Covariate-based constrained randomization of group-randomized trials

Abstract

Group-randomized study designs are useful when individually randomized designs are either not possible, or will not be able to estimate the parameters of interest. Blocked and/or stratified (for example, pair-matched) designs have been used, and their properties statistically evaluated by many researchers. Group-randomized trials often have small numbers of experimental units, and strong, geographically induced between-unit correlation, which increase the chance of obtaining a “bad” randomization outcome. This article describes a procedure – random selection from a list of acceptable allocations – to allocate treatment conditions in a way that ensures balance on relevant covariates. Numerous individual- and group-level covariates can be balanced using exact or caliper criteria. Simulation results indicate that this method has good frequency properties, but some care may be needed not to overly constrain the randomization. There is a trade-off between achieving good balance through a highly constrained design, and jeopardizing the appearance of impartiality of the investigator and potentially departing from the nominal Type I error.

Get full access to this article

View all access options for this article.

References

1. Cook TD , Campbell DT . Quasi-experimentation: design and analysis issues for field settings. Chicago, IL: Rand-McNally, 1979, 354–356.

2. Donner A , Klar N . Statistical considerations in the design and analysis of community intervention trials. J Clin Epidemiol 1996; 49: 435–439.

3. Fortmann SP , Flora JA , Winkleby MA , Schooler C , Taylor CB , Farquhar JW . Community intervention trials: reflections on the Stanford Five-City Project Experience. Am J Epidemiol 1995; 142: 576–586.

4. Gail MH , Mark SD , Carroll RJ , Green SB , Pee D .On design considerations and randomization-based inference for community intervention trials. Stat Med 1996; 15: 1069–1092.

5. Diehr P , Martin DC , Koepsell T , Cheadle A . Breaking the matches in a paired t-test for community interventions when the number of pairs is small. Stat Med 1995; 14: 1491–1504.

6. Donner A . Sample size requirements for stratified cluster randomization designs. Stat Med 1992; 11: 743–750.

7. Freedman LS , Green SB , Byar DP . Assessing the gain in efficiency due to matching in a community intervention study. Stat Med 1990; 9: 943–952.

8. Martin DC , Diehr P , Perrin EB , Koepsell TD . The effect of matching on the power of randomized community intervention studies. Stat Med 1993; 12: 329–338.

9. Pocock SJ , Simon R . Sequential treatment assignment with balancing for prognostic factors in the controlled clinical trial. Biometrics 1975; 31: 103–115.

10.

10. Signorini DF , Leung O , Simes RJ , Beller E , Gebski VJ , Callaghan T . Dynamic balanced randomization for clinical trials. Stat Med 1993; 12: 2343–2350.

11.

11. Berry DA , Eick SG . Adaptive assignment versus balanced randomization in clinical trials: a decision analysis. Stat Med 1995; 14: 231–246.

12.

12. Rosenbaum PR . The central role of the propensity score in observational studies for causal effects. Biometrika 1983; 70: 41–55.

13.

13. Graham JW , Flay BR , Johnson CA , Hansen WB , Collins LM . Group comparability: A multiattribute utility measurement approach to the use of random assignment with small numbers of aggregated units. Evaluation Review 1984; 8: 247–260.

14.

14. Morris C , Hill J . The Health Insurance Experiment: design using the finite selection model. In Morton SC , Rolph JE eds. Public policy and statistics: case studies from RAND. New York: Springer-Verlag, 2000, 29–53.

15.

15. Royall RM . Current advances in sampling theory: implications for human observational studies. Am J Epidemiol 1976; 104: 463–474.

16.

16. Murray DM . Design and analysis of group-randomized trials. New York: Oxford University Press, 1998.

17.

17. Donner A , Klar NS . Design and analysis of cluster randomisation trials in health research. London: Arnold, 2000.

18.

18. Kalish LA , Begg CB . The impact of treatment allocation procedures on nominal significance levels and bias. Control Clin Trials 1987; 8: 121–135.

19.

19. Chen H. Effect of ignoring randomization constraints in the analysis of clinical trials. PhD dissertation. Baltimore: The Johns Hopkins University, 1997.

20.

20. Bailey R . Restricted randomization. Biometrika 1983; 70: 183–198.

21.

21. Fisher RA . The design of experiments, fourth edition (first 1935). New York: Hafner-Publishing Co. Inc., 1947.

22.

22. Bailey RA . Restricted randomization: a practical example. J Am Statist Assoc 1987; 82: 712–719.

23.

23. Gentleman JF . Algorithm AS 88: generation of all NCR combinations by simulating nested DO loops. Appl Statist 1975; 24: 374–376.

24.

24. Aptech Systems Inc. The GAUSS system. Maple Valley, WA: Aptech Systems Inc., 1999.

25.

25. Tukey J . Tightening the clinical trial. Control Clin Trials 1993; 14: 266–285.

26.

26. Moulton LH , O'Brien KL , Kohberger R et al. Design of a group-randomized Streptococcus pneumoniae vaccine trial. Control Clin Trials 2001; 22: 438–452.