Clinical Trials

Clinical Trials: A Practical Guide ■❚❙❘The answers to some of these questions – such as the types of recruiting centers,the numbers recruited by each center, and whether the project was commerciallyfunded – are often found in the appendices or notes. Issues such as commercialfunding are also worth knowing since such studies tend to have higher recruitmentrates than non-commercially funded projects.The time between the end of follow-up and the time of publication can also be ofinterest. Most high-impact journals publish items 6–12 months after acceptance,unless fast tracked. Fast-tracked sections are available in some journals for trialsaddressing very topical or key issues, or for a major breakthrough. Editors of thesesections aim to review and publish a report in as little as 2 months.On the other hand, studies with results that are difficult to interpret, or withnegative results, might be submitted or accepted for publication much laterthan positive, high-profile studies. The delay could be due to many suggestionsfrom the reviewers – asking for additional work to improve the publication beforeacceptance – or to a number of previous journals considering and then rejectingthe paper, each consideration taking 4–8 weeks, since authors are restrictedto submitting the article to only one journal at a time. A journal might providereviewers’ comments on a web site, which can provide additional insights aboutthe reasons for a lag in publication time, such as requests for additional patientnumbers or reanalysis using other statistical methods.5. Are the observed treatment differences due to systematicerror (bias) or confounding?In a clinical trial, the observed treatment effect regarding the safety and efficacyof a new drug can appear to be clinically and statistically significant and yet mightbe due to the result of systematic error or bias within the study [10]. Even the mostcareful measurement and elegant statistical analysis can not salvage a biasedclinical trial, although learning about the mechanism of the bias might in itself beof scientific merit. The most common types of bias in clinical research are thoserelated to subject selection and outcome measurement. A reader should thereforereview a report with a question about whether such bias might have beenprevented during the design and conduction. For example, did the patients in thetrial get recruited in a clinical setting unrepresentative of the wider patientpopulation – eg, patients only recruited in one country with highly developedmedical services? This would then cause a geographical bias such that the resultsmight not be reproducible in a different national setting. In addition, exclusion ofsubjects from statistical analysis because of noncompliance or missing data couldbias an estimate of the true benefit of a treatment, particularly if more patients431