## Introduction

Office blood pressure (BP) measurement is the gold standard for the management of hypertension [1,2,3]. Physicians often observe fluctuations of office BP between visits, and these fluctuations are considered “noise,” resulting in inaccurate BP values. Therefore, the measurement of out-of-office BP, i.e., ambulatory and home monitored BP, has been recommended in international hypertension guidelines and is widely used in clinical practice because of superior results, reproducibility, and evaluation of the effects of antihypertensive drugs, compared with those based on office BP measurement. As a result, many studies reported that ambulatory BP monitoring or home measurement had prognostic value for cardiovascular morbidity and mortality greater than that based on office BP measurement in the hypertensive population. Thus, until recently, the importance of out-of-office BP measurement has been emphasized for the management of hypertension. However, Rothwell et al. [4] reported the clinical impact of visit-to-visit BP variability in 2010, and the Systolic Blood Pressure Intervention Trial (SPRINT) demonstrated that strict control of systolic BP (<120 mmHg) using office-based measurement compared with standard treatment of systolic BP (<140 mmHg) in high-risk populations without diabetes or prior stroke significantly reduced cardiovascular events [5]. The measurement of office BP subsequently regained prominence in clinical and research settings.

### BP variability observed in the office

Rothwell et al. [4, 6] simultaneously published three interesting reports on visit-to-visit BP variability in 2010. First, the United Kingdom Transient Ischemic Attack Aspirin (UK-TIA) trial reported that the top decile group of visit-to-visit systolic BP (SBP) variability in a population with prior cerebrovascular events was a strong predictor of stroke incidence, independent of mean SBP level, compared with that in the lowest decile group (hazard ratio [HR]: 6.22, 95% confidence interval [CI]: 4.16–9.29, P < 0.0001) [4]. In addition, the top decile of maximum SBP during follow-up was also associated with stroke incidence, independent of mean SBP level (HR: 12.08, 95% CI: 7.40–19.72, P < 0.0001). The Anglo-Scandinavian Cardiac Outcomes Trial Blood Pressure Lowering Arm (ASCOT-BPLA) study reported that residual visit-to-visit variability of SBP, i.e., BP variability under antihypertensive treatment, was associated with stroke incidence. Although ASCOT-BPLA was randomized into amlodipine (plus perindopril) and atenolol (plus bendroflumethiazide) groups for the investigation of class effects of antihypertensive drugs on cardiovascular events in hypertensive patients with more than three cardiovascular risk factors, a second paper by Rothwell et al. [6] reported that within-individual BP variability and visit-to-visit BP variability were lower in the amlodipine group than in the atenolol group in ASCOT-BPLA. That report concluded that the effect of amlodipine-based treatment for the reduction of BP variability compared with atenolol-based treatment was associated with a reduced event rate. In addition, the report demonstrated that the effect of amlodipine-based treatment for reduction of interindividual BP variability was greater than that using atenolol-based treatment alone (Fig. 1). Finally, a third paper reported that the effect of a calcium-channel blocker (CCB) on reduction of interindividual BP variability was greatest among antihypertensive drugs tested, based on the results of a meta-analysis [7]. These three papers introduced three terms used to describe BP variability, i.e., visit-to-visit BP variability, within-individual BP variability, and interindividual BP variability. The authors concluded that interindividual variability was strongly associated with intraindividual variability, leading to some confusion among physicians. The figure demonstrates the concept of BP variability using the three indexes. Thus, interindividual BP variability is significantly different from visit-to-visit BP variability and within-individual BP variability. Interindividual BP variability refers to the variability of response to BP reduction by an antihypertensive drug. For instance, we previously reported a comparison between the effects of amlodipine and valsartan monotherapy for BP reduction using ambulatory BP monitoring in untreated hypertensive patients [8]. That study revealed that both amlodipine and valsartan monotherapy significantly reduced 24-h baseline BP levels through the end of the treatment period of 8 to 16 weeks, but the effect of BP reduction was higher with amlodipine than with valsartan. Interestingly, although amlodipine reduced not only mean 24-h BP values but also standard deviation of mean 24-h BP values, valsartan increased standard deviation of mean 24-h BP values after the follow-up period. The standard deviation value refers to interindividual BP variability under treatment with an antihypertensive drug. Therefore, we cannot statistically compare differences in interindividual BP variability for two drugs.

### Methodology of evaluation of visit-to-visit BP variability

Most methods used for the evaluation of visit-to-visit BP variability use standard deviation (SD) when calculating mean BP values, because it is easier to calculate and probably more practical than other methods. However, higher SD is usually correlated with higher average values. Therefore, the coefficient of variation (CV), calculated dividing the mean value by the SD, has also been used as a standard method for the evaluation of visit-to-visit BP variability. Average real variability (ARV) of BP is calculated by summation of the values obtained for the absolute difference between a BP value and the immediately preceding BP value. If BP measurements are continuously taken at similar time intervals, the ARV may provide more clinically useful data than SD or CV. However, when BP is measured at different intervals, the clinical significance of ARV may be diluted. Another BP measure, i.e., variability independent of the mean (VIM), has been developed by Rothwell et al. [4].

The calculation of VIM is as follows. First, the formula of non-linear regression is made by using standard deviation (SD) of BP [SDi] and mean of BP [Mi] of each subject as follow.

$${\mathrm{SD}}_i = \beta _0 \times M_i^{\beta _1},\beta _0 = {\mathrm{constant}}$$, β1 = power (parameter of regression)

A log-transformed linear regression is also applied instead of a non-linear regression.

$$\ln \left( {{\mathrm{SD}}_i} \right) = \beta _0 + \beta _1 \times \ln \left( {M_i} \right)$$, β0 = constant, β1 = regression coefficient

Then, change the formula above into the following formula

$$\frac{{{\mathrm{VIM}}_i}}{{\left( {\bar M} \right)^{\beta _1}}} = \frac{{{\mathrm{SD}}_i}}{{\left( {M_i} \right)^{\beta _1}}}$$
$${\mathrm{VIM}}_i = \left( {\bar M} \right)^{\beta _1} \times \frac{{{\mathrm{SD}}_i}}{{\left( {M_i} \right)^{\beta _1}}}$$
$$\bar M = {\mathrm{mean}}\,{\mathrm{of}}\,M_i$$

VIM has been considered a better index of BP variability than other indexes, because it has no association with average BP level. However, there is a significant difference between VIM and other indexes of BP variability (SD, CV, and ARV). For example, if a patient has values of 10 mmHg, 6%, and 12 mmHg in SD, CV, and ARV, respectively, the values of these indexes never change. However, when a patient is included in a different database set for analysis, VIM varies depending on each database set, because VIM is calculated based on the non-linear regression analysis of each data set. Therefore, clinicians should separate VIM and other indexes of BP variability depending on the situation. When we evaluate BP variability in a more scientific light, VIM may be more useful than other BP indexes. When we evaluate BP variability in a more practical light, BP indexes other than for VIM may be acceptable.