Home blood pressure (BP) is a useful indicator to detect masked or white coat hypertension, and to evaluate the effect and duration of antihypertensive treatment as well as patient adherence to the treatment as compared with clinic BP. Home BP has increasingly gained popularity with widening availability of home BP measurement devices among patients. However, in daily clinical practice, physicians may hesitate to adjust treatment in their hypertensive patients when home BP and clinic BP show a different trend, because most evidence for hypertension were based on clinic BP and little is known whether on-treatment home BP in treated hypertensive patients can predict cardiovascular risk. Some studies have compared the prognostic association between home BP and clinic BP. The majority of studies, such as Ohasama study , Self-Measurement of Blood Pressure at Home in the Elderly: Assessment and Follow-up study , Finn-Home study , and the Hypertension Objective Treatment Based on Measurement by Electrical Devices of Blood Pressure study  reported that home BP was superior to clinic BP, whereas the Pressioni Arteriose Monitorate e Loro Associazioni study  and the Didima study  did not find a significant difference between them. These studies have their own limitations: populations in a single community were analyzed [1,5,6]; nonhypertensive individuals were included [1,3,5,6]; a small number of events resulted in a lack of statistical power [1,5,6]; only baseline BP was assessed [1–3,5,6], or on-treatment BP was only measured at baseline and at the end of follow-up . The International Database on Home blood pressure in relation to Cardiovascular Outcomes (IDHOCO) database , which integrated five cohorts, including the Ohasama study, the Finn-Home study, and the Didima study showed that only home BP was significantly associated with cardiovascular events in treated hypertensive patients, but yet again only baseline BP was assessed.
In most of those studies above, the prognostic significance of BP was discussed only based on relative risk, such as hazard ratios per 1 mmHg increase in BP. However, another study argues that prognostic significance should not be discussed only based on relative risk .
Against this background, we investigated the prognostic significance of clinic SBP (CSBP) and morning home SBP (MHSBP) measured at several time points, from multiple aspects [i.e. hazard ratios per 1 mmHg increase, concordance index (c-index), net reclassification improvement (NRI), and the 2-year absolute risk by Cox proportional hazards model] using data from the Home blood pressure measurement with Olmesartan Naive patients to Establish Standard Target blood pressure (HONEST) study that followed more than 20 000 hypertensive patients for 2 years.
The HONEST study was a prospective observational study with a 2-year follow-up by 30 September 2012. The aims and protocol  and the main results  have already been reported. The protocol was approved by the Ethical Committee of Daiichi Sankyo Co., Ltd., the review boards of the participating institutions at their discretion, and Ministry of Health, Labour, and Welfare of Japan before study commencement. This study was carried out at registered medical institutions in compliance with Good Postmarketing Study Practice and each institution's internal regulations for clinical studies.
Olmesartan (OLM)-naive outpatients with essential hypertension (physician reported, no BP range) had recorded their morning home BP on at least 2 days within 28 days prior to taking OLM. After providing written informed consent and being prescribed OLM, patients were registered. Baseline clinic BP was also measured on at least 1 day within 28 days prior to taking OLM. Patients were excluded mainly if they had a history of recent cardiovascular events (e.g. myocardial infarction, stroke, cardiovascular interventions, and hospitalization for heart failure) or a planned cardiovascular intervention. With the exception of prior use of OLM, no restriction was placed on prior antihypertensive drug treatment or the use of combination antihypertensive drug treatment during the study. The selection of target clinic BP and home BP was left to the discretion of individual physicians. Collected data were patient characteristics (e.g. disease history and complications), clinic BPs, home BPs, clinical laboratory test values, cardiovascular events, and adverse events.
