Research highlights, 14 January 2011

Research questions “Research highlights” is a weekly round-up of research papers appearing in the print BMJ. We start off with this week’s research questions, before providing more detail on some individual research papers and accompanying articles.

Treatment for scorpion sting in rural India
Bawaskar and Bawaskar’s randomised head to head trial on treating scorpion sting in rural India is not a typical BMJ paper. But editorialists Edward Mills and Nathan Ford highlight its worth as “a reminder that global health researchers often neglect conditions that matter to large impoverished communities.”

Faced with Mesobuthus tamulus sting as a common and potentially life threatening presentation in their rural practice, the researchers tried to clarify the role of treatment with scorpion antivenom. They found that when used in combination with the standard treatment, prazosin, antivenom greatly shortened patients’ recovery times compared with prazosin alone.

The preliminary, specialised, and localised nature of the research might have ruled it out for publication in the BMJ, but reviewers and editors were impressed by the trialists’ efforts to address an important public health problem with very limited resources. A shortcoming was that the primary outcome, time to recovery, was a subjective one assessed by the (unblinded) investigators themselves. But this was the best available option under the circumstances, and certainly not a simple one to implement – the researchers sat by victims’ bedsides for hours monitoring symptoms.

Given the importance of this problem to some readers, and its paradoxical neglect – which may in part explain why antivenom remains expensive and inaccessible – we concluded that the findings deserved an audience. We also allowed the researchers to register the trial retrospectively: when they started the trial, Clinical Trials Registry – India didn’t exist, and the authors weren’t aware of international journals’ requirements for registration.

Biased reporting of outcomes in clinical trials
Many journals, including the BMJ, will welcome a trial with “negative” results as long as the study is robust and adequately powered. Avoiding bias against such studies helps to balance the evidence base, and it may be important to know that an intervention didn’t work.

Editors’ decisions are not, however, the only cause of publication bias; authors also skew the scientific record by failing to report fully all the study outcomes they had planned to measure. Rosalind Smyth and colleagues contacted authors of 268 randomised controlled trials (183 identified in Cochrane systematic reviews as probably reporting outcomes in a biased way and a further 85 randomly selected from PubMed ). In all, 161 authors replied and, eventually, 59 took part in phone interviews to discuss the reporting of their work in protocols and papers. More than a quarter said that they had analysed but chosen not to report some prespecified trial outcomes, and in all but one case this made the conclusions more positive than they should have been. It was just as common to measure and then not analyse some outcomes.

Smyth and colleagues found only one case of outright manipulation, however, and mostly attribute the rest to researchers’ lack of awareness that they should report everything accurately and fully. Their own study’s low response rate introduces some bias too, but it’s unlikely that only the responders had something to admit.

Sven Trelle and colleagues’ network meta-analysis of the cardiovascular safety of seven non-steroidal anti-inflammatory drugs came up against the same problem. They analysed unpublished data for the trials of celecoxib and lumiracoxib, but were denied access to unpublished safety data from trials of rofecoxib and etoricoxib. Many of the included outcome data were reported incompletely and erratically, with important discrepancies in the reported number of events between different sources of information for major trials. Naproxen seemed least harmful but the authors warn that “evidence is lacking to suggest that any of the investigated drugs are safe in cardiovascular terms.”

Improving chlamydia screening for men
In England, twice as many women are screened for chlamydia as men. Katy Turner and colleagues have modelled data reported by the National Chlamydia Screening Programme in 2008-9 to determine how best to tackle this gender inequity – either by increasing coverage of primary screening in men or increasing the efficacy of partner notification.

Increasing male coverage of screening to the same level as female coverage caught almost 40000 extra cases, whereas doubling the efficacy of partner notification picked up far fewer: around 19000 additional diagnoses. However, upping partner notification reduced the cost per infection treated by £57, but higher male coverage increased the cost by £22. Increasing the efficacy of partner notification also reduced the sex ratio (female:male) of treatment of infected individuals from 2.0:1 to 1.4:1.

Overall, improving the efficacy of partner notification would add an extra £3.3m to the £46.3m cost of the programme, compared with a further £22.9m if screening of men was increased. The authors conclude that: “the additional resources required to increase male screening coverage to reach equity with females would be more effectively employed in achieving high partner notification efficacy among those who test positive.”
Research online: For this and other new research articles see

A new approach to screening for Down’s syndrome
Current techniques for prenatal diagnosis of trisomy 21, which causes most cases of Down’s syndrome, are invasive and associated with a small risk of miscarriage. Rossa Chiu and colleagues have looked into whether measuring the amount of fetal chromosome 21 in the blood of pregnant women is a suitable alternative .

The authors tested 2-plex and 8-plex massively parallel sequencing of DNA molecules in plasma from 753 women to determine the proportion of fetal DNA molecules that originated from chromosome 21; cases where this proportion was above a certain threshold were classed as having a trisomy 21 fetus. When they validated their findings against full karyotyping of prospectively collected or archived samples, the researchers found that the 2-plex protocol detected trisomy 21 fetuses with 100% sensitivity and 97.9% specificity.

This approach might be used to rule out trisomy 21 among high risk pregnancies, reducing the number of cases that require amniocentesis or chorionic villus sampling.