Development and external validation of a nomogram to predict adverse outcomes following delayed PCI in STEMI

Study population

Between June 2019 and August 2022, Tongji Hospital, affiliated with Tongji Medical College of Huazhong University of Science and Technology, received 198 consecutive patients diagnosed with STEMI who subsequently underwent delayed PCI [Figure 1]. Delayed PCI was defined as a procedure performed at least 48 h after the initial onset of ischemic symptoms.

Figure 1. Flowchart of patient enrollment and study design.

The diagnosis of STEMI was established according to current guideline criteria^[7], including: (1) Chest pain caused by insufficient blood flow to the heart that lasts more than 20 min and does not go away even after taking nitroglycerin, and (2) Electrocardiograph (ECG) signs indicating ST-segment elevation in at least two neighboring leads (measuring up to 0.2 mV in chest leads or 0.1 mV in limb leads), or the sudden appearance of a complete left or right bundle branch block.

Eligible patients were required to meet the following inclusion criteria: confirmed STEMI with symptom onset ≥ 48 h at the time of PCI. The exclusion criteria were: (1) abnormalities of the heart's critical valves; (2) acute pulmonary edema or advanced heart failure (New York Heart Association (NYHA) class III or IV upon entry, or heart-related shock); (3) people undergoing urgent angioplasty or bypass surgery for massive heart attacks; (4) heart attacks that do not exhibit the usual ST-segment changes; and (5) other health issues that could significantly reduce life expectancy, such as cancer or severe dysfunction of the liver, kidneys, lungs, endocrine system, nerves, or blood. The ethics review committees of Tongji Hospital, Tongji Medical College of Huazhong University of Science and Technology (Wuhan, China), and The Second Hospital of Lanzhou University (Lanzhou, China) approved this study, which was conducted in accordance with the Declaration of Helsinki.

PCI procedure

Cardiac catheterization was performed with 5-Fr Judkins catheters through radial or femoral access, as per normal protocol. Prior to the intervention, patients were treated according to normal coronary artery disease guidelines and dual antiplatelet regimens. Culprit lesions were identified based on coronary angiographic findings in conjunction with electrocardiographic and echocardiographic assessments. All procedures were performed by experienced interventional cardiologists. Based on the distinct anatomical characteristics of each coronary blockage, the interventional team selected between balloon angioplasty and coronary stent implantation; the doctor made the ultimate decision.

In cases of intraoperative slow-flow or no-reflow, prompt rescue treatments were administered as clinically indicated, including intracoronary pharmacological therapy and/or mechanical interventions. Coronary artery disease was classified as serious when there was a 70% or greater stenosis in arteries other than the left main, or when the left main artery had a 50% or more occlusion^[8].

Myocardial perfusion was measured using the Thrombolysis In Myocardial Infarction (TIMI) flow classification system^[9], and evaluated using the Rentrop grading scheme^[10]. No-reflow was defined as a significant decrease in coronary blood flow following the procedure, even when there was no major heart artery obstruction, arterial lining tears, muscle constriction, or blood clot formation at the initial site. TIMI flow grades 0-1 were classified as no-reflow, whereas TIMI flow grade 2 was defined as slow-flow^[11,12].

Follow-up process

The primary outcome of this study was event-free survival (EFS)^[13], measured as the interval between initial coronary catheterization and the first composite adverse outcome. The composite outcome included all-cause mortality, nonfatal myocardial infarction, or progression to advanced heart failure (NYHA class IV), assessed up to the end of the study period or the final follow-up date. Subsequent assessments were conducted during scheduled phone calls and standard clinic check-ups at 1, 3, and 6 months after hospital discharge. After that first time, monitoring continued every six months. The median observation time was 20 months (range: 1–51 months), with data collection ending in September 2023.

Healthcare practitioners identified key incidents by meticulously reviewing hospital documentation, analyzing doctor ratings from routine check-ups, and maintaining open lines of communication with patients and their families. All clinical endpoints were independently adjudicated by two experienced cardiologists. All subjects completed the study.

Statistical analysis

Associations between continuous variables and EFS were first explored to determine their functional relationships. Optimal cutoff values for continuous variables included in the nomogram were identified using maximally selected rank statistics implemented in the survminer package in R, thereby enhancing clinical interpretability.

