. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Blood pressure and the uncertainty of prediction using hazard ratio Christian Torp-Pedersen 1,2,3 *, Rikke Nørmark Mortensen 1 , Jørgen Jeppesen 4 , and Thomas Alexander Gerds 5 1 Unit for Epidemiology and Biostatistics, Aalborg University Hospital, Denmark; 2 Department of Cardiology, Aalborg University Hospital, Denmark; 3 Department of Health, Science and Technology, Aalborg University, Denmark; 4 Department of Medicine, Amager Hvidovre Hospital Glostrup, University of Copenhagen, Denmark; and 5 Department of Biostatistics, University of Copenhagen, Denmark Online publish-ahead-of-print 28 November 2018 This commentary refers to ‘Blood pressure variability and risk of cardiovascular events and death in patients with hypertension and different baseline risks’, by M. H. Mehlum et al., 2018;39:2243–2251. Mehlum et al. 1 report that ‘Higher visit-to-visit systolic blood pres- sure variability is associated with increased risk of cardiovascular events in patients with hypertension, irrespective of baseline risk of cardiovascular events’. Unfortunately, the conclusions of this study are not supported by the analyses presented. The first problem with this study is conditioning on the future which is well known to be inappropriate. The main analysis is based on all available blood pressure values after 6 months and all events after 6 months and includes blood pressure measurements after start of time. The second problem is the way the standard deviation of the blood pressure measurements was calculated. The number of meas- urements and the time between the measurements was ignored. Therefore, it is not possible to derive a sound interpretation, and the results are potentially misleading. The third problem relates to prediction. The study seems to indi- cate that blood pressure variability provides predictive power be- yond blood pressure, but the study only provides hazard ratios and thus does not address prediction. A significant change of the hazard rate may or may not translate into a significant change of the pre- dicted risks. 2 A time horizon and an appropriate method to compare at the horizon is necessary. This was not provided, yet methods to examine this are available. A natural question would be: how much higher is the predictive accuracy of the model when blood pressure variability is added to a model that includes average blood pressure. This can be illustrated graphically and the difference reported. 2 A statistical evaluation of any improvement is made by evaluation whether a proper score 3 such as area under the receiver operator curve 4 is improved by adding the extra parameter. Conflict of interest: none declared. References 1. Mehlum MH, Liestøl K, Kjeldsen SE, Julius S, Hua TA, Rothwell PM, Mancia G, Parati G, Weber MA, Berge E. Blood pressure variability and risk of cardiovascular events and death in patients with hypertension and different baseline risks. Eur Heart J 2018;39:2243–2251. 2. Mortensen RN, Gerds TA, Jeppesen JL, Torp-Pedersen C. Office blood pressure or ambulatory blood pressure for the prediction of cardiovascular events. Eur Heart J 2017;38:3296–3304. 3. Gneiting T, Raftery AE. Strictly proper scoring rules, prediction, and estimation. J Am Stat Assoc 2007;102:359–378. 4. Gerds TA, Cai T, Schumacher M. The performance of risk prediction models. Biom J 2008;50:457–479. * Corresponding author. Tel: þ45 24453790, Email: ctp@heart.dk Published on behalf of the European Society of Cardiology. All rights reserved. V C The Author(s) 2018. For permissions, please email: journals.permissions@oup.com. European Heart Journal (2018) 39, 4219 DISCUSSION FORUM doi:10.1093/eurheartj/ehy595 Downloaded from https://academic.oup.com/eurheartj/article/39/47/4219/5212705 by guest on 23 October 2022