RESULTS from clinical trials are being seriously compromised by missing data yet the problem is receiving little attention from the research community, according to a prominent group of researchers.
“Substantial instances of missing data are a serious problem that undermines the scientific credibility of causal conclusions from clinical trials”, according to an article published in the New England Journal of Medicine. (1)
The article, which summarises the key findings from a US National Research Council report on missing data in clinical trials, was written by members of a panel assembled to address gaps in US Food and Drug Administration guidelines in this area. (2)
In the NEJM report, the authors wrote that participants in randomised controlled trials who discontinued treatment were a major source of missing data. An assumption that analysis methods could compensate for this could not be justified, they said.
Researchers should instead take more steps to prevent participants from discontinuing treatment, and put more effort into gaining consent for the collection of data on treatments and outcomes on those who did drop out, they wrote.
They provided 16 ideas for trial design and conduct to limit missing data.
In another article, published in Pediatrics, a group of researchers have also flagged the problem of missing data, in this case in trials registered at US clinical research website ClinicalTrials.gov. (3)
The researchers were studying the prevalence of paediatric trials compared to non-paediatric trials and found that children were underrepresented in clinical research.
They concluded that strategies were needed to decrease the amount of missing data, standardise the definition of certain variables, and improve the completeness of study results entered into ClinicalTrials.gov to “advance pediatric clinical trial infrastructure and methodology and improve child health”.
Professor Chris Del Mar, professor of public health at Bond University, agreed with the concerns. He said he had first-hand experience of the missing data problem through his involvement with a Cochrane review on the efficacy of Tamiflu when he was unable to obtain the primary data from the drug’s manufacturer Roche. (4)
“This is selective reporting and it makes us think that perhaps medical journals are becoming the marketing arm of pharma, because they are basically advertising the drugs but readers don’t have access to all the data”, he told MJA InSight.
“Missing data is terribly important and there are lots of causes of it and it can bias the results. This issue hasn’t had enough attention. It’s like a disease. It’s nasty and we need to vaccinate against it”, he said.
Professor John Eisman, a senior researcher with Garvan Institute of Medical Research, said the NEJM article mainly sought to address the problem of large dropout rates.
The more participants who remained in the study, the surer you could be of the relevance of the findings for the wider community, Professor Eisman said.
“The issue is how do you deal with dropouts — can you adjust for it? You can, but only sub-optimally”, he said.
However, one potential outcome of placing too much emphasis on minimising dropouts was a tendency to select very healthy people for a study, which again reduced the study’s relevance to the general population.
“I would say that there are other things to be concerned about. I honestly don’t believe this is such a major issue”, Professor Eisman said.
“The reality is that everything you do is going to be a compromise that relates to your capacity to do the study, the interest of the patient in the study, the cost of doing the study and a patient’s ethical right to withdraw when they choose.”
– Amanda Bryan
1. N Engl J Med 2012; 367:1355-1360
2. US National Research Council. The prevention and treatment of missing data in clinical trials. National Academies Press 2010
3. Pediatrics 2012; Online 1 October
4. Cochrane Summaries 2012; Online 14 March
Posted 8 October 2012