Summary  Background It is known that two‐thirds of patients who develop clinical metastases following treatment of a primary cutaneous melanoma initially present with locoregional metastases and one‐third initially present with distant metastases. However, few reports in the literature give detailed figures on different metastatic pathways in cutaneous melanoma.

Objectives The aim of the present study was to perform a detailed analysis of the different metastatic pathways, the time course of the development of metastases and the factors influencing them.

Methods In a series of 3001 patients with primary cutaneous melanoma at first presentation, 466 subsequently developed metastasis and were followed‐up over the long term at the University of Tuebingen, Germany between 1976 and 1996. Different pathways of metastatic spread were traced. Associated risk factors for the different pathways were assessed. Differences in survival probabilities were calculated by the Kaplan–Meier method and evaluated by the log‐rank test.

Results In 50·2% of the patients the first metastasis after treatment of the primary tumour developed in the regional lymph nodes. In the remaining half of the patient sample the first metastasis developed in the lymphatic drainage area in front of the regional lymph nodes, as satellite or in‐transit metastases (21·7%) or as direct distant metastases (28·1%). Anatomical location, sex and tumour thickness were significant risk factors for the development of metastasis by different pathways. The most important risk factor appeared to be the location of the primary tumour. The median intervals elapsing before the first metastasis differed significantly between the different metastatic pathways. The direct distant metastases became manifest after a median period of 25 months, thus later than the direct regional lymph node metastases (median latency period, 16 months) and the direct satellite and in‐transit metastases (median latency period, 17 months). In patients who developed distant metastases the period of development was independent of the metastatic route. The time at which the distant metastases developed was roughly the same (between 24 and 30 months after the detection of the primary tumour), irrespective of whether satellite or in‐transit metastases, lymph node metastases or distant metastases were the first to occur.

Conclusions The time course of the development of distant metastasis was more or less the same irrespective of the metastatic pathway; this suggests that in patients with in‐transit or satellite metastasis or regional lymph node metastasis, haematogenic metastatic spread had already taken place. Thus, the diagnostic value of sentinel lymph node biopsy and the therapeutic benefit of elective lymph node dissection may be limited, as satellite and in‐transit metastases or direct distant metastases will not be detected and haematogenous spread may already have taken place when the intervention is performed.