Home blood pressure measurement (HBPM) has become increasingly popular and is now gaining more and more acceptance by patients and clinicians. Several factors can account for this phenomenon. First of all, reliable automatic devices for HBPM have become available, and their performance characteristics can easily be retrieved via the Internet (1). Secondly, HBPM is less liable to observer bias and the white coat effect than office blood pressure measurements (OBPM) (2). In addition, with HBPM more measurements can be obtained within a limited period of time. Finally, HBPM data correlate better with daytime values of ambulatory blood pressure measurement (ABPM) (3).

However, the introduction of HBPM into clinical practice has also faced us with new problems. These include questions such as: how many measurements are needed to estimate a patient’s usual blood pressure (BP), what are normal values for HBPM, how does antihypertensive treatment affect HBPM, and to what extent can HBPM predict prognosis? The intention of this review, therefore, is to evaluate current knowledge about HBPM to find answers to these questions. To this end, we performed a systematic review of the literature on HBPM.

We performed systematic searches of HBPM in PubMed, EMBASE, and the Cochrane database. The search strategy used six key words in various combinations: BP, hypertension, self-measurement, home measurement, ambulatory measurement, and compliance. Additional studies were found from reference lists of identified articles and reviews.

Because the first protocol with guidelines for validation of BP monitoring devices dates back from July 1990 (4), we wanted to include in our review only those papers that have been published after that time. Allowing for some lag time before the guidelines were fully implemented, we took 1992 as the starting year for our literature search. Decisions on inclusion of studies into the systematic review were based on three criteria:

Two investigators screened the full text of all potentially relevant articles. When more than one paper by the same author or research group was available, these were included for analysis only when it was likely that a different patient sample had been used.

Papers that fulfilled the selection criteria and provided both OBPM and HBPM data were collected. When multiple drug treatment studies were performed using the same population, only the results of the first study mentioned were used for analysis. Studies that combined subjects with and without antihypertensive drug treatment, without the possibility to distinguish both groups, were excluded for analysis.

To test for differences between OBPM and HBPM, we pooled the results weighted with inverse variances (direct pooling) (7). In case the variance was not reported, we imputed the highest variance of the study that was included. Changes in home and office BP, obtained before and after treatment, were compared conservatively by considering these outcomes as unpaired data. To investigate whether age and gender could explain heterogeneity in differences between OBPM and HBPM, we performed linear regression analysis with age and gender (as proportion of men) as independent variables and the difference between OBPM and HBPM as the dependent variable. All statistical calculations were performed using SPSS (SPSS Inc., Chicago, Illinois).

To determine the number of measurements that are needed to obtain a reliable estimate of a patient’s usual BP when applying HBPM, several studies assessed the reproducibility and/or accuracy of the technique (8- 11). The results from these studies are summarized in (Table 1). From all studies, which addressed the number of measurements required to obtain an accurate impression of BP for a period of at least one month (8- 11), we may conclude that the minimum number of measurements should be two duplicates a day (twice in the morning and twice in the evening) for three consecutive days with exclusion of BP readings taken at the first day.

(Table 2) provides an overview of studies that have addressed the issue of normal values. Various methods were used to obtain normal values (12- 18), namely:

Taken together, the data from these studies suggest that the upper limit for normal HBPM values should be set at 135 mm Hg systolic and 85 mm Hg diastolic.

A pooled analysis of untreated subjects in the studies (13,17,19- 32) is listed in (Table 3) and shows that the mean overall difference between OBPM and HBPM was 6.9 mm Hg (95% confidence interval [CI] 6.6 to 7.2, p < 0.001) for systolic BP (SBP), and 4.9 mm Hg (95% CI 4.7 to 5.1, p < 0.001) for diastolic BP (DBP). Differences between OBPM and HBPM increase with age for SBP but not for DBP (Figure 1). Bland-Altman analysis of the data shows that the difference between OBPM and HBPM also increases significantly with higher SBP (Figure 2) but not DBP. Finally, a higher percentage of men in a study population resulted in larger differences between OBPM and HBPM for SBP (p < 0.01).

Studies among treated subjects (20- 21,24,26,28- 35) are summarized in (Table 4) and show a mean overall difference between OBPM and HBPM of 5.3 mm Hg (95% CI 5.1 to 5.6, p < 0.0001) for SBP, and 3.1 mm Hg (95% CI 2.9 to 3.3, p < 0.0001) for DBP. In this group no relationship of these differences with age, gender distribution, or the height of DBP emerged. Although differences tended to be greater with higher pressures for SBP (Figure 2), the relationship failed to reach statistical significance. The OBPM-HBPM difference was significantly smaller in treated than in untreated patients.

(Table 5) presents the studies that investigated drug efficacy by using both OBPM and HBPM (21,24,26,29- 32). After pooled analysis, differences in OBPM appeared to be 20.1 mm Hg (95% CI 19.6 to 20.7) and 13.6 mm Hg (95% CI 13.3 to 14.0) for SBP and DBP, respectively, while for HBPM differences were 13.9 mm Hg (95% CI 13.4 to 14.4) and 9.1 mm Hg (95% CI 8.8 to 9.4), respectively (Figure 3). The falls in pressure after treatment were significantly larger for OBPM than for HBPM (p < 0.0001 for SBP and DBP).

