Ambulatory Blood Pressure Monitoring and Blood Pressure Self-Measurement in the Diagnosis and Management of Hypertension

Methods

Our principal strategy for identifying relevant articles was to search the MEDLINE data base using the MeSH headings, blood pressure determination and blood pressure monitoring, covering the period 1966 through early 1992. Articles published before 1966 were identified through review of the reference sections of other papers. Articles considered for review included original research papers as well as position papers, consensus statements, and review articles. Those selected for detailed review documented original research pertaining to the diagnosis, treatment, or prognosis of hypertension using self-measured blood pressure, noninvasive ambulatory blood pressure, and measurements by nonphysicians. Data were extracted and results were tabulated for selected key issues. The reproducibility of methods for article identification and data extraction were not formally assessed; however, qualitative comparisons of our results to those published in other reviews suggest high reliability of our methods.

Results

Blood Pressure Measurement Techniques

Techniques for measuring blood pressure include standard measurement of resting blood pressure in the office by physicians, nurses, or technicians using a sphygmomanometer; measurement by automated but stationary devices; self-measurement by patients or by relatives or friends; and measurement by automated ambulatory blood pressure devices. No one approach is optimal, and each has its advantages and limitations.

Office blood pressure is the standard on which most of the literature is based. Typically, a mercury or aneroid sphygmomanometer is used and the appearance and disappearance of Korotkoff sounds are recorded (auscultatory technique). Ideally, before making a diagnosis of hypertension and before initiating (or adjusting) drug treatment, multiple readings at two or more visits are obtained while the patient rests comfortably in the seated position. Limitations of office blood pressures include inadequate maintenance and calibration of the sphygmomanometer, inappropriate technique, observer bias, and imprecision as a result of few observations [13-16]. An additional concern relates to the effect of different types of observers (physician versus nurse or technician) on office blood pressure.

Stationary blood pressure devices are often available in public buildings, pharmacies, supermarkets, and airports. For many individuals, they provide a convenient means to obtain measurements. However, the accuracy of these devices is often unsatisfactory [17-19]. In addition, the commercial setting may be unrepresentative of an individual's usual environment. Overall, these devices may serve to enhance patient awareness but are unlikely to contribute substantially to hypertension management decisions.

Self-measurement devices allow for repeated measurements outside the medical environment and are popular among patients. A variety of mercury, aneroid, and electronic devices are available. Recognizing the widespread interest in this technology, the National High Blood Pressure Education Program has published two reports describing potential roles for self-measurement devices [20, 21]. These reports suggest that such devices may 1) encourage patients to participate more actively in their own treatment; 2) improve adherence to antihypertensive drug therapy; 3) aid physicians in evaluating the efficacy of treatment and in simplifying medication regimens; and 4) obviate the need for frequent office visits.

Despite the potential benefits, however, several concerns exist. First, many commercially available devices are inaccurate [22-24]. Second, not all individuals are able to obtain satisfactory readings [25, 26]. Third, most patients require training [23]. Results of the Tecumseh Blood Pressure Study, for example, showed that the average person required 20 minutes of training before being able to measure blood pressure satisfactorily using a manual aneroid device [26]. Because physicians often have a peripheral role in the selection, purchase, and use of these devices, it is doubtful that patients receive uniformly adequate training. An additional concern relates to whether self-measured blood pressure is representative of a patient's usual blood pressure. Self-measurements are usually taken at convenient times, for example, at home in the morning or evening but less commonly at work. Finally, evidence that self-measured blood pressure can lead to improved blood pressure control or medication adherence is inconclusive [27-31].

Ambulatory blood pressure monitoring allows for automatic blood pressure measurements at frequent intervals throughout the day and night. Measurements are recorded outside the medical environment without the need for a manual observer. Most ambulatory blood pressure devices use either an auscultatory or oscillometric technique. The former technique is similar to traditional sphygmomanometry and involves placement of a microphone to detect the appearance and disappearance of Korotkoff sounds. Oscillometry involves use of device-specific algorithms to analyze patterns of cuff oscillations from which blood pressure is estimated. Devices using the auscultatory method consist of a lightweight console, a cuff placed on the nondominant arm, a microphone positioned over the brachial artery, and a cable connecting the console to the cuff and microphone. Some of these devices require placement of electrocardiographic (ECG) leads. Oscillometric devices, similarly, have a console, cuff, and connecting tube but do not require placement of a microphone or ECG leads.

Several technical issues deserve attention. First, the accuracy, precision, patient acceptability, and mechanical reliability of ambulatory blood pressure devices remain of concern [32-35]. Until recently, most evaluations focused on assessment of accuracy. Typically, simultaneous or sequential measurements are recorded by the ambulatory device and a standard technique (usually indirect sphygmomanometry but occasionally direct intra-arterial measurement) on individuals seated in a quiet room. Few investigations have assessed accuracy while participants engage in ambulatory activities [36, 37]. Performance of the devices over time, patient acceptability, and mechanical reliability have received scant attention.

