There is growing interest in news releases from scientific conferences and medical journals. Studies with accompanying press releases are better represented in the lay media and considered more newsworthy by health journalists [2], [3]. Yet, many caution that press releases do not serve as a precise medical reference. They frequently report incomplete information or omit basic study data [4]. In addition, an editorial process to confirm accuracy of the results is usually lacking.

No standard format to evaluate press releases exists. However, a previous study developed a coding scheme to record quality measures relevant to journal press releases (4 and Woloshin, personal communication, 2005). We modified this tool with what we believe to be important descriptors of press release quality based on other methodological tools used for critical appraisal [7]. These include descriptors such as reporting of study size, study design, study subjects, duration of follow-up and mention of harm or limitations of the research (Appendix S2). We also documented if results were quantified as relative or absolute risks. We performed PubMed and MEDLINE searches until August 2007 (minimum 20 months) to identify subsequent publication of abstracts presented at scientific meetings and categorized them based on their Institute for Scientific Information 2007 impact factor [8]. Journals were classified as “high-impact” if their impact factor was among the top 10 in their relevant category.

More than half (59%) of the press releases reported results presented as abstracts at a scientific meeting (Table 1). Some (20%) described studies published in peer-reviewed journals but many (38%) did not provide full reference to the corresponding journal article. The remainder of releases (21%) did not mention the source of original data at all.

Basic study details were frequently reported. Of those reporting study design, 67% were randomized controlled trials (RCTs), 14% were uncontrolled trials and the remainder (9%) included survey and cohort studies. Study size was frequently recorded. The majority (60%) were medium-sized studies (31–1000 people) involving human subjects (99%). Follow-up time and quantification of study results was included in 71% and more common in press releases of RCTs. When quantified, 27% provided a base rate while 73% presented a ratio measure, most commonly as a relative risk.

Harms or adverse events were cited in about three-quarters (76%) of releases but study limitations (6%) were rarely noted, with the majority (62%) reporting the results as preliminary. Almost one-half of releases (47%) quoted a study author who was usually the principal investigator. The authors' comments typically emphasized the benefit of the intervention but only 10% described limitations of the research. Additional sources of study funding were also reported (42%) and typically involved partnership with another pharmaceutical company.

Of the 138 scientific meeting-related press releases, 55 (41%) abstracts, primarily of RCTs, were subsequently published in high-impact journals, 17 (12%) in low-impact journals and 64 (47%) remained unpublished when evaluated at a minimum of 20 months after the scientific conferences (Figure 1).

Our study shows that industry press releases are generally good at providing basic study details, but may provide incomplete information needed for a reader to gauge a study's clinical implications. Many press releases focused on research presented at scientific conferences, often considered the forefront of current knowledge. However, this information is frequently based on preliminary data without peer-review and may omit limitations such as small study size, uncontrolled study design or short duration of follow up [10], [11]. In turn, this can mislead readers about the validity of results [12]. Furthermore, several releases were not explicit about the source of original data and only a fraction noted any study limitations. A third of releases did not quantify study results, and of those that did, results were presented using a ratio measure - a format known to exaggerate the perceived magnitude of effect [13].

It is not surprising that randomized controlled trials were overrepresented in press releases as these studies have considerable influence on clinical practice. In addition, widespread illnesses such as heart disease, cancer and HIV/AIDS were the most cited medical disciplines. Previous work has shown that medications used to treat chronic disorders are the focus of considerable advertising efforts by pharmaceutical companies and may garner more media attention [14], [15].

We found no other work evaluating press releases from pharmaceutical companies. However, Woloshin et al. examined the quality of press releases issued by medical journals [4]. Similar to our data, they found that releases do not routinely highlight limitations and often use formats that exaggerate study findings.

Our study has several limitations. First, we only examined archived releases for the year 2005 available in electronic format on corporate websites. This led to the exclusion of fifth-ranked Merck & Co. While results may not be generalizable to smaller drug companies, we selected the world's most successful pharmaceutical manufacturers who all have strong media presence and public relations practices. Second, we did not corroborate press release content with subsequent publication, as our aim was to report study characteristics rather than test data accuracy. This approach may be justified as previous work has shown that abstracts often differ from published peer-reviewed results. In a study examining transition of scientific meeting abstracts to full-length journal article, 41% of publications exhibited significant discrepancies when compared to the original abstract data [16]. Third, the findings are predicated on our construct of what constitutes a good quality press release. Our instrument was based on previous work and established guides but results may differ with other assessments tools. Lastly, while we know that journalists frequently rely on press releases as a source of health topic ideas, the association between releases and ensuing coverage in newspaper, radio or television stories was not investigated [17].

Our findings suggest that the quality of pharmaceutical press releases needs improvement. Simple ways of enhancing quality would include referencing original data to confirm press release content, quantifying study results in clinically meaningful ways, and identifying important limitations so that the generalizability of results across different populations and settings can be appreciated.

It is unlikely that the pharmaceutical industry will change its reporting of original research. Thus, journalists and newsmakers should be aware of the shortcomings of press release data and learn to scrutinize this information before adoption into health news for public consumption. Creating a standardized “checklist” of quality indicators is one possible appraisal tool for readers to use.

Appendix S1