Is it difficult to calculate the sample size for a SMART?


No. In fact, in most cases it is quite easy. For the three most common primary aims in a SMART, the sample size formula/calculation is either the same one used to size a standard 2-arm RCT, or it requires only a very small and intuitive adjustment this to this calculation. 

Consider the generic prototypical SMART shown in the following schematic.

If the primary aim of the SMART is to test first-stage intervention 1 vs first-stage intervention 2 (this is the “main effect” of first-stage treatment) on an end of study outcome, then the sample size calculation is identical to the calculation in a 2-arm randomized trial.  

If, instead, the primary aim of the SMART is to test the effect of modest augmentation vs vigorous augmentation among non-responders to first-stage intervention, then the SMART’s total sample size calculation can be obtained easily in two easy steps: (1) use a standard 2-arm randomized trial calculator to get the total number of non-responders needed, (2) augment this number by the anticipated non-responder rate to get the total sample size for the SMART. 

If the primary aim of the SMART is to compare two adaptive interventions starting with different first-stage treatments, the sample size is again very easy to calculate in two steps: (1) use a standard 2-arm randomized trial calculator, (2) augment this number by the “SMART inflation factor” 

Clustered SMARTs and SMARTs with a repeated measures outcome have additional adjustment factors. And, as with any randomized trial, the study’s total sample size should be further adjusted to account for study drop-out or other reasons for missing data in the primary research outcome.




Join the d3center Mailing List

Keep up to date with the latest news, events, software releases, learning modules, and resources from the d3center.