Standards for evaluating ambulatory blood pressure monitors are in flux. The Association for the Advancement of Medical Instrumentation developed standards for automated monitors, in general, but did not address issues specific to ambulatory devices [38]. In 1990, the British Hypertension Society published a protocol that extends the Association for the Advancement of Medical Instrumentation recommendations to include: 1) observer training; 2) pre-use interdevice variability assessment in a laboratory setting; 3) in-use [field] assessment; 4) post-use interdevice variability assessment; 5) device validation; and 6) unbiased reporting of the complete evaluation [34]. Evaluations following the British Hypertension Society protocol have documented that the performance of ambulatory blood pressure devices is, indeed, quite variable [39-42].

Second, the collection of satisfactory ambulatory blood pressure data requires the services of a skilled technician who understands the operating characteristics of the device, can set certain parameters (for example, the frequency of measurements), perform a calibration test, choose an appropriately sized cuff, and properly place the microphone and ECG leads (if applicable) [43]. Of equal importance, the technician must instruct the patient on how to wear the monitor (for example, the need to keep the arm still during inflations), offer suggestions on how to minimize inconvenience, and instruct the patient on how to keep a diary that will permit an evaluation of the relationships between blood pressure and activity. On return of the monitor, the technician uploads data onto a personal computer or downloads them onto hard copy using device-specific hardware.

Third, data management and analysis present a number of challenges. The raw data must be edited to remove biologically implausible readings. At present, procedures for this task are not standardized, and several approaches have been proposed, including manual editing of data, deletion of measurements above or below specified limits, and database methods [44]. The most commonly reported summary statistics are average daytime, nighttime, and 24-hour ambulatory blood pressure. In addition, some investigators have proposed use of blood pressure loads, defined as the percentage of measurements above certain limits [45, 46]. There is no consensus about which summary measure should be used in clinical decision making. Diary data, if collected, must be integrated with measurements either as free-form descriptions or through entry of specified items into preset fields.

Finally, reference ambulatory blood pressure values that differentiate normotension from hypertension remain uncertain. A panel of ambulatory blood pressure authorities recently recommended that a mean daytime ambulatory blood pressure above 135/85 mm Hg may indicate hypertension [47], a threshold based in part on findings from ambulatory blood pressure studies conducted in healthy, normotensive persons [48]. Overall, nonstandardized methods of data management and analysis complicate the interpretation of ambulatory blood pressure recordings and hinder their use in clinical decision making.

Comparisons of Blood Pressure Levels by Different Techniques

Studies comparing office to self-measured blood pressure have shown variable and often substantial differences between these techniques (Figures 1 and 2) [8,25,49-73]. With few exceptions, the mean office blood pressure was higher than self-measured blood pressure, particularly at high blood pressures (range of mean differences, 1.2 to 21.6 mm Hg for systolic and 6.5 to 12.7 mm Hg for diastolic blood pressure). Within studies, differences between self-measured and office blood pressures varied enormously between patients. For example, in a study of hypertensive outpatients, Ayman and Goldshine reported blood pressure differences (office minus self-measured) of 6 to 78 mm Hg for systolic blood pressure and 10 to 38 mm Hg for diastolic blood pressure [3]. Twenty-nine percent of study participants had a systolic blood pressure difference of more than 30 mm Hg, and 24% had a diastolic blood pressure difference of 20 mm Hg or more.

Studies that compared all three measurement techniques found office blood pressure to be higher than both self-measured blood pressure and 24-hour or daytime ambulatory blood pressure (Table 2) [56, 57, 60-62, 71-73]. Mean self-measured blood pressure tended to be higher than corresponding 24-hour and daytime ambulatory blood pressure. The nocturnal decline in blood pressure undoubtedly accounts for the greater disparity between self-measured blood pressure and 24-hour ambulatory blood pressure than between self-measured blood pressure and daytime ambulatory blood pressure. Table 3 shows correlations between office and ambulatory blood pressure as well as mean blood pressure differences in 16 studies [49, 56, 62, 71, 72, 74-83, 87]. Correlations (r) between office and ambulatory blood pressures varied considerably (r = 0.19 to 0.84 for diastolic office and daytime ambulatory blood pressures) but often were quite high (r > 0.7 in 4 of 11 studies correlating diastolic office and daytime ambulatory blood pressure). Overall, data presented in Tables 2 and 3 indicate that differences between office blood pressure and both ambulatory and self-measured blood pressures were frequently substantial but not systematic.

