Chapter  63:  Utility of Blood Pressure Monitoring Outside of the Clinic Setting: Evidence Report/Technology Assessment Number 63

Objectives

Ambulatory BP (ABP) and self-measured BP (SMBP) monitoring are two techniques that record frequent BP outside of the clinic setting. The overall objective of this report was to summarize evidence on the clinical utility of ABP and SMBP monitoring.

Search Strategy

Electronic searches were completed of MEDLINE^®, Cochrane Collaboration CENTRAL Register of Controlled Trials, and HealthSTAR. Hand searching was completed of key journals, conference proceedings and references lists. Electronic searching was completed to March 2001, and hand searching was completed to May 2001.

Selection Criteria

Articles were included in this evidence synthesis if they were English-language reports of original data that addressed one of the specific research questions in nonpregnant adults.

Main Results

Eighteen studies compared clinic BP, SMBP, and/or ABP. For both systolic and diastolic BP, clinic measurements exceeded SMBP and ABP. Few studies compared SMBP and ABP. Sixteen studies determined the prevalence of white coat hypertension (WCH). Overall, WCH prevalence was approximately 20 percent among hypertensives but varied considerably by definition. Few studies assessed the reproducibility of WCH (two studies) or the reproducibility of differences between clinic BP and either ABP (one study). In cross-sectional studies of BP with left ventricular mass and/or albuminuria (25 studies), ABP levels were directly associated with both measurements; also, left ventricular mass was less in individuals with WCH than in those with sustained hypertension. Ten prospective studies assessed the relationship of ABP with subsequent clinical outcomes. In each study, at least one dimension of ABP predicted outcomes. WCH predicted a reduced risk of CVD events compared to sustained hypertension. However, data were inadequate to compare the risk associated with WCH to the risk associated with normotension. A nondipping or inverse dipping pattern predicted an increased risk of clinical outcomes. The literature was insufficient to determine whether absolute SMBP levels or WCH based on SMBP was associated with left ventricular mass or proteinuria (just one study) or whether SMBP measurements predicted subsequent CVD (just one study). In both cross-sectional and prospective studies, the poor or uncertain quality of clinic measurements precluded a satisfactory comparison of SMBP and ABP with clinic BP. Twelve trials assessed whether use of SMBP had an impact on BP control. In half of these studies, including two trials that tested contemporary devices, use of SMBP was associated with reduced BP. The availability of just two ABP trials limited inferences about the utility of ABP to guide BP management. In general, few studies reported enrollment of African-Americans. Studies infrequently reported results stratified by gender. The only notable subgroup finding was a higher prevalence of WCH in women than men.

Conclusions

In cross-sectional studies, ABP levels and ABP patterns were associated with BP-related target organ damage. Likewise, in prospective studies, higher ABP, sustained BP, and a nondipping ABP pattern were associated with an increased risk of subsequent CVD events. Few studies examined corresponding relationships for SMBP. An inadequate number of clinic BP measurements, as well as the poor or uncertain quality of these measurements, precluded satisfactory comparisons of risk prediction based on ABP or SMBP with risk prediction based on clinic BP. In aggregate, these findings provide some evidence that ABP monitoring is useful in evaluating prognosis. However, evidence was insufficient to determine whether the risks associated with WCH are sufficiently low to consider withholding drug therapy in this large subgroup of hypertensive patients. For SMBP, available evidence suggested that use of SMBP can improve BP control; however, further trials that evaluate contemporary SMBP devices are needed.

Overview

Elevated blood pressure (BP), also termed hypertension, is a common, powerful, and independent risk factor for cardiovascular diseases (CVD) and kidney disease. Approximately 25 percent of the adult U.S. population, about 50 million persons, has hypertension, defined as current use of anti-hypertensive medication, a systolic BP >140 mmHg, and/or diastolic BP > 90 mmHg.

In view of the epidemic of high BP and its complications, prevention and control of high BP continues to be a major national health priority. Governments, institutions, health care providers, insurers, private industry, and non-profit organizations have committed substantial resources to prevent and treat hypertension. Still, hypertension control rates have been unsatisfactory.

Measuring BP to diagnose hypertension and to monitor therapy is problematic. Concomitantly, the enormous scope of the BP problem, the high aggregate costs of hypertension care, and the potential for medication side effects have spawned efforts to target therapy more effectively. This entails identifying lower risk individuals who might be candidates for less aggressive therapy and higher risk individuals who should receive more aggressive therapy. Measurement of BP outside of the office or clinic setting by ambulatory BP (ABP) monitoring and self-measured BP (SMBP) monitoring might accomplish these objectives.

Clinic Blood Pressure Measurements

BP as recorded in the office or clinic setting is the standard technique recommended for measurement of BP in routine medical care. The standard technique includes use of a mercury sphygmomanometer (or a calibrated aneroid device or validated electronic device) and an appropriate-sized cuff. Prior to measurement, patients should rest quietly in the seated position for several minutes. At each visit, at least two readings should be obtained. Except for those individuals with extremely high BP, the diagnosis of hypertension and adjustments in medication should then be based on the average of readings across two or more visits.

Clinic BP measurements have several limitations, even if they are measured according to established guidelines. First, clinic BP measurements exhibit enormous variability, which hinders accurate classification and which frustrates providers and patients. Another limitation is that BP measured in the clinic may not be a representative estimate of usual BP outside the clinic setting. Commonly, BP rises in the clinic setting, in response to the observer and/or other aspects of the medical environment. The difference between measurements obtained in and outside the clinic setting leads to confusion about the diagnosis of hypertension and the need to start or modify therapy. Unfortunately, there are additional limitations because clinic measurements often do not conform to established guidelines. Specific limitations include lack of observer training, inadequate rest period prior to initial measurement, use of wrong-sized cuffs, rapid deflation of cuff, incorrect position of patients, and awkward position of the observer and/or manometer.

Over the past several years, stationary automated devices and aneroid devices have increasingly replaced mercury sphygmomanometers in the clinic setting. Aneroid devices are inexpensive but still require an individual, typically a health care provider, to manually inflate a cuff and record the appearance and disappearance of Korotkoff sounds. In contrast, fully automated devices require minimal technical skills, that is, only placement of a cuff and initiation of a reading. An additional reason leading to greater use of aneroid and automated devices stems from concerns over mercury toxicity.

Self-measured Blood Pressure (SMBP)

SMBP devices include mercury sphygmomanometers, aneroid manometers, semiautomatic devices, and fully automatic electronic devices. Automatic devices measure BP using an oscillometric technique in which systolic and diastolic BP are estimated from the pattern of vibrations in the cuff as it is deflated. Fully automated devices are popular because the patient does not have to inflate the cuff or listen for the appearance and disappearance of Korotkoff sounds. Although numerous, perhaps hundreds, of SMBP devices are on the market, very few have been independently validated.

SMBP devices provide an opportunity to record BP at home, outside of the artificial setting of the medical office or clinic. Ideally, the patient is trained to record BP using a standard technique. Occasionally, physicians may observe the patient recording a BP measurement in the clinic and then perform a cross check of readings. The presentation of SMBP data is extraordinarily variable. Commonly, patients at their own initiative provide written lists of readings to their physicians at office visits. However, recent innovations have greatly enhanced the potential utility of SMBP devices to synthesize and present data. Contemporary SMBP devices have the capacity to store and download readings via phone or computer. Data can then be synthesized and reports can be generated and sent to the patient and/or physician.

SMBP has several potential uses. Repeated measurements, if averaged, should provide a more precise estimate of usual BP than occasional measurements obtained in the clinic. As a substitute for clinic BP, SMBP monitoring could then be used to adjust anti-hypertensive drug therapy and thereby reduce the need for frequent clinic visits and their associated costs and inconvenience. The extent to which physicians, or patients, use SMBP data to adjust medication is unclear. In addition, self-measurement of BP has also been proposed as a means to improve adherence with treatment.

Self-measurement of BP theoretically provides a means to diagnose white coat hypertension (WCH), also termed non-sustained or office hypertension. This pattern refers to an elevation of clinic BP in the hypertensive range but normal or low BP outside the clinic setting. Individuals with WCH may be at comparatively low risk for BP-related complications in comparison to individuals with sustained hypertension. An important issue is whether the risk of WCH exceeds that of nonhypertensives.

