Some problems and some solutions in research on psychotherapeutic intervention in cancer.

Several requirements of psychotherapeutic intervention (PI) research are presented. Some major problems are discussed. First, confidence in experimental results may be diminished by interference of possible confounders. A large number of these exist, both physical, perhaps more easily measurable, and psychological or psychosocial. Second, the value and limitations of randomization are presented, with special reference to small N. Validity and reliability are addressed, with emphasis on things that may dilute the strength of both of these measures. Natural remedies are proposed. Internal consistency and reliability are compared and cautions are given regarding their use and possible invalid substitution of consistency for reliability. Comparison of different PIs is commented on, with some relevant examples. Finally, a matter of special interest is discussed: the inconsistency between Spiegel's control survival curve and that produced from local population normative data.