Several trials have addressed the relationship between home measurements and target organ damage (TOD) and found that HBPM correlated better with TOD, in particular left ventricular hypertrophy than OBPM (33,36). It correlated as well with left ventricular mass index as ABPM did (37). Moreover, HBPM appears to be a better prognostic indicator with respect to cardiovascular mortality (38- 39) and cardiovascular events (39) than OBPM.

Recently, results from the Ohasama study (40) indicated that one should obtain as many measurements as possible, although with any number of measurements HBPM was already superior to OBPM in predicting future risk.

Because it is possible that BP at home differs from BP at work due to job stress or other factors, BP recorded at work might, in such situations, give a better indication of the overall BP value (10). Despite some discrepancies in the results with regard to morning and evening HBPM, the fact remains that the early morning surge has immense prognostic potential (41- 44). For that reason, HBPM in the morning is probably more valuable as compared to evening HBPM in terms of assessing cardiovascular prognosis. On the other hand, Brook (45) showed that the accuracy of HBPM does not depend on a particular monitoring schedule parameter such as number of measurements per day, replications per measurement session, or total duration of home monitoring.

Taking all available information together, it seems that the minimum number of measurements should be two duplicates a day (twice in the morning and twice in the evening) for three consecutive days. Measurements on the first day are persistently higher than subsequent measurements, and for that reason the data from the first day should be discarded (9).

Normal values for HBPM have been established on the basis of an international database composed by Thijs et al. (15). Because the 95th percentile method for normotensive subjects was used for obtaining normal values, these normal data were not directly dependent upon OBPM values. In our view, the proclamation of HBPM normal data according to OBPM values, as has been done in other studies, is not a priori justified because OBPM values are highly variable and subject to observer bias as well as the white coat effect.

Preferably, normal values should be based on prognostic studies such as those from Tsuji et al. (16), who defined hypertension as a BP value ≥137/84 mm Hg and normal BP as SBP <137 mm Hg and DBP between 66 and 83 mm Hg. Fortunately, these values are in line with the reference values from the international database (namely 137/85 mm Hg). In accordance with recent guidelines (46- 47), it is fair to set the upper limit of normal values at 135 mm Hg systolic and 85 mm Hg diastolic.

Our analysis shows that the difference in SBP between OBPM and HBPM in untreated patients tends to be larger in men than in women and to rise with age and the level of BP. However, whether age and the degree of hypertension are independent predictors of the difference between OBPM and HBPM cannot be derived from the available data. We further found that the difference between OBPM and HBPM is greater before than during antihypertensive treatment. Although the difference during treatment still tended to increase with the height of pressure, the relationship was no longer statistically significant. Differences in DBP were more or less constant, regardless of age, gender, severity of hypertension, or treatment status.

The most important clinical consequence of the above is that HBPM can largely eliminate the white coat effect, which is defined as an elevation of BP measured in the office compared to either ABPM or HBPM (48). Detection of the white coat effect is relevant to avoid overestimation of daily BP and unnecessary drug prescription. Because the white coat effect also increases with age (49) and is more related to untreated than to treated hypertension (50), HBPM seems to be particularly suitable to assess elderly people with hypertension. Although HBPM may not detect the same patients with the white coat effect as ABPM would (2), it serves at least the purpose of being a reasonable screening test for that phenomenon.

Still, a word of caution is in order because most studies that addressed HBPM in relation to OBPM used different devices for both methods; OBPM was usually assessed with manual mercury sphygmomanometers (auscultatory method), which carries the risk of observer bias due to, for example, hearing loss, impaired ability to react, and digit preference (31,51- 52). On the other hand, HBPM was commonly performed with automatic or semiautomatic devices that used an oscillometric method to record the pressure. This may at least in part explain some of the differences between OBPM and HBPM. Additionally, most studies from (Table 5) that addressed drug efficacy using HBPM and OBPM showed a larger decline in OBPM after treatment as compared with HBPM. Such differences could be due to the physician expecting a decline in BP after drug administration. This assumption is supported by the study of Vaur et al. (31) who showed a decline in OBPM after treatment while the patient was using a placebo whereas this effect was absent in HBPM. Therefore, in order to exclude measurement bias in future clinical trials, OBPM and HBPM should be performed with the same device, preferably an automatic one (53). In the ongoing Home versus Office blood pressure MEasurements: Reduction of Unnecessary treatment Study (HOMERUS) trial (54), this approach has already been adopted.

The possibility of HBPM to obtain multiple measurements within a relatively short period of time makes the technique particularly useful to determine drug efficacy (21,30,32). As an index of efficacy, the morning-to-evening ratio has been introduced as an alternative for the “trough-to-peak ratio,” which is used with ABPM (55). With respect to the morning-to-evening ratio, one assumes that if medication is taken at a 24-h interval, the trough is reached just before the new medication is taken after 24 h in the morning, while 12 h earlier in the evening, the full effect of the drug can be expected (the peak).

