Intended for healthcare professionals


Will eradication of Helicobacter pylori improve symptoms of non-ulcer dyspepsia?

BMJ 2000; 320 doi: (Published 29 April 2000) Cite this as: BMJ 2000;320:1208

Studies included in meta-analysis had heterogenous, not homogenous, results

  1. Chris Cates, general practitioner (
  1. Manor View Practice, Bushey, Hertfordshire WD2 2NN
  2. Department of Primary Care and General Practice, University of Birmingham, Birmingham B15 2TT
  3. Centre for Digestive Diseases, University of Leeds, Leeds LS2 9JT
  4. Systematic Review Programme, ICRF/NHS Centre for Statistics in Medicine, Oxford OX3 7LF
  5. Cochrane Upper GI and Pancreatic Diseases Review Group, Leeds LS2 9LN
  6. Centre Hospitalier Universitaire Vaudois, CH-1011 Lausanne, Switzerland
  7. Department of Medicine, University of Sydney, Clinical Sciences Building, Nepan Hospital, PO Box 63, Penrith, NSW 2751, Australia
  8. Division of Gastroenterology, Triemli Hospital, CH-8063 Zurich, Switzerland
  9. Department of Gastroenterology, Adelaide and Meath Hospitals, Tallaght, Dublin 24
  10. Department of Family and Community Medicine, University of Toronto, Sunnybrook and Women's College Health Sciences Centre, Toronto, Ontario, Canada
  11. Inner City Health Research, St Michael's Hospital, Toronto
  12. Department of Family Medicine, Center for Evidence-Based Practice, State University of New York Health Science Center at Syracuse, Syracuse, NY, USA

    EDITOR—Jaakkimainen et al's meta-analysis concludes that an improvement in dyspeptic symptoms occurred among patients with non-ulcer dyspepsia in whom Helicobacter pylori was eradicated.1 Unfortunately, there is a small but crucial problem at the heart of the analysis. The authors report that the summary estimates are statistically homogenous, but this is incorrect. In the observational studies the P value of <0.001 indicates massive heterogeneity between the results of the studies included. In the therapeutic trials the P value of 0.046 also indicates heterogeneity.

    Meta-analysts faced with such heterogeneity have three choices: they may ignore the heterogeneity and pool the results with a fixed effects model; they may use a random effects model, which takes the heterogeneity into account; or they may decide not to pool the results. In this instance the authors chose to use a fixed effects model despite the heterogeneity. In consequence the confidence intervals of the pooled estimates are narrow and significance is imputed. A random model might well lead to a summary estimate that does not reach significance. The data extracted from the individual trials have not been published in so it is not possible to check this.

    The authors have carried out a sensitivity analysis of the therapeutic trials and noted that the single regimen trials measured only short term outcomes and were of lower methodological quality. In contrast, the two trials of triple treatment measured outcomes at 12 months and were of higher quality; these trials showed a much smaller summary estimate of eradication, which barely reaches significance (odds ratio 1.4, 95% confidence interval 1.0 to 2.3). There is clinical as well as statistical heterogeneity between these two groups of studies, so a summary estimate that combines both is of doubtful meaning. Thus in my opinion the best option is not to pool …

    View Full Text

    Log in

    Log in through your institution


    * For online subscription