Missing data were present across several variables; however, the proportion of missing values did not exceed 30% for any variable. The pattern of missingness for all study variables is presented in Supplementary Figure 1. To avoid exclusion of variables with incomplete data, missing values were imputed using the k-nearest neighbor (KNN) algorithm.

Prior to multivariable modeling, univariate Cox proportional hazards regression analyses were performed to evaluate the associations between candidate variables and EFS, allowing for an inclusive initial screening. Given the limited number of outcome events, inclusion of all candidate variables in a multivariable model could have resulted in overfitting. Therefore, a parsimonious multivariable Cox proportional hazards regression model was constructed by including not only variables with statistical significance in univariate analyses (P < 0.05) but also clinically relevant factors identified by experienced clinicians.

Based on the results of the multivariable analysis, nomograms were developed to estimate 6-month and 12-month EFS probabilities. Model discrimination was assessed using the concordance index (C-index). Calibration was evaluated by comparing predicted and observed outcomes using calibration plots generated from 1,000 bootstrap resamples. The clinical utility of the nomogram was further examined using decision curve analysis (DCA).

Patients were stratified into different risk groups according to nomogram-derived risk scores, and Kaplan-Meier survival curves were generated to compare EFS across risk strata. Differences between groups were assessed using the log-rank test.

All statistical analyses were conducted using SPSS software (version 26.0; IBM Corp., Chicago, IL, USA) and R software (version 4.3.1; R Foundation for Statistical Computing, Vienna, Austria). Nomograms and calibration plots were generated using the rms package. DCA was performed with the dcurves package, time-dependent receiver operating characteristic (ROC) curves were constructed using the survival ROC package, and Kaplan-Meier analyses were conducted using the survival and survminer packages. Internal validation was performed using bootstrap resampling with 1,000 iterations. All statistical tests were two-sided, and a P value < 0.05 was considered statistically significant.

Patient characteristics

A total of 198 patients with STEMI who underwent delayed PCI were included in the study. Their demographic, clinical, laboratory, and procedural details are neatly outlined in Table 1 (refer to Figure 1 for more information). The mean age was 62 years (range, 49-75). Of the 198 patients, 146 (73.7%) were male. Continuous variables were categorized according to predefined optimal cutoff values. Overall, patients exhibited substantial heterogeneity in baseline clinical and laboratory parameters, reflecting the broad spectrum of disease severity among STEMI patients undergoing delayed PCI. Detailed distributions of laboratory indices, hemodynamic parameters, and cardiac biomarkers are presented in Table 1.

The detailed cutoff values for continuous variables are provided in Supplementary Table 1.

Nomogram construction based on clinical characteristics and hospitalization status

Four key factors emerged as independent predictors of poor outcomes in STEMI patients through both univariate and multivariable Cox proportional hazards regression analyses: (1) heart rates exceeding 83 beats per minute; (2) the need for mechanical ventilation; (3) absence of statin therapy; and (4) occurrence of intraoperative slow-flow or no-reflow phenomenon (all with P-values < 0.05; see Table 2). While left ventricular ejection fraction showed a significant correlation with EFS in the initial single-variable assessment, it failed to maintain its independent prognostic value when all variables were considered together, leading to its exclusion from the final predictive model.

To create a predictive tool for 6-month and 12-month EFS in patients receiving delayed PCI, we integrated these four key factors into a nomogram (depicted in Figures 2 and 3). Within this visual aid, each predictor was given a points value that reflected its individual impact on risk assessment, allowing clinicians to tally up these scores to gauge the likelihood of EFS. The higher the cumulative score, the steeper the downhill trajectory toward potential complications.

Figure 2. Hazard ratios (HRs) and 95% confidence intervals (CIs) of the four independent prognostic factors for EFS.

Figure 3. Nomogram for predicting 6-month and 12-month EFS in patients with STEMI undergoing delayed PCI.

Goodness of fit and discrimination of the nomogram

The nomogram’s discriminatory ability was assessed using the C-index. It showed robust discrimination, with a C-index of 0.782 (95% confidence interval [CI]: 0.690-0.874).