Home blood pressure measurements
Patients used their own electrical upper arm devices for home BP based on the cuff-oscillometric principle. All such devices available in Japan have been validated and approved by the Ministry of Health, Labour, and Welfare of Japan, and are in accordance with the United States [Association for the Advancement of Medical Instrumentation (AAMI)]  or European standards . At the time of obtaining informed consent, physicians instructed patients to measure home BP twice consecutively in the morning and twice consecutively in the evening on 2 different days for each measurement point, according to the methods recommended by the 2009 guidelines of the Japanese Society of Hypertension (JSH 2009)  (in the morning: within 1 h of waking, after urination, before taking morning medications, before eating breakfast, and 1–2 min of rest in a sitting position; at bedtime: before retiring and after 1–2 min of rest in a sitting position). Measurement points of home BP and pulse rate were at 1, 4, 16 weeks, and 6, 12, 18, and 24 months. Before registration to the study (on at least 2 days within 28 days prior to taking OLM), patients might possibly have measured home BP using different methods to those defined in the protocol. Home BP was measured both in the morning and in the evening. However, only morning home BP was mandatory at the time of enrollment. In addition, cardiovascular events tend to occur most frequently in the morning, along with a rise in BP . Therefore, we used morning home BP in the present analysis.
Clinic blood pressure measurements
After the initiation of the study, the measurement points of clinic BP were at 4 and 16 weeks, and 6, 12, 18, and 24 months. Given that this study was conducted in the setting of daily medical practice, clinic BP was measured at least once at each measurement point according to the usual methods of each institution; no recommendations or training were provided. This was also true during the period within 28 days prior to taking OLM before patient registration to the study.
Assessment of blood pressure measurements
Three types of BP measurements were evaluated: baseline BP, BP during follow-up, and the achieved BP. The baseline BP was defined as the mean of BP measurements obtained in the 28-day period before starting OLM therapy (including the starting day of OLM treatment). The BP during follow-up was defined as the mean of all available BP measurements after the start of OLM treatment until the last reported value or the preceding day of the first cardiovascular event. The achieved BP was defined as the BP measurement at the last time point during the follow-up period or the preceding day of the first cardiovascular event.
We established three event review committees: cerebrovascular, cardiovascular, and other events. Each event review committee consisted of two or more specialists who identified all events according to the predetermined definitions of the events . Additional information was requested if necessary by each committee.
In this analysis, we used the composite cardiovascular event defined as cerebral infarction, intracerebral hemorrhage, subarachnoid hemorrhage, unclassified stroke, myocardial infarction, coronary revascularization procedures for angina pectoris, or sudden death.
Eligible patients included in the analyses were those who received OLM at least once during the treatment period and excluding patients who were proven ineligible after data collection. Quantitative data were expressed as mean ± standard deviation, and qualitative data were expressed as percentages unless otherwise stated.
To compare prognostic association of each BP measurement for cardiovascular events, we first estimated hazard ratios per 1 mmHg increase in each BP variable using Cox proportional hazards model adjusted for sex, age, family history of cardiovascular diseases, dyslipidemia, diabetes mellitus, history of cardiovascular diseases, and smoking. Although in the analysis mentioned above, it was assumed that there would be a linear relationship between the BP increase and cardiovascular risk increase, a nonlinear relationship may have existed. Thus, we conducted a likelihood ratio test to assess whether the addition of a quadratic term of BP improves the goodness-of-fit to the model, and consequently, the addition of a quadratic term was almost statistically significant for the on-treatment BP (follow-up BP and achieved BP), but not for baseline BP. Therefore, we decided to add a quadratic term of the on-treatment BP into the model in the subsequent analysis.
We calculated c-index to quantitatively compare the discrimination ability of each BP variable using Cox proportional hazards model  adjusted for the abovementioned confounders. In addition, we also calculated the category-based NRI with the threshold at 2.5 and 5.0% to compare how correctly each BP variable (MHSBP and CSBP) classify patients based on the 2-year predictive risk of cardiovascular event, using the Cox proportional hazards model [16,17] that included confounders and both clinic BP and morning home BP. Confidence intervals (CIs) for c-index and NRI were estimated using the bootstrap method with 1000 replications.