Ambulatory Blood Pressure (ABP) Measurement

ABP monitoring is a noninvasive, fully automated technique in which BP is recorded over an extended period of time, typically 24 hours. The required equipment includes a cuff, a small monitor (attached to a belt), and a tube connecting the monitor to the cuff. Usually, a trained technician places the device on the patient, provides instructions to the patient, and then downloads data from the device when the patient returns. Most ABP devices use an oscillometric technique. Compared to SMBP, relatively few ABP devices are on the market. However, in contrast to SMBP devices, most currently available ABP devices have undergone validation testing, as recommended by the American Association of Medical Instrumentation (AAMI) or the British Hypertension Society (BHS).

During a typical ABP monitoring session, BP is measured every 15 to 30 minutes over a 24-hour period (including both awake and asleep hours). The total number of readings usually varies between 50 and 100. BP data are stored in the monitor and then downloaded into device-specific computer software. The raw data can then be synthesized into a report that provides mean values by hour and period (daytime [awake], nighttime [asleep], and 24-hour BP), both for systolic and diastolic BP. The most common output used in decisionmaking are absolute levels of BP, that is, mean daytime, nighttime, and 24-hour values. Because of the expense of ABP equipment (up to $5,000 for a monitor, cuff set and software), the requirement for technicians, the inconvenience and logistics of placing and removing ABP devices, and, until recently, the lack of reimbursement, it is uncommon for ABP monitoring to be done frequently. However, use of ABP will likely increase as a result of the decision by the Centers for Medicare and Medicaid Services (CMS) to cover ABP in selected settings, namely, the identification of WCH.

In addition to mean absolute levels of ABP, certain ABP patterns may predict BP-related complications. The patterns of greatest interest are white coat hypertension and nondipping BP. Using both daytime and nocturnal ABP, one can identify individuals, termed nondippers, who do not experience the decline in BP that occurs during sleep hours. Usually, nighttime (asleep) BP drops by 10 percent or more from daytime (awake) BP. Research has suggested that individuals with a nondipping pattern (less than 10-percent BP reduction from night to day) may be at increased risk of BP-related complications compared to those with a normal dipping pattern.

Although ABP could be used to monitor therapy, the most common application is diagnostic, that is, to ascertain an individual's usual level of BP outside the clinic setting and thereby identify individuals with WCH. In addition to detection of WCH, ABP devices may be used to identify individuals with a nondipping BP pattern and to evaluate apparent drug resistance, hypotensive symptoms to medications, episodic hypertension, and autonomic dysfunction. Use of ABP monitoring has been controversial. First, few prospective studies have determined whether this technology predicts cardiovascular disease outcomes and whether this technology provides additional information beyond that of routine clinic measurements. Second, insurers have been concerned that health care providers might overutilize ABP. Third, it has been unclear whether SMBP monitoring is a satisfactory and less expensive alternative to ABP monitoring. Accordingly, health insurers have been reluctant to reimburse for ABP monitoring.

Reporting the Evidence

The utility of BP monitoring outside of the clinic setting was a topic nominated to the Agency for Healthcare Research and Quality (AHRQ) by a group of experts in BP measurement. In September of 2000, AHRQ awarded a contract to the Johns Hopkins Evidence-based Practice Center (EPC) to prepare an evidence report on this topic. The Johns Hopkins EPC established a team and work plan to develop a report that would identify and synthesize the best available evidence on BP monitoring. One of the first tasks was the identification of an appropriate partner. In December 2000, the National High Blood Pressure Education Program (NHBPEP) of the National Heart, Lung, and Blood Institute (NHLBI) of the National Institutes of Health (NIH) hosted a working meeting. The NHBPEP includes representatives from national professional and voluntary organizations as well as from Federal agencies. Arising from that meeting was an agreement from the NHBPEP Coordinating Committee to partner with the Johns Hopkins EPC on this project.

A core group of five clinically and/or methodologically oriented technical experts advised the EPC team at key points in the project. This group included experts in ABP monitoring, SMBP monitoring, clinic BP measurement, clinical hypertension, and diagnostic test evaluation. These individuals reviewed draft research questions. Also, this core group along with additional experts in BP measurement and hypertension provided early input at an ad hoc meeting convened by the NHBPEP. The target population consisted of nonpregnant adults with BP in the nonhypertensive or hypertensive range. These individuals are candidates for BP monitoring, and many are candidates for anti-hypertensive drug therapy.

Key Questions

After an extensive deliberative process and with input from the technical experts, the following questions were developed:

• Comparison of clinic, ambulatory, and SMBP readings:
      1a. What is the distribution of the BP differences between clinic, ambulatory, and SMBP readings? If there are differences, are these differences reproducible?
      1b. What is the prevalence of WCH as defined by SMBP? Is this pattern reproducible?
      1c. What is the prevalence of WCH as defined by ABP measurement? Is this pattern reproducible?

• SMBP levels and WCH based on SMBP as related to clinical outcomes:
      2a. Is SMBP more or less strongly associated with BP-related target organ damage than clinic BP measurements?
      2b. Does SMBP predict subsequent clinical outcomes?
      2c. What is the incremental gain in prediction of clinical outcomes from use of self-measurement devices beyond prediction from clinic BP alone?
      2d. What is the effect of treatment guided by SMBP in comparison to treatment guided by clinic BP, in terms of:
  

  1. BP-related target organ damage

  2. symptoms

  3. use of anti-hypertensive drug therapy

  4. BP control

• ABP levels and WCH based on ABP as related to clinical outcomes:
      3a. Is ambulatory blood pressure more or less strongly associated with BP-related target organ damage than clinic BP measurements?
      3b. Does ambulatory blood pressure predict subsequent clinical outcomes?
      3c. What is the incremental gain in prediction of clinical outcomes from use of ambulatory devices beyond prediction from clinic BP alone?
      3d. What is the effect of treatment guided by ABP in comparison to treatment guided by clinic BP, in terms of:
  

  1. BP-related target organ damage

  2. symptoms

  3. use of anti-hypertensive drug therapy

  4. BP control

• Does the evidence for the above questions vary according to a patient's age, gender, income level, race/ethnicity, and clinical subgroups (e.g., hypertensive/normotensive, diabetic, renal transplant status)?

Methodology

Searching the literature included identifying reference sources, formulating a search strategy for each source, and executing and documenting each search. A comprehensive search plan was developed that include electronic and hand searching. Several electronic databases were searched and a separate strategy was developed for each. First searched was MEDLINE^®, which was accessed through PubMed^®. Searches using PubMed^® were completed in January 2001 and March 2001. The Cochrane CENTRAL Register of Controlled Trials was searched once (Issue 1, 2001). HealthSTAR was searched in February 2001.

Hand searching for possibly relevant citations took several forms. First, priority journals were identified through an analysis of the frequency of citations per journal in the database of search results as well as through discussions amongst the EPC team. Fifteen specialty and general journals were identified. The January to May 2001 issues of these journals were searched. For the second form of hand searching, a database of reference material, identified through an electronic search for relevant guidelines and reviews, through discussions with experts, and through the article review process, was created in the reference management software, ProCite. A listing of titles and abstracts from this database, the BP References Database, was reviewed by the principal investigator to identify key articles. The reference lists of these articles were then reviewed to identify possibly relevant citations. Finally, proceedings from recent conferences were also reviewed.

Abstract and Article Review Process

Specific inclusion and exclusion criteria were applied at each of three levels of review (two levels of abstract review, then article review). Inclusion criteria became more stringent at each level. The titles and abstracts were reviewed for each article identified. During the abstract review process, emphasis was placed on identifying all articles that may possibly have original data pertinent to the questions. For the first-level abstract review, titles and abstracts for all articles retrieved by the literature search were printed on an abstract form and distributed to two reviewers. Because of the extensive volume of literature, a second level abstract review, at which additional exclusion criteria were applied, was necessary. Citations deemed eligible for full article review based on the initial abstract review were printed onto the second level abstract form and distributed to two reviewers.

The purpose of the article review was to confirm the relevance of each article to the research questions, to determine methodological characteristics pertaining to study quality, and to collect evidence that addressed the research questions. Because of the large number of citations that remained eligible for full article review even after the second level abstract review, additional exclusion criteria were applied at the article review level. The final full list of exclusion criteria differed by question. For instance, for question 1a, a comparison of BP by the different techniques, the criterion of more than 1 day of measurement for clinic BP was added because an average clinic BP based on just 1 day of measurements (typically just one to three readings) is extremely imprecise and could lead to a biased comparison with ABP or SMBP.

