Repetitive transcranial magnetic stimulation (rTMS) is an established treatment for major depressive disorder (MDD), yet treatment outcomes show substantial interindividual variability. Predicting individual response to rTMS has been explored by examining several feature domains including clinical and quantitative electroencephalography (QEEG). While research in the field moves towards elaborate machine learning (ML) models based on QEEG, the presumed predictive superiority of QEEG over clinical features remains empirically untested. The present study compared clinical and spectral QEEG features for rTMS response prediction within the same ML framework and sample using data from 119 MDD participants in the TDBRAIN repository. A cross-validated grid search was used for hyperparameter tuning, producing separate models for each feature set. The best clinical and spectral QEEG models obtained an AUC of 0.75 and 0.64 on the held-out test set, respectively. Whereas only the clinical model performed above chance, the two models did not differ significantly. These results challenge the presumed advantage of QEEG for rTMS response prediction, and future work should emphasize methodological standardization, transparency, and the integration of clinical and EEG domains to establish the true additive value of QEEG in ML-based prediction of rTMS treatment response in MDD.