Prognosis of Isolated Systolic and Isolated Diastolic Hypertension as Assessed by Self-Measurement of Blood Pressure at Home:  The Ohasama Study FREE

THE DIAGNOSIS of hypertension is usually based on blood pressure (BP%) criteria for both systolic and diastolic BP (SBP and DBP, respectively).^1- 2 However, some subjects fulfill only the SBP or only the DBP criteria. These subjects are categorized as having isolated systolic hypertension (ISH) or isolated diastolic hypertension (IDH). While the prognostic significance of ISH has been established by longitudinal intervention trials and meta-analysis,^3- 6 only 2 longitudinal studies investigated the prognostic significance of IDH.^7- 8 In these studies, the diagnosis of hypertension was based on casual or clinic measurements of BP. Casual BP measurement is known to have several limitations, such as poor reproducibility, the presence of a "white coat effect," observer bias. Recently, the use of self-measurement of BP at home (hereafter referred to as "home BP measurement"), which has been reported to have a better reproducibility^9- 11 and prognostic value than casual BP measurement,¹² has been acknowledged. The Sixth Report of the Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure¹ and the 1999 World Health Organization-International Society of Hypertension Guidelines for the Management of Hypertension² have also emphasized its usefulness. Since 1987, we have been conducting a prospective cohort study to investigate the relationship between home BP measurements and survival in the general population of Ohasama, Japan.^12- 15 The objective of this substudy was to clarify the prognostic significance of ISH and IDH, assessed using home BP measurements, in estimating the cardiovascular outcome for this general population.

This study is part of a longitudinal observational study of subjects who have been participating in a BP measurement project in Ohasama, Japan, since 1987. The socioeconomic and demographic characteristics of this region and full details of the project have been described elsewhere.^12- 15 The study protocol was approved by the Institutional Review Board of Tohoku University School of Medicine, Sendai, Japan, and by the Department of Health of the Ohasama Town Government.

The selection of study subjects has been described previously.¹⁵ Briefly, the subjects were aged 40 years or older and were residents of 3 of the 4 regions of Ohasama (N = 2716). Hospitalized persons (n = 121) and persons with dementia or who were bedridden (n = 31%) were excluded from the study. Individuals who worked out of town (n = 575%) were also excluded because the project involved ambulatory BP monitoring. Informed consent to participate in the study was given by 1957 of 1989 eligible subjects. Home BP measurements were obtained from 1913 subjects who collected their own data on more than 3 occasions (3 days) during the 4-week study period. This criterion was based on our previous observation that the average BP value for the first 3 days did not differ significantly from values obtained during the entire study period.¹³ Therefore, the study population consisted of 1913 individuals (mean age, 61.0 years; men–women ratio, 40:60) representing 90% of the total eligible population. We have previously confirmed that these subjects were representative of the total population.¹⁵

In our study, we used the following procedure to ascertain the accuracy of home BP measurements. First, physicians and public health nurses conducted a health education class to inform the population about home BP recording and to teach them how to measure their BP. Then, we checked whether they were able to measure their BP correctly. Eighty percent of households in this town attended the class, and public health nurses visited all of the remaining households to provide instruction on home BP measurement and to check whether they were able to measure their BP correctly. These procedures were also described in detail in our previous articles.^13- 14 The subjects were then asked to measure their BP every morning and to record the results for 4 weeks. Measurements were obtained within 1 hour of awaking and before breakfast, with the subject seated and having rested for at least 2 minutes. In subjects receiving antihypertensive drugs, home BP was measured before taking the drugs. The home BP of an individual was defined as the mean of all measurements obtained for that person. The mean (±SD) number of home BP measurements was 20.8 + 8.3 (range, 3-38).

Free annual health checkups, including BP measurements, are available for all persons aged 40 years or older. Using a semiautomatic device, a nurse or technician took 2 consecutive BP measurements with the subjects seated, after at least 2 minutes of rest. The screening BP was defined as the average of these 2 readings. Of the 1913 study subjects, 1789 (90%) had a health checkup during the year in which they were taking home BP measurements. There was no statistically significant difference in mean age, sex distribution, or mean home measurements of SBP and DBP between those who participated in screening and those who did not.

