Random Forest Classification to Predict Response to High-Definition Transcranial Direct Current Stimulation for Tinnitus Relief: A Preliminary Feasibility Study.
Emilie Cardon, Laure Jacquemin, Martin Schecklmann and 7 others
PMID 35666538WHAT IT FOUND
A computer model predicted tDCS response in 99 patients with 86% accuracy, but failed when tested on 32 patients from a different clinic.
Without a sham control, it is impossible to know if the model predicts treatment benefit or natural fluctuation.
Key findings
01The random forest classifier achieved 85.71% accuracy on unseen data from the primary cohort, with 100% sensitivity and 81.48% specificity.
02When tested on an external dataset of 32 patients receiving traditional tDCS, the model's performance did not differ significantly from chance (AUC 0.635, P = 0.071).
03The study lacked a sham control group, so it is currently not possible to distinguish naturally occurring fluctuations of tinnitus severity from specific effects of the HD-tDCS.
STILL TO COME
How it was doneWhat they foundWhat it means for SLPs
Read the rest of this summary
You get three full summaries a month, free, and we do not ask for a card. Search, the TL;DRs and your library stay unlimited either way.
What it does not show
The study lacked a sham control group, so it is impossible to know if the model predicts a true treatment effect or just natural variation in tinnitus severity. The model failed to generalize to an external dataset, performing no better than chance on 32 patients from a different clinic. The sample size for the external validation was small, with only 6 responders, which makes the performance estimates unstable. The external validation used traditional tDCS and different outcome measures (THI instead of TFI), which may not be directly comparable to the HD-tDCS protocol used to train the model.
Declared interests
The authors declared no conflicts of interest. The study was supported by non-U.S. government funding.
The easy way to misread this
Do not interpret the 86% accuracy in the primary group as evidence that this tool can predict who will benefit from tDCS. The model failed to perform above chance on an external dataset, and the absence of a sham control means the 'response' it predicted could simply be natural fluctuation of tinnitus.