Poor oxygenation of solid tumors is a major indicator of adverse prognosis after standard treatment, e.g. radiotherapy. This observation founded on intratumoral pO(2) electrode measurements has been supported more recently by studies of injected hypoxia markers (pimonidazole, EF5) or hypoxia-related proteins (hypoxia-inducible factor-1alpha, carbonic anhydrase IX) detected immunohistochemically. Alternative approaches include imaging of tumor hypoxia by nuclear medicine studies and the measurement of hypoxia-related proteins (osteopontin) in patient plasma. Low oxygen levels as found in tumors are rarely observed in normal tissues. The presence of hypoxic tumor cells is therefore regarded not only as an adverse prognostic factor but as an opportunity for tumor-specific treatment. Classic approaches to normalize tumor oxygenation involve the breathing of modified gas mixtures and pharmacologic modification of blood flow as in the "accelerated radiotherapy, carbogen, nicotinamide" (ARCON) scheme. Specific killing of hypoxic tumor cells can potentially be achieved by hypoxia-selective cytotoxins (model substance tirapazamine), which has shown promise in head and neck cancer. Direct targeting of hypoxia-related molecules such as hypoxia-inducible factor-1alpha, the central regulator of the hypoxic response in tumor cells, is an attractive approach currently tested in preclinical models. For clinical applications, the appropriate combination of hypoxia detection for patient selection with a hypoxia-specific treatment is essential. A therapeutic benefit has been suggested for the selection of patients by plasma osteopontin level and treatment with the hypoxic radiosensitizer nimorazole in addition to radiotherapy, for selection by F-misonidazole positron-emission tomography (PET) and treatment with tirapazamine in addition to chemoradiation and for selection by pimonidazole immunohistochemistry and ARCON treatment, all in head and neck cancer.