Methodological bias in cluster randomised trials

Medical Research Collaborating Center, Seoul National University College of Medicine, 2nd Floor Cancer Research Institute Building, 28 Yongon Dong, Chongno Gu, Seoul 110-744, Korea. <>
BMC Medical Research Methodology (Impact Factor: 2.27). 02/2005; 5(1):10. DOI: 10.1186/1471-2288-5-10
Source: PubMed


Cluster randomised trials can be susceptible to a range of methodological problems. These problems are not commonly recognised by many researchers. In this paper we discuss the issues that can lead to bias in cluster trials.
We used a sample of cluster randomised trials from a recent review and from a systematic review of hip protectors. We compared the mean age of participants between intervention groups in a sample of 'good' cluster trials with a sample of potentially biased trials. We also compared the effect sizes, in a funnel plot, between hip protector trials that used individual randomisation compared with those that used cluster randomisation.
There is a tendency for cluster trials, with evidence methodological biases, to also show an age imbalance between treatment groups. In a funnel plot we show that all cluster trials show a large positive effect of hip protectors whilst individually randomised trials show a range of positive and negative effects, suggesting that cluster trials may be producing a biased estimate of effect.
Methodological biases in the design and execution of cluster randomised trials is frequent. Some of these biases associated with the use of cluster designs can be avoided through careful attention to the design of cluster trials. Firstly, if possible, individual allocation should be used. Secondly, if cluster allocation is required, then ideally participants should be identified before random allocation of the clusters. Third, if prior identification is not possible, then an independent recruiter should be used to recruit participants.

Download full-text


Available from: David Torgerson,
  • Source
    • "In CRTs, achieving balance on individual characteristics between treatment groups can sometimes be difficult. Indeed, clusters are sometimes randomized before patient inclusion, which increases the risk of selection bias because recruiters are no longer blinded to the treatment allocation which compromises allocation concealment [2] [3] [4] [5] [6] [7]. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Despite randomization, selection bias may occur in cluster randomized trials. Classical multivariable regression usually allows for adjusting treatment effect estimates with unbalanced covariates. However, for binary outcomes with low incidence, such a method may fail because of separation problems. This simulation study focused on the performance of propensity score (PS)-based methods to estimate relative risks from cluster randomized trials with binary outcomes with low incidence. The results suggested that among the different approaches used (multivariable regression, direct adjustment on PS, inverse weighting on PS, and stratification on PS), only direct adjustment on the PS fully corrected the bias and moreover had the best statistical properties. Copyright © 2014 John Wiley & Sons, Ltd.
    Statistics in Medicine 09/2014; 33(20). DOI:10.1002/sim.6185 · 1.83 Impact Factor
  • Source
    • "In the effort of minimizing the risk of bias caused by confounding parameters, randomized trials are needed. In a randomized trial, it may be methodologically preferable randomizing patients individually [40]. However, in an attempt to assess the impact of the implementation of recommendations made at a departmental level, it may be difficult to conduct organizational interventions at an individual level. "
    [Show abstract] [Hide abstract]
    ABSTRACT: ObjectiveIn hospital and health care organizational factors may be changed to reduce postoperative mortality. The aim of this study is to evaluate a possible association between mortality and ‘length of hospital stay’, ‘priority of surgery’, ‘time of surgery’, or ‘surgical delay’ in hip fracture surgery.DesignObservational cohort study.SettingProspectively and consecutively reported data from the Danish Anaesthesia Database were linked to The Danish National Registry of Patients and The Civil Registration System. Records on vital status, admittance, discharges, codes of diagnosis, anaesthetic and surgical procedures were retrieved.Participants6143 patients aged more than 65 years undergoing hip fracture surgery.Main Outcome MeasuresAll-cause mortality.ResultsThe one year mortality was 30% (28–31%, 95% Confidence interval (CI)). In a multivariate model ‘length of hospital stay’ less than 10 days and more than 20 days are associated with mortality with hazard ratios of 1.34 (1.20–1.53 CI, p<0.001) and 1.27 (1.06–1.51 CI, p<0.001), respectively. ‘Priority of surgery’ categorized as ‘non-scheduled’ is associated with mortality with a hazard ratio of 1.31 (1.13–1.50 CI, p<0.001). Surgical delay and time of surgery are not significantly associated with mortality.ConclusionNon-scheduled surgery and length of hospital stay were associated with increased mortality. Confounding by indication may bias observational studies evaluating early and late discharge as well as priority; therefore cluster randomized clinical trials comparing different clinical set ups may be warranted evaluating health care organizational factors.
    PLoS ONE 06/2014; 9(6):e99308. DOI:10.1371/journal.pone.0099308 · 3.23 Impact Factor
  • Source
    • "Randomization of PHCCs will be done once the caregivers eligible for the study have been chosen, so that recruitment by health professionals is not influenced by their allocation to a study group [28,29]. "
    [Show abstract] [Hide abstract]
    ABSTRACT: In Spain, family is the main source of care for dependent people. Numerous studies suggest that providing informal (unpaid) care during a prolonged period of time results in a morbidity-generating burden. Caregivers constitute a high-risk group that experiences elevated stress levels, which reduce their quality of life.Different strategies have been proposed to improve management of this phenomenon in order to minimize its impact, but definitive conclusions regarding their effectiveness are lacking.Methods/design: A community clinical trial is proposed, with a 1-year follow-up period, that is multicentric, controlled, parallel, and with randomized allocation of clusters in 20 health care centers within the Community of Madrid. The study's objective is to evaluate the effectiveness of a standard care intervention in primary health care (intervention CuidaCare) to improve the quality of life of the caregivers, measured at 0, 6, and 12 months after the intervention.One hundred and forty two subjects (71 from each group) >=65 years, identified by the nurse as the main caregivers, and who provide consent to participate in the study will be included.The main outcome variable will be perceived quality of life as measured by the Visual Analogue Scale (VAS) of EuroQol-5D (EQ-5D). The secondary outcome variables will be EQ-5D Dimensions, EQ-5D Index, nursing diagnosis, and Zarit's test. Prognostic variables will be recorded for the dependent patient and the caregiver.The principle analysis will be done by comparing the average change in EQ-5D VAS value before and after intervention between the two groups. All statistical tests will be performed as intention-to-treat. Prognostic factors' estimates will be adjusted by mixed-effects regression models. Possible confounding or effect-modifying factors will be taken into account. Assistance for the caregiver should be integrated into primary care services. In order to do so, incorporating standard, effective interventions with relevant outcome variables such as quality of life is necessary. Community care nurses are at a privileged position to develop interventions like the proposed one.Trial registration: This trial has been registered in under code number NCT 01478295.
    BMC Nursing 01/2014; 13(1):2. DOI:10.1186/1472-6955-13-2
Show more