Zettle, R. D. (2003). Acceptance and commitment therapy (ACT) versus systematic desensitization in treatment of mathematics anxiety. The Psychological Record, 53, 197-215.
This study compared ACT to systematic desensitization. The recent meta-analysis of the ACT literature included it and you can download the study from this website. Rob found that ACT worked better than a control condition, as did desensitization. On measures of math anxiety, ACT and desensitization were not statistically significantly different, but the d was in the wrong direction, and on general levels of overall anxiety, desensitization was better.
In 1999 ACT book we said this about fitting a fairly challenging intervention like ACT to the problem: "Before ACT begins, the client must be prepared for it. It can be an intensive intervention and the clients should not be subjected to such interventions lightly." In line with that, it seems possible that Math Anxiety is too low level of a problem to warrant ACT. Interestingly, in Zettle's study highly experientially avoidant participants did better in ACT than those who were not avoidant -- while the same did not hold true for desensitization.
This study has since been replicated and the results were a bit different:
Brown, L. A., Forman, E. M., Herbert, J. D., Hoffman, K. L., Yuen, E. K. and Goetter, E. M. (2011). A randomized controlled trial of acceptance-based behavior therapy and cognitive therapy for test anxiety: A pilot study. Behavior Modification, 35, 31-53.
It was a very small RCT (N = 16) for test anxiety comparing ACT (with mindfulness elements) and Beck’s cognitive therapy. Much like the Zettle study, there were similar outcomes on self-reports. In this study however ACT participants did better on actual test scores in school. So perhaps the Zettle result was not just due to a focus on a more minor population.