Paul Glasziou: How many journals do you need to read?

Do you regularly read Chronobiology International? No, me neither. But that was the source journal for the article we read at a recent GP journal club I attended in Perth (Western Australia, not Scotland). It reported the 5.6 year follow up results of a 2,000 patient randomised trial that compared evening to morning dosing of antihypertensives[1]. That’s a clinically important question for GPs. The rather surprising result was that night time dosing led to a statistically significant halving of mortality. However the intermediate effects—of slightly lower daytime blood pressure and less nocturnal dipping—did not seem enough to explain the impact. The triallists were probably “lucky” and the reduction found was an overestimate. Nevertheless, given the lack of harms, we agreed it was reasonable to recommend evening dosing (and fitted in with how folk should take statins).

While the trial result was surprising, equally surprising was the source journal. Though none of the GP group had previously heard of Chronobiology International, this study seemed very relevant to primary care management. Why was such a trial not in one of the general medical or primary care journals we might read?* A good example of Bradford’s law of scatter, which says that the last third of relevant studies will be widely scattered (in a ratio of 1:n:n2 across the 3rds).

So how many journals does a clinician need to read to find all the trials in their speciality? We recently looked at a dozen specialities, and neurology appeared the worst [2]. There were 2770 neurology relevant trials in 2009 which were scattered across 896 journals (about one sixth of the 5,632 journals currently indexed for MEDLINE). Of course, no neurologist could, or would want to, read all 2,770 trials, but to find even quarter of these would require an estimated 29 journals, and to find half would require 114 journals. And this includes neurology specialist journals, general medical journals, and journals of related specialities such as rehabilitation.

Because systematic reviews summarize a number of trials, they are less scattered: 547 reviews in 292 journals. But still many more journals than the typical two to six journals a clinician will receive (and maybe read some of). So the traditional process of reading the local general medical journal and a couple of speciality ones will mean many important developments are missed. A couple of options are McMaster’s ACCESSSS—a free but carefully filtered site that provides updates on recent high quality studies; another evolving option is PubMedHealth, which aims to have in one place, in a single search, most systematic reviews (not restricted to Medline-indexed journals). Each clinician and speciality group will need to organise their own processes to address this scatter of research information, but the first step is awareness there is a problem.

Paul Glasziou is professor of evidence based medicine at Bond University and a part-time general practitioner.

* PS Dr Hermida informed me that it was rejected by other journals because “nobody else ever reported similar findings,” and a confirmatory larger trial is ongoing.

1. Hermida RC, Ayala DE, Mojón A, Fernández JR. Influence of circadian time of hypertension treatment on cardiovascular risk: results of the MAPEC study. Chronobiol Int. 2010;27(8):1629-51.

2. Hoffmann T, Erueti C, Thorning S, Glasziou P. The scatter of research: cross sectional comparison of randomised trials and systematic reviews across specialties. BMJ. 2012;344:e3223