Diagnostic Performance and Reproducibility of the CARS Endoscopic Score in Achalasia: A Systematic Review and...

Achalasia's insidious onset and nonspecific presentation contribute to diagnostic delays, often exceeding 20 months from symptom onset to confirmation. The four-domain CARS endoscopic score (Content, Anatomy, Resistance, Stasis) may expedite risk stratification. We performed a meta-analysis to assess diagnostic accuracy, interobserver reliability, and treatment-monitoring potential.

We searched PubMed, Embase, Web of Science, the Cochrane Library, and ClinicalTrials.gov through June 2025. Study-level 2 × 2 contingency data were then synthesized using bivariate random-effects models to derive pooled sensitivity, specificity, positive predictive value, and negative predictive value, generate HSROC curves, and construct Fagan nomograms at 5% and 25% pre-test probabilities.

We screened 49 studies and ultimately included five studies encompassing 1112 patients with a mean age of 54.5 years and overall achalasia prevalence of 31.7%. For a CARS threshold ≥ 4, pooled sensitivity was 0.73 (95% CI 0.69-0.78), specificity 1.00 (0.99-1.00), positive predictive value 0.99, and negative predictive value 0.88; employing a dual-threshold strategy (CARS = 0 to rule out; ≥ 4 to rule in) further improved sensitivity to 0.95 (0.92-0.98), specificity to 0.99 (0.97-1.00), positive predictive value to 0.99, and negative predictive value to 0.96. The HSROC AUC approached 0.99, and moreover interobserver agreement was almost perfect with a pooled Cohen's κ = 0.84 (95% CI 0.76-0.91).

CARS reliably stratifies patients into low-risk (CARS = 0), moderate-risk (1-3), and high-risk (≥ 4) groups facilitating deferred testing, targeted manometry, or prompt invasive evaluation while demonstrating excellent interobserver agreement, underscoring its clinical utility.

PROSPERO number: CRD420251007005.