Comparing Non-Laboratory-Based and Laboratory-Based Cardiovascular Risk Predictions: Systematic Review and Meta-Analysis.

INTRODUCTION

Cardiovascular disease (CVD) remains the leading cause of global morbidity and mortality. This study assesses the agreement between non-laboratory-based and laboratory-based CVD risk equations across diverse settings.

METHODS

PubMed, Scopus, Web of Science, ProQuest Dissertations and Theses Global, and Google Scholar were systematically searched for studies published up to March 4, 2025. The protocol was registered with PROSPERO (CRD42021291936). Studies comparing laboratory-based and non-laboratory-based CVD risk equations were included, excluding those with participants who had CVD at baseline. A meta-analysis was conducted using mixed-effects meta-regression. The agreements for each study item (unit of analysis) were measured using the Spearman correlation coefficient and Kappa statistics.

RESULTS

A total of 33 studies, including 243,587 participants and nine CVD risk equations, were identified. The pooled Spearman correlation between the non-laboratory-based and laboratory-based equations was 0.954 (95% CI: 0.928-0.971, I2 = 100%, p < 0.0001), and the pooled kappa was 0.64 (95% CI: 0.61-0.67, I2 = 99%, p < 0.0001). Correlation was higher in studies conducted before 2000 (0.974; 95% CI: 0.970-0.978) compared to those conducted after 2000 (0.953; 95% CI: 0.948-0.958; p < 0.0001). Studies from high-income settings had higher correlations (0.967; 95% CI: 0.963-0.970) than those from low-income settings (0.945; 95% CI: 0.933-0.955). Equations predicting fatal outcomes had higher correlations (0.979; 95% CI: 0.977-0.982) than those predicting both fatal and non-fatal outcomes (0.942; 95% CI: 0.937-0.947).

CONCLUSION

Non-laboratory-based CVD risk equations demonstrate strong correlation and substantial agreement with laboratory-based equations. Non-laboratory-based CVD risk equations show strong concordance with laboratory-based equations in many settings; however, strong correlation and substantial agreement do not necessarily indicate predictive equivalence, and their interchangeability and implementation should be considered context-specific and require external validation and recalibration.