Fifty-seven articles (3,934 patients; 3,944 sentinel node procedures) were included in the review. Sample sizes ranged from 8 to 500, with 10 studies containing more than 100 patients. All studies were prospective, with a valid reference test (histology), and clear sentinel node criteria and detection protocols. Twenty articles met at least three of the five remaining quality criteria.
The identification rate using the sentinel node procedure was 92.4% (94.6% in studies with more than 100 patients) and this was more successful in studies using ex vivo procedures (93.7% versus 89.2% in those using in vivo procedures; 18 studies). Pooled sensitivity was 69.6% (95% CI 64.7 to 74.6; 57 studies; 3,934 patients); the false negative rate was 30.4% (95% CI 25.5 to 35.3); and the overall pooled accuracy of the sentinel node procedure was 88.2% (95% CI 86 to 90.3).
Sensitivity was significantly higher: in patients when four or more sentinel nodes were identified (10 studies); in those with colon carcinomas (31 studies); and in those patients with early stage disease (six studies). Method of sentinel node detection (in vivo/ex vivo) when colloid was added to blue dye, type of tracer used, and presence of high quality studies did not show any statistically significant influence on the findings.
The mean upstaging rate was 18.9% ±11.8 (range 0 to 50%; 46 studies). Further results from subgroups of studies were also reported.