Adherence to the CONSORT guideline in papers w rit ten by professional medical w riters

Background Many papers in the biomedical literature are drafted not by those who did the research, but by professional medical writers. CONSORT guidelines give specifi c recommendations for items that should be included in publications of randomised controlled trials. This study investigated whether papers written by professional medical writers were more compliant with the CONSORT guidelines than other papers. Findings All randomised clinical trials published in the journal Current Medical Research and Opinion between October 2004 and August 2009 were included in this study. Data were abstracted by two researchers, both of whom were blind to the objectives of the study; one recorded whether each CONSORT item was absent, present but incompletely described, or completely described and the other checked each paper for whether a medical writer had been acknowledged and whether the paper had industry sponsorship. The mean number of completely described guidelines was compared between papers written by a medical writer and those written by others. The secondary analysis was to compare industry-sponsored papers with those that did not declare industry sponsorship. 241 papers were included, 93% of which were industry sponsored; 63% acknowledged assistance from a medical writer. Papers that acknowledged medical writers complied with more CONSORT items (17 of 22) than those that did not (16 of 22; difference between groups 0.75 items completed, 95% CI 0.07 to 1.43, P = 0.03). Too few non-industry-sponsored papers were found to allow a meaningful comparison of industry and non-industry-sponsored papers. Conclusions Papers that acknowledged assistance from professional medical writers were more likely to comply with the CONSORT guidelines than papers that did not. However, the difference was small, and the practical importance of the difference is unknown. Introduct ion Many papers in the biomedical literature are drafted not by those who did the research, but by professional medical writers. Many professional medical writers receive training in how to write papers, and write papers and other medical documents as a full-time job. It might therefore be hypothesised that they are better qualifi ed to write papers than most researchers, for whom writing the paper is often simply an unfortunate extra chore that needs to be done at the end of a piece of research. However, despite the theoretical benefi ts of assistance from professional medical writers, there are almost no data to show whether those benefi ts are realised in practice. In a systematic review in 2003, Lagnado only found anecdotal evidence that professional medical writers improve the quality and readability of papers, and concluded “I did not fi nd fi rm evidence to support these reported benefi ts.” [1] Measuring the writing quality in published papers is hard to do, as many aspects of writing quality are subjective. However, the CONSORT guidelines give specifi c recommendations for items that should be included in publications of randomised controlled trials, with a 22-item checklist [2]. The extent to which papers of randomised trials comply with the CONSORT guidelines could be considered a measure of the completeness with which the research is documented, which is one measure of writing quality, albeit a measure of only one dimension of a complex multi-dimensional concept. The aim of this study was to determine whether papers written by professional medical writers were more compliant with the CONSORT guidelines than other papers. An updated version of the CONSORT guidelines has recently been published [3]; however, this research pre-dates the publication of those guidelines and therefore used the 2001 version. Methods The primary objective of this study was to determine whether papers written by professional medical writers are more likely to comply with the recommendations of the CONSORT guideline than papers that were not written by professional medical writers. A secondary objective was to determine whether industry sponsorship of papers was associated with compliance with the CONSORT guideline. Involvement of professional medical writers and industry sponsorship are often considered as a single issue, although in reality they are two quite distinct concepts. All randomised clinical trials published in the journal Current Medical Research and Opinion between October 2004 and August 2009 were included in this study. That journal was selected because it has a high proportion of papers written by professional medical writers and was therefore expected to yield a suffi cient number of such papers for Vol. 19, No. 3, 2010 The Write Stuff The Journal of the European Medical Writers Association 197 Adherence to the CONSORT guideline in papers written by professional medical writers analysis. A previous pilot study (unpublished) in a wider range of journals failed to yield useful results because the number of papers acknowledging professional medical writers was too small to allow meaningful comparisons. The date range was chosen for pragmatic reasons, as we had had a subscription to the journal since October 2004 and therefore had full text articles available since that date. The instructions to authors of Current Medical Research and Opinion had recommended that manuscripts of randomised controlled trials comply with the CONSORT guideline since April 2005. Data were abstracted by two interns, both of whom were blind to the objectives of the study to avoid any bias in collecting the data. Both interns were science graduates and received brief training in the methods of the study. One intern (VM), who was not aware that the study was designed to compare papers written by professional medical writers with those that were not, compared each paper with each item in the CONSORT checklist, and recorded whether the item was absent, present but incompletely described, or completely described. The other intern (AM), who was not aware that the study was designed to assess compliance with the CONSORT checklist (or indeed any other measure of quality), checked each paper for whether a professional medical writer had been acknowledged (rated as yes, no, or unclear), and whether the paper had industry sponsorship. Although it was not always easy to infer the nature of any writing assistance from often vague statements in acknowledgements, we attempted to defi ne the involvement of a professional medical writer as someone who had had a role in drafting the manuscript, and if it was clear that only editing of an already complete manuscript was being acknowledged, we did not count that as writing assistance. A total score was calculated for each paper as the sum of the items that were completely described (minimum = 0, maximum = 22). If an item was not completely applicable, a full point was awarded if the paper described the parts that were applicable and contained suffi cient information to be sure other parts were not applicable. The primary analysis was a t-test of the difference in those scores between papers written by a professional medical writer and those that either were not or were unclear. A secondary analysis was done to compare industry-sponsored papers with those that did not declare industry sponsorship. As a sensitivity analysis, the total score was recalculated with the addition of half a point for each item that was present but incompletely described. As a further sensitivity analysis, the odds of completion of CONSORT items were investigated by logistic regression. Because items within a specifi c paper would be expected to be correlated, a random effects logistic regression model was used in which the paper was included as a random effect, and acknowledgement of a professional medTable 1. Characteristics of the included papers Source of funding No medical writer acknowledged Acknowledgement unclear Medical writer acknowledged Total Industry 60 (27%) 17 (8%) 147 (66%) 224 (100%) Other 9 (53%) 3 (18%) 5 (29%) 17 (100%) Total 69 (29%) 20 (8.3%) 152 (63%) 241 (100%) ical writer and the number of the CONSORT item were included as fi xed effects. Exploratory analyses were done to calculate the odds ratios and their confi dence intervals for completion of each CONSORT item individually. Results 241 papers were included in the study. Details of industry sponsorship and acknowledgement of professional medical writers are shown in Table 1. As expected for a journal that focuses on industry-sponsored research, the overwhelming majority of papers were industry sponsored, and a little over half clearly acknowledged assistance from a professional medical writer. Most CONSORT items were at least partially described in almost all papers, although some were less well described (Figure 1). Items that were particularly poorly described by both groups of writers were items 9 (concealment of random allocation), 10 (implementation of randomisation), and 14 (dates of recruitment and follow up periods). The frequency of reporting of each CONSORT item by medical writers and other writers is given in Table 2. Papers that acknowledged professional medical writers complied with more CONSORT items than those that did not (Table 3). The difference between groups was statistically signifi cant for the primary measure of counting only complete CONSORT items (difference between groups 0.75 items completed, 95% CI 0.07 to 1.43, P = 0.03) but not for the secondary measure in which half points were counted if items were present but incompletely described (difference between groups 0.53 items completed, 95% CI –0.02 to 1.07, P = 0.06).