To the Editor: We agree that the "what", "how" and "who" of guideline development all deserve equal, explicit and systematic attention.1

A fundamental task for architects of consensus guidelines is to get the "what" right first. Agreement about the importance of the topic and the objective of the exercise is crucial to its ultimate success. Edmonds and colleagues state that "formulation of precise indications for the use of NSAIDs [non-steroidal anti-inflammatory drugs] rather than CSIs [COX-2-specific inhibitors] (or vice versa) would generate interminable controversy".2 The foundation for this assertion is not clear and the authors do not present data about the level of agreement on this by the experts initially assembled.

The NSW Therapeutic Assessment Group (NSW TAG) believes that providing timely, independent and evidence-based guidance to clinicians about the place in therapy for such new drugs is extremely important. The membership of NSW TAG identified this as a priority soon after the marketing of celecoxib in Australia, and agreed unanimously to develop evidence-based recommendations on indications for the use of this drug. Our consensus development process involved a wide variety of experts in therapeutics and was successfully completed without generating "interminable controversy".3 We wonder whether our different experiences may be partly related to a difference in the initial level of consensus on the importance of the chosen topic.

The "how" of the process followed by Edmonds et al is not described in sufficient detail to enable systematic evaluation of its validity. How systematic was the search for evidence or the process for inclusion or exclusion of studies? What was the level of evidence on which final recommendations were based?

Importantly, high quality guideline development processes require a "balance of healthcare disciplines in the guideline development group".4 Getting the right "who" is a prerequisite for getting the "how" right. Edmonds et al state that membership was arbitrary, with predominant representation from rheumatologists and relevant pharmaceutical companies. Given the problems associated with physician–industry interactions,5 it has been suggested that authors with significant conflicts of interest should be excluded from participating in guideline development.6 The rationale for arbitrary selection of members and inclusion of members from the pharmaceutical industry is not explicitly stated.

These issues may have contributed to the difficulties the group experienced, and may detract from the validity of their recommendations. Future trips down the "road to consensus" should run more smoothly after careful consideration of the "what", "how" and "who" at the outset — no "ifs and buts" about it.