Design: Systematic review and diagnostic accuracy meta-analysis
Population: 222 adults with left thoracoabdominal penetrating trauma, four studies
Primary outcome: Sensitivity and specificity of CT for diaphragmatic injury
Effect: Sensitivity 0.70 (95% CI 0.56 to 0.80); specificity 0.94 (0.89 to 0.97)

This diagnostic accuracy meta-analysis pooled four studies (222 patients) comparing CT with a surgical reference standard in adults with left thoracoabdominal penetrating trauma. Seven more studies were summarised narratively.

Pooled CT sensitivity was 0.70 (95% CI 0.56 to 0.80; I² = 36%) and specificity 0.94 (0.89 to 0.97; I² = 0%). Individual study sensitivities ranged from 50% to 82%. Diaphragmatic injury was found in 23% to 54% of patients in the additional studies, and therapeutic laparoscopy was usually feasible.

A positive CT is reliable, but a negative one is not. Missed left-sided diaphragm injuries can present months or years later with herniation and strangulated bowel, when the repair is harder and the mortality higher. For a stable patient with a left thoracoabdominal stab wound and a negative CT, diagnostic laparoscopy remains the way to exclude injury. The evidence base is small, so the pooled figures carry real uncertainty.

Do not use a negative CT alone to exclude diaphragm injury after a left thoracoabdominal penetrating wound.
Offer diagnostic laparoscopy to haemodynamically stable patients with a left thoracoabdominal wound, even with a normal CT.
Repair a diaphragm defect found at laparoscopy during the same procedure where feasible.
Document the wound site precisely, since right-sided wounds carry a different risk of later herniation because of the liver.
Warn patients managed without laparoscopy about delayed symptoms such as left chest pain, vomiting or obstruction.

Why it matters

A missed left diaphragm injury can turn into strangulated bowel years later, so a falsely reassuring scan carries a long tail of harm.

The statistics, in plain English

Sensitivity of 0.70 means CT picked up about 70 of every 100 real injuries. Specificity of 0.94 means it was rarely positive without an injury. Moderate heterogeneity in sensitivity (I² = 36%) and only four studies mean the true sensitivity could be anywhere from about 56% to 80%.