The ROC curve analysis demonstrating time dependence further validates the nomogram's predictive precision. The areas under the ROC curve (AUCs) were 0.816 for 6-month EFS and 0.775 for 12-month EFS, indicating good short- and intermediate-term prognostic performance [Figure 4].

Figure 4. Time-dependent ROC curves of the nomogram for predicting EFS. (A) ROC curve for 6-month EFS; (B) ROC curve for 12-month EFS.

The nomogram incorporates four independent prognostic factors: heart rate, mechanical ventilation, statin therapy, and intraoperative slow-flow or no-reflow phenomenon. The aggregate score results from adding up the points for each predictor and reflects the projected likelihood of EFS occurring at 6 and 12-month intervals.

Net benefit and predictive power of the nomogram

The calibration curve estimates the closeness between the risk predicted by the nomogram and the observed risk. As shown in Figure 5, the nomogram demonstrates good calibration for predicting 6-month and 12-month EFS. The nomogram's net advantage and forecasting prowess were evaluated through decision curve and calibration curve analyses^[14]. In Figure 6, we can clearly see that the nomogram outshines the "treat-all" and "treat-none" approaches. This is thanks to its superior ability to refine risk categorization and support tailored post-discharge care, yielding a higher overall net benefit across a broad spectrum of threshold probabilities. In conclusion, these findings suggest that our nomogram exhibits strong performance in predicting EFS in STEMI patients.

Figure 5. Calibration curves of the nomogram for predicting 6-month (A) and 12-month (B) EFS. Dots represent deciles of predicted risk. The diagonal line indicates ideal agreement between predicted and observed outcomes, and the solid line represents the bootstrap-corrected calibration curve based on 1,000 bootstrap resamples.

Figure 6. Decision curve analysis (DCA) of the nomogram for predicting 6-month (A) and 12-month (B) EFS in patients with STEMI.

The decision curves clearly demonstrate the nomogram's overall clinical utility when evaluated against various threshold probability scenarios. In this visualization, the red line indicates the outcome of adopting a blanket treatment approach for every patient, while the green line depicts the results of implementing a wait-and-see strategy that withholds treatment from all individuals. The net advantage, which is determined from the ratio of true positives to false positives, is displayed on the vertical axis.

External validation

To validate our findings, we set up an independent external validation group using data from 599 STEMI patients who received delayed PCI treatment at the Second Hospital of Lanzhou University. These individuals were followed for a median of 20 months, during which we documented a total of 100 composite adverse outcomes. The participant recruitment process for this validation group is shown in Supplementary Figure 2. The training and external validation cohorts were compared across key demographic and clinical parameters. As evidenced in Supplementary Table 2, no statistically meaningful distinctions emerged between these groups, which indicates that their baseline characteristics were suitably comparable for our analysis. The nomogram showed stable predictive prowess within the external validation group. We conducted analyses involving time-dependent ROC curves, calibration graphs, and decision curve analysis^[15], which showed good discrimination, calibration, and clinical utility, respectively, and were comparable to the results observed in the training cohort [Supplementary Figure 3]. These findings support the robustness and generalizability of the nomogram in an independent patient population.

Risk stratification based on nomogram

To evaluate how well the nomogram could stratify patients by risk level, we calculated each patient’s predicted probability of EFS. The median value of these predictions (62.11%) was used as the cutoff to divide patients into low-risk and high-risk categories.

Kaplan-Meier survival analysis showed clearly different outcomes between the two groups. The high-risk patients had significantly worse EFS than their low-risk counterparts (log-rank P < 0.0001; Figure 7).

Figure 7. Kaplan-Meier curves for EFS according to nomogram-based risk stratification. Patients were stratified into low- and high-risk groups using the median predicted probability of EFS (62.11%) as the cutoff.

Conclusion

This study introduces a user-friendly nomogram leveraging standard clinical parameters and inpatient metrics that effectively predicts individualized Event-Free Survival in STEMI patients undergoing delayed Percutaneous Coronary Intervention. If further validated in large-scale, multicenter prospective studies, this nomogram may serve as a practical and accessible prognostic tool in clinical practice, supporting individualized risk stratification and informed post-discharge management and decision-making.