Article review forms were developed to collect data in a standardized fashion. This process was complex and time consuming due to the heterogeneity of the literature and the diverse questions being addressed. These forms then guided article review. For each of the articles deemed potentially eligible after second-level abstract review, two reviewers read the article, confirmed eligibility status, abstracted key information, and assessed study quality on several dimensions. Because of heterogeneity in study design, data collection forms and elements differed by research question.

Presentation of Results

Evidence tables that summarize aspects of study quality, characteristics of the study population, and features of BP measurement were constructed. For most research questions, these summary tables were similar. However, the evidence tables that display study results differed substantially by research question. Qualitative summaries were prepared which synthesized the evidence and included, to a limited extent, a quantitative assessment (for example, the number/percent of studies with significant associations, and occasionally by relevant study characteristics).

A draft version of the report was distributed to the partner, the technical advisory group, and other peer reviewers. All substantive comments were collated, the responses of the EPC team summarized, and edits were made to the report as appropriate.

Findings

• Key question 1. Comparison of clinic BP, SMBP, and ABP readings.

  • Question 1a. Distribution of BP differences.
    A total of 18 studies addressed the distribution of BP differences. BP levels measured outside the clinic setting differed from those obtained in the clinic. For both systolic and diastolic BP, clinic measurements exceeded SMBP, daytime ABP, nighttime ABP and 24 hour ABP. In the few studies that compared SMBP and ABP, daytime ABP and SMBP appeared similar, while nighttime ABP was consistently lower than SMBP. The literature was insufficient to determine whether these BP differences are reproducible.

  • Question 1b. Prevalence of WCH based on SMBP.
    A total of four studies addressed this issue. Hence, the literature was insufficient to determine the prevalence of WCH by SMBP.

  • Question 1c. Prevalence of WCH based on ABP.
    A total of 16 studies addressed this issue. Prevalence varied by WCH definition and study population. Overall, the prevalence was approximately 20 percent among patients with hypertension. Only two studies addressed the reproducibility of WCH. Hence, the literature was insufficient to determine whether WCH based on ABP is reproducible.

• Key question 2. The relationship of SMBP levels and WCH based on SMBP to clinical outcomes.

  • Question 2a. Associations of SMBP with target organ damage.
    Only one study addressed this issue. Hence, the literature was insufficient to determine the associations of absolute SMBP levels or WCH as determined by SMBP with left ventricular mass or proteinuria.

  • Question 2b. Associations of SMBP with clinical outcomes in prospective studies.
    Only one study addressed this issue. Hence, the literature was insufficient to determine whether absolute SMBP levels or WCH based on SMBP predicts subsequent CVD.

  • Question 2c. Comparison of risk prediction from SMBP and clinic BP.
    Only one study addressed this issue. The dearth of studies combined with the poor or uncertain quality of clinic BP measurements precluded an answer to this question.

  • Question 2d. Effect of treatment guided by SMBP.
    Twelve trials addressed this issue, but the evidence was inconsistent. In half of these trials, interventions that included SMBP led to reduced BP. Two trials used contemporary SMBP technology which can store and synthesize SMBP measurements and which can generate BP reports. In both of these trials, the SMBP intervention led to reduced BP.

• Key question 3. The relationship of ABP levels and WCH based on ABP to clinical outcomes.

  • Question 3a. Cross-sectional associations of ABP with target organ damage.
    A total of 25 studies addressed these issues. Left ventricular mass and albuminuria were positively associated with ABP.

  • Question 3b. Associations of ABP with clinical events in prospective studies.
    A total of 10 studies addressed this issue. In each study, at least one dimension of ABP predicted subsequent clinical events, primarily CVD. In two of these studies, WCH was associated with a reduced risk of CVD relative to the risk associated with sustained hypertension. No prospective study adequately compared the risk associated with WCH relative to the risk associated with non-hypertension. In four of five studies, a nondipping or inverse dipping pattern predicted an increased risk of adverse events.

  • Question 3c. Comparison of risk prediction from ABP and clinic BP.
    A total of nine prospective studies addressed this issue, but only two studies assessed incremental gain, that is, whether ABP provided additional information that was predictive of risk beyond that of clinic BP. However, the poor or uncertain quality of clinic BP measurements precluded a satisfactory comparison of risk prediction from ABP and clinic BP.

  • Question 3d. Effect of treatment guided by ABP.
    Only two trials addressed this issue. Hence, the literature was insufficient to determine the effects of treatment guided by ABP.

• Key question 4. Findings according to subgroups.

  • The vast majority of studies included both men and women, but few studies reported results separately by gender.

  • Few studies reported enrollment of African-Americans, and race-stratified data were rarely presented.

  • The only notable subgroup finding was a higher prevalence of WCH in women than in men.

In summary, ABP levels and ABP patterns were associated with BP-related target organ damage in cross-sectional studies. Likewise, in prospective studies, higher ABP, sustained hypertension, and a nondipping ABP pattern were associated with an increased risk of subsequent CVD events. Few studies examined corresponding relationships for SMBP. An inadequate number of clinic BP measurements, as well as the poor or uncertain quality of clinic BP measurements, precluded satisfactory comparisons of risk prediction based on ABP or SMBP with risk prediction based on clinic BP. In aggregate, these findings provide some support for use of ABP monitoring in evaluating prognosis. However, evidence was insufficient to determine whether the risks associated with WCH are sufficiently low to consider withholding drug therapy in this large subgroup of hypertensive patients. For SMBP, available evidence from several trials suggested that use of SMBP can improve BP control; however, further trials that evaluate contemporary SMBP devices are needed.

Future Research

The optimal approach to measure BP remains uncertain. In view of the high prevalence of uncontrolled hypertension, the continuing epidemic of BP-related diseases, and the potential for alternative measurement techniques to improve diagnosis and target therapy, there is a need for comparative studies that assess the relative efficacy, feasibility, and costs of ABP, contemporary SMBP technology, and clinic BP. Specific types of research needs are as follows:

• Prospective observational studies that include SMBP, ABP, and clinic BP. Specific research questions include:

  • What is the repeatability of WCH?

  • What are the risks associated with WCH? In particular, is the risk associated with WCH sufficiently low to justify non-treatment? If yes, in which patients?

  • Does WCH as assessed by SMBP carry the same risk as WCH as assessed by ABP?

  • What are the risks associated with nondipping status?

  • Is nondipping status a surrogate for some other variable that might be measured more easily, that is, without ABP?

  • What is the incremental gain from use of SMBP or ABP over clinic BP alone?

• Clinical trials that test whether contemporary SMBP technology, compared to conventional management by clinic BP, can improve BP control and health outcomes. An additional comparison group might include BP management by ABP. These trials should also compare the aggregate costs of these approaches.

• Decision analyses that determine the costs and effects of strategies that integrate clinic BP, SMBP, and ABP.

• Synthesis of evidence on BP measurements in clinic setting, including issues related to the accuracy and performance of different devices (mercury, aneroid, automated BP) and different observers (physicians, nurses, technicians).

In future research, clinic BP should be measured appropriately by trained observers using validated equipment; measurements should be obtained at several visits. Also, because of the dearth of large-scale, high-quality studies, there is a clear need for government sponsorship of key studies.

To improve the quality of ABP and SMBP publications, standardized methods should be disseminated to researchers and authors. Also, journals should require standardized approaches for presenting ABP data. For published articles, full copies of protocols should be made available, perhaps on the Web. This is especially important because the intense pressure from editors to shorten manuscripts typically leads to reductions in the methods section.

Availability of the Full Report

The full evidence report from which this summary was taken was prepared for the Agency for Healthcare Research and Quality (AHRQ) by the Johns Hopkins Evidence-based Practice Center (EPC), Baltimore, MD, under contract number 290-97-006. It is expected to be available in fall 2002. At that time, printed copies may be obtained free of charge from the AHRQ Publications Clearinghouse by calling 800-358-9295. Requesters should ask for Evidence Report/Technology Assessment No. 63, Utility of Blood Pressure Monitoring Outside of the Clinic Setting. In addition, Internet users will be able to access the report and this summary online through AHRQ's Web site at www.ahrq.gov.

