Sixteen studies were included (total number of participants unclear).
Fourteen studies were prospective and eight of these recruited consecutive patients. Three studies reported blind assessment of the reference standard or sonography.
Studies that included axillae with both palpable and non-palpable nodes.
There were 5 studies in which the criterion for positivity was based on size. The sensitivity ranged from 66 to 73% and the pooled sensitivity was 69% (95% CI: 63, 75); there was no evidence of heterogeneity (p=0.95). The specificity ranged from 44 to 98% and the pooled specificity was 75% (95% CI: 70, 80); there was strong evidence of heterogeneity (p<0.001).
There were 6 studies in which the criterion for positivity was based on morphology. The sensitivity ranged from 55 to 92% and the specificity from 80 to 97%. The pooled sensitivity and pooled specificity were 71% (95% CI: 65, 76) and 86% (95% CI: 83, 89), respectively; there was strong evidence of heterogeneity in both measures (p<0.003).
Studies that included axillae with non-palpable nodes only.
There were 4 studies in which the criterion for positivity was based on size. The sensitivity ranged from 49 to 87% and the specificity from 78 to 97%. The pooled sensitivity and pooled specificity were 61% (95% CI: 55, 67) and 77% (95% CI: 72, 82), respectively; there was strong evidence of heterogeneity in both measures (p<0.001).
There were 4 studies in which the criterion for positivity was based on morphology. The sensitivity ranged from 26 to 76% and the specificity from 89 to 98%. The pooled sensitivity and pooled specificity were 44% (95% CI: 65, 76) and 92% (95% CI: 89, 95%), respectively; there was strong evidence of heterogeneity in both measures (p<0.02).
Sonographically guided biopsy.
There were 8 studies of only needle biopsy. The sensitivity ranged from 25 to 95% and the pooled sensitivity was 75% (95% CI: 70, 79); there was strong evidence of heterogeneity (p<0.001). Specificity was 97% in 1 study and 100% in all other studies.
There were 5 studies of all axillae. The sensitivity ranged from 6 to 63% and the pooled sensitivity was 45% (95% CI: 40, 51); there was strong evidence of heterogeneity (p<0.001). Specificity was 100% in all studies.