To assess the flexible relationship between cardiovascular risk and each BP variable, the 2-year absolute risk of cardiovascular event for each level of BP variable was estimated using the same Cox proportional hazards model, standardizing other risk factors than BP to the average of the HONEST study.
All statistical tests were two-sided with a significance level of 0.05, and SAS version 9.2 (SAS Institute Inc., Cary, North Carolina, USA) was used.
As previously reported , 22 373 patients were enrolled in the HONEST study from across Japan. By excluding patients who were proven ineligible after data collection, 21 591 patients were included in the analysis. During the follow-up period (2.02 ± 0.50 years), 425 patients withdrew consent (but agreed to the use of data obtained until then), and 190 patients died. Of the patients who agreed to continue to participate in the study, contact with 1950 patients (9.0%) was lost during the follow-up period (<21 months). The mean age was 64.9 years; 50.6% were female; and 50.4% were receiving antihypertensive drugs. Concomitant morbidities at baseline were dyslipidemia (44.6%), diabetes mellitus (20.5%), and chronic kidney disease (20.1%). A total of 280 cardiovascular events (6.46 events per 1000 person-years; 95% CI 5.75–7.27) occurred during follow-up.
MHSBP and CSBP were 151.2 ± 16.3 (n = 21 588) and 153.6 ± 19.0 mmHg (n = 21 589) at baseline, 135.2 ± 10.8 (n = 21 001) and 135.2 ± 11.5 mmHg (n = 21 346) during follow-up, and 132.1 ± 13.2 (n = 21 001) and 133.3 ± 14.8 mmHg (n = 21 346) at the completion of the study. The percentage of patients who measured home BP according to the JSH 2009 was 82.1% at baseline, 92.2% after 12 months, and 93.9% after 24 months.
Association between blood pressure and cardiovascular risk
Figure 1a shows hazard ratios per 1 mmHg increase in CSBP and MHSBP at baseline and during follow-up when each BP variable was included in a separate model. Baseline MHSBP (hazard ratio 1.011; 95% CI 1.004–1.019; P = 0.0020), follow-up MHSBP (hazard ratio 1.039; 95% CI 1.029–1.049; P < 0.0001), and follow-up CSBP (hazard ratio 1.026; 95% CI 1.016–1.036; P < 0.0001) were significantly associated with cardiovascular risk, whereas no association was found with baseline CSBP (hazard ratio 1.006; 95% CI 1.000–1.012; P = 0.0594).
Figure 1b shows hazard ratios when two BP variables (MHSBP and CSBP at each time point) were included in the same model. Although baseline MHSBP (hazard ratio 1.011; 95% CI 1.002–1.020; P = 0.0147) and follow-up MHSBP (hazard ratio 1.033; 95% CI 1.020–1.046; P < 0.0001) were significantly associated with cardiovascular risk, no association was found with baseline CSBP (hazard ratio 1.001; 95% CI 0.993–1.008; P = 0.8805) and follow-up CSBP (hazard ratio 1.008; 95% CI 0.996–1.020; P = 0.1720).
Figure 1c shows hazard ratios when two BP variables (baseline SBP and follow-up or achieved SBP) were included in the same model. In the model that included baseline and follow-up MHSBP, a significant association was found with the follow-up MHSBP (hazard ratio 1.043; 95% CI 1.031–1.055; P < 0.0001) but not with the baseline MHSBP (hazard ratio 0.994; 95% CI 0.985–1.003; P = 0.1852). Similarly, in the model that included baseline and follow-up CSBP, a significant association was found with the follow-up CSBP (hazard ratio 1.027; 95% CI 1.016–1.038; P < 0.0001) but not with the baseline CSBP (hazard ratio 0.998; 95% CI 0.991–1.005; P = 0.5719).
Figure 1d shows hazard ratios when four BP variables were included in the same model. In the model that included baseline and follow-up MHSBP and CSBP, only the follow-up MHSBP was significantly associated with cardiovascular risk (hazard ratio 1.036; 95% CI 1.022–1.050; P < 0.0001).