Although only a few studies have addressed the relationship between HBPM and TOD, they are consistent in their results. Clearly, HBPM is superior to OBPM and comparable to ABPM in predicting cardiac as well as renal damage (33,36). As far as long-term cardiovascular prognosis is concerned, data also indicate that HBPM is the better predictor of outcome when compared to OBPM. For instance, the Ohasama study (38) clearly established a relationship between HBPM and mortality risk. When home BP values and screening BP values from this population study were simultaneously incorporated into a Cox model, only the average of multiple (more than three) home systolic BP values was significantly and strongly related to cardiovascular mortality risk. The average of the two initial home BP values was also better related to mortality risk than were the screening BP values. More recently, the same investigators presented their ten-year follow-up data (40), which indicated that the predictive value of HBP increases progressively with the number of measurements. When at least 14 measurements are obtained, HBPM shows a 35% increase in the risk of stroke per 10 mm Hg elevation in BP. Nevertheless, even if one considers only initial HBP values (one measurement), stroke risk is better predicted by HBPM than by OBPM.

Despite some limitations of this study (other risk factors were ignored in the analysis, there was a higher overall mortality from cerebrovascular disease in Ohasama than in the rest of the country, and an arbitrarily chosen end point was used for determining acceptable risk), this is an important long-term follow-up study linking HBPM to cerebrovascular risk. The study also suggested that the correlation between home DBP and mortality is nonlinear (U-shaped curve) and that a very low DBP may be associated with increased mortality as well.

Although more research is needed, several arguments already speak in favor of implementing HBPM into daily clinical practice. First of all, it eliminates the white coat effect and allows identification of patients with white coat hypertension. Second, it offers the possibility to obtain multiple readings under standardized conditions with little measurement variability. This increases knowledge of usual BP value in such conditions as borderline hypertension (28,56), type II diabetes mellitus (57- 58), and older age (35,59- 61). In addition, it allows better judgment of drug efficacy (21,24,26,29- 32). Third, HBPM data correlate better than OBPM values with TOD, in particular left ventricular hypertrophy (33,36- 37,58), with cardiovascular events (39), and with cardiovascular mortality (38- 39). The predictive power of HBPM increases with the number of measurements and should perhaps be based on the average of at least 14 data points (40). Finally, HBPM may increase patients’ awareness of hypertension and compliance with drug treatment, potentially leading to reduced mortality and costs (62). In line with this supposition, a recent meta-analysis by Cappuccio et al. (63) showed that subjects using HBPM had lower BP values and were more likely to achieve their target BP value than subjects without HBPM.

In practice, many factors can influence HBPM. It is important, therefore, that established guidelines for procedures of self-monitoring of BP (64- 65) are meticulously followed and that patients receive extensive instruction from a well-trained nurse or physician (66). Recordings must be taken with devices (67- 75) that have been validated according to AAMI (5), BHS (4), and/or ESH (6) standards (Table 6). Moreover, BP devices should be memory-equipped in order to prevent reporting bias (76). Because manual devices, which require the patient to inflate the cuff and/or to determine BP himself, are subject to the same forms of measurement bias as office recordings, automated ones are to be preferred.

Notwithstanding all advantages, HBPM also has some limitations. First of all, the HBPM technique, like ABPM, is less suitable in subjects with large arm sizes for whom no appropriate cuff in available, in those with an irregular pulse, or when there is reason to suspect vascular stiffening. Indeed, almost all validated HBPM devices employ the oscillometric technique, which may yield results that differ substantially from BP readings taken with a sphygmomanometer. This is particularly true in elderly patients and those with diabetes (77- 78). Second, ABPM is still superior for measuring BP at predetermined times without any influence of the patient, to record BP during daily routine or during the night and to ascertain whether a drug is effective during the early morning surge. However, because HBPM is less expensive and less inconvenient for the patient, it can serve as a reliable addition to OBPM, although the latter should not be abandoned yet (23,79). Moreover, there is already an HBPM device available that is able to measure BP during sleep at predetermined times (80). Third, HBPM should not be recommended for subjects with pre-eclampsia because both the auscultatory and oscillometric methods have shown to be inappropriate in this situation (81). Fourth, it should be realized that certain aspects of HBPM need further research, especially because the recently published Treatment of hypertension based on Home or Office blood Pressure (THOP) trial showed that adjustment of antihypertensive treatment based on HBPM instead of OBPM led to less intensive drug treatment but also to less BP control (82). Therefore, until this subject has been investigated further, treatment decisions based on HBPM alone should be taken cautiously. Finally, because regular subjects require at least 20 min of instruction (28) before understanding the procedure, HBPM may not be appropriate for every patient because of its complexity. Additionally, HBPM should also be discouraged when it causes anxiety or induces self-modification of treatment.

Still, due to an ever-increasing workload for physicians, it seems to be only a matter of time before people measure their BP at home and transmit it through the internet to the hospital, instead of visiting the doctor at the clinic. This modern approach of hypertension management is already put into practice at the HOMED-BP trial (83).(7)