Machine Learning Methods in Assessing the Risks of Target Organ Damage in Masked Hypertension

Aim To develop models for predicting the risk of target organs damage (TOD) in different phenotypes of “masked” arterial hypertension (MAH) based on methods of machine learning (ML).
Material and methods A retrospective cohort analysis was performed for 284 clinical records of patients (261 males, 23 females; median age, 38 years). Group 1 included 125 patients with grade 1-2 arterial hypertension (AH) and low or moderate risk; group 2 included 159 subjects with normal “office” blood pressure (BP) exposed to chronic professional stress. The 24-h BP monitoring (24-h BPM) and ultrasound examination of the heart and carotid arteries were performed; glomerular filtration rate (GFR) was estimated using the СКD-EPI formula. MAH was phenotyped by clustering 24-h BPM data, and the risk of TOD was predicted by analysis of odd ratios (OR) and with the ML methods, random forest (RF) and artificial neural networks (ANN). Data were analyzed using the R language in the RStudio environment.
Results According to results of the 24-h BPM and cluster analysis, 121 (76.1 %) subjects of group 2 had MAH. The MAH phenotypes were identified as follows: systolic-diastolic (SDMAH) (43.8 %); isolated systolic (ISMAH) (35.5 %), and isolated diastolic (IDMAH) (20.7%). As compared to stable AH, subjects with different MAH phenotypes showed both increases and decreases in individual 24-h BPM indexes. Thus, in subjects with IDMAH, mean 24-h values of systolic and diastolic BP were significantly lower than with AH while in SDMAH, they were considerably higher. The OR analysis demonstrated that odds of differently located TOD were associated with definite MAH phenotypes. With that, ISMAH was associated with the highest risk of glomerular hyperfiltration; IDMAH was associated with reduced GFR and vascular remodeling; and SDMAH was associated with left ventricular myocardial hypertrophy. The developed models for predicting the risk of TOD based on the RF and ANN methods showed a high accuracy, which was provided by multistep procedures of selecting the predictors and cross-validation.
Conclusion Modern ML technologies enhance the risk stratification of patients with different clinical variants of AH.

1. Geltzer B.I., Kotelnikov V.N., Vetrova O.O., Karpov R.S. Masked arterial hypertension: prevalence, pathophysiological determinants and clinical significance. Rossiiskii kardiologicheskii zhurnal.. 2019 24 (9): 92-98. (In Russ) DOI: 10.15829/1560-4071-2019-9-92-98

2. Pickering TG, Eguchi K, Kario K. Masked hypertension: a review. Hypertens Res. 2007; 30(6): 479-488. DOI: 10.1291/hypres.30.479.

3. Booth JN 3rd, Muntner P, Diaz K.M., Viera AJ, Bello NA, Schwartz JE, et al. Evaluation of criteria to detect masked hypertension. J. Clin. Hypertens. 2016; 18(11):1086-1094. DOI: 10.1111/jch.12830.

4. Peacock J, Diaz KM, Viera AJ, Schwartz JE, Shimbo D. Unmasking masked hypertension: prevalence, clinical implications, diagnosis, correlates, and future directions J Hum Hypertens. 2014 September; 28(9): 521–528. DOI:10.1038/jhh.2014.9.

5. Tientcheu D., Ayers C., Das S.R., McGuire D.K., de Lemos J.A., Khera A., Kaplan N., Victor R., Vongpatanasin W. Target organ complications and cardiovascular events associated with masked hypertension and white-coat hypertension: analysis from the Dallas heart study. J. Am. Coll. Cardiol. 2015; 66 (20): 2159-2169. DOI: 10.1016/j.jacc.2015.09.007.

6. Hänninen M.R., Niiranen T.J., Puukka P.J., Kesäniemi Y.A., Kähönen M., Jula A.M. Target organ damage and masked hypertension in the general population: the Finn-Home study. J. Hypertens. 2013; 31 (6): 1136-43. DOI: 10.1097/HJH.0b013e32835fa5dc.

7. Manios E, Michas F, Stamatelopoulos K, Koroboki E, Stellos K, Tsouma I, et al. Association of Isolated Systolic, Isolated Diastolic, and Systolic-Diastolic Masked Hypertension With Carotid Artery Intima-Media Thickness. The Journal of Clinical Hypertension. 2015. 17(1): 22-26. DOI: 10.1111/jch.12430.

8. Sharman JE, Hare JL, Thomas S, Davies JE, Leano R, Jenkins C, Marwick TH. Association of masked hypertension and left ventricular remodeling with the hypertensive response to exercise. Am J Hypertens 2011; 24 (8):898–903. DOI: 10.1038/ajh.2011.75.

9. Williams B, Mancia G, Spiering W, Agabiti Rosei E, Azizi M, Burnier M, et al.; ESC Scientific Document Group. 2018 ESC/ESH Guidelines for the management of arterial hypertension. Eur Heart J. 2018; 39 (33):3021-3104. DOI: 10.1093/eurheartj/ehy339.

10. Sitkova E.S., Mordovin V.F., Pekarsky S.E., Ripp T.M., Falkovskaya A.Yu., Ryabova T.R. et al. Blood pressure variability as a factor in the best cardioprotective efficacy of renal denervation. // Rossiiskii meditsinskii zhurnal. 2018; 33 (2): 9-15. (In Russ) DOI: 10.29001 / 2073-8552-2018-33-2-9-15.

11. McInnes L, Healy J, Saul N, Grossberger L. UMAP: Uniform Manifold Approximation and Projection The Journal of Open Source Software. 2017; 3(29):861 ·DOI: 10.21105/joss.00861

12. Campello RB, Moulavi D, Sander, J. Density-Based Clustering Based on Hierarchical Density Estimates. Proceedings of the 17th Pacific-Asia Conference on Knowledge Discovery in Databases, PAKDD 2013, Lecture Notes in Computer Science 7819, p. 160. DOI: 10.1007/978-3-642-37456-2_14

13. Günther F, Fritsch S. neuralnet: Training of Neural Networks. The R Journal. 2010; 2/1: 30-38. DOI: 10.32614/RJ-2010-006

14. Krittanawong С, Zhang Н, Wang Z, Aydar M, Kitai T. Artificial Intelligence in Precision Cardiovascular Medicine. Journal of the American College o f Cardiology. 2017. 69 (21): 2657-2664. DOI: 10.1016/j.jacc.2017.03.571.

15. Motwani M, Dey D, Berman DS, Germano G, Achenbach S, Al-Mallah MH, et al. Machine learning for prediction of all-cause mortality in patients with suspected coronary artery disease: a 5-year multicentre prospective registry analysis. Eur Heart J 2017; 38:500-507. DOI: 10.1093/eurheartj/ehw188.

16. Guidi G, Pettenati MC, Melillo P, Iadanza E. A machine learning system to improve heart failure patient assistance. IEEE J Biomed Health Inform 2014;18:1750–1756. DOI: 10.1109/JBHI.2014.2337752.