How is the timing of clinical trial registration related to clinical trial characteristics and risk of bias?




Poster session 1


Sunday 16 September 2018 - 12:30 to 14:00

All authors in correct order:

Tan A1, Jiang I1, Askie L1, Hunter K1, Seidler AL1
1 NHMRC Clinical Trials Centre, University of Sydney, Australia
Presenting author and contact person

Presenting author:

Anna Lene Seidler

Contact person:

Abstract text
Prospective trial registration aims to reduce selective outcome reporting and publication bias and improve accountability and transparency. It has been proposed that systematic reviews should only include prospectively registered trials. A recent review of fertility treatment trials advised against this, noting high rates of unregistered trials, and that registration is not always a good predictor of risk of bias. It is currently unclear how registration status and timing are associated with trial characteristics and risk of bias across all areas of health research.

The aims of this study were to:
1) determine adherence to trial registration policies in health research;
2) determine whether trial registration varies depending on trial characteristics; and
3) analyse the relation between registration and risk of bias.

We included all trials published from January to June 2017 in 28 high- and low-impact factor general and specialty medicine journals. We extracted detailed information on registration status and timing (prospective registration, retrospective before or after completion of enrolment, or unregistered) and study characteristics (including sample size, industry funding, and studied health condition) from all studies. We assessed risk of bias in a stratified random sample of these trials using the Cochrane 'Risk of bias' tool.

We identified 373 trials, and assessed risk of bias in a sample of 183. Registration rates were high (Figure 1): 95% of trials were registered prospectively or retrospectively before completion of recruitment. Larger sample size, industry funding, and multiple recruitment sites were predictors of earlier registration. Prospectively registered trials had a significantly lower risk of bias across all domains compared to unregistered and retrospectively registered trials (Figure 2).

Our results suggest an increase in registration rates, compared to previous studies. Using a more detailed classification of registration status enabled us to create a thorough picture of the relation between trial characteristics, registration, and risk of bias. This will be useful in informing the debate about risk of bias of unregistered and retrospectively registered trials, and their inclusion in systematic reviews.

Patient involvement:
None. Addressing risk of bias can ultimately improve healthcare.


Relevance to patients and consumers: 

Bias in health research can over-estimate true treatment effects, thereby potentially misguiding evidence-based clinical practice guidelines. Detecting and minimising risk of bias can ultimately improve healthcare by enabling researchers and guideline developers to conduct higher quality systematic reviews.