Author: Amira Mandour
Biostatistician | Clinical Trials & Statistical Modeling
Expert
Introduction:
In this study, a 1:2 matched case–control design was employed to explore the relationship between selected clinical and behavioral predictors—smoking status, systolic blood pressure (SBP), and electrocardiogram (ECG) abnormalities—and the likelihood of myocardial infarction.
To appropriately account for the matched design, conditional logistic regression was applied to estimate the odds ratios for MI while controlling for the matching factors (age, race, sex, and hospital status).
Objectives:
Primary Objective: To assess the association between smoking status, systolic blood pressure, and ECG abnormalities with the occurrence of myocardial infarction using conditional logistic regression.
Secondary Objectives:
To quantify the strength of association (odds ratios) for each risk factor after controlling for the matching variables.
To evaluate the relative importance of clinical versus behavioral predictors in the risk of MI.
Study Design and Data Description:
Study Type: 1:2 Matched Case–Control Study
Total Sample Size: 117 subjects (39 matched strata)
Matching Factors: Age, race, sex, and hospital status
Outcome Variable:
Predictor Variables:
Smoking status (1 = current smoker, 0 = non-smoker)
Systolic blood pressure (continuous, measured in mmHg)
ECG: ECG abnormality (1 = abnormal, 0 = normal)
Each stratum contains one MI case and two matched controls, ensuring comparability within strata and controlling for the effects of matching variables.
Results and Interpretation:
A conditional logistic regression model was fitted to evaluate the association between smoking status, systolic blood pressure (SBP), and electrocardiogram (ECG) abnormalities with the odds of myocardial infarction (MI), accounting for the 1:2 matched design on age, sex, race, and hospital status. The dataset consisted of 117 subjects grouped into 39 matched strata (each containing one MI case and two matched controls).
Model Fit:
The model demonstrated statistically significant overall fit (Likelihood ratio test: χ²(3) = 22.2, p < 0.001; Wald test: χ²(3) = 13.68, p = 0.003; Score test: χ²(3) = 19.68, p < 0.001), suggesting that the predictors collectively differentiate MI cases from controls. The model’s concordance statistic (C = 0.737, SE = 0.076) indicates good discriminative ability — meaning that in approximately 74% of matched sets, the model correctly assigns higher risk to the MI case than to its controls.
The pseudo R² value was approximately 0.17, which indicates that about 17% of the variability in MI occurrence is explained by the included predictors. Although modest, this is common for epidemiologic models focused on estimating associations rather than predicting outcomes.
Interpretation of Key Findings:
After adjusting for the matching factors (age, sex, race, and hospital status): Systolic Blood Pressure (SBP) emerged as a statistically significant risk factor for myocardial infarction (p = 0.0028). Each incremental increase in SBP substantially elevated the odds of experiencing an MI, highlighting the importance of blood pressure control in cardiovascular prevention.
ECG Abnormalities were associated with a large increase in the odds of MI (OR ≈ 5), but the wide confidence interval (0.93–26.36) suggests considerable uncertainty, likely due to the limited sample size.
Smoking Status showed a positive but non-significant relationship with MI (OR = 2.07, p = 0.194). While consistent with established clinical evidence linking smoking to cardiovascular risk, the lack of statistical significance here may reflect limited statistical power rather than absence of effect.
Table 1: Conditional Logistic Regression Analysis of Risk Factors for Myocardial Infarction:
| Table 1: Conditional Logistic Regression of Risk Factors for Myocardial Infarction | |||
| Characteristic1 | OR1,2 | 95% CI1,2 | p-value1 |
|---|---|---|---|
| Smoking Status | 2.07 | 0.69, 6.23 | 0.194 |
| Systolic Blood Pressure (mmHg) | 1.05 | 1.02, 1.08 | 0.003 |
| ECG Abnormality | 4.95 | 0.93, 26.4 | 0.061 |
| 1 Conditional logistic regression; strata = Match | |||
| 2 OR = Odds Ratio, CI = Confidence Interval | |||
Model Evaluation and Interpretation:
Although the pseudo R² value (0.173) indicates that the model explains about 17% of the variability in MI risk, this level of explanatory power is typical for biomedical data where disease risk depends on multiple unmeasured biological and lifestyle factors. Importantly, the model’s statistically significant likelihood ratio test and reasonable concordance (0.74) suggest that it effectively captures meaningful associations between the predictors and MI risk.
Overall, the model supports SBP as a key independent risk factor for myocardial infarction, while smoking and ECG abnormalities demonstrate positive but less certain associations. These findings are consistent with established cardiovascular literature and highlight the clinical relevance of maintaining normal blood pressure and monitoring cardiac electrical abnormalities in preventing MI.
Forest Plot of Odds Ratios for Risk Factors Associated with Myocardial Infarction:
Figure 1 displays the odds ratios and 95% confidence intervals for risk factors associated with myocardial infarction from the conditional logistic regression model. Higher systolic blood pressure (OR = 1.047, p = 0.003) is significantly associated with increased odds of MI. ECG abnormalities show a strong but borderline significant association (OR = 4.95, p = 0.061), while current smoking shows a non-significant increase in odds (OR = 2.07, p = 0.194).
Figure 1. Forest plot showing estimated odds ratios and 95% confidence intervals for myocardial infarction risk factors. Significant predictors are marked with stars (* for p<0.05). Systolic blood pressure was significant, while smoking and ECG abnormalities were not.
Predicted Risk of Myocardial Infarction by Systolic Blood Pressure, Smoking Status, and ECG Abnormality:
Figure 2 illustrates the predicted probability of myocardial infarction based on the conditional logistic regression model, showing how risk varies with systolic blood pressure, smoking status, and presence of ECG abnormalities. The figure highlights that higher SBP, current smoking, and abnormal ECG are associated with increased predicted risk of MI.
Figure 2. Predicted probability of myocardial infarction across systolic blood pressure, stratified by smoking status and ECG abnormality.
Conclusion:
The conditional logistic regression model showed that higher systolic blood pressure was a statistically significant risk factor for myocardial infarction. Although the pseudo R² (0.173) indicates modest explained variability, this is typical for epidemiologic models aimed at estimating associations rather than prediction. The model was statistically significant overall (Likelihood ratio test p < 0.001), with acceptable precision of estimates (standard errors smaller than coefficients). Smoking and ECG abnormalities showed positive but non-significant associations. Odds ratios derived from exponentiated coefficients provided interpretable measures of relative risk within the matched case–control framework.