Because treatment of distant melanoma metastases is not very effective, nodal spread should be diagnosed early so that therapeutic lymphadenectomy can be started as early as possible. Physical examination alone often does not detect nodal metastases and palpable nodes cannot be clasified unambiguously. Whether lymph-node ultrasonography-an inexpensive procedure-improves detection of nodal invasion during the initial staging and follow-up of patients with melanoma is controversial. We used meta-analysis techniques for diagnostic tests to assess the merit of ultrasonography and palpation in detection of nodal invasion in patients with melanoma. Five databases were screened until December, 2003. 12 studies, including 6642 patients and 18?610 paired palpation and ultrasound examinations, were eligible. The main limitations were variations in the definition of false negatives, and verification bias. Ultrasonography had a higher discriminatory power (odds ratio 1755; 95% CI 726-4238) than did palpation (21 [4-111]; p=0.0001). Furthermore, positive-likelihood ratios were 41.9 (95% CI 29-75) for ultrasonography and 4.55 (2-18) for palpation; negative-likelihood ratios were 0.024 (0.01-0.03) and 0.22 (0.06-0.31), respectively. Our results showed clearly that ultrasonography detects lymph-node invasion more accurately than palpation, and should therefore probably be used routinely in patients with melanoma.