Home BP was measured using a commercially available device (HEM401C or 701C model; Omron Life Science Co Ltd, Tokyo, Japan), automatic devices which use the cuff-oscillometric method to generate a digital display of SBP and DBP.¹⁶ On the basis of the cuff-oscillometric principle, SBP is measured when the continuous minimum oscillation of the cuff pressure during cuff inflation is determined. The BP measuring device used in this study does not take a measurement when the cuff is inflated insufficiently, and displays an error code. The screening BP was measured using a fully automatic device (USM-700F; UEDA Electronic Works Co Ltd, Tokyo, Japan), based on the Korotkoff sound technique (a microphone method). The circumference of the arm was less than 34 cm in most cases, so we used a standard arm cuff for both BP measurements. The device used to measure home BP has previously been validated¹⁶ and satisfied the criteria of the Association for the Advancement of Medical Instrumentation.¹⁷

We classified the subjects into 4 groups according to their home BP levels as follows: systolic-diastolic hypertension (SDH): SBP, 137 mm Hg or higher and DBP, 84 mm Hg or higher; ISH: SBP, 137 mm Hg or higher and DBP, lower than 84 mm Hg; IDH: SBP, lower than 137 mm Hg and DBP, 84 mm Hg or higher; and normotension (NT): SBP, lower than 137 mm Hg and DBP, lower than 84 mm Hg. The cut-off level of 137/84 mm Hg to define hypertension for home BP measurements is based on our previous observations of prognostic significance.¹⁵ We treated the subjects whose body mass indexes were 25.0 kg/m² or higher as obesity.¹⁸

Residence in the town of Ohasama on October 31, 1999, was confirmed using the residents' registration cards, which were considered accurate and reliable because they are the basis for payment of pension and Social Security benefits in Japan.

Mortality from cardiovascular disease (CVD mortality) was defined as death from disease of the circulatory system (International Classification of Diseases, 10th Revision (ICD-10) code I, CVD). The association between a particular type of hypertension and CVD mortality was investigated using the Cox proportional hazards regression model.¹⁹ The dependent variable in this analysis was the number of days from the first home BP measurement to either the date of death from CVD or the censoring of survivors on October 31, 1999. Subjects who died of noncardiovascular causes were also censored. In all analyses, we included age, sex, smoking status, obesity, use of antihypertensive medication, and a history of CVD, hypercholesterolemia, or diabetes mellitus as possible confounding variables in the multivariate models. This information was obtained from questionnaires sent to each subject at the time of home BP measurement, medical records of annual health checkups, and medical records held at Ohasama Prefectural Hospital. In all analyses, we treated the subjects with NT as a reference group.

An estimate of the relative hazard (RH) for each variable, along with its 95% confidence interval (95% CI), was derived from the coefficient of the Cox proportional hazards regression model and its SE, using the SAS PHREG procedure.²⁰ Variables were compared using the χ²test, t test, or analysis of variance, as appropriate. Differences at P<.05 were considered statistically significant. All data are expressed as mean + SD.

The mean duration of follow-up was 8.6 years (maximum follow-up, 11.7 years). There were 93 deaths (4.8%) owing to CVD and 172 deaths (9.0%) of a non-CVD cause. Of the 93 deaths owing to CVD, 55 (59%) were due to stroke and 38 (41%) were due to heart disease. Twenty-two subjects (1.2%) moved away from the region and were lost to follow-up.

Of the 1913 study subjects, 402 (23%) were classified as current or ex-smokers, 301 (22%) were classified as obese, and 598 (31%) were taking antihypertensive medication. A history of CVD, hypercholesterolemia, or diabetes mellitus was identified in 74 subjects (4%), 224 subjects (13%), and 229 subjects (13%), respectively. Mean home BP measurements were 125.5 + 11.3 mm Hg for SBP and 75.1 + 10.0 mm Hg for DBP. The mean home pulse pressure (PP) was 50.3 + 10.3 mm Hg.

