3. What's New History.

January 2004 (Issue 2, 2004)	This review now reflects published and unpublished evidence up to January 2004. In this update there are 7 new studies (Dose ratio 1:1: Currie 2002; Kuna 2003*; Dose ratio 1:2: Backman 2001*; Ige 2002; Kannisto 2002; Majer‐Teboul 2001; Szefler 2002). This gives a total of 57 trials included in the review. This review now incorporates data from fully published articles from two trials previously included as abstracts: Berend 1998 (now Berend 2001); de Benedictis 1998 (now de Benedictis 2001*). Both trials contribute data to the dose ratio 1:2. Following correspondence with GSK there are data from an unpublished trial previously included as an abstract: Lundback 1997*; dose ratio: 1:1. Impact upon findings Dose ratio 1:2. The confidence intervals tightened around mean differences for lung function variables, exacerbations and side‐effects with the addition of new data. In certain instances new data also led to significant heterogeneity which we have explored. Dose ratio 1:1. There is now evidence that FP is more effective than BDP/BUD on certain measures of lung function most notably FEV1, FVC and PEF. However, the effects on exacerbations have become more equivocal with the addition of new data and the level of heterogeneity in this outcome may reflect the combination of data from studies with varying baseline risks. *Denotes unpublished data made available for this review.
January 2005 (Issue 2, 2005)	This review now reflects published and unpublished evidence up to January 2005. There are two new trials in this update (Ferguson 2002; Molimard 2005). Furthermore, the reviewers have attempted to incorporate estimates from studies previously unable to contribute data due to inadequate data analysis in prior versions of RevMan. We have also excluded one study comparing HFA‐BDP with FP (Aubier 2001) as this is now subject to a different systematic review. The new data increased the statistical power of many of the analyses with regard to lung function and data which were measured on continuous scales (i.e. where average scores are achieved). The new evidence does not alter the conclusions of the review.