Expression of Concern regarding the paper by Viani et al. Issa Dahabreh, Medical School, National University of Athens, Greece 10 October 2007 Sir: We would like to express our concerns regarding the meta-analysis by Vianni et al, entitled “Breast-conserving surgery with or without radiotherapy in women with ductal carcinoma in situ: a meta-analysis of randomized trials”, published in the Radiation Oncology Journal (Radiat Oncol 2007;2:28). We critically read the manuscript by Viani et al and noticed that the events used in the analysis are substantially fewer than those reported in the most recent follow-up reports of the eligible studies. As an example, for the NSABP-B17 study, regarding breast cancer metastases, Viani et al have reported one event per arm (Figure 5) when in the latest follow up of this study (Semin Oncol 2001;28:400-418) there are 10 events in the radiotherapy arm and 7 in the control arm. This observation compelled us to investigate all data presented and analyzed and led to the discovery of several major mistakes that render the findings of Viani et al irrelevant. We list the most important here:1.Viani et al. extracted data from older publications of the following studies for ALL outcomes: a. NSABP-B17: they used NEJM 1993;328:1581-1586 rather than Semin Oncol 2001;28:400-418,b. EORTC-10853: they used Lancet 2000;355:528-33 rather than J Clin Oncol 2006;24:3381-3387. A meta-analysis of published studies is expected to provide a quantitative summary of all available evidence. The authors obviously failed to do that by ignoring thousands of patient-years of follow-up for no good reason.2.Obvious mistakes in data extraction:a. In the outcome “ipsilateral invasive” the number of patients used for the NSABP-B17 and EORTC-10853 studies are wrong.b. The number of patients used for odds ratio estimation for the NSABP-B17 trial is wrong (for ALL outcomes).c. For the outcome “overall mortality rate” the events abstracted from the SweDCIS trial are in fact breast-cancer specific deaths.d. For the outcome “ipsilateral invasive cancers” the authors used ipilateral non-invasive cancers for the SweDCIS and NSABP-B17 trials.e. The number of events and the number of patients used for the UKCCR trial are wrong for ALL outcomes. Viani et al state that the UKCCR trial design is “complex”. Although the handling of factorial trials in meta-analysis is somewhat controversial, Viani et al, by using the numbers reported in Table 1 of the UKCCR manuscript, have reduced a randomized trial to an observational study! To make things easier for readers the authors of this trial have noted in the table legend that “Data for descriptive purposes only. For comparison of treatment effects, factorial analysis as shown in tables 2 and 3 should be used.” Viani et al. have completely disregarded this suggestion; consequently their data abstraction is heavily biased. Given the above examined methodological flaws, we wonder how the paper by Viani et al was accepted for publication. We believe the paper should be retracted or, at least, associated with an official editorial expression of concern. Yours sincerely,Issa J. Dahabreh (firstname.lastname@example.org)Konstantinos P. Economopoulos (email@example.com) National University of Athens,Athens, Medical School, Greece Competing interests We submitted, in ignorance of the work of Viani et al, an abstract on the same subject to the ECCO conference. Our work was entitled “Radiotherapy after breast conserving surgery for ductal carcinoma in situ: an overview of randomized trials”, abstract number PP-21. We have no other competing interests.