Background

Elevated blood pressure (BP), also termed hypertension, is a common, powerful, and independent risk factor for cardiovascular diseases (CVD) and kidney disease. BP-related CVD include cerebrovascular disease (or stroke), coronary heart disease (CHD), heart failure, and peripheral artery disease. The risk relationships are progressive and graded such that the risk of these diseases rises throughout the range of BP including BP in the non-hypertensive range.^1,2

Approximately 25 percent of the adult U.S. population, about 50 million persons, has hypertension, defined as current use of anti-hypertensive medication, a systolic BP >140 mmHg, and/or diastolic BP < 90 mmHg.³ Less than half of adults have optimal BP defined as systolic BP < 120 mmHg and DBP < 80 mmHg. Hypertension disproportionately affects certain subgroups, particularly African-Americans and older-aged persons. With increasing age, the prevalence of hypertension rises such that over 50 percent of U.S. adults ages 60 years and older have hypertension. While hypertension affects both genders, men have a higher prevalence than women at younger ages, but the opposite is true at later ages (> approximately 50 years).

A compelling body of evidence from clinical trials has documented that drug therapy not only lowers BP but also prevents stroke, CHD and heart failure.^4,5 A complementary strategy to drug therapy for hypertension is non-pharmacologic, lifestyle therapy. A substantial body of research has documented that lifestyle modification can lower BP and prevent hypertension in non-hypertensive individuals who are not candidates for drug therapy but who nonetheless remain at risk for BP-related complications.⁶

In view of the epidemic of high BP and its complications, prevention and control of high BP continues to be a major national health priority. Governments, institutions, health care providers, insurers, private industry and non-profit organizations have committed substantial resources to research aimed at prevention and treatment of hypertension. Professional organizations and governmental bodies have developed guidelines to screen, diagnose, prevent and treat hypertension.⁷ Health insurance companies typically cover the costs of anti-hypertensive care, including, to a variable extent, medication costs. Still, hypertension control rates have been unsatisfactory. In response, performance guidelines have been developed as a means to monitor and improve hypertension control.⁸

Despite this ongoing and massive effort to prevent BP-related complications, the most appropriate technique to measure BP remains uncertain, both to diagnose hypertension and to monitor therapy. Concomitantly, the enormous scope of the BP problem, the high aggregate costs of hypertension care, and the potential for medication side effects have spawned efforts to target therapy more effectively. Specifically, attention has focused on identification of lower risk individuals who might be candidates for less aggressive therapy and higher risk individuals who should receive more aggressive therapy. Measurement of BP outside of the office or clinic setting has been proposed as an alternative to traditional BP measurements. Ambulatory BP (ABP) monitoring and self-measured BP (SMBP) monitoring are two measurement techniques that can record BP outside of the clinic setting and that might accomplish the above objectives.

Clinic Blood Pressure Measurements

BP as recorded in the office or clinic setting is the standard technique recommended for measurement of BP in routine medical care.⁷ Such measurements have been used in the major observational studies that documented risk relationships between BP and clinical events and in most clinical outcome trials that documented the benefits of anti-hypertensive therapy. Ideally, the observer is trained and then retrained periodically. The standard technique includes use of a mercury sphygmomanometer (or a calibrated aneroid device or validated electronic device) and an appropriate size cuff. Prior to measurement, patients should rest quietly in the seated position for several minutes. At each visit, at least two readings should be obtained. Typically, BP measurements at a given visit are then averaged. Except for those individuals with extremely high BP, the diagnosis of hypertension and adjustments in medication should then be based on the average of readings across two or more visits. Numerous national and international professional organizations have prepared guidelines for measurement of clinic BP.⁷

Clinic BP measurements have several limitations, even if they are measured according to established guidelines.⁹ First, clinic BP measurements exhibit enormous variability, which hinders accurate classification and which frustrates providers and patients. Contributing to this variability are short-term variability (within clinic visit), diurnal variability (within the same day), and long-term variability (across an extended period of time, days or weeks). One solution is to measure BP across several visits, spaced several days or weeks apart. Another limitation is that BP measured in the clinic may not be a representative estimate of usual BP outside the clinic setting.¹⁰ Commonly, BP rises in the clinic setting, in response to the observer and/or other aspects of the medical environment. An alerting reaction appears to trigger this response. The difference between measurements obtained in and outside the clinic setting leads to confusion over the diagnosis of hypertension and the need to start or modify therapy. The problem is exacerbated by the practical requirement for cutpoints to diagnose and treat hypertension despite the fact that BP is a continuous, unimodal distribution. In the end, because of misclassification, there is potential both for undertreatment of persons with high blood pressure and overtreatment of those with low blood pressure. Unfortunately, there are additional limitations because clinic measurements often do not conform to established guidelines.¹¹ Specific limitations include lack of observer training, inadequate rest period prior to initial measurement, use of inappropriate sized cuffs, rapid deflation of cuff, incorrect position of patients, insufficient number of BP measurements and visits, and awkward position of the observer and/or manometer.

Over the past several years, stationary automated devices and aneroid devices have increasingly replaced mercury sphygmomanometers in the clinic setting. Aneroid devices are inexpensive but still require an individual, typically a health care provider, to manually inflate a cuff and record the appearance and disappearance of Korotkoff sounds. In contrast, fully automated devices require minimal technical skills, that is, only placement of a cuff and initiation of a reading. The convenience of automated readings and the potential to avoid training and retraining of technicians has made automated readings extremely popular. An additional reason leading to greater use of aneroid and automated devices stems from concerns over mercury toxicity.¹² Specifically, to reduce the amount of mercury released into the environment and to minimize the risk of accidental mercury exposure, government officials have encouraged health care officials to eliminate mercury from health care settings.

Self-measured Blood Pressure (SMBP)

SMBP devices include mercury sphygmomanometers, aneroid manometers, semi-automatic devices, and fully-automatic electronic devices. Automatic devices measure BP using an oscillometric technique in which systolic and diastolic BP are estimated from the pattern of vibrations in the cuff as it is deflated. This technique is quite different from the usual auscultatory technique in which systolic BP is estimated as the point of appearance of Korotkoff sounds and diastolic BP as the point of disappearance. Fully automated devices are popular because the patient does not have to inflate the cuff, listen for the appearance and disappearance of Korotkoff sounds, and read measurements off a column or dial. Hence, these devices appeal to individuals with hearing or visual impairments, or limited dexterity. Although numerous, perhaps, hundreds of SMBP devices are on the market, very few have been independently validated. In a recent review of published validation studies, only 23 devices had undergone validation testing; of these, only five were recommended by the European Society of Hypertension.¹³

SMBP devices provide an opportunity to record BP during awake hours, outside of the artificial setting of the medical office or clinic. Ideally, the patient is trained to record BP using a standard technique. Occasionally, physicians may observe the patient recording a BP measurement in the clinic and then perform a cross check of readings. While the medical literature has documented that patients can record BP accurately, there have been concerns about the accuracy of readings, the completeness of reports submitted to physicians, and the potential for biased readings based on selective reporting.¹⁴

The presentation of SMBP data is extraordinarily variable. Commonly, patients at their own initiative provide written lists of readings to their physicians at office visits. However, recent innovations have greatly enhanced the potential utility of SMBP devices to synthesize and present data. Contemporary SMBP devices have the capacity to store and download readings via phone or computer. Data can then be synthesized from which reports are generated and then transmitted to the patient and/or physician.

SMBP has several potential uses.¹⁴ Repeated measurements, if averaged, should provide a more precise estimate of usual BP than occasional measurements obtained in the clinic. As a substitute for clinic BP, SMBP monitoring could then be used to adjust anti-hypertensive drug therapy and thereby reduce the need for frequent clinic visits and their associated costs and inconvenience. The extent to which physicians, or patients, use SMBP data to adjust medication is unclear. Self-measurement of BP has also been proposed as a means to improve adherence with treatment. In addition, self-measurement of BP theoretically provides a means to diagnose 'white coat hypertension (WCH)', also termed 'non-sustained' or 'office' hypertension. This pattern refers to an elevation of clinic BP in the hypertensive range but normal or low BP outside the clinic setting. Individuals with WCH may be at comparatively low risk for BP related complications in comparison to individuals with sustained BP. An important issue is whether the risk of WCH exceeds that of non-hypertensives.¹⁰

Ambulatory Blood Pressure (ABP) Measurement

ABP monitoring is a non-invasive, fully automated technique in which BP is recorded over an extended period of time, typically 24 hours. The required equipment includes a cuff, a small monitor (attached to a belt), and a tube connecting the monitor to the cuff. Usually, a trained technician places the device on the patient, provides instructions to the patient, and then downloads data from the device when the patient returns. Most, but not all, ABP devices use an oscillometric technique. Compared to SMBP, relatively few ABP devices are on the market. However, in contrast to SMBP devices, most currently available ABP devices have undergone validation testing, as recommended by the American Association of Medical Instrumentation (AAMI) or the British Hypertension Society (BHS). In a review of validation studies by O'Brien et al, 24 devices had undergone validation testing and 16 were recommended.¹³

During a typical ABP monitoring session, BP is measured every 15-30 minutes over a 24 hour period including both awake hours and asleep hours. The total number of readings usually varies between 50 and 100. BP data are stored in the monitor and then downloaded into device-specific computer software. The raw data can then be synthesized into a report that provides mean values by hour and period [daytime (awake), nighttime (asleep), and 24 hour BP], both for systolic and diastolic BP. The most common output used in decision making are absolute levels of BP, that is, mean daytime, nighttime, and 24 hour values. Because of the expense of ABP equipment (up to $5,000 for a monitor, cuff set and software), the requirement for technicians, the inconvenience and logistics of placing and removing ABP devices, and until recently, the lack of reimbursement, it is uncommon for ABP monitoring to be done frequently.

