The risk of visit-to-visit blood pressure variability in chronic kidney disease

In 2010, Rothwell et al.[1] reported in Lancet that visit-to-visit variability (VVV) in SBP is a strong predictor of stroke, independent of mean SBP. The report was based on the retrospective analysis of patients after a transitory ischaemic attack from several cohorts (United Kingdom Transient Ischaemic Attack and three more) and from hypertensive patients of the Anglo-Scandinavian Cardiac Outcomes Trial-Blood Pressure Lowering Arm. In this trial, a significant relationship with coronary heart disease was also found in addition to the relationship with risk of stroke. The report raises interest in an issue that has traditionally been regarded as a barrier to the accurate measurement of blood pressure (BP), the high variability of BP. From the beginning, the concept was challenged in terms of whether the highest VVV is a cause or consequence of the underlying high risk. Moreover, interest has grown in understanding the mechanism underlying the association. The answer to these questions is not easy, in part, related to the absence of prospective studies. After the pioneer study, the potential impact of the VVV was retrospectively analysed in a large number of population cohorts, clinical trials and observational studies, and evidence supporting the Rothwell hypothesis has been reported. In parallel, studies looking for the determinants of VVV, the relationship with other parameters of variability and the hypothesis about explanations of the association were also a matter of interest resulting in a large number of publications. From the initial report, VVV appears as a key descriptor in 268 entries in PubMed. A great majority of the reported studies identified a positive relationship between VVV and mortality risk and/or other cardiovascular events, mainly stroke and less frequently coronary heart disease. In fact, secondary analysis from clinical trials of hypertension [2–5], dyslipidemia [6–8], and population-based studies [9–11] has been associated with risk of mortality [3–7], stroke [2,3,9,10], and coronary heart disease [3,6]. Furthermore, in diabetic patients, both mortality [12,13] and cardiovascular diseases [12–15] have been associated with VVV also. Similarly, in patients after a first event of coronary heart disease [16,17], stroke [18,19], or acute heart failure [20], the higher the VVV the higher is the risk for a second event or hospitalization. Similarly, high VVV of BP was associated with endothelial damage [21], increased arterial stiffness [22], progression of coronary artery calcification [23], and left ventricular diastolic dysfunction [24], all of which may contribute to risk of death and cardiovascular outcomes. All the reported data, however, were retrospective and the association to one kind of event or another sometimes depends on the age of population, underlying risk, length of follow-up, and number of events. In the present issue of the Journal of Hypertension, a retrospective study by Chang et al. in stage 3 and 4 chronic kidney disease (CKD) patients from a large community-based cohort of the Kaiser Permanente reports the impact of VVV in cardiovascular and renal events [24]. Among the 114 900 adults and using three different sets of VVV estimates, the highest versus the lowest quintile of the coefficient of variation was associated with higher rates of death and haemorrhagic stroke. However, VVV was inconsistently associated with heart failure, and not significantly associated with acute coronary syndrome, ischaemic stroke, or end-stage renal disease, perhaps because of the relatively low incidences of the last outcome. The results were consistent across the three VVV estimates, coefficient of variation, standard deviation, and average real variability. Previous studies have analysed the association of VVV and events in patients with diverse CKD stages [25,26], as well as in patients with end stage renal disease in renal replacement therapy, peritoneal [27] or haemodialysis [28]. Among the outcomes, surrogate endpoints such as changes in renal blood flow or albuminuria, and hard endpoints such as total and cardiovascular mortality were assessed. Overall, risk of total mortality [25], progression to end stage renal disease [28], and cardiovascular disease [25] associated with high VVV was observed. Similarly, high VVV was also associated to progression of albuminuria and worsening of renal function mainly in diabetic patients [15,29–31], although a few studies reported no association [32]. The study of Chang and co-workers [24] was based on real-life data collected in the electronic health records (EHR) of a large healthcare organization, which allowed combined research and surveillance programs that encompass whole populations. The EHR provides unique opportunities to enhance our understanding of the distribution and control of the underlying cardiovascular risk factors and other modifying conditions which impact patient outcomes [33]. When EHR are used, the quality of the data and potential selection bias are important issues to consider. In the present study, it seems that the BP readings were performed automatically following similar protocol in all the centres, although some bias during the sampling process can increase the possibility of having positive results. Patients with more BP measurements could be those with higher risk because of necessity of frequent visits for multimorbidities, and the higher the number of BP measurement the higher VVV because of a random effect. In fact, there was a high prevalence of comorbid conditions, as well as lower values of estimated glomerular filtration rate and higher prevalence of proteinuria, in the highest quintiles of VVV of BP compared with the lowest quintile. The question is whether the highest VVV is a cause or consequence of the higher risk condition, and in parallel to the higher number of BP measurements. In the present study of Chang et al.[24], information on the influence of the kind of antihypertensive drugs in the VVV was also presented. Baseline use of angiotensin-converting enzyme inhibitors or angiotensin II receptor blockers, β-blockers, and nitrates were associated with high BP VVV, whereas calcium channel blockers (CCB) and diuretics (DIU) were associated with low VVV. The impact of different classes of antihypertensive drugs has been assessed in several studies with controversial results [34–40], whereas some found differences among the classes of antihypertensives [35–38], in others, they were not found [34,39,40]. Pooling the studies, there was a trend to link CCB with less VVV [37] whereas angiotensin-converting enzyme and angiotensin II receptor blockers were linked with high VVV when they are used in monotherapy. In fixed-dose combinations, CCBs seem to be associated with low VVV compared with DIU [36]. No studies were performed previously in patients with CKD, such as the Chang study, a group of patients in which renal function influences the rate of BP control, the frequency of nondipping phenomenon, and the drug pharmacokinetics. In this regard, it is not unexpected that DIU, with water and sodium handling, and CCB with a lineal pharmacokinetics noninfluenced by renal function were those associated with the lowest VVV. Despite the impact of the potential antihypertensive class in the VVV during the antihypertensive treatment, what is more relevant is the treatment adherence to the drug therapy. Based on self-report or pill count in African-Americans with hypertensive CKD, lower medication adherence was associated with higher VVV [41], a conclusion that was supported by a study in which the association between VVV and outcomes was observed in the active arm of the study but not in the placebo arm [4]. Another study, however, suggests that low antihypertensive medication adherence explains only a small proportion of VVV [42]. The question is, if the VVV is a proxy of the better BP control. The fact that by losing weight and reducing salt intake VVV was reduced [38] also indicates that the VVV may be a proxy. Despite the presence of the present and other previous studies, more research is required to develop standardized definitions of VVV of BP, to confirm the value of VVV as a prognostic indicator, and to ascertain whether efforts to reduce VVV of BP in addition to mean BP will improve outcomes. ACKNOWLEDGEMENTS Conflicts of interest There are no conflicts of interest.