As shown in Fig. 1a and c, the achieved MHSBP showed a similar prognostic association with the follow-up MHSBP. However, the results of the achieved CSBP differed from those of the follow-up CSBP.
Analyses of DBP and BP before sleep were also performed, and showed the same tendency (data not shown).
Table 1 shows c-index values of each BP variable. In the present study, c-index was above 0.7 for all BP variables, showing that all BP variables are useful in terms of the discrimination ability, although the mean BP during follow-up of MHSBP showed the highest ability.
Table 2 shows NRI values when MHSBP was added into the model including CSBP and confounders, and when CSBP was added into the model including MHSBP and confounders. Follow-up MHSBP had better reclassification ability than follow-up CSBP. The same trend was found in achieved BP, but both BP variables at baseline did not improve classification.
Figure 2a–c shows 2-year absolute risk of baseline, follow-up, and achieved CSBP and MHSBP. The curve showing the relationship between the absolute risk and each BP measurement was steeper for follow-up MHSBP (Fig. 2b) than others.
We investigated the prognostic significance of MHSBP and CSBP at baseline and during follow-up in on-treatment hypertensive patients from multiple aspects. SBP during follow-up (as compared with SBP at baseline), particularly MHSBP (as compared with CSBP) had better prognostic significance in predicting cardiovascular events in Japanese patients with essential hypertension during a 2-year clinical study. In terms of the hazard ratios, the association of SBP during follow-up was stronger than that of SBP at baseline, and the association of MHSBP was stronger than that of CSBP. In addition, in terms of discrimination ability, c-index of SBP during follow-up was higher than that of SBP at baseline, and c-index of MHSBP was as high as or higher than that of CSBP. The results of NRI analysis showed that follow-up MHSBP had better reclassification ability than follow-up CSBP. These results were consistent with the result of the comparison of hazard ratios. To our knowledge, no study has compared the discrimination ability between baseline BP and mean BP during the follow-up. Our findings are novel in showing that SBP during follow-up (as compared with SBP at baseline) and MHSBP (as compared with CSBP) had better prognostic significance in predicting cardiovascular events, using data from a large-scale 2-year observational study involving over 20 000 Japanese patients with essential hypertension.
Comparison of morning home blood pressure and clinic blood pressure
From the result of c-index, both CSBP and MHSBP have reasonable discrimination ability for cardiovascular events; from the result of NRI, follow-up and achieved MHSBP have better reclassification ability than those of CSBP; and from the result of hazard ratio per 1 mmHg increase, MHSBP have stronger association with cardiovascular risk. Our previous report  has shown that masked hypertension is significantly associated with cardiovascular risk. Home BP can be a stronger predictor of cardiovascular risk than clinic BP.
In the IDHOCO study , both home and CSBP were a significant predictor for cardiovascular death and cardiovascular events in all study participants; however, in on-treatment patients, only home SBP was a significant predictor, as consistent with the results of the present study. Moreover, in the Self-measurement of blood pressure at Home in the Elderly: Assessment and Follow-up study  involving only hypertensive patients, home BP at baseline was a more useful predictor for cardiovascular events.
Comparison of blood pressure at baseline and during follow-up
In the present study, prognostic significance of BP during the follow-up was better than that of BP at baseline. This indicates that strict BP control by antihypertensive treatment can prevent cardiovascular events for a short term of 2 years in hypertensive patients in clinical practice, regardless of BP at baseline.
There have been few reports that investigated the prediction of cardiovascular events using BP both at baseline and during treatment. In the Hypertension Objective Treatment Based on Measurement by Electrical Devices of Blood Pressure study , the prognostic significance of BP at baseline and at one point during treatment (either BP at the study completion or BP within 6 months before cardiovascular events) was investigated; however, there was no direct comparison of BP at baseline and during treatment.
This study was not a randomized controlled trial but an observational study in daily clinical practice. Patients with various complications and history of cardiovascular disease were included, with no restrictions placed on prior antihypertensive treatment, type of antihypertensive treatment or BP value at baseline. Therefore, BP during follow-up had better prognostic significance than baseline BP in predicting cardiovascular risk, and the results of NRI analysis of baseline BP differ from those for follow-up BP and achieved BP.