Table 1 lists the characteristics of the subjects in each group, classified on the basis of home BP measurements. Subjects with IDH were significantly younger, while subjects with ISH were significantly older than all of the study subjects. Men and smokers were more preponderant in the IDH group, while women and nonsmokers were more preponderant in the ISH group. Mean SBPs were higher in subjects with SDH (148.7 + 9.6 mm Hg) or ISH (144.9 + 7.9 mm Hg) than in those with IDH (130.1 + 4.5 mm Hg%) or NT (118.3 + 9.7 mm Hg), while mean DBPs were higher in subjects with SDH (91.4 + 6.1 mm Hg) and subjects with IDH (87.4 + 4.0 mm Hg) than in subjects with ISH (76.5 + 5.1 mm Hg) and subjects with NT (70.9 + 7.1 mm Hg). Mean home PPs were higher in subjects with ISH (68.4 + 8.9 mm Hg) or SDH (57.3 + 9.4 mm Hg) than those with IDH (42.7 + 5.2 mm Hg) or NT (47.5 + 7.4 mm Hg).

We investigated the relationship between the type of hypertension and CVD mortality among the entire study population. Cardiovascular disease death rates per 100 person-years were higher in subjects with SDH (1.11) or ISH (2.04) than in subjects with NT (0.33) or IDH (0.26). The RHs for CVD mortality were also significantly higher in subjects with SDH (RH = 1.85; 95% CI, 1.07-3.21; P = .03) or ISH (RH = 2.35; 95% CI, 1.41-3.90; P = .001) than in subjects with IDH (RH = 1.09; 95% CI, 0.33-3.59; P = .89) or NT.

Since antihypertensive treatment would have affected the prognosis, we analyzed the relationship between the type of hypertension and CVD mortality in subjects who were taking antihypertensive medication (treated subjects%) and those who were not (untreated subjects), separately. The results demonstrated a similar tendency to those observed in the analysis of the entire study population.

Among untreated subjects, CVD death rates per 100 person-years were higher in subjects with SDH (1.01) or ISH (1.64) than in subjects with NT (0.24) or IDH (0.14). The RHs for CVD mortality were also higher in subjects with SDH (RH = 1.94; 95% CI, 0.85-4.44; P = .12%) or ISH (RH = 2.42; 95% CI, 1.09-5.37; P = .03) than in subjects with IDH (RH = 1.57; 95% CI, 0.20-12.27; P = .67) or NT.

Treated subjects showed almost the same tendency for the relationship between the type of hypertension and CVD risk, ie, CVD death rates per 100 person-years were higher in subjects with SDH (1.18) or ISH (2.32) than in subjects with NT (0.65) or IDH (0.41). The RHs for CVD mortality were also higher in subjects with SDH (RH = 1.86; 95% CI, 0.87-3.97; P = .11) or ISH (RH = 2.42; 95% CI, 1.22-4.79; P = .001) than in subjects with IDH (RH = 0.98; 95% CI, 0.22-4.37; P = .75) or NT.

Since the subjects with NT and IDH were younger than those with SDH and ISH, we also analyzed the 2 age groups separately, ie, middle-aged (40-64 years) and elderly (≥65 years) subjects. A similar tendency was also observed when we analyzed the relationship between CVD mortality and the type of hypertension in middle-aged subjects and elderly subjects. For the middle-aged group, the CVD mortality rates per 100 person-years were higher in subjects with SDH (0.49) or ISH (0.67) than in those with NT (0.10) or IDH (0.10). Similarly, among the elderly subjects, CVD mortality rates per 100 person-years were higher in subjects with ISH (2.67) or SDH (1.86) than in subjects with IDH (1.19) and NT (1.00).

Systolic-diastolic hypertension and ISH were associated with a higher CVD mortality rate per 100 person-years than IDH and NT in both men and women (men/women with SDH, 0.85/1.55; men/women with ISH, 3.93/0.98; men/women with IDH, 0.00/0.81; and men/women with NT, 0.46/0.27).

