Only after the publication of this article by Alia El-Kadiki and Alexander J Sutton (BMJ 2005;330:871, 16 Apr ) did the authors and editors become aware that doubts had been raised about the validity of three of the trials included in this systematic review.
The BMJ and the authors agreed that further analysis would be helpful, and the authors have therefore conducted a sensitivity analysis excluding data from the three questionable trials.This is now published as a supplement to the original paper (see http://bmj.com/cgi/content/full/bmj.38399.495648.8F/DC2).
The effect on the three outcome measures is as follows: (a) mean difference in number of days spent with infection: only the three questionable studies met the inclusion criteria, so the originally published beneficial difference of 17.5 (95% confidence interval 11 to 24) days is now completely discounted; (b) odds ratio of at least one infection in the study period: no change from published meta-analysis; (c) incidence rate ratio for the difference in infection rates: exclusion of the one questionable trial that was relevant to this outcome means that the pooled incidence rate is now 1.00 (0.85 to 1.17), not 0.89 (0.78 to 1.03) as published.
If the allegations that these three studies are not reliable are true then the remaining evidence base suggests no benefit for the use of multivitamins for preventing infections in elderly people.