95% CI: 2.45 (2.23 to 2.70)), NTproBNP (1.73 (1.60 to 1.87)) and CRP (1.73 (1.58 to 1.90)) were significantly associated with ACS. Furthermore, all 12 ceramides levels and three ceramides ratios were significantly associated with ACS with all p<0.001 except Cer(d18:1/24:0) with p=0.025 (table 3).

Multivariable logistic regression analysing independently significant predictors of ACS are shown in table 4. Model 1 included traditional risk factors such as hypertension, current smoker, CRP, LDL-C and NTproBNP (all p<0.001 except for hypertension p=0.033). Model 2 only included hs-cTnT (2.454 (2.239 to 2.701), p<0.001). Model 3 included traditional risk factors and hs-cTnT. Current smoker, CRP, LDL-C, NTproBNP as well as hs-cTnT were independently significant in detecting patients with ACS. Model 4 included traditional risk factors and ceramides levels, with OR (95% CI) of Cer(d18:1/24:1(15Z))/Cer(d18:1/24:0): 2.960 (2.156 to 4.072), Cer(d18:1/14:0): 1.439 (1.007 to 2.062) and Cer(d18:1/22:0): 2.079 (1.315 to 3.288). Model 5 included traditional risk factors, hs-cTnT (2.190, (1.939 to 2.485)) and ceramides levels (Cer(d18:1/24:1(15Z))/Cer(d18:1/24:0): 4.455 (3.112 to 6.396), Cer(d18:1/14:0): 1.73 (1.166 to 2.605) and Cer(d18:1/22:0): 2.301 (1.371 to 3.866), all p<0.01). Ceramides were shown to be independently associated with ACS after adjusting for traditional factors and hs-cTnT.

Discrimination and reclassification of ACS with various multivariable logistic regression models

As shown in table 5, the model including traditional factors (hypertension, current smoker, LDL-C, CRP and NT-proBNP level) had an area under the curve (AUC) of 0.780 (0.750 to 0.811) in diagnosing ACS. AUC of hs-cTnT in predicting ACS was 0.800 (0.767 to 0.833). After combining traditional factors and hs-cTnT, the AUC became 0.808 (0.776 to 0.841) (figure 2A). The AUC of model including traditional factors and ceramides was 0.811 (0.784 to 0.837). Notably, when combining traditional factors, hs-cTnT and ceramides, the AUC of the model rose to 0.865 (0.840 to 0.889) (figure 2B). We also compared discrimination and reclassification power of these models. Results showed that ceramides significantly improved the AUC of the model with traditional factors and hs-cTnT (0.865 (0.840 to 0.889) vs 0.808 (0.776 to 0.841), p<0.001). NRI analyses also showed that the model including traditional factors and hs-cTnT was significantly improved by the inclusion of ceramides (NRI=0.511 (0.388 to 0.635), p<0.001).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Receiver operating characteristic (ROC) curves of multivariable models for the discrimination of ACS. (A) Traditional risk factors (hypertension, current smoker, LDL-C, CRP and NT-proBNP) and high-sensitive troponin are included in the model. (B) Traditional risk factors (hypertension, current smoker, LDL-C, CRP and NT-proBNP), high-sensitive troponin, Cer(d18:1/14:0), Cer(d18:1/22:0) and Cer(d18:1/24:1(15Z))/Cer(d18:1/24:0) are included in the model. ACS, acute coronary syndrome; AUC, area under the curve; CRP, C reactive protein; LDL-C, low-density lipoprotein cholesterol; NT-proBNP, N-terminal pro-brain natriuretic peptide .