Evaluation of absolute risk
The IDHOCO study , which investigated 10-year cardiovascular disease risk using baseline BP, showed similar results to the present study in that the association with the cardiovascular disease risk was stronger with home BP than clinic BP. Although it is generally considered that the cardiovascular risk at CSBP 140 mmHg was equivalent to the risk at MHSBP 135 mmHg, the difference in BP was smaller in the present study. This may be because of the shorter follow-up period of the present study (2 years), and the difference would be bigger if the follow-up period was longer.
Comparison of achieved blood pressure and blood pressure during follow-up
Although we did not make direct comparison of prognostic significance between BP during follow-up and achieved BP, different results were found in MHSBP and in CSBP. The difference may result from the difference of number of time points. MHSBP was less affected by number of time points and have higher accuracy to guess ‘true’ BP for predicting cardiovascular risk than CSBP. Accordingly, if it is difficult to calculate MHSBP during follow-up in daily clinical practice, achieved MHSBP can be a substitute for mean MHSBP during follow-up.
The HONEST study was not a randomized controlled trial but a prospective observational study supported by general practitioners in daily clinical practice. The findings of the present study are limited by the study design that intended to reflect real-world clinical practice.
Moreover, the present study included patients who already had cuff-oscillometric home BP-measuring devices, but types of devices used to measure home BP were not specified. It is considered that patients used the same devices from baseline until the end of the follow-up period; however, the possibility that patients changed devices cannot be denied. Nevertheless, home BP-measuring devices available in Japan have been validated and approved by the Ministry of Health, Labour, and Welfare of Japan, and are in accordance with the United States (AAMI)  or European standards . As done in most of the previous studies [1–3,5,6], home BP measurements were recorded by patients and reported to their physician, so the possibility of biased reporting by patients and physicians cannot be excluded completely. The percentages of patients who measured home BP according to the instructions in JSH 2009 differed between at baseline and during the follow-up. Moreover, clinic BP was also measured by the methods of individual medical institutions, validation methods of BP-measuring devices were not specified, and were done by attending physicians. However, automated sphygmomanometer devices available in Japan have been validated in accordance with the United States (AAMI)  or European standards . However, we consider their influence to be limited because of the nature of the study design that used both home BP and clinic BP with a large sample size.
In the present study, participating physicians were not advised by which types of BP measurement they adjust the treatment. However, JSH 2009 , published in the year the study started, emphasized the importance of home BP. Moreover, the study protocol stated that the aim was to evaluate BP using home BP as an indicator, the participating physicians were thought to be aware of the importance of home BP.
Finally, the mean follow-up period of the present study was relatively short (2.02 years). However, as the present study involved a large number of patients, aggregated data for the events per patient year were similar to or greater than those reported in previous studies.
We gratefully acknowledge the numerous study investigators, fellows, nurses, and research coordinators at each of the study institutions, who have participated in the HONEST study. We also gratefully acknowledge the contribution of the patients to the study.
Some of the findings described in this manuscript were presented at the 37th Annual Scientific Meeting of the Japanese Society of Hypertension (Yokohama, Japan; 17–19 October 2014).
This study was supported by funding for data collection and statistical analysis by Daiichi Sankyo Co., Ltd. Statistical analyses were done by EPS Corporation (Tokyo, Japan) under the direction of the sponsor and the authors. Medical editorial assistance was provided by Nature Japan K.K. (Macmillan Medical Communications, Tokyo, Japan), and funded by Daiichi Sankyo Co., Ltd.
Conflicts of interest
K.S., K.K., T.K., S.T., and I.S. received honoraria from Daiichi Sankyo Co., Ltd. N.Z., Y.I., and Y.O. are employees of Daiichi Sankyo Co., Ltd.