In addition to mean absolute levels of ABP, certain ABP patterns may predict BP-related complications. The patterns of greatest interest are 'white coat hypertension' and 'non-dipping' BP. Using both daytime and nocturnal ABP, one can identify individuals, termed 'non-dippers', who do not experience the decline in BP that occurs during sleep hours. Usually, nighttime (asleep) BP drops by 10 percent or more from daytime (awake) BP. Research has suggested that individuals with a 'non-dipping' pattern (less than 10 percent BP reduction from night to day) may be at increased risk of BP-related complications compared to those with a normal dipping pattern.¹⁵

Although ABP could be used to monitor therapy, the most common application is diagnostic, that is, to ascertain an individual's usual level of BP outside the clinic setting and thereby identify individuals with WCH. In addition to detection of WCH, ABP devices may be used to identify individuals with a 'non-dipping' BP pattern and to evaluate apparent drug resistance, hypotensive symptoms to medications, episodic hypertension, and autonomic dysfunction.⁷ Use of ABP monitoring has been controversial. First, few prospective studies have determined whether this technology predicts cardiovascular disease outcomes and whether this technology provides additional information beyond that provided by routine clinic measurements.¹⁶ Second, insurers have been concerned that health care providers might overutilize ABP. Third, it has been unclear whether SMBP monitoring is a satisfactory and less expensive alternative to ABP monitoring. Accordingly, health insurers have been reluctant to reimburse for ABP monitoring. Recently, however, the Centers for Medicare and Medicaid Services has decided to cover use of ABP to diagnose WCH.

Scope and Purpose of Report

This evidence report summarizes and examines the evidence supporting the clinical utility of non-invasive ABP and SMBP monitoring. Although these technologies have been proposed for use in several settings, the focus of this report was the evaluation and management of adults with elevated BP. Patient populations included in this report were non-pregnant adults with BP in the non-hypertensive or hypertensive range.

Recruitment of Technical Experts and Peer Reviewers

Experts were sought who could provide content and/or methodological guidance. The five technical experts were chosen to cover several domains: hypertension management, SMBP, ABP, clinic BP, and evaluation of screening and diagnostic tests. Input was sought from the partner and technical experts through ad hoc correspondence as well as through more formal requests for feedback during the project. Specific requests for feedback were made for key decisions, such as selection and refinement of the questions.

Comprehensive feedback on the draft report was sought from the partner, the technical experts, and other reviewers. Reviewers included members of the NHBPEP Coordinating Committee selected through discussions with the partner. (See appendix A for list of organizations represented by reviewers from which comments were received.)

Patient Population

The search was not limited by age, gender or any other patient characteristic. However, because of the extensive volume of literature, the review did not synthesize evidence for all types of populations. For instance, it was felt that the use of blood pressure monitoring during pregnancy was a distinctive application of these technologies that was beyond the scope of this report. Likewise, articles that focused exclusively on populations of children (less than 20 years of age) were not reviewed.

Questions

The original questions provided by AHRQ included several descriptive questions that were more appropriately addressed as background text in Chapter 1. The EPC team refined the remaining questions and requested feedback from the technical experts and from the partner. When the large volume and heterogeneity of the literature became apparent, the EPC team refined the questions further. Listed below are the questions addressed in this report.

• Comparison of clinic, ambulatory, and SMBP readings:
      1a. What is the distribution of the BP differences between clinic, ambulatory and SMBP readings? If there are differences, are these differences reproducible?
      2a. What is the prevalence of WCH as defined by SMBP? Is this pattern reproducible?
      3a. What is the prevalence of WCH as defined by ABP measurement? Is this pattern reproducible?

• SMBP levels and WCH based on SMBP as related to clinical outcomes:
      2a. Is SMBP more or less strongly associated with BP-related target organ damage than clinic BP measurements?
      2b. Does SMBP predict subsequent clinical outcomes?
      2c. What is the incremental gain in prediction of clinical outcomes from use of self-measurement devices beyond prediction from clinic BP alone?
      2d. What is the effect of treatment guided by SMBP in comparison to treatment guided by clinic BP, in terms of:
  

  1. BP-related target organ damage

  2. symptoms

  3. use of anti-hypertensive drug therapy

  4. BP control

• ABP levels and WCH based on ABP as related to clinical outcomes:
      3a. Is ambulatory blood pressure more or less strongly associated with BP-related target organ damage than clinic BP measurements?
      3b. Does ambulatory blood pressure predict subsequent clinical outcomes?
      3c. What is the incremental gain in prediction of clinical outcomes from use of ambulatory devices beyond prediction from clinic BP alone?
      3d. What is the effect of treatment guided by ABP in comparison to treatment guided by clinic BP, in terms of:
  

  1. BP-related target organ damage

  2. symptoms

  3. use of anti-hypertensive drug therapy

  4. BP control

• Does the evidence for the above questions vary according to a patient's age, gender, income level, race/ethnicity, and clinical subgroups (e.g., hypertensive/normotensive, diabetic, renal transplant status)?

Causal Pathway

Figure 1. Conceptual Framework

   Figure 1. Conceptual Framework

Figure 1

Literature Search Methods

Searching the literature included the steps of identifying reference sources, formulating a search strategy for each source, and executing and documenting each search.

Abstract Review

Specific inclusion and exclusion criteria were applied at each of three levels of review, with criteria becoming more stringent as the process moved from searching, to the review of abstracts and to the review of articles. After identifying a citation, its title and abstract were reviewed, and articles were included or excluded from the article review on this basis.

Article Review

The purpose of the article review was to confirm relevance of each article to the research questions, to determine methodological characteristics pertaining to study quality, and to collect evidence that addressed the research questions. Where articles described more than one study, reviewers were instructed to complete the eligibility assessment (i.e., comparison to inclusion and exclusion criteria), quality assessment and data abstraction for each study separately. For each question, publications of the same information from the same study were also excluded. These apparent duplicate publications were reviewed on a per case basis. Multiple publications were kept if they reported on different results (i.e., different outcomes). Otherwise, the article with a more comprehensive reporting of the data reviewed.

Because of the large number of citations that remained eligible for full article review even after the second level abstract review, additional exclusion criteria were applied at the article review level. The final full list of exclusion criteria differed by question.

Exclusion criteria applied to all articles during article review:

• does not include human data

• not in English

• no original data

• meeting abstract (no full article for review)

• article does not apply to any of the research questions

• article does not include ambulatory or self-measured blood pressure

• article included <50 patients OR addressed reproducibility and included <20 patients

• device evaluation was the primary purpose of the study

• study population is exclusively pregnant women

• study population is exclusively children (<20 years of age)

• article addresses research question, but does not present data in an abstractable format

• article addresses only the prevalence of dipping versus non-dipping and no other research questions

Additional exclusion criteria for articles addressing question #1:

• article provided data for clinic blood pressure AND ambulatory blood pressure, or clinic blood pressure AND self-measured blood pressure but did not include a formal within-person comparison of measurements (e.g., no p-value, standard error, standard deviation, confidence intervals or only correlation coefficient(s) provided)

• clinic blood pressure measurement used in analyses was completed on one day only
  The criterion of more than one day of measurement for clinic blood pressure was added because an average clinic blood pressure based on just one day of measurements (typically just one to three readings) is extremely imprecise and could lead to a biased comparison with ambulatory or self-measured blood pressure. This criterion was not applied to articles addressing questions 2-4.