Differences among office, self-measured, and ambulatory blood pressure measurements have attracted considerable attention because of the potential for misclassification of hypertensive status based on office blood pressure measurements alone. Pickering and colleagues, for example, identified a subset of patients (termed white coat or office hypertensives), whose diastolic office blood pressure, measured by a physician, was above 90 mm Hg but whose blood pressure outside of the office appeared normal [4]. In a study of outpatients attending a tertiary care facility, these investigators classified individuals by physician-measured diastolic blood pressure into three groups: 1) established hypertensive persons [diastolic office blood pressure greater than 105 mm Hg]; 2) borderline hypertensive persons [90 to 104 mm Hg]; and 3) normotensive volunteers (< 90 mm Hg). They also classified them as hypertensive or normotensive based on their mean awake ambulatory blood pressure, using the 90th percentile of the distribution of awake ambulatory blood pressure in the normotensive group to define the upper limits of normal (90% of normotensive persons had an awake ambulatory blood pressure of 134/90 mm Hg or less). Twenty-one percent of borderline and 5% of established hypertensive persons had evidence of the white coat phenomenon, namely high office blood pressure but an apparently normal awake ambulatory blood pressure. Less emphasized was the observation that office blood pressures measured by a technician tended to be similar to awake ambulatory blood pressures, at least among hypertensive patients. These findings lend further support to the existence of an alerting reaction associated with physician measurement [84, 85].

The true prevalence of office hypertension is unknown. A recent community-based study found that nearly 25% of patients with untreated, newly diagnosed hypertension had a daytime diastolic ambulatory blood pressure of less than 90 mm Hg [86]. In contrast, Pearce and colleagues [87] questioned the existence of the white coat phenomenon. In a methodologically rigorous, community-based study, these investigators compared office blood pressure (two technician-measured office blood pressures at each of five visits) to a single 24-hour or daytime ambulatory blood pressure in screenees from the Multiple Risk Factor Intervention Trial. Of the 50 participants, two thirds were normotensive and one third were receiving treatment for hypertension. Correlations between ambulatory and office blood pressures were high under these circumstances (r = 0.83 to 0.90 and 0.76 to 0.79 for systolic and diastolic blood pressure, respectively). Also, mean ambulatory blood pressure was actually higher than mean office blood pressure. Previous studies (Table 3) had shown lower correlations between office and ambulatory blood pressures and higher awake and 24-hour ambulatory blood pressure than office blood pressure, at least in hypertensive individuals. Pearce and colleagues concluded that ambulatory blood pressure monitoring is unlikely to improve estimation of usual blood pressure beyond that achievable by careful, repeated measurements of office blood pressure and that the white coat phenomenon is probably rare. An accompanying editorial [88] suggested that the rarity of the white coat phenomenon in this study may reflect the inclusion of few hypertensive individuals and differences in patient expectations for the office visit, that is, research objectives in Pearce's study but hypertension management in clinic-based studies.

The observed prevalence of office hypertension may depend in part on regression to the mean, a statistical phenomenon in which repeated measurements tend to be lower in persons initially selected on the basis of high blood pressure. This may be the case, for example, in the studies by Pickering and colleagues [4] and Hoegholm and colleagues [86], in which hypertensive individuals, selected on the basis of office blood pressure, subsequently had ambulatory blood pressure recorded. However, a number of studies with replicate measures of both office blood pressure and ambulatory blood pressure at two or more visits [61, 71, 89] showed that mean differences between ambulatory blood pressure and office blood pressure persist, even though the magnitude of the difference may decrease.

Table 4 summarizes the results of five studies that compared office blood pressures recorded by physicians to office measurements by nurses or technicians [4, 90-93]. These studies all showed that physicians, on average, recorded higher office blood pressures than nonphysicians (range of mean differences, 3 to 16 mm Hg for systolic and 1 to 18.9 mm Hg for diastolic blood pressure, respectively). Such findings are consistent with a study by Mancia and colleagues [85], who showed that physicians during hospital visits elicited a greater increase in intra-arterial systolic/diastolic blood pressure than did nurses (24/14 versus 12/7 mm Hg, respectively). Conversely, another study, which did not provide data, reported nonsignificant differences among physician, nurse, and researcher measurements [94]. Overall, despite the lack of standardized measurement techniques and the possibility of order effects, these studies consistently showed that physician-measured office blood pressure tends to be higher than nurse- or technician-recorded office blood pressure.

The influence of different types of observers on office blood pressure is of particular concern because of the potential for misclassification of hypertensive status. In this regard, it should be noted that most epidemiologic studies and clinical trials have relied on blood pressure measured by technicians or nurses.

Relationships between End-Organ Damage and Blood Pressure

Most studies that have examined associations between left ventricular hypertrophy and blood pressure have found stronger correlations with ambulatory blood pressure than with office blood pressure (Table 5) [49, 83, 87, 95-112]. The classic paper by Sokolow and colleagues [49], published in 1966, concluded that left ventricular hypertrophy (measured by ECG criteria) and heart size on radiograph were more strongly correlated with mean systolic and diastolic ambulatory blood pressures than with corresponding casual office measurements. Most other studies, relying on echocardiographic measurement of left ventricular structure, have concluded that left ventricular mass is more closely correlated with systolic than with diastolic blood pressure and with daytime or 24-hour ambulatory blood pressure than with office blood pressure.

