Article

Meta-analyses involving cross-over trials: Methodological issues

Medical Statistics Unit, London School of Hygiene and Tropical Medicine, Keppel Street, London WC1E 7HT, UK.
International Journal of Epidemiology (Impact Factor: 9.18). 03/2002; 31(1):140-9.
Source: PubMed

ABSTRACT

Meta-analysis of randomized controlled trials (RCTs) is usually based on trials where patients are randomized individually into two different, parallel, treatment groups. This paper concentrates on RCTs of a different design-two-period, two-treatment cross-over trials.
The characteristics of these trials are outlined, with detailed examples of methods for analysis for both continuous and binary data. These case studies are then extended into the context of a meta-analysis. The Cochrane Library was surveyed to assess current practice for synthesis.
Methods are described for continuous and binary data for use both when the necessary paired data are given and also when they need to be calculated or imputed, and some suggestions are provided to help people wishing to synthesize data from cross-over trials into meta-analyses. The survey suggested that about 8% of the trials in the Cochrane library were cross-over trials and 18% of the reviews referred to such trials, although there was no consistent approach to their inclusion into the reviews.
Methods do exist for including valuable information from two-period, two-treatment cross-over trials into quantitative reviews. However, poor reporting of cross-over trials will often impede attempts to perform a meta-analysis using the available methods.

Download full-text

Full-text

Available from: Helen V Worthington
  • Source
    • "Sometimes, some clusters may remain empty when the person responsible does not include patients.32 These situations jeopardize the comparability of groups and lead to selection bias.33 The application of this type of study is limited to the clinical development of MDs. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Clinical evidence available for the assessment of medical devices (MDs) is frequently insufficient. New MDs should be subjected to high quality clinical studies to demonstrate their benefit to patients. The randomized controlled trial (RCT) is the study design reaching the highest level of evidence in order to demonstrate the efficacy of a new MD. However, the clinical context of some MDs makes it difficult to carry out a conventional RCT. The objectives of this review are to present problems related to conducting conventional RCTs and to identify other experimental designs, their limitations, and their applications. A systematic literature search was conducted for the period January 2000 to July 2012 by searching medical bibliographic databases. Problems related to conducting conventional RCTs of MDs were identified: timing the assessment, eligible population and recruitment, acceptability, blinding, choice of comparator group, and learning curve. Other types of experimental designs have been described. Zelen's design trials and randomized consent design trials facilitate the recruitment of patients, but can cause ethical problems to arise. Expertise-based RCTs involve randomization to a team that specializes in a given intervention. Sometimes, the feasibility of an expertise-based randomized trial may be greater than that of a conventional trial. Cross-over trials reduce the number of patients, but are not applicable when a learning curve is required. Sequential trials have the advantage of allowing a trial to be stopped early depending on the results of first inclusions, but they require an independent committee. Bayesian methods combine existing information with information from the ongoing trial. These methods are particularly useful in situations where the number of subjects is small. The disadvantage is the risk of including erroneous prior information. Other types of experimental designs exist when conventional trials cannot always be applied to the clinical development of MDs.
    Full-text · Article · Sep 2014 · Medical Devices: Evidence and Research
  • Source
    • "Furthermore, most studies employed small sample sizes (i.e., less than 20 subjects); only five studies employed sample sizes larger than 30 subjects. So, effect sizes should be interpreted with caution as they could be biased due to the small sample size of most of the trials (Elbourne et al., 2002; Durlak 2009). Furthermore, most studies included only male patients or a disproportionally small number of female patients. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Adults with Attention Deficit Hyperactivity Disorder (ADHD), especially teenagers and young adults, show important car driving impairments, including risky driving, accidents, fines and suspension of driver’s license. We systematically reviewed the efficacy of stimulant and non-stimulant drugs on driving performance of ADHD patients. We searched several databases for randomized controlled trials (RCTs) published through March, 2013. Fifteen RCTs (the majority with crossover design) evaluated methylphenidate (MPH) immediate-release (MPH-IR), MPH osmotic-controlled oral system (MPH-OROS), MPH transdermal system (MTS), extended-release mixed amphetamine salts (MAS-XR); atomoxetine (ATX) and lisdexamfetamine (LDX). Methods varied widely; including simulators and/or cars and different courses and scenarios. Various outcomes of driving performance, including a ‘composite’ or ‘overall’ driving score were considered. In general, stimulants improved driving performance in ADHD patients (either in RCTs conducted in simulators and/or cars). MPH-OROS improved driving performance compared with MAS-XR, placebo, or no-drug conditions. Although MPH-OROS and MPH-IR produced similar improvements during the day, MPH-IR lost its efficacy in the evening. MAS-XR also improved driving performance, but worsened driving performance in the evening. MTS (one study) showed a positive effect, but drug compliance varied widely across patients. LDX had positive effect on driving (two studies with the same sample). Studies with ATX report conflicting results. Improvement was more consistent in teenagers and young adults. In general, treatment with psychostimulants or ATX in therapeutic dosages had no negative impact on driving performance of ADHD patients. To conclude, treatment with stimulants in therapeutic doses improves driving performance in ADHD patients, especially teenagers and young adults.
    Full-text · Article · Sep 2014 · European Neuropsychopharmacology
  • Source
    • "LUTS progresses relatively slowly, long-term follow-up is not required and a carry-over effect of the intervention (changing one's position) seems not likely. Therefore, the use of a cross-over design is an adequate study design, not prone to result in methodological bias [29], [37]. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Background It is suggested that the body posture during urination can influence urodynamic parameters in patients with Lower Urinary Tract Symptoms (LUTS) to an extent approaching pharmacological interventions. In this article, the influence of body position during micturition on maximum urinary flow rate (Qmax), voiding time (TQ) and post-void residual volume (PVR) in healthy males and patients with LUTS is analyzed by means of a systematic review and meta-analysis. Evidence Acquisition A systematic search was conducted in 14 medical databases. Studies comparing urodynamic parameters in standing versus sitting position were eligible for inclusion. Studies were stratified according to health status of included male participants: healthy individuals and patients with LUTS. Standardized mean differences for Qmax, TQ and PVR were pooled in a random effects model. Results Eleven articles were included. In men with LUTS, a significantly lower PVR (−24.96 ml; 95%CI −48.70 to −1.23) was shown in sitting position compared to standing. In accordance, Qmax was increased (1.23 ml/s; 95%CI −1.02 to 3.48), and TQ was decreased (−0.62 s; 95%CI −1.66 to 0.42) in sitting position, although these differences did not reach statistical significance. In healthy men, Qmax (0.18 ml/s; 95% CI −1.67 to 2.02), TQ (0.49 s; 95%CI −3.30 to 4.27) and PVR (0.43 ml; 95%CI −0.79 to 1,65) were similar in sitting and standing position. Conclusion For healthy men, no difference is found in any of the urodynamic parameters. In patients with LUTS, the sitting position is linked with an improved urodynamic profile.
    Full-text · Article · Jul 2014 · PLoS ONE
Show more

