Editor's Choice | This Week in BMJ | Press releases

BMJ No 7109 Volume 315 This week in brief Saturday 13 September 1997 A graphical test detects bias in meta-analysis Covert duplicate publication leads to overestimation of drug efficacy Delayed publication of negative results may bias reviews Total purchasing is not yet total Genital cancers are linked to human papillomavirus infection Fast track admission in sickle cell crises produces faster pain relief A graphical test detects bias in meta-analysis Systematic reviews and meta-analyses are widely used to inform decision making in clinical practice, but many have expressed reservations about this technique. The debate has been fuelled by the publication of several meta-analyses whose findings were later contradicted by large randomised trials. Egger et al (p 629) developed a simple test based on the funnel plot (a scatter plot of effect estimates against sample size) for detecting bias in meta-analysis. This test predicted discordance of results when meta-analyses were compared with single large trials of the same intervention. The authors found bias in 38% of meta-analyses published in four leading journals but in only 13% of meta-analyses from the Cochrane Database of Systematic Reviews. They argue that systematic reviews and meta-analyses should be examined routinely for bias. Covert duplicate publication leads to overestimation of drug efficacy While undertaking a systematic search for published randomised trials of ondansetron as a postoperative antiemetic, TramÍ4r et al found that 14 out of 84 trials (17%) were duplicates (p 635). Data from 11 980 patients receiving ondansetron appeared in the literature although only 8645 (72%) were original. Only one duplicate mentioned the matching report. The number needed to treat to prevent vomiting with intravenous ondansetron 4 mg compared with placebo was 9.5 (95% confidence interval 6.9 to 15) in 16 reports which were never duplicated, but 3.9 (3.3 to 4.8) in three reports which were duplicated. The number needed to treat with data from all original reports was 6.4 (5.3 to 7.9); with all the duplicates, as published, it fell to 4.9 (4.4 to 5.6). So an unaware reader would have overestimated the antiemetic efficacy of ondansetron by 23%. Delayed publication of negative results may bias reviews Trials with significant results are more likely to be published than those with negative results. On p 640 Stern and Simes report the outcome in a cohort of clinical studies approved by an Australian hospital over 10 years. They confirmed that studies with significant, or positive, results were more likely to be published and also to be published earlier than negative studies. This applied particularly to clinical trials. These results provide a strong argument for the need to register prospectively clinical trials and to select studies for systematic reviews from trial registries. Total purchasing is not yet total Several pilot projects are underway in Britain to assess the concept of "total purchasing" of health services by general practitioners. These extend existing fundholding schemes and allow groups of general practitioners to hold larger budgets and purchase a wider range of secondary and community services for their patients. On p 652 Mays et al assess the pilots; these showed wide variation in the scope and ways of working. Most of the 53 pilots are in suburban or rural areas; they showed a range of management costs, from \P0.26 per head to \P8.05 with a mean of \P3.00. The authors conclude that the NHS is likely to be organised around GP purchasing in future but that several issues remain to be resolved. Genital cancers are linked to human papillomavirus infection Human papillomavirus has an established association with cervical cancer and has also been implicated in other anogenital cancers. On p 646 Bjørge et al report the results of a nested case-control study of serologically diagnosed infection with the main oncogenic types of human papillomavirus and the subsequent risk of developing non-cervical anogenital cancers. Two large serum banks and the nationwide cancer registries in Finland and Norway were used for this prospective seroepidemiological evaluation. Infection with human papillomavirus type 16 increased the risk of genital cancers, particularly vulvar and vaginal cancers. Fast track admission in sickle cell crises produces faster pain relief Interviews with patients with sickle cell disease showed that accident and emergency staff were inexperienced in dealing with sickle cell crises and pain relief was thus delayed. Fertleman et al therefore established a fast track procedure where children went straight to the ward where they were seen by a nurse who could administer preprescribed doses of pethidine and naloxone (p 650). Parents who had used both systems preferred the new, and the median time to pethidine administration fell from 38 minutes to 5 minutes. Home | Current issue | Past issues | Classified ads | Career Focus | Feedback Collections | About this site | About the BMJ | BMA | Medline