Several studies deserve special comment. Devereux and colleagues [95] found that ambulatory blood pressure measured on a work day was more strongly associated with left ventricular mass than ambulatory blood pressure measured on a non-work day. The conclusion that ambulatory blood pressure measured during work may be a better predictor of risk than ambulatory blood pressure at home must be considered tentative, however, because participants, rather than the investigators, using a randomization schedule, chose the type of day (work or nonwork) for ambulatory blood pressure monitoring. Verdecchia and colleagues [108] found that nighttime ambulatory blood pressure was more closely correlated with left ventricular mass than daytime ambulatory blood pressure and that the extent of nocturnal decrease in blood pressure was inversely related to left ventricular structural abnormalities. Hence, the failure of blood pressure to fall during sleep may also be a determinant of cardiovascular risk. Finally, White and colleagues [45] found that most indices of left ventricular function and structure were more closely associated with blood pressure load (the percentage of readings above a specified level of blood pressure) than with mean ambulatory blood pressure.

A case-comparison study by the same investigator [113] provides intriguing results on the relationship between blood pressure and cardiac structure and function. This study classified participants into three groups: 1) office hypertensive persons (high office blood pressure [>140/90] and normal awake ambulatory blood pressure [<130/80]); 2) sustained hypertensive persons (high office blood pressure [>140/90] and high awake ambulatory blood pressure [>140/90]); and 3) normotensive controls (normal office blood pressure [<135/85] and normal awake ambulatory blood pressure [<130/80]). Individuals with sustained hypertension had larger left atria, greater left ventricular mass, and worse left ventricular function (both systolic and diastolic) than office hypertensive individuals, despite the fact that both groups had virtually identical office blood pressures; findings in office hypertensive persons were similar to those who were normotensive. Overall, this clinical study suggests that individuals with office hypertension are similar to normotensive persons in having less cardiac dysfunction and structural abnormalities than those with sustained blood pressure elevations.

At variance with these findings are the results from the population-based Tecumseh Blood Pressure Study [114]. This study of young adults classified participants by office blood pressure and self-measured blood pressure into three groups: 1) sustained borderline hypertension [high office and self-measured blood pressures]; 2) white coat borderline hypertension [high office but normal self-measured blood pressure]; and 3) normotension. Seven percent of the total sample and 58% of borderline hypertensive persons were classified as being white coat borderline hypertensive. The white coat and sustained borderline hypertensive groups had similar risk factor profiles for atherosclerosis (insulin, high-density lipoprotein cholesterol, and triglycerides) and for hypertension (weight, parental history of hypertension, and past high blood pressure); both groups had worse profiles than the normotensive group. The findings of this study suggest that low self-measured blood pressure does not necessarily identify a low-risk group of borderline hypertensive persons.

The only study correlating left ventricular hypertrophy with office, self-measured, and ambulatory blood pressure found correlation coefficients with end-diastolic wall thickness of 0.22, 0.45, and 0.26 for systolic blood pressure and 0.07, 0.40, and 0.24 for diastolic blood pressure, respectively [56]. Only the correlations between self-measured blood pressure and left ventricular hypertrophy were statistically significant (P < 0.01).

Similarly, only one study has assessed relationships between end-organ damage (left ventricular mass index and microalbuminuria) and blood pressure measured by different types of observers (physicians and nurses) [93]. Each end point was more highly correlated with nurse measurements than with physician measurements. For example, the correlations between left ventricular mass index and systolic blood pressure measured by nurses and physicians were 0.30 (P < 0.05) and 0.19 (P > 0.2), respectively. These findings should be interpreted with caution, however, because the nurses were trained researchers, leading to differences in measurement technique as well as observer type.

Sokolow and colleagues [49] and Parati and colleagues [115] used target-organ damage scores as a summary measure of end-organ effects. Both studies found stronger relationships between target organ damage and ambulatory blood pressure than between target organ damage and office blood pressure. In Parati's study, persons with lower mean 24-hour ambulatory blood pressure had a lower prevalence of organ damage and lower target organ damage within most strata of patients classified by office blood pressure. Furthermore, at any level of mean 24-hour ambulatory blood pressure, greater blood pressure variability was associated with more target organ damage.

The four studies that have examined the relationship of measurement technique to indices of renal dysfunction (micro-albuminuria and excretion of N-acetyl-B-D-glucosaminidase) provide inconsistent results [116-119]. In three studies, renal dysfunction was more highly correlated with ambulatory blood pressure than with office blood pressure [116, 117, 119], whereas the converse was observed in one study [118].

Only two studies have examined correlations between measurement techniques and retinopathy. Sokolow and colleagues [49] found that fundus grade (Keith-Wagener criteria) was more closely correlated with ambulatory blood pressure than with office blood pressure, whereas Dimmitt [118], using retinal photographs, was unable to show any difference in the strength of associations.