For articles addressing questions #2a and #3a, the following specific exclusion criteria were applied:

• article described cross-sectional/retrospective study and did not include comparison with clinic measurement

• article described cross-sectional study but outcome was not left ventricular mass (by echocardiography) or proteinuria/albuminuria

Several endpoints were considered to compare the ability of clinic, self-measured, and ABP monitoring to assess target organ damage caused by hypertension. Left ventricular mass and protein/albumin excretion were included in the report because they are frequently used in the clinic setting to assess the severity and prognosis of hypertension, they are frequently used in hypertension research studies, and there are standard methods available that may allow for some comparability across studies. Other echocardiographic indices of left ventricular enlargement, such as septal thickness or posterior wall thickness, are not consistently reported, and were not considered in this report. Other markers of target organ damage, such as other echocardiographic determinations of left ventricular function, retinopathy, brain MRI findings, carotid intima-media thickness, were not considered in this report.

Because a relatively small number of articles were expected and the abstraction would be quite different, prospective studies (questions #2b or #3b), studies of reproducibility (question #1 a, b, c) and trials examining the impact of treatment guided by clinic versus that guided by ambulatory (question #3d) or self-measurement (question #2d), were tagged during the initial article review. A separate review was then completed for each of these questions including the following additional or modified exclusion criteria.

For articles addressing reproducibility (#1 a, b, c) the additional or modified exclusion criteria were:

• article included < 20 patients

• article does not include reproducibility of white-coat hypertension.

An initial review of articles did not identify any articles addressing reproducibility of the differences between clinic, ambulatory and/or self blood pressure measurements (question #1a). A separate review form for this question was, therefore, not developed. However, the review form used for articles addressing reproducibility was designed to identify articles addressing reproducibility of differences for future consideration.

Additional exclusion criteria for prospective or longitudinal studies (question #2b or #3b) was outcome not of interest.

For articles concerning effect of treatment guided by ambulatory or self measured blood pressure (question #2d or #3d), the additional criterion applied was non-random allocation of participants.

Peer Review

Throughout the project, feedback was sought from the technical experts through ad hoc and formal requests for guidance. A draft of the completed report was sent to the technical experts, as well as to the partner, AHRQ, and other peer reviewers. Substantive comments were entered into a database. Revisions were made to the evidence report, as warranted, and a summary of the comments and their disposition was submitted to AHRQ with the final report.

Literature Search and Abstract Review Process

Table 1: Summary of search and abstract review results (more...)

Table 1: Summary of search and abstract review results

				Number of Citations
Source	Search Strategy	Date	Retrieved	Unique (included in abstract review process)	Eligible for Second Level Abstract Review	Eligible for Article Review Process
PubMed	PubMed	Jan 11, 2001	4,426	4,426	843	529
CENTRAL Issue 1 2001	CENTRAL	Feb 15, 2001	1,382	232	9	6
HealthSTAR	HealthSTAR	Feb 16, 2001	34	34	0	0
PubMed	PubMed	Mar 23, 2001	162	160	50	33
Handsearch	table of contents of priority journals	to May 31, 2001	9	NA	6	6
Handsearch	reference lists of key reviews	July 20, 2001	181	NA	29	22
		TOTAL	6,194	4,852	937	596

Of the 6,194 citations retrieved by the search methods, 4,852 were uniquely identified; that is, not previously included in the Blood Pressure Citations database. Of the 4,852 citations, 902 (19 percent) were classified as eligible for second level abstract review. Citations were excluded at this level if they did not address any of the research questions (37 percent), met any exclusion criteria (26 percent) or a combination of the above. Reviewers did not need to agree on what exclusion criterion applied. The most frequent exclusion criterion applied was that the article did not include ABP or SMBP (used by one or both reviewers to delete 1,256 citations). Other major exclusion criteria were a sample size of less than 20 patients (963 citations) and no original data provided (348 citations).

The 902 citations deemed eligible from the first abstract review were imported into a new database and the 35 citations identified by the hand searching efforts were added. Of the 937 citations reviewed at the second level abstract review, 596 (64 percent) were deemed eligible for full article review. As for the first review, the reviewers did not need to agree on a reason for deleting the citation. Of the 341 citations deleted, reviewers agreed that 186 (55 percent) citations included less than 50 patients, that 29 (8 percent) described cross-sectional studies that addressed only question #2 or #3 and did not contain comparison to clinic measurement, that 28 (8 percent) did not address any of the research questions, and that 24 (7 percent) described cross-sectional studies with outcomes other than left ventricular mass or proteinuria/albuminuria. The remainder of the citations were deleted for other reasons or based on a combination of reasons.

Article Review Process

From the abstract review process, 596 citations were identified for inclusion in the article review phase. We were unable to retrieve, and, therefore, unable to complete article review of three articles.^18-20

Of the 593 articles reviewed, one article described two studies. Each study was assessed and abstracted separately so there were 594 studies for which a review was completed. An initial scan was completed to identify articles with less than 100 patients. These 223 citations were excluded from the general review but were reviewed, as appropriate, for the study questions addressing reproducibility (#1a-c), prediction of clinical outcomes (#2b and #3b - prospective studies) and effect of treatment guided by self or ambulatory blood pressure measurement (#2d and #3d - trials); the minimum sample size for the reproducibility studies was 20, while the minimum sample size for the prospective studies and clinic trials was 50.

General Review

Table 2: Reasons for exclusion at article review level (more...)

Table 2: Reasons for exclusion at article review level (n=252)

Reason for Exclusion	Number of Articles (percent of excluded)
article did not include human data	0 (0%)
article was not in English	2 (0.8%)
article contained no original data	12 (4.8%)
meeting abstract (no full article for review)	1 (0.4%)
article did not apply to any of the research questions	34 (14%)
article did not include ambulatory or self-measured blood pressure	5 (2%)
article had ≤50 patients or addressed reproducibility and had ≤20 patients	3 (1.2%)
device evaluation was the primary purpose of the study	3 (1.2%)
study population was exclusively pregnant women	18 (7%)
study population was exclusively children (< 20 years of age)	13 (5.2%)
article addressed research question but did not present data in abstractable format	17 (6.7%)
article addressed only the prevalence of dipping versus non-dipping and no other research questions	6 (2.4%)
other reasons (e.g., duplicate publication)	22 (8.8%)
Exclusions specific to articles only addressing Question #1
article provided data for clinic blood pressure AND ambulatory blood pressure, or clinic blood pressure AND self-measured blood pressure but did not include a formal within-person comparison of measurements (e.g. no p-value, standard error, standard deviation, confidence intervals or only correlation coefficient(s) provided)	34 (14%)
clinic blood pressure measurement used in analyses was completed on one day only	60 (24%)
Exclusions specific to articles only addressing Questions #2a and #3a
article described cross-sectional/retrospective study and did not include comparison with clinic measurement	13 (5%)
article described cross-sectional study but outcome was not left ventricular mass (by echocardiography) or proteinuria/albuminuria	9 (3.6%)
	252

After the exclusion of 223 articles with under 100 patients, there were 370 articles (representing 371 studies) included in the general review. At the article review level, 252 (68 percent) articles were excluded (representing 253 studies). The primary reasons for exclusion were that the article addressed question #1 only and clinic blood pressure measurement used in analyses was completed on one day only (24 percent of excluded articles) and that the article did not include formal comparison of measurements (14 percent). (See Table 2 for list of exclusions.)

The articles determined to be eligible for review were tagged as addressing the following questions: comparison of readings (question #1) 33 studies, association of SMBP with LV mass or proteinuria/albuminuria (question #2a) one study, and association of ABP with LV mass or proteinuria/albuminuria (question #3a) 27 studies.

As part of the general review process articles were tagged if they addressed issues not being covered in this evidence report and if they addressed any of the other questions being reviewed in separate processes. Articles were tagged as addressing the following issues not included in this review: incremental gain of SMBP (question #2c) (0 studies) or ABP (question #3c) (0 studies) over clinic BP, and the association of dippers with left ventricular mass (six studies) or proteinuria/albuminuria (three studies).

Description of the Literature

The identified literature addressing BP measurement outside of the office setting was vast and heterogeneous. Most ABP and SMBP studies have been published in specialty journals, primarily those in the field of hypertension. From the 596 articles that were eligible for review, the following journals published ten or more articles (ordered from highest to lowest number of publications): Journal of Hypertension (71 articles), American Journal of Hypertension (67 articles), Journal of Human Hypertension (51 articles), Hypertension (48 articles), Blood Pressure Monitoring (36 articles), Journal of Hypertension - Supplement (33 articles), American Journal of Cardiology (11 articles), and Clinical/Experimental Hypertension (11 articles). In contrast, publications in general medical journals were relatively uncommon. For example, the Annals of Internal Medicine published just two articles, the Archives of Internal Medicine five articles, and the Journal of the American Medical Association nine articles.