Although there was little difference in SBP between subjects with SDH and subjects with ISH, the RH in subjects with ISH was higher than that in subjects with SDH. Furthermore, the risk of subjects with IDH who had the minimal PP was similar to that in subjects with NT despite the higher SBP in subjects with IDH. This finding suggests that home PP measurement is also an important prognostic factor for CVD. Therefore, we analyzed the relationship between CVD mortality and home PP as a continuous variable, instead of stratifying PP by the type of hypertension. As a result, each 10–mm Hg increase in PP was significantly related to an increase in the risk of CVD mortality (RH = 1.37; 95% CI, 1.14-1.65; P = .001). There were only small differences between the RH of home PP among treated subjects (RH = 1.31; 95% CI, 1.01-1.71; P = .046), untreated subjects (RH = 1.43; 95% CI, 1.04-1.85; P = .007), and all subjects.

We also analyzed the relationship between CVD mortality and the type of hypertension assessed by screening BP measurements. For this analysis, we classified the subjects into 4 groups according to their screening BP levels, using 140/90 mm Hg as criteria to define hypertension.

Isolated systolic hypertension tended to be associated with a higher RH (RH = 1.49; 95% CI, 0.89-2.47; P = .13) and CVD mortality rate (0.94 per 100 person-years) than NT (mortality rate = 0.35 per 100 person-years). However, in subjects with SDH, the RH for CVD mortality (RH = 1.10; 95% CI, 0.53-2.30; P = .79) and mortality rate (0.49 per 100 person-years) were no different from those with NT (mortality rate = 0.35 per 100 person-years). This relationship was also observed when the subjects with IDH (RH = 1.20; 95% CI, 0.16-8.96; P = .86, mortality rate = 0.39 per 100 person-years) was compared with subjects with NT. Furthermore, screening PP values were not significantly related to the risk of CVD mortality (RH = 1.09; 95% CI, 0.94-1.27, P = .24, per 10–mm Hg increase).

This study was part of a longitudinal observation of a representative sample of the general population of a rural Japanese community. When assessed by home BP measurements, the result demonstrated that, compared with subjects with NT, subjects with ISH or SDH have a significantly higher risk of CVD mortality, while those with IDH do not. These findings were observed both in treated and untreated subjects, in young and elderly subjects, and in men and women.

Only 2 other studies have investigated IDH as a risk factor for CVD. Fang et al⁷ followed up 1560 middle-aged subjects with diastolic hypertension, and demonstrated that the risk of acute myocardial infarction was significantly higher in subjects with SDH than in those with IDH.⁷ Petrovitch et al⁸ followed up 7590 Japanese-American men among the general population of Hawaii and showed that, for middle-aged persons, the risk of a first symptomatic stroke was lower in subjects with IDH or NT than in subjects with ISH or SDH.⁸ Although these subjects were classified according to casual BP measurements, the same result, ie, a lower risk of CVD morbidity in subjects with IDH than in subjects with SDH, was observed as in this study, during which home BP measurements were used.

In the present analysis, we confirmed that the risk of CVD mortality was somewhat higher in subjects with ISH than in subjects with SDH, while IDH was associated with a lower risk than SDH or ISH. Thus, our results indicate that SBP, but not DBP, is a strong predictor of CVD mortality when assessed by home BP measurements. The prognostic value of SBP (as assessed by repeated casual BP measurements) has recently been confirmed by an investigation in an elderly Italian population.²¹

In our analysis based on home BP measurements, subjects with ISH or SDH, who also had an increased home PP than subjects with NT or IDH who had lower PP, were at higher risk of CVD mortality than the subjects with NT or IDH. In contrast, the risk in IDH did not differ significantly from that in NT. Systolic blood pressure was significantly higher in subjects with IDH than in subjects with NT; however, PP was somewhat lower in subjects with IDH, although the risks of CVD mortality were no different. Similarly, the CVD mortality risk was somewhat higher in subjects with ISH than in subjects with SDH, although SBP in ISH was similar to that in SDH. Furthermore, home PP were also significantly related to the mortality risk when analyzed as a continuous variable, independently of other risk factors for CVD mortality. These results are consistent with those of several previous studies investigating the relationship between PP and CVD mortality,^7,22- 27 and thus reconfirm the importance of PP in predicting the prognosis of hypertension. To our knowledge, this study is the first to demonstrate the clinical significance of PP as assessed by home BP measurement.