In 1990, Shimada and colleagues [120] reported that occult cerebrovascular lesions, specially low-intensity foci (lacunae), and periventricular hyperintense lesions on magnetic resonance imaging were more closely correlated with ambulatory blood pressure than with office blood pressure.

Study Limitations

All the previously mentioned studies correlating blood pressure with end-organ damage were cross-sectional in design and, therefore, have limited value with respect to causal inferences [121]. Also, most had small sample sizes and enrolled convenient, rather than representative, samples. Equally important, the variety of techniques used to measure office, self-measured, and ambulatory blood pressures probably obscured relationships. In this regard, most studies did not describe the office blood pressure technique, the type of observer, and the number of measurements. Often, antihypertensive medications were discontinued only a few weeks before blood pressure measurement and target organ assessment; because blood pressure may not increase immediately after discontinuation of medications, the relationship between blood pressure and target organ disease may be blurred by residual treatment effects, especially if office blood pressure and ambulatory blood pressure are differentially affected. Finally, and most importantly, interpretation of correlations between blood pressure and target organ disease is difficult in view of the widely varying numbers of blood pressure readings used to calculate mean office and mean ambulatory blood pressures (often 2 to 10 readings for office blood pressure versus 80 to 100 for ambulatory blood pressure). The larger number of ambulatory blood pressure measurements undoubtedly resulted in enhanced precision and probably contributed to ambulatory blood pressure's apparent superiority.

Prediction of Cardiovascular Mortality and Morbidity

Most evidence on the superiority of ambulatory blood pressure over office blood pressure in predicting cardiovascular risk is indirect and depends on the stronger association of ambulatory blood pressure with left ventricular mass which, in turn, is an important, independent predictor of cardiovascular morbidity and mortality in both hypertensive and community samples. In hypertensive persons, left ventricular mass index was a stronger predictor of future cardiovascular events than either systolic or diastolic blood pressure [122]. In the Framingham Heart Study, left ventricular mass was strongly predictive of both fatal and nonfatal cardiovascular events [123, 124].

Only a single prospective study has directly compared the predictive value of noninvasive ambulatory blood pressure and office blood pressure. This study, initially reported by Perloff and colleagues in l983 and reanalyzed in 1989, collected baseline and follow-up data on 1076 referred hypertensive patients who were characterized at baseline by both office blood pressure and ambulatory blood pressure (measured by a patient-activated semiautomatic device) [75, 125]. Participants were then followed for a mean of 5 years. Those individuals with higher than predicted ambulatory blood pressure (derived from the regression of ambulatory blood pressure on office blood pressure) had significantly higher rates of fatal and nonfatal cardiovascular events than those with lower than predicted ambulatory blood pressure. This trend applied to both systolic and diastolic blood pressure and was most apparent in those patients with less severe hypertension (systolic blood pressure <160 mm Hg or diastolic blood pressure <105 mm Hg) and those with no previous cardiovascular event. In the subsequent report, the authors, using a Cox proportional-hazards model to address concerns over potential confounding, noted that residual ambulatory systolic pressure (the difference between observed and predicted systolic ambulatory blood pressure) remained a significant predictor of event-free survival even after controlling for absolute levels of office blood pressure and systolic ambulatory blood pressure, age, sex, ECG evidence of left ventricular hypertrophy, retinopathy, and drug therapy.

Although this study provides important evidence suggesting that ambulatory blood pressure is a better predictor of cardiovascular risk than office blood pressure, methodologic issues continue to plague its interpretation. Paramount are concerns about the representativeness of the study sample, lack of information on the extent of blood pressure control during follow-up, the absence of information on other important risk factors such as smoking and serum cholesterol [126, 127], the quality and quantity of office and ambulatory blood pressures, and the possibility of biased ascertainment of end points. Especially for soft outcomes, such as angina and transient ischemic attacks, ascertainment could have been influenced by knowledge of ambulatory blood pressure results or by differential follow-up of individuals with high ambulatory blood pressure.

Use of Ambulatory and Self-Measured Blood Pressures during Treatment

Blood pressure responses to antihypertensive medications have traditionally been assessed by office blood pressures. A study by Waeber and colleagues [128] and subgroup analyses from a few clinical trials of antihypertensive drug therapies [129, 130] suggest that ambulatory blood pressure may be able to identify patients with poorly controlled office blood pressure who are likely to respond to changes in medical therapy. In Waeber's study, daytime ambulatory blood pressure was measured in 34 treated hypertensive persons who had office blood pressures above 95 mm Hg on two consecutive visits. Changes were then made in medication regimens in an effort to achieve an office blood pressure of less than 90 mm Hg. Patients whose initial diastolic ambulatory blood pressure was below 90 mm Hg (n = 17) did not experience significant decreases in ambulatory blood pressure following changes in treatment, whereas those with higher ambulatory blood pressures did. These findings suggest that intensifying treatment in patients with elevated office blood pressure but normal ambulatory blood pressures may increase medication side effects and costs with little or no improvement in blood pressure control. These intriguing results need to be verified in larger trials that include steps to avoid the statistical artifact of regression to the mean [131].

