Fifteen studies were included in the review (79,526 patients, range 80 to 34,069); 12 were prospective and three retrospective. Six studies enrolled consecutive patients. Eight studies avoided partial verification bias. Six studies avoided differential verification bias. Eight studies avoided progression bias. Eleven studies avoided incorporation bias. Five studies blinded interpreters of the reference standard and seven studies blinded interpreters of the index test. Four studies did not report on uninterpretable results. Nine studies explained withdrawals. Where reported, loss to follow-up was up to 18.6%. Radiography was conducted as part of the reference standard in 45% to 100% of patients.
Canadian C-spine rule (nine studies): Sensitivity ranged from 90% to 100%. Specificity ranged from 1% to 77%. The fraction of false negatives ranged from 0% to 0.11%. Median LR- was 0.18 (IQR 0.03 to 0.24) and median LR+ was 1.69 (IQR 1.57 to 1.81).
NEXUS (seven studies): Sensitivity ranged from 83% to 100%. Specificity from 13% to 46%. The fraction of false negatives ranged from 0% to 1.0%. Median LR- was 0.30 (IQR 0.19 to 0.41) and median LR+ was 1.44 (IQR 1.14 to 1.52).
One direct comparison reported that the Canadian C-spine rule had better accuracy than NEXUS and reduced imaging rates by 44% compared to 36% for NEXUS. Results were reported for prospective studies that used the rule unadapted.