In our study (Deist et al. 20181) we collected 12 datasets from previous studies on post-radiotherapy outcomes (e.g., survival, toxicity etc.). The dataset outcomes were modelled with 6 classifiers using the other dataset parameters. In 1 out of 12 datasets we incorrectly identified the outcome parameter column. The manuscript needs minor updates in text and updates for all figures for correctness, but the overall conclusion remains completely unchanged. The dataset in question (set E in the published manuscript) concerned non-small cell lung cancer patients and the correct outcome was 2-year overall survival. However, we incorrectly built prediction models using vital status at last follow-up as the outcome. The effects on the manuscript are described below, followed by an erratum list. Additionally, an updated version of the manuscript (with Track Changes) is provided as a supplementary file. The effects on the results are negligible and the conclusion remains completely unchanged. However, minor updates throughout the manuscript are necessary for correctness. The erroneous processing of the dataset (originally set E) has been corrected and the analysis as described in the manuscript has been repeated. For this dataset this means that 500 new models have been generated in the experiment (5 folds * 100 repetitions), resulting in updated performance metrics for this set. In the manuscript we describe individual and aggregate results over all datasets and therefore several numbers in text and tables, and all figures need updating. Additionally, since we order the anonymized sets based on discrimination performance the erroneously processed set (originally set E) is now labelled set I, consequently causing changes in the labels of the other sets as well. With the correct outcome being modelled, the Areas Under the Curve (AUCs) drop for all classifiers for the dataset in question. Random forest (rf) remains the best ranking classifier, and elastic net logistic regression (glmnet) still ranks second. The ranking of the other 4 classifiers is now different for this dataset. In the published manuscript, we reported best classification performance across 12 datasets for rf and glmnet. This conclusion remains unchanged after the corrigendum. No effect on the median ranking of the other classifiers is observed, but the interquartile ranges have changed slightly as illustrated in the updated scatter-boxplot below. Furthermore, since we report results on a large number of datasets the effect of the erroneous processing of one dataset is limited, which is a confirmation of the robustness of the analysis performed and reported in the manuscript. Figure 2. [Color figure can be viewed at wileyonlinelibrary.com] Figure 3. Original [Color figure can be viewed at wileyonlinelibrary.com] Figure 4*. [Color figure can be viewed at wileyonlinelibrary.com] Figure. 5*. [Color figure can be viewed at wileyonlinelibrary.com] *Since we order the anonymized sets based on discrimination performance the erroneously processed set (originally set E) is now labelled set I, consequently causing changes in the labels of the other sets as well. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.

Purpose Machine learning classification algorithms (classifiers) for prediction of treatment response are becoming more popular in radiotherapy literature. General Machine learning literature provides evidence in favor of some classifier families (random forest, support vector machine, gradient boosting) in terms of classification performance. The purpose of this study is to compare such classifiers specifically for (chemo)radiotherapy datasets and to estimate their average discriminative performance for radiation treatment outcome prediction. Methods We collected 12 datasets (3496 patients) from prior studies on post‐(chemo)radiotherapy toxicity, survival, or tumor control with clinical, dosimetric, or blood biomarker features from multiple institutions and for different tumor sites, that is, (non‐)small‐cell lung cancer, head and neck cancer, and meningioma. Six common classification algorithms with built‐in feature selection (decision tree, random forest, neural network, support vector machine, elastic net logistic regression, LogitBoost) were applied on each dataset using the popular open‐source R package caret. The R code and documentation for the analysis are available online (https://github.com/timodeist/classifier_selection_code). All classifiers were run on each dataset in a 100‐repeated nested fivefold cross‐validation with hyperparameter tuning. Performance metrics (AUC, calibration slope and intercept, accuracy, Cohen's kappa, and Brier score) were computed. We ranked classifiers by AUC to determine which classifier is likely to also perform well in future studies. We simulated the benefit for potential investigators to select a certain classifier for a new dataset based on our study (pre‐selection based on other datasets) or estimating the best classifier for a dataset (set‐specific selection based on information from the new dataset) compared with uninformed classifier selection (random selection). Results Random forest (best in 6/12 datasets) and elastic net logistic regression (best in 4/12 datasets) showed the overall best discrimination, but there was no single best classifier across datasets. Both classifiers had a median AUC rank of 2. Preselection and set‐specific selection yielded a significant average AUC improvement of 0.02 and 0.02 over random selection with an average AUC rank improvement of 0.42 and 0.66, respectively. Conclusion Random forest and elastic net logistic regression yield higher discriminative performance in (chemo)radiotherapy outcome and toxicity prediction than other studied classifiers. Thus, one of these two classifiers should be the first choice for investigators when building classification models or to benchmark one's own modeling results against. Our results also show that an informed preselection of classifiers based on existing datasets can improve discrimination over random selection.