The diagnostic performance of CT and MRI for detecting extranodal extension in patients with head and neck squamous cell carcinoma: a systematic review and diagnostic meta-analysis.

To review the diagnostic performance of CT and MRI for detecting extranodal extension (ENE) in head and neck squamous cell carcinoma (HNSCC) patients.
MEDLINE and EMBASE databases were searched up to October 7, 2019. Studies that evaluated the diagnostic performance of CT and/or MRI for detecting ENE in HNSCC patients were included. A 2 × 2 table was reconstructed for each study. Pooled sensitivity and specificity were calculated using the bivariate model and hierarchical summary receiver operating characteristic (HSROC) model. Subgroup analyses were performed according to HPV status and radiological features. Pooled correlation coefficient for interobserver agreement was calculated.
Twenty-two studies including 2478 patients were included. The pooled sensitivity and specificity for detecting ENE were 73% (95% CI, 62-82%) and 83% (95% CI, 75-89%), respectively, for CT, and 60% (95% CI, 49-70%) and 96% (95% CI, 85-99%), respectively, for MRI. There was substantial heterogeneity for both CT and MRI. A threshold effect was present for MRI. On subgroup analysis, the pooled specificity of CT was significantly lower in patients with HPV+ OPSCC than in patients with HPV‑ oral cavity cancer or all HNSCC (74% vs. 87%; p = 0.01). Central node necrosis showed significantly higher pooled sensitivity (81% vs. 51%; p = 0.02), while infiltration of adjacent planes showed significantly higher pooled specificity (94% vs. 65%; p = 0.03). The pooled correlation coefficient was 0.72 (95% CI, 0.60-0.81).
Both CT and MRI show reasonable diagnostic performance for detecting ENE in HNSCC patients and interobserver agreement was substantial.
• Pooled sensitivity and specificity were 73% and 83% for CT and 60% and 96% for MRI without significant difference. • Pooled specificity was lower for HPV+ OPSCC than for HPV‑ oral cavity cancer or all HNSCC (74% vs. 87%, p = 0.01), likely due to central node necrosis. • Central node necrosis showed higher sensitivity (81% vs. 51%; p = 0.02), while infiltration of adjacent planes showed higher specificity (94% vs. 65%; p = 0.03).