Of these 596 articles, the vast majority of articles (445 articles, 75 percent) were published between 1990 and 1999; 72 articles (12 percent) were published in 2000 or 2001, and another 73 articles (12 percent) between 1980 and 1989. A similar pattern of journal types and of publication years was evident for the articles that were abstracted for this report.

For the majority of the studies, a funding source could not be identified. Approximately 20 percent of studies cited a government source of funding. Of the 89 studies abstracted, 18 percent were completed in the United States, while 54 percent were completed in European countries.

Question #1

Comparison of clinic, ambulatory, and SMBP readings:

Question #1a. What is the distribution of the BP differences between clinic, ambulatory, and SMBP readings?

A total of 18 studies addressed the distribution of BP differences among clinic BP, ABP, and SMBP and met the inclusion criteria, which included a minimum sample size of 100 and a requirement for at least 2 visits of clinic BP measurements. Among these, six studies compared clinic BP and SMBP,^21-26 12 studies compared clinic BP and ABP,^22,25,27-36 and 3 studies compared SMBP and ABP.^25,37,38 One study compared all three types of BP measurements.²⁵

Two studies reported gender-stratified analyses.^28,33 For both men and women, clinic BP exceeded daytime and 24 hour BP, but the differences appeared somewhat greater in women than men. The same pattern was evident for both systolic and diastolic BP.

Only three studies compared SMBP and ABP (Evidence Tables 9 and 10). There were no significant differences between SMBP and daytime ABP for either systolic or diastolic BP. In contrast, for both systolic and diastolic BP, SMBP was substantially greater than nighttime ABP in the one study that reported differences and was also greater than 24 hour BP in two studies.

In summary, for both systolic and diastolic BP, clinic BP measurements exceed SMBP, daytime ABP, nighttime ABP and 24 hour ABP. Few studies compared SMBP and ABP levels.

Question #1b. What is the prevalence of WCH as defined by SMBP?

Question #1c. What is the prevalence of WCH as defined by ABP measurement?

For studies using ABP monitoring as the method for comparison to clinic BP, the prevalence of WCH ranged from 11 percent to 67 percent. The exceptionally high prevalence of WCH seen in the latter study is noteworthy for several reasons.⁴⁶ The study sample was composed of persons receiving medication for the treatment of hypertension. Thus, the extent to which individual blood pressure medications and/or their dosing schedules influenced the results is unknown. Moreover, the participants in this study were enrolled from a tertiary referral center for management of drug resistant hypertension, a population that may exhibit a higher prevalence of WCH. Excluding the highest and lowest estimates for the prevalence of WCH, the prevalence of WCH ranged from 11.9 to 39 percent. The largest study estimated the prevalence of WCH at 19 percent.⁴⁷ The study that utilized the greatest number of clinic BP measurements (n=9) for use in comparison to ABP estimated the prevalence of WCH at 23 percent.³⁹ Finally, in each study that presented prevalence estimates by gender, the prevalence of WCH was higher in women compared to men. In one study, the prevalence of WCH was statistically higher in women than in men, but no gender-specific prevalence estimates were provided.⁴⁵

In summary, the prevalence of WCH is difficult to ascertain due the lack of standard definitions for both clinic and non-clinic blood pressures. Most studies were relatively small and the populations studied were quite heterogeneous. Nevertheless, the prevalence of WCH from the available evidence is estimated to be between 11 and 69 percent. However, the largest study and the study that utilized the greatest number of clinic blood pressure measurements in its analysis, place the estimate closer to approximately 20 percent. A similar range was observed for WCH as determined by SMBP. Finally, in studies that examined prevalence of WCH by gender, women consistently had a higher prevalence of WCH than men.

Question #1a-c. Reproducibility of differences in readings and WCH

Question #2

The relationship of mean blood pressure levels and WCH as defined by SMBP to clinical events.

Question #2a. Is SMBP more or less strongly associated with BP-related target organ damage than clinic BP measurements?

In summary, only a single study compared SMBP and clinic BP with target organ damage. In this study, SMBP was a better predictor of left ventricular mass than clinic BP. Correlations of albumin excretion with SMBP and clinic BP were similar. Although the study was methodologically sound, the added prognostic information provided by self-measured blood pressure with respect to clinic measurements on target organ damage remains uncertain. No study compared the levels of target organ damage in normotensives, white coat hypertensives, and sustained hypertensives as determined by self-measured blood pressure.

Question #2b. Does SMBP predict subsequent clinical outcomes?

Two articles, both published from the same prospective observational study, addressed the issue of whether SMBP can predict subsequent BP-related events.^60,61 In one article, the outcome variables were total mortality and CVD mortality.⁶⁰ In the other article, fatal and non-fatal stroke was the outcome.⁶¹

Neither clinic systolic BP nor clinic diastolic BP was significantly associated with any of the three outcomes in a progressive, dose-response fashion. However, for stroke, the RRs associated with the highest quintile of clinic systolic and diastolic BP were significant. For SMBP, the RR associated with the fifth quintile of diastolic was significant.⁶¹ In the original publication, the relationship between systolic SMBP and stroke was non-linear, that is, J-shaped.⁶¹ For CVD mortality and for total mortality, systolic SMBP but none of the other BP measurements was significantly associated with these outcomes.⁶⁰

Neither study explicitly tested whether SMBP was superior to clinic BP for predicting outcomes or whether SMBP provided additional prognostic information (incremental gain) beyond that of clinic BP.

In summary, the published literature is insufficient to provide a definitive answer to this research question. The only cohort study that has assessed whether SMBP can predict outcomes documented a linear, progressive relationship of systolic SMBP with total and CVD mortality but a non-linear, J-shaped relationship with stroke. Neither study reported comparative analyses on risk prediction by SMBP and clinic BP.

Question #2c: What is the incremental gain in prediction of clinical outcomes from use of self-measurement devices beyond prediction from clinic BP alone?

Please see discussion for Question #2b.

Question #2d. What is the effect of treatment guided by SMBP in comparison to treatment guided by clinic BP.

All 12 trials had a parallel group design (eight with two groups, two with three groups, one with four groups, and one with five groups). In nine of the trials, SMBP was the only component of the active intervention arm, except for BP reports to patients and/or physicians in three studies. Other dimensions of the active intervention groups were an activated significant other (trained and encouraged to measure in BP) in one trial, telephone evaluation of adherence in one trial, and a multi-component behavioral treatment program in one trial. Two of the 12 trials used telemetry as part of the active intervention program.^66,70 One trial used ABP as the outcome variable while all others used clinic BP measurements.⁷⁰

Initiation and use of medication was reported in three trials. In two trials,^62,68 including the one trial in which BP rose, medication use at the end of follow-up was higher in the control group compared to the SMBP group. In one other trial, medication use was similar.⁶⁹ One trial, that included SMBP as well as telemetric transmission of data and a multi-factorial intervention, documented improved adherence in this group.⁶⁶ One trial documented that SMBP reduced costs of hypertension care.⁷¹

The interpretation of SMBP trial results is complex. First, because SMBP is a diagnostic technology used to assist in BP management, the impact of SMBP is indirect, that is, mediated through changes in BP therapies, both pharmacologic and non-pharmacologic. Hence, an evaluation of SMBP must include an assessment of the approach to therapy in both active and control groups. Unfortunately, none of the papers explicitly stated whether and how SMBP guided therapy. Second, SMBP can be used to adjust BP medications for two distinct problems, that is, to improve BP control in those with inadequately controlled hypertension or to reduce the intensity of BP therapy in persons with apparently low BP. Hence, the lack of BP reduction from SMBP in some studies may reflect a mixed effect, namely, downward titration of medications in some patients and upward titration of medications in other patients. Third, while all trials used SMBP, many of the trials combined SMBP with other interventions, often as a means to improve adherence with therapy. Fourth, SMBP technology is undergoing rapid advances that should influence its effectiveness, specifically, the development of integrated systems that not only synthesize SMBP readings but also can transmit reports to patients and physicians with feedback including advice on therapy. While such advances should, in general, improve the utility of SMBP, there is the potential for inadvertently recording and synthesizing data from multiple individuals (e.g., spouse).

