guideline on this topic noted that no benefit could be proven in either survival or quality of life, so no recommendation could be made for or against second-line chemotherapy (ASCO, 1997). Without evidence to discourage it, the aggressive use of chemotherapy is common. In a survey of practices for dying patients, more than 50 percent of ASCO members stated that they would give second-line chemotherapy to a 43-year-old woman with progressive non-small-cell lung cancer, nearly 20 percent would give third-line chemotherapy, and a substantial number would give fourth-line chemotherapy (Ezekiel Emanuel, personal communication to Tom Smith, 1998).
Multiple guidelines on the same topic are not always consistent. Breast cancer screening guidelines, for example, differ in the recommended age at which routine mammography screening should begin. That there are differences in practice guidelines is not surprising given the many factors that can affect the ultimate conclusions of a guideline. Guideline recommendations may vary because of differences in the composition of guideline panels, the way evidence from the literature is reviewed, the interpretation of evidence, and the procedures to reach agreement on recommendations. The interpretation of data depends on values and perspectives. Some patients and practitioners will think that a small and statistically non-significant improvement in six-month disease-free survival associated with the use of high-dose chemotherapy in lymphoma is sufficient evidence of clinical benefit, whereas others will not.
Guideline panel membership appears to have a strong effect on shaping recommendations. There is, for example, evidence of a user bias in which physicians who perform a given intervention are more likely to judge it as beneficial. When faced with the same evidence from the literature on the effectiveness of carotid endarterectomy, a panel of surgeons rated 70 percent of indications as appropriate, whereas a multidisciplinary group found only 38 percent of indications to be appropriate (Leape et al., 1992). Panels in the United Kingdom and the United States with the same physician mix came to different conclusions when assessing the appropriateness of treatment for coronary disease, in part because the U.K. panelists seemed to require a higher standard of scientific evidence than did their U.S. counterparts (Brook et al., 1988).
One of the hallmarks of a good guideline is having clear documentation of the methods used to arrive at recommendations. The ASCO text that supports the guideline on the provision of follow-up services for patients with breast cancer can be critiqued for not documenting the basis of the guideline's recommendations. Evidence from a well-designed randomized clinical trial indicated that outcomes for women with breast cancer were similar whether follow-up care was provided by a general practice doctor or a specialist surgeon, but the ASCO guideline reinforced the ASCO board's policy that all patients with cancer have the right to see a cancer specialist at all times (Grunfeld et al., 1996). In this case, group opinion and a desire to be consistent with ASCO policy took precedence over the available literature in framing the recommendation, but the guideline does not make this explicit.
The guideline process is a dynamic one, which must incorporate new information as it becomes available. New evidence from the largest clinical trial on follow-up care for individuals with colorectal cancer will, for example, have to be considered as available guidelines are revised. The current NCCN guidelines, for example, recommend annual chest x-ray, annual colonoscopy, and regular computed axial tomography scans, all strategies that do not improve survival according to new evidence (Schoemaker et al., 1998).
How successful guidelines are in informing and changing practice depends in large part on how they are developed, disseminated, and implemented. Personal involvement in the process of change encourages adherence, and evidence suggests that physicians are more likely to ascribe