Improper analysis of trials randomised using stratified blocks or minimisation.
read more
Citations
Targeted Temperature Management at 33°C versus 36°C after Cardiac Arrest
When and how should multiple imputation be used for handling missing data in randomised clinical trials – a practical guide with flowcharts
Bevacizumab for newly diagnosed pleural mesothelioma in the Mesothelioma Avastin Cisplatin Pemetrexed Study (MAPS): a randomised, controlled, open-label, phase 3 trial
References
Practical statistics for medical research
CONSORT 2010 Explanation and Elaboration: updated guidelines for reporting parallel group randomised trials
Sequential treatment assignment with balancing for prognostic factors in the controlled clinical trial.
Related Papers (5)
Sequential treatment assignment with balancing for prognostic factors in the controlled clinical trial.
The risks and rewards of covariate adjustment in randomized trials: an assessment of 12 outcomes from 8 studies
Frequently Asked Questions (17)
Q2. How many simulations would estimate a confidence interval for coverage to within 0.5%?
Eight thousand replications were used so that if the true coverage was 95% then 8000 simulations would estimate a confidence interval for coverage to within ±0.5%.
Q3. What is the effect of balancing variables in the randomisation process?
as more balancing variables are used in the randomisation process, the within-stratum correlation will increase (assuming the balancing variables are associated with the outcome).
Q4. How many trials used centre as a balancing variable?
Centre was used as a balancing variable in 35 trials (54%), and 24 trials (37%) used at least one patient-level prognostic factor as a balancing factor.
Q5. What is the way to determine which balancing variables are not related to the outcome?
Given that the only effect of adjusting for a stratification variable which is unrelated to outcome is the loss of a degree of freedom, it is recommended that when stratification or minimisation has been used all analyses which estimate the treatment effect are pre-specified to be adjusted for balancing factors.
Q6. How many trials were excluded from the analysis?
Of these, seven were excluded: three were cluster-randomised trials, one was a crossover trial, two were single arm studies, and one was a secondary analysis of a study that had been previously reported within the review period.
Q7. How many blocks were used in the second set of simulations?
in order to investigate the effect of block size on the results, the second set of simulations used block sizes of 2, 8, and 32.
Q8. What is the way to determine which balancing variables are not related to outcome?
Determining which balancing variables are not related to outcome within a trial will rely on post-hoc analysis or preliminary testing (where the method of analysis depends on the results of a preliminary significance test).
Q9. What is the way to adjust for balancing variables?
Given that balancing variables should only be chosen if they are highly prognostic, it is likely that the effect sizes seen in their simulations would be similar to those seen in actual trials using stratification or minimisation.
Q10. What is the effect of ignoring balancing variables in the analysis?
This shows that ignoring balancing variables in the analysis will lead to confidence intervals that do not have nominal coverage.
Q11. What is the effect of ignoring the stratification or minimisation variables in the analysis?
A number of simulation studies have shown that ignoring the stratification or minimisation variables in the analysis may lead to invalid tests of significance[27, 28, 29, 30, 31].
Q12. What does the term "Negligible within-stratum correlation" mean?
In practice, this means the authors should always expect non-negligible within-stratum correlation in stratified trials, as variables should only be used in balancing if they are expected to be related to outcome.
Q13. What is the definition of a well conducted trial?
Well conducted randomised controlled trials are considered the gold standard for unbiased comparison of treatments as they ensure there are no systematic differences between treatment groups.
Q14. What is the reason why some people are mistrustful of adjusted analyses?
Some people are inherently mistrustful of adjusted analyses as they feel the investigators may have used a variable selection technique that leads to biased results, or performed several different analyses and only presented those which were most favourable.
Q15. What are the issues that lead to a loss of interpretability?
These issues lead to a loss of interpretability: the authors can no longer interpret a 95% confidence interval as nominal, or p-values as the probability of observing a result as or more extreme under the null hypothesis.
Q16. What was the effect of the adjusted analyses on coverage rates?
With small sample sizes for binary or time-to-event outcomes, adjusted analyses resulted in coverage rates that were too low (Table 2).
Q17. What is the basic method of allocating patients to a treatment?
The most basic method of allocating patients to a treatment is simple randomisation[1], where the probability of being assigned to either treatment is the same for all patients.