To the Editor:
We thank Drs. Patel and Ahmad for their interest in and comments on our recently published prediction model for short-term cardiac resynchronization therapy (CRT) response.(1)
We agree that an accurate prediction of an outcome is paramount for shared decision-making.(2) We also agree that substantial differences in cardiac electrophysiology between men and women (3) appropriately raised the question of whether CRT candidates would benefit from a sex-specific prediction of CRT response. The development of a binary prediction model requires having data of adequate size, estimated based on the squared root of the mean squared prediction error and mean absolute prediction error of the model.(4) Assessment of the out-of-sample predictive performance of the model should consider three factors: the number of predictors, the total sample size, and the events fraction. We wholeheartedly agree with Drs. Patel and Ahmad that sex disparities in the CRT field are unacceptable and should be eliminated. A relatively small sample of women (as compared to men) in our study is an important limitation that should be taken into consideration.
Notably, our machine learning (ML) prediction model is fully open, with all coefficients reported. In contrast, the model developed by Cai et al.(5) is a “black-box” ML model with yet identical to our model discrimination performance (ROC AUC of 0.76).
Lastly, we thank Drs. Patel and Ahmad for highlighting an important feature of our study – focus on a short-term CRT response, opening up an opportunity for CRT optimization, targeting modifiable predictors of CRT response, and improving CRT delivery.
Funding :
This work was supported in part by HL118277, Medical Research Foundation of Oregon and OHSU President Bridge funding (LGT).
Disclosures:
SMART AV trial was sponsored by Boston Scientific.
Footnotes
Clinical trial registration—ClinicalTrials.gov Identifier: NCT00677014
References:
- 1.Howell SJ, Stivland T, Stein K, Ellenbogen KA, Tereshchenko LG. Using Machine-Learning for Prediction of the Response to Cardiac Resynchronization Therapy: The SMART-AV Study. JACC Clinical electrophysiology 2021. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Kattan MW, Cowen ME. Encyclopedia of medical decision making. Thousand Oaks, Calif.: SAGE Publications, 2009. [Google Scholar]
- 3.Howell SJ, German D, Bender A et al. Does sex modify an association of electrophysiological substrate with sudden cardiac death? The Atherosclerosis Risk in Communities (ARIC) study. Cardiovascular Digital Health Journal 2020;1:80–88. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.van Smeden M, Moons KG, de Groot JA et al. Sample size for binary logistic prediction models: Beyond events per variable criteria. Stat Methods Med Res 2019;28:2455–2474. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Cai C, Tafti AP, Ngufor C et al. Using ensemble of ensemble machine learning methods to predict outcomes of cardiac resynchronization. J Cardiovasc Electrophysiol 2021;32:2504–2514. [DOI] [PubMed] [Google Scholar]