Five clinical trials have tested whether self-measured blood pressure, used alone or combined with other strategies, improves medication adherence or blood pressure control [27-31]. Of these, only one study, conducted in 38 poorly controlled, noncompliant hypertensive persons, documented a significant reduction in blood pressure (5 mm Hg) and an improvement in medication compliance [27]. In two other studies, clinically relevant improvements in adherence [28] and blood pressure control [29] were noted but did not achieve a conventional level of statistical significance. Thus, in terms of enhancing compliance and blood pressure control, evidence supporting use of self-measured blood pressure is promising but inconclusive.

Safety

The risks associated with noninvasive ambulatory blood pressure monitoring are relatively minor. Case reports have identified upper extremity thrombophlebitis [132], the Rumpel-Leede sign (petechial rash, ecchymoses and edema distal to the cuff) [133], and dermatitis [134] as potential complications. More common are arm discomfort and the inconvenience associated with ambulatory blood pressure monitor use, including difficulty in sleeping and interference with usual activities [135].

Costs

The net cost of ambulatory blood pressure monitoring will depend on the unit cost of a test, patterns of use, and its effect on costs associated with downstream treatment decisions. Charges for ambulatory blood pressure monitoring range from $100 to $450 per test [unpublished survey of selected insurers conducted by SpaceLabs, Inc. and Suntech, Inc.]. The resource cost of a test, however, is approximately $120 based on the following assumptions: 1) purchase of two $5000 monitors amortized over 3 years; 2) purchase of a $500 printer amortized over 5 years; 3) an annual maintenance contract of $500 per monitor; 4) $15 000 for a half-time technician; 5) throughput of 300 tests per year; 6) a physician's interpretation fee of $30 per test; and 7) an institutional overhead rate of 40% (applied to noncapital expenses). This resource cost per test is especially sensitive to the number of procedures performed in a laboratory per year, the technician's salary, and the institutional overhead rate.

The costs associated with blood pressure self-measurement are modest. Most devices cost $75 or less [23, 136, 137], and initial training and subsequent refresher sessions by nurses or technicians should add no more than $25 per year. Assuming that a device lasts 3 years, total costs for blood pressure self-measurement are less than $50 per year.

Discussion

Available evidence, coupled with a sound rationale for measuring blood pressure outside the medical environment, strongly suggest, but do not unequivocally establish, useful clinical roles for ambulatory and self-measurement devices. Table 6 summarizes our impressions of the potential benefits, harms, and costs of using ambulatory and self-measurement devices to diagnose hypertension and to monitor therapy. Estimates are presented both for the magnitude of effects and the degree of certainty provided by published information. We did not combine evidence formally because of heterogeneity in study designs and the dearth of prospective data.

Benefits

The most important potential benefits of ambulatory and self-measurement devices would be more accurate classification of blood pressure, better prediction of cardiovascular risk, improved ability to monitor blood pressure responses to treatment, and improved adherence with therapy. Judicious use of ambulatory blood pressure monitoring and, to a lesser extent, blood pressure self-measurement should improve the accuracy of blood pressure classification. Office blood pressures, as typically measured by physicians, have been shown to overestimate usual blood pressure in substantial numbers of persons. Undoubtedly, many patients with high office blood pressure but considerably lower blood pressure outside the medical environment are diagnosed as hypertensive and treated with drug therapy. Ambulatory and self-measurement devices provide convenient methods to identify such individuals. Alternatively, indirect evidence suggests that standardized office measurements recorded by well-trained nonphysicians at multiple visits may achieve the same end.

The claim that ambulatory blood pressure monitoring improves cardiovascular risk prediction depends on the results of a large number of cross-sectional studies that have demonstrated that left ventricular mass is more strongly associated with ambulatory blood pressure than with office blood pressure (Table 5), coupled with the fact that left ventricular hypertrophy is an independent risk factor for cardiovascular morbidity and mortality [122-124]. Only one prospective study corroborates these conclusions [75, 125]. The consistent nature of these findings, despite methodologic shortcomings, suggests that ambulatory blood pressure should be a better predictor of cardiovascular risk than office blood pressure.

Evidence supporting use of ambulatory devices to monitor treatment responses is limited to one small trial [128] and subgroup analyses of clinical trials of antihypertensive drug therapy [129, 130]. Possible advantages include a reduction in unnecessarily intensive treatment (with associated side effects and costs) and the ability to titrate medications to achieve uniform reduction in blood pressure throughout the day and night.

Use of self-measurement devices also has the unique advantages of involving patients in their own treatment and offering the potential for frequent monitoring of blood pressure at home. Evidence documenting improved medication adherence or blood pressure control as a result of self-monitoring, however, is sparse.

