The authors of this paper by Claire M Jenkinson and colleagues have told us of a small error in their analysis (BMJ 2009;339:b3170, doi:10.1136/bmj.b3170). In the statistical analysis section of the Methods, the penultimate sentence in the print version (p 607) should have read: “The time course of treatment effects was estimated by a generalised estimating equations linear model incorporating time × treatment interactions with an unstructured correlation matrix to allow for the repeated measurements nature of the data.” As a consequence, in the primary outcomes section of the Results (p 608), the effect of exercise at 24 months should have been given as 0.91 (−1.60 to −0.23; P = 0.009) [not −0.91 (−1.66 to −0.17; P=0.016), as was stated]—that is, slightly more significant.
These changes also affect the full online version of the article.
Cite this as: BMJ 2010;340:c2088