1. Ohkubo T, Imai Y, Tsuji I, Nagai K, Kato J, Kikuchi N, et al. Home blood pressure
measurement has a stronger predictive power for mortality than does screening blood pressure
measurement: a population-based observation in Ohasama, Japan. J Hypertens
2. Bobrie G, Chatellier G, Genes N, Clerson P, Vaur L, Vaisse B, et al. Cardiovascular prognosis of ‘masked hypertension
’ detected by blood pressure
self-measurement in elderly treated hypertensive patients. JAMA
3. Niiranen TJ, Hänninen MR, Johansson J, Reunanen A, Jula AM. Home-measured blood pressure
is a stronger predictor of cardiovascular risk than office blood pressure
: the Finn-Home study. Hypertension
4. Noguchi Y, Asayama K, Staessen JA, Inaba M, Ohkubo T, Hosaka M, et al. Predictive power of home blood pressure
and clinic blood pressure
in hypertensive patients with impaired glucose metabolism and diabetes. J Hypertens
5. Sega R, Facchetti R, Bombelli M, Cesana G, Corrao G, Grassi G, Mancia G. Prognostic value of ambulatory
and home blood pressures compared with office blood pressure
in the general population: follow-up results from the Pressioni Arteriose Monitorate e Loro Associazioni (PAMELA) study. Circulation
6. Stergiou GS, Baibas NM, Kalogeropoulos PG. Cardiovascular risk prediction based on home blood pressure
measurement: the Didima study. J Hypertens
7. Niiranen TJ, Asayama K, Thijs L, Johansson JK, Ohkubo T, Kikuya M, et al. Outcome-driven thresholds for home blood pressure
measurement: International Database of HOme blood pressure
in relation to Cardiovascular Outcome. Hypertension
8. Pepe MS, Janes H, Longton G, Leisenring W, Newcomb P. Limitations of the odds ratio in gauging the performance of a diagnostic, prognostic, or screening marker. Am J Epidemiol
9. Saito I, Kario K, Kushiro T, Teramukai S, Zenimura N, Hiramatsu K, et al. Rationale, study design, baseline characteristics and blood pressure
at 16 weeks in the HONEST Study. Hypertens Res
10. Kario K, Saito I, Kushiro T, Teramukai S, Ishikawa Y, Mori Y, et al. Home blood pressure
and cardiovascular outcomes in patients during antihypertensive therapy: primary results of HONEST, a large-scale prospective, real-world observational study
11. Association for the Advancement of Medical Instrumentation. American National Standard. Electronic or Automated Sphygmomanometers. NSI/AAMI SP 10-1992. Arlington, VA: AAMI; 1993:40.
12. European Committee for Standardization. Non-invasive Sphygmomanometers. Part 3. Supplementary Requirements for Electromechanical Blood Pressure
Measuring Systems. British Standard BS EN 1060–3: 1997. European Standard EN 1060–3 1997. Brussels: ECS; 1997.
13. Ogihara T, Kikuchi K, Matsuoka H, Fujita T, Higaki J, Horiuchi M, et al. The Japanese Society of Hypertension
Guidelines for the Management of Hypertension
(JSH 2009). Hypertens Res
14. Muller JE, Tofler GH, Stone PH. Circadian variation and triggers of onset of acute cardiovascular disease. Circulation
15. Pencina MJ, D’Agostino RB Sr, Song L. Quantifying discrimination of Framingham risk functions with different survival C statistics. Stat Med
16. Pencina MJ, D’Agostino RB Sr, D’Agostino RB Jr, Vasan RS. Evaluating the added predictive ability of a new marker: from area under the ROC curve to reclassification and beyond. Stat Med
17. Pencina MJ, Steyerberg EW, D’Agostino RB Sr. Extensions of net reclassification improvement calculations to measure usefulness of new biomarkers. Stat Med
18. Nomura K, Asayama K, Thijs L, Niiranen TJ, Lujambio I, Boggia J, et al. Thresholds for conventional and home blood pressure
by sex and age in 5018 participants from 5 populations. Hypertension