Harms

The most important potential harm of relying on ambulatory or self-measured blood pressure to diagnosis hypertension would be increased cardiovascular risk, if treatment were inappropriately withheld from patients who might have benefited. Evidence from a recent meta-analysis of clinical trials of antihypertensive drug therapy suggests that this risk should be small [2]. This study found that drug treatment of hypertensive individuals reduced the risk of fatal and nonfatal strokes by 42% and coronary heart disease events by 14%. These relative risk reductions correspond to absolute risk reductions of 22.5 strokes and 11.8 coronary events averted per 10 000 person-years of treatment. It is noteworthy that this meta-analysis combined data from clinical trials that enrolled severely hypertensive persons (entry diastolic blood pressures greater than 115 mm Hg) with trials that enrolled less severely hypertensive persons (entry diastolic blood pressures less than 110 mm Hg). Hence, the absolute risk of withholding treatment in office hypertensive persons with a diastolic office blood pressure of 90 to 99 mm Hg should be less, especially in the absence of other cardiovascular risk factors and target organ damage.

Other harms of ambulatory blood pressure monitoring and blood pressure self-measurement appear to be minimal. For ambulatory devices, these include inconvenience and arm soreness from multiple cuff inflations. For self-measurement devices, increased anxiety is a consideration. Of greater concern is the accuracy of these techniques during routine use. Even when the devices are accurate during product evaluation, regular calibration and careful adherence to standard measurement techniques are essential.

Costs

Widespread and routine use of ambulatory blood pressure monitoring would add considerably to the direct costs of hypertension measurements, especially if it were used to monitor treatment responses as well as to diagnose hypertension. On a societal level, the additional direct cost would be approximately $6 billion per year, if each of the estimated 50 million hypertensive persons in the United States were to have one ambulatory monitoring test per year (at a cost of $120 per test). A comparable estimate for routine use of self-measurement devices at a unit cost of $50 per year would be $2.5 billion. Naturally, if ambulatory blood pressure monitoring were applied only to a subset of hypertensive persons, direct costs would be much lower. Downstream cost savings from less frequent or less intensive drug treatment could be substantial but also are a matter of speculation.

Policy Options

The following three broad policy options should be considered: 1. At one extreme, policy makers could endorse use of both technologies as standard practice based on the apparently favorable balance between benefits and risks. This would be consistent with previous decisions on other diagnostic technologies. Such an endorsement would, however, ignore concerns over the high cost of widespread use of ambulatory blood pressure monitoring and expected difficulties in controlling its utilization.

2. At the other extreme, policy makers could argue that existing evidence is insufficient to support any application of these technologies. This position would effectively discount the value of published evidence. For ambulatory blood pressure monitoring, such a decision would implicitly emphasize the overriding importance of costs, implying a cost-effectiveness rather than a net benefits standard.

3. A third option, intermediate between these extremes, would be to endorse these technologies only for indications that are best supported by existing evidence. For ambulatory monitoring, one indication would be the initial evaluation of certain low-risk patients with suspected hypertension. For self-measurement devices, a wider range of applications could be endorsed, accepting a lower standard of scientific proof because of lesser cost consequences. From a practical point of view, approval of ambulatory blood pressure monitoring for a few applications would require either restricting insurance coverage (for example, to a single test per person in any 12-month period) or limiting its use to certain health care delivery systems (such as health maintenance organizations and Department of Veterans Affairs facilities), clinical researchers, and self-paying patients.

On balance, we favor this intermediate position, which gives credence to the large body of evidence supporting the value of these technologies but simultaneously acknowledges residual uncertainties over their efficacy and the risk of overuse. At the same time, we advocate a broad research agenda aimed at verifying the benefits, harms, and costs of using ambulatory and self-measurement devices to classify individuals and to guide treatment decisions.

Interim clinical applications should target those individuals who are most likely to benefit, and least likely to be harmed, from decisions guided by these technologies. For this reason, we believe their use in clinical decision making should be limited to certain individuals at comparatively low risk for cardiovascular disease, namely a subset of those patients with stage 1 diastolic hypertension (that is, a diastolic blood pressure of 90 to 99 mm Hg) [138]. Figure 3 presents one plausible strategy that integrates these technologies and nonphysician measurements in the diagnosis of office hypertension in persons with a physician-measured diastolic blood pressure of 90 to 99 mm Hg. An identical strategy could be applied to diagnose office hypertension in individuals with stage 1 systolic hypertension (that is, a systolic blood pressure of 140 to 159 mm Hg).