Questions & Answers about this publication

  • Tea Reljic added an answer in RevMan:
    Can anyone help me to introduce data from a dichotomous outcome of a cross-over trial into Revman?

    We are conducting a meta-analysis that includes primary studies which compare different support surfaces for the prevention of pression ulcers. We have come across a cross-over trial and we are not quite sure of how to introduce data into Revman. We are thinking of adding up results from both periods and making the study similar to a parallel trial but we don´t know if this approach is correct.

    Tea Reljic

    Generally it is discouraged to add up the two periods and treat cross-over trials as a parallel study since the confidence intervals are likely to be too wide, and the study will receive too little weight. However the error will be on the conservative side (the cross-over studies will be under-weighted). Elbourne et al discuss methods for including cross-over studies in reviews.

    • Source
      [Show abstract] [Hide abstract]
      ABSTRACT: Meta-analysis of randomized controlled trials (RCTs) is usually based on trials where patients are randomized individually into two different, parallel, treatment groups. This paper concentrates on RCTs of a different design-two-period, two-treatment cross-over trials. The characteristics of these trials are outlined, with detailed examples of methods for analysis for both continuous and binary data. These case studies are then extended into the context of a meta-analysis. The Cochrane Library was surveyed to assess current practice for synthesis. Methods are described for continuous and binary data for use both when the necessary paired data are given and also when they need to be calculated or imputed, and some suggestions are provided to help people wishing to synthesize data from cross-over trials into meta-analyses. The survey suggested that about 8% of the trials in the Cochrane library were cross-over trials and 18% of the reviews referred to such trials, although there was no consistent approach to their inclusion into the reviews. Methods do exist for including valuable information from two-period, two-treatment cross-over trials into quantitative reviews. However, poor reporting of cross-over trials will often impede attempts to perform a meta-analysis using the available methods.
      Full-text · Article · Mar 2002 · International Journal of Epidemiology

    + 1 more attachment