The ability to reliably predict cancer outcome could tailor therapy to the aggressiveness of the tumour to achieve the best results in terms of loco-regional control, overall survival and quality of life. Retrospective and prospective clinical trials involving large series of patients have validated some predictive clinical and pathological factors, whereas the utility of many other prognostic factors has not been established. This has led to some confusion in clinical practice. In order to clarify the significance, role and cost of these prognostic factors we carried out a Medline search of all papers published between 1993 and 1998 concerning the reliability and cost of markers with prognostic significance, in head and neck squamous cell carcinoma, and assessed the results according to a number of criteria relating to reliability and cost. Regarding reliability we classified prognostic factors into: (1) those with a proven significance based on the fact that they were unanimously reported as having an independent statistical correlation with outcome and prognosis; and (2) those for which results were not unanimous, and which significance is still controversial. Cost analysis showed a substantial difference between validated tests which are of low cost and experimental tests which are expensive. Based on these data regarding both the reliability and cost of each prognostic factor, we propose guidelines for their use in clinical practice in the year 2000.