Alan Talevi

Eduardo Franco

eLife Assessment
This valuable study reports machine learning models derived from large-scale data to predict the risk of post-stroke epilepsy. The evidence supporting the conclusions is convincing, although there are some validation issues (lack of cross-validation, possible bias in external validation results). The study may be of interest in the field of clinical neurology

anonymous

Reviewer #1 (Public review):
Summary:
This is a large cohort of ischemic stroke patients from a single centre. The author successfully set up predictive models for PTS.
Strengths:
The design and implementation of the trial are acceptable, with the credibility of the results. It may provide evidence of seizure prevention in the field of stroke treatment.
Weaknesses:
My concerns are well responded to.

Reviewer #2 (Public review):
Summary
The authors present multiple machine-learning methodologies to predict post-stroke epilepsy (PSE) from admission clinical data.
Strengths
The Statistical Approach section is very well written. The approaches used in this section are very sensible for the data in question.
Typos have now been addressed and improved interpretability has been added to the paper, which is appreciated.
Weaknesses
The authors have clarified that the first features available for each patient have been used. However, they have not shown that these features did not occur before the time of post-stroke epilepsy. Explicit clarification of this should be performed.
The likely impact of the work on the field
If this model works as claimed, it will be useful for predicting PSE. This has some direct clinical utility.
Analysis of features contributing to PSE may provide clinical researchers with ideas for further research on the underlying aetiology of PSE.

Reviewer #3 (Public review):
Summary:
The authors report the performance of a series of machine learning models inferred from a large-scale dataset and externally validated with an independent cohort of patients, to predict the risk of post-stroke epilepsy. Some of the reported models have very good explicative performance, and seem to have very good predictive ability.
Strengths:
The models have been derived from real-world large-scale data.
Performances of the best-performing models seem to be very good according to the external validation results.
Early prediction of risk of post-stroke epilepsy would be of high interest to implement early therapeutic interventions that could improve prognosis.
Code is publicly available. The authors also stated that the datasets used are available on request.
Weaknesses:
The writing of the article may be significantly improved.
Although the external validation is appreciated, cross-validation to check robustness of the models would also be welcome.
External validation results may be biased/overoptimistic, since the authors informed that "The external validation cohort focused more on collecting positive cases 80 to examine the model's ability to identify positive samples", which may result in overoptimistic PPV and Sensitivity estimations. The specificity for the external validation set has not been disclosed.

Predictive models for secondary epilepsy within 1 year in patients with acute ischemic stroke: a multicenter retrospective study

Figures and data

Single factor significant analysis results

Univariable logistic regression results

Multivariable logistic regression results

The AUC,Accuracy, Sensitivity, Specificity, F1-score, positive predictive value (PPV) and negative predictive value (NPV) of ML models of train|test groups

Be the first to read new articles from eLife