Pathologists play an essential role in predicting which patients with cancer are likely to develop metastases. Therapies to treat or prevent metastases are currently based on tumor type, tumor grade, tumor stage, and a few immunohistochemical and molecular findings. Most of this information comes from the primary tumor. Early-stage tumors are composed of billions of cells, and since a tiny fraction of the whole tumor--perhaps a single cell--can be sufficient for establishing a metastasis, it would seem that identifying traits within the primary tumor that predict metastasis could be like the proverbial search for a needle in a haystack. Circulation of tumor cells within the bloodstream must be an important intermediate event in metastasis, and metastasis is responsible for the lethality of most cancers. By restricting attention to the perhaps 0.0000001% of tumor cells that actually make their way into the bloodstream, it seems likely that we can improve our ability to predict prognosis and guide therapy. In addition, study of such "circulating tumor cells" (CTCs) can expose basic mechanisms for invasion and metastasis. Even in patients with advanced metastatic cancer, CTCs are in the range of only a few cells per milliliter of blood. Enumerating CTCs requires identifying roughly 1 in a billion total cells, or 1 in a million nucleated cells. Such a sensitive assay faces serious technical challenges, and our understanding of the obstacles is incomplete. We don't know the conditions in which normal epithelial cells may contribute to false-positive results (eg, becoming introduced into the circulation in association with an ulceration or trauma); there is little known about the proportion of living to dead circulating tumor cells; and we don't know much about putative circulating normal "stem cells." Furthermore, it is known from animal studies that most injected tumor cells fail to produce a metastasis, (1) so the task of deciphering which rare event is important seems even more daunting than finding that needle in the haystack.