Brain Stimulation Effects on Food Cravings and Consumption: An Update on Lowe et al. (2017) and a Response to Generoso et al. (2017)
There were two limitations of this original meta-analysis: 1) we did not account for correlation between pretreatment and posttreatment measures (because no studies reported these data), thereby effectively treating within-subjects designs as between-subjects designs, and 2) in the approximately 12 months since the article was accepted for publication, findings from four additional studies have emerged involving neuromodulation for food cravings, including one of our own. We were initially concerned that our methodological choice coupled with the emergence of new data might affect the pattern and interpretation of our overall findings. Accordingly, we recalculated all of our effects using an imputed value of r = .50 to account for pretest to posttest correlations among dependent measures in within-subjects designs. A correlation of r = .50 provides a conservative estimate of correlation among dependent measures, given that most pretest to posttest correlations are greater than .50 (2). In addition, we included findings from the four new studies (one unpublished) with single-session neuromodulation data (3–5) and one newly published study from which we originally drew partial unpublished data (6); these significantly augment the study database from which we conducted the original meta-analysis (7–17).
Findings from the revised and expanded analyses are presented in Tables 1 and 2. We again found a significant overall neuromodulation effect on cravings in single-session format, with a replicated significant effect for rTMS, but not tDCS. We also added a laterality analysis for the craving effects (enabled now by the additional studies), which showed apparent superiority for neuromodulation applied to the left- over the right-sided application. Of additional novel interest is the fact that we also found a significant effect of neuromodulation on food consumption, which seems to confirm our conjecture that in the original meta-analysis the null findings were due to low statistical power.
Together, these updated findings largely support our three original conclusions: 1) there is a significant effect of single-session neuromodulation on cravings, of moderate magnitude, 2) the effect is significant for rTMS but not tDCS, and 3) statistical power may have interfered with our ability to detect a significant single-session neuromodulation effect on consumption.
In their “Letter to the Editor” in response to our original meta-analytic review, Generoso and colleagues (18) critique our decision to include cTBS study data, given that it is not an excitatory technique. The authors present a revised analysis focusing on rTMS methods, excluding cTBS but including data from two publications (19,20) not originally included in our original meta-analysis; the authors conclude that the effects of rTMS are no longer significant in their new analysis.
The authors' suggestion that excitatory rTMS is of more direct clinical relevance is well taken, and we agree that quantifying the effects of excitatory techniques in isolation is appropriate when considering the possible clinical use of rTMS for reducing cravings.