EDITOR, - Douglas G Altman highlights the need for improvement in research studies and publications.1 Editors of the BMJ have been active in trying to tighten the standards of published articles,^2,3 and formal statistical appraisal is now part of the peer review process. However, four papers chosen haphazardly from last year's BMJ provide examples of sloppy use of the literature, which could be tackled by explicit guidelines and perhaps formal appraisal of the literature. The BMJ's current instructions to authors include guidance only on referencing style. We suggest three guidelines for discussion.

Firstly, authors should adequately review the literature by using available bibliographic databases and should provide documentation of search methods for referees. In particular, statements beginning “To our knowledge no previous reports have...” need support by documentation of searches made by the authors. For example, Kelly et al claimed that “no previous reports have assessed illness in diabetic patients in relation to social deprivation,”5 but a literature search elicits numerous papers that would be relevant. These include reports of the influence of social status on glycaemic control,6 the development of diabetic complications,7 and the link between socioeconomic status and vascular disease in diabetic patients.8

Russell et al claimed that “no study has investigated the nature of backache after childbirth.”8 This is an ambitious claim as backache after childbirth has been reported and examined before, although it was not the main subject of the studies.^10,11

Secondly if systematic reviews pertaining to subject studied exist these should normally be referenced. For example, the National Perinatal Epidemiology Unit published a systematic review of controlled perinatal trials12 which Day and Primhak did not reference in their review of current practices in neonatal intensive care,13 although it was important and appropriate to their study. …