Initially, repeated blood pressures would be obtained by well-trained nurses or technicians at three separate office visits. These measurements should reduce any alerting reactions elicited during physician measurement [85] and would be similar to the techniques used in most clinical trials of antihypertensive drug therapy. Next, individuals with office blood pressures of 90 to 99 mm Hg (measured by a nonphysician) would be stratified on the basis of target-organ disease (myocardial infarction, stroke, left ventricular hypertrophy, and nephropathy) and other cardiovascular risk factors (diabetes, smoking, and hypercholesterolemia), using information from the history, physical examination, and routinely collected laboratory studies [138]. Those with evidence of blood pressure-related, target organ damage, or other cardiovascular risk factors would receive nonpharmacologic and pharmacologic treatments, as deemed appropriate; those with no evidence of target organ damage and no risk factors would be classified by self-measured blood pressure (if the patient is willing, interested, and able) or otherwise awake ambulatory blood pressure. Drug treatment would be considered for those patients with elevated self-measured blood pressure or ambulatory blood pressure. A parallel strategy could be used to document office hypertension in individuals with apparent drug resistance, after problems with medication adherence and other causes of refractory hypertension have been explored [139]. Finally, self-measurement devices could be useful in monitoring patients in other clinical settings (for example, during drug titration).

We can only speculate on how implementation of these strategies would affect the benefits, risks, and costs of hypertension treatment. From a clinical point of view, they have little downstream risk as long as individuals, in whom drug treatment is withheld because of the office hypertension phenomenon, receive follow-up at regular intervals. On a societal level, these strategies place ambulatory monitoring at the end of a decision-making process and thereby limit its use to a small, comparatively low-risk subset of hypertensive persons. Net costs will depend primarily on the extent to which the direct costs of ambulatory monitoring and blood pressure self-measurement are offset by cost savings from reduced frequency and intensity of antihypertensive drug therapy.

Several practical issues need to be addressed before implementation of our proposed strategies. Agreement is needed on the operational definitions of high ambulatory and self-measured blood pressure. A daytime ambulatory blood pressure of 135/85 mm Hg has been suggested as a threshold for initiating or intensifying pharmacologic treatment [47], but this level has yet to be widely endorsed. Furthermore, it is unclear whether the same threshold should apply to self-measured blood pressure. We believe that these decision thresholds are best established by consensus, and preferably by a broad-based expert panel such as the Joint National Committee on the Detection, Treatment, and Evaluation of High Blood Pressure convened by the National High Blood Pressure Education Program. This group could also provide recommendations on whether average ambulatory blood pressure or blood pressure load should guide clinical decisions; whether 24-hour or daytime ambulatory blood pressure should be used; and whether blood pressures on work days, nonwork days, or a combination should be obtained. Pending such decisions, we recommend that practitioners apply a relatively low blood pressure level, perhaps 135/85 mm Hg or less, as the threshold for initiating or intensifying drug therapy based on mean self-measured or mean daytime ambulatory blood pressure.

Our proposed strategies effectively operationalize certain recommendations of the Joint National Committee and its working group on ambulatory monitoring, specifically those pertaining to the diagnosis of office hypertension and the evaluation of drug resistance [12, 138]. These reports also mention other clinical settings (Table 1) in which ambulatory monitoring might be useful (for example, evaluating nocturnal blood pressure changes, episodic hypertension, hypotensive symptoms, carotid sinus syncope, and pacemaker syndromes). Although the pathophysiology of these conditions leads us to believe that ambulatory blood pressure monitoring might provide useful information, evidence on the actual effectiveness of this technology in these clinical settings is limited.

Research Agenda

The need for further research is unequivocal. Well-designed clinical studies are needed to do the following:

1. Provide unbiased evaluations of ambulatory and self-measurement devices under field and controlled conditions;

2. Compare ambulatory, self-measured, and office blood pressures (both physician and nonphysician measurements) in representative primary care and community populations. Such studies will provide better estimates of the prevalence of office hypertension and will address the issue of whether these techniques provide qualitatively unique information or merely a more precise estimate of office blood pressure;

3. Assess the roles of ambulatory blood pressure monitoring and blood pressure self-measurement in cardiovascular disease risk prediction;

4. Determine which environmental settings (work, home, sleep, 24-hour) and which ambulatory blood pressure measures (mean ambulatory blood pressure, blood pressure load) are optimal for cardiovascular disease risk prediction and clinical decision making;

5. Determine whether treatment guided by ambulatory or self-measured blood pressure is superior to treatment guided by office blood pressure in terms of cardiovascular outcomes (left ventricular mass and, if feasible, clinical cardiovascular events) and adverse treatment effects. Two European multicenter trials with these objectives are underway [140];

6. Examine the effects of ambulatory monitoring and blood pressure self-measurement on the costs and cost-effectiveness of hypertension management. Studies addressing these issues deserve high priority and perhaps should be jointly funded by the National Institutes of Health, Agency for Health Care Policy and Research, and manufacturers of ambulatory and self-measurement devices.

Summary

Limited clinical applications for ambulatory monitoring and blood pressure self-measurement appear warranted based on available research and clinical experience supporting the relevance of office hypertension. Although epidemiologic data suggest that absolute cardiovascular risk might increase due to resultant underdiagnosis and undertreatment, the magnitude of such effects should be extremely small if, as proposed, drug treatment is withheld only from comparatively low-risk individuals. Endorsement of ambulatory blood pressure for routine clinical use in other clinical settings, however, should await more convincing evidence of its clinical effectiveness.