When we estimated the risk of CVD mortality in subjects with SDH, ISH, IDH, or NT, as classified by screening BP measurements, the risk tended to be higher in subjects with ISH than in those with NT, while subjects with IDH or SDH showed no significant difference from subjects with NT. Similarly, no significant association between CVD mortality and screening PP was observed. The present results obtained on the basis of casual BP were in contrast to previous reports.^7- 8,25 Such inconsistency may be attributable to the small sample size and short-term observation period compared with previous prospective studies involving large numbers of subjects, which reported a statistically significant association between casual BP and CVD risk. If we had conducted observations for a longer period using a sample, as was done in the previous study, a significant association between the type of hypertension and CVD mortality might have been observed even on the basis of screening BP. However, we would like to emphasize again that home BP measurement did predict the CVD mortality despite the short observational period in this sample size. Furthermore, since our previous observations had demonstrated that the predictive power for mortality of the initial 2 home BP measurements among 21 consecutive measurements was higher than that of casual BP even though the numbers of measurements were the same,¹² it is suggested that not only reproducibility but also other factors such as the absence of a white coat effect, observer bias, environmental influence, and so on may be associated with the better predictive power of a home BP measurement than that of a casual BP measurement. Thus, home BP measurements have better predictive power than casual BP measurements. However, as the predictive power of multiple casual BP measurements is reported to be as high as that of ambulatory BP,²⁸ it is difficult to obtain consecutive and multiple casual BP measurements within a certain period at screening or in daily clinical practice. We conclude that home BP measurement is beneficial because it facilitates multiple measurements over a certain period.

We demonstrated (1) that IDH is associated with a low risk of CVD mortality (similar to that observed in NT) while ISH carries a similarly high risk to SDH, and (2) that PP is a strong independent predictor of CVD mortality. These findings were made using home BP measurements, which have been confirmed to be a more reliable method of estimating the prognosis of hypertension in the general population than casual BP measurements. These results suggest that the prognosis of hypertension would be improved by treatment focused on SBP rather than on DBP. However, they do not mean that IDH is harmless, since it is possible that a younger adult with IDH may convert to ISH or develop SDH with increasing age. Further observational and interventional studies are, therefore, needed to clarify the clinical significance of IDH.

Accepted for publication June 7, 2000.

This work was supported by research grants from the Miyagi Prefectural Kidney Association, Sendai, Japan, the Takeda Medical Foundation, Osaka, Japan, Taisya-ijou Chiryou Kenkyu Kikin, 1999, Tokyo, Japan, and Mitsui Life Social Welfare, as well as a research grant titled "Evaluation of the Effect of Drug Treatment on Hypertension and Other Chronic Disease Conditions in the Elderly," Rojin Hoken Jigyo, 1998 and 1999, Tokyo, a Health Science research grant on Health Services, Junkanki-byo Itaku Kenkyu 11C-5, 1999, Tokyo, and Rojin-hoken Jigyou-hi tou Hojokin, 1999, Tokyo, from the Ministry of Health and Welfare, and by a Research Grant for Scientific Research (10470102) and a Research Grant for JSPS Research Fellows (01180) from the Ministry of Education, Science and Culture of Japan, Tokyo.

Corresponding author: Atsushi Hozawa, MD, Department of Public Health, Tohoku University School of Medicine, 2-1 Seiryo-machi, Aoba-ku, Sendai 980-8575, Japan (e-mail: hozawa@mail.cc.tohoku.ac.jp).