In summary, interventions that included SMBP improved BP control in six of 12 trials. In view of major design limitations, particularly suboptimal measurement of the outcome variable, it is possible that additional studies would have documented benefits had they used a more satisfactory outcome measurement technique. Few published trials used contemporary technologies that automatically synthesize SMBP data over time and that allow for telemetric transmission of SMBP measurements. Of the two trials that used this technology, both documented reduced BP from intervention that included this technology.

Question #3

The relationship of mean levels and WCH as defined by ABP measurement to clinical events.

Question #3a. Is ABP more or less strongly associated with BP-related target organ damage than clinic BP measurements?

For each type of BP measurement assessed (clinic, 24 hour, daytime, or nighttime), the correlations of diastolic BP with LV mass index were in general lower than those of systolic BP with LVMI. Twenty four hour diastolic BP correlations with LV mass index were consistently higher than clinic diastolic BP correlations, with the exception of the normotensive group in the study by Schulte et al.⁹³ Also, daytime and nighttime diastolic BP measurements tended to correlate better with LV mass index than clinic diastolic BP, although not as strongly correlated as 24 hour diastolic BP.

Although the correlation of LV mass and protein excretion with BP tended to be larger for ABP (particularly 24 hour and daytime) than for clinic BP, the poor quality of clinic BP determinations in the majority of studies precludes a satisfactory comparison with clinic BP as recommended by guidelines. The impact of WCH, as determined by ambulatory monitoring, on target organ damage was also evaluated. White coat hypertensives had intermediate levels of LV mass between normotensives and sustained hypertensives as determined by ABP. However, normotensives and white coat hypertensives had similar levels of protein excretion, and only sustained hypertensives had clearly elevated values. These studies were also limited by the poor overall quality of clinic BP measurements, and by the lack of adjustment for potential confounders when comparing normotensives, white coat, and sustained hypertensives.

Question #3b. Does ABP predict subsequent clinical outcomes?

A total of 14 articles from 10 prospective observational studies addressed the issue of whether ABP can predict subsequent BP-related events.^32,94-106 Of the 10 studies, one study published three articles that covered different aspects of this research question,^98-100 two other studies each published two relevant articles,^{32,95,104,105} and the remaining seven studies published only one article. Unless otherwise stated, this section will report and enumerate by 'study' rather than by 'article'.

All but one study documented the type of ABP device that was used.⁹⁷ A SpaceLabs device was used in six studies,^{32,94,95,102,104-106} a Diasys device in one study,⁹⁶ a Nippon Colin device in two studies,^98-100,103 and a Remler device in one study.¹⁰¹ Accordingly, the most common technique to record BP was oscillometric. In six studies, the ABP devices had been validated according to criteria of the BHS or the AAMI.^{32,94-96,102,104-106} In three other studies, the devices had undergone validation studies prior to widespread use of the BHS or AAMI criteria.^98-101,103 In most studies, a fixed time period was used to define 'daytime' and 'nighttime' BP, while in one study,^98-100 'awake' and 'asleep' were defined by actual participant reports. The interval between readings ranged from 15 to 30 minutes (4 readings to 2 readings per hour) for daytime BP and from 15 to 60 minutes (4 readings to 1 reading per hour) for nighttime BP.

Limited information is available on the type and number of clinic BP measurements. Four of the ten studies did not provide any information on clinic measurements.^94,96,97,106 Of the remaining six studies, four used a mercury device,^{94,101,102,104,105} one used an automated device,^98-100 and one additional study did not mention the type of device.⁹⁵ In four studies, the type of observer was mentioned; a technician or nurse measured clinic BP in three studies, while a physician measured BP in one study.^104,105 Clinic BP was recorded on just one day in three studies^{98-100,103-105} and on three days in another three studies.^{32,95,101,102} In these six studies, the total number of BPs contributing to average clinic BP ranged from two to nine. In one study, 'clinic' BP measurements were taken at home by medical personnel.^98-100

Evidence Tables 46 and 47 present risk estimates as the relative risk, or hazard ratio, of the outcome by change in BP (a continuous variable, mmHg) or by category of BP. Cutpoints for the categories of BP were conventional cutpoints (e.g., systolic BP of 140 mmHg), convenience values, or values of the BP distribution (e.g., quintiles). For this report, the reference category was the lowest level of BP. Because these studies commonly displayed risk relationships in other formats, relative risk estimates were, in several instances, calculated from data presented in the articles,^{95,99,101,104,106} including an article in which the reference category was not the lowest BP category.⁹⁹

As displayed in Evidence Tables 46 and 47, a total of eight prospective studies (nine articles) reported the relationship between absolute levels of systolic ABP and subsequent outcomes,^{32,94,96,99-103,105} while four studies (five articles) reported corresponding relationships for diastolic ABP.^{94,99-101,103} For systolic BP, at least one study outcome was significantly related to clinic BP in two of five articles,^101,105 to daytime ABP in four of seven articles,^32,100-102 to nighttime ABP in four of five studies,^{32,94,100,103} and to 24 hour ABP in five of six articles.^{32,96,100,103,105} For diastolic BP, at least one study outcome was significantly related to daytime ABP in two of five articles,^100,101 nighttime ABP in two of four articles,^100,103 and 24 hour ABP in one of three articles.¹⁰³ Clinic diastolic BP was significantly associated with outcomes in the anticipated direction in one of five studies¹⁰¹ and in an inverse direction in another study;⁹⁴ the latter finding may have resulted from the study population, namely, dialysis patients in whom a lower diastolic BP may be related to excess risk. Overall, absolute level of ABP (mean daytime, nighttime or 24 hour BP, systolic or diastolic) predicted outcomes in each of eight studies that examined this issue, while clinic BP predicted outcomes in two of five studies.

Table 3: Number of significant associations / number (more...)

Table 3: Number of significant associations / number of number of studies, by outcome and type of BP measurement (clinic BP; day, night and 24 hr ABP, systolic and diastolic; WCH and Dipping Status)

	Systolic BP				Diastolic BP				ABP Patterns
	Clinic	Day	Night	24 Hr	Clinic	Day	Night	24 Hr	WCH	Dipping
Total mortality	0/2	0/2	1/2	1/2	0/1	0/1	0/1	0/1	0	2/2
CVD mortality	0/3	1/3	2/3	1/3	1 a /2	1/2	0/2	0/2	0	1/2
CVD morbidity and mortality	2/4	3/4	2/2	4/4	1/2	1/2	1/1	1/1	1/1	3/3
Stroke	0/2	2/2	2/2	2/2	0/1	1/1	1/1	0	1/1	3/3
Cardiac morbidity and mortality	0/1	1/1	1/1	1/1	0	0	0	0	1/1	0
Dialysis	0	0	0	0	0	0	0	0	0	1/1

a

Significant inverse association

Nine of 14 articles compared prediction of outcomes by ABP to prediction by clinic BP. Of these nine studies, just two studies^32,101 assessed 'incremental gain', that is, whether ABP provided additional information that was predictive of risk beyond that of clinic BP. To assess incremental gain, one study used a residual method to determine whether ABP predicted the residual variance left after regression of outcomes on clinic BP,¹⁰¹ and one presented regression analyses with both clinic BP and ABP in the same model.³² The other seven studies compared prediction by clinic BP and ABP without determining whether ABP provided additional information beyond clinic; of these, six studies used stepwise regression techniques^{97,99,100,102,103,105} and one used discriminant function analyses.⁹⁶ ABP was a better predictor of outcomes than clinic BP in each of the seven studies that compared prediction of outcomes by clinic BP and ABP. In the two other studies, ABP provided incremental gain in information beyond that of clinic BP.

In summary, ABP predicted BP-related clinical outcomes. In each of ten prospective studies (14 articles), at least one dimension of ABP predicted one or more clinical outcomes. Absolute ABP levels (mean daytime, nighttime or 24 hour BP, systolic or diastolic) predicted outcomes in each of eight studies, WCH predicted a reduced risk of outcomes compared to sustained hypertension in each of two studies, and non-dipping or inverse dipping predicted an increased risk in four of five studies.

However, available data were insufficient to compare prediction of outcomes by ABP and clinic BP. Absolute clinic BP levels predicted outcomes in two studies in the anticipated direction, in one study in an unanticipated opposite direction, and did not predict outcomes in two other studies; five studies did not report whether clinic BP predicted outcomes. Although ABP was a better predictor of outcomes than clinic BP in most studies and even provided 'incremental gain' in outcome prediction in two studies, measurement of clinic BP and the types of comparative analyses were suboptimal. Hence, it is unclear whether the apparent superiority of ABP over clinic BP resulted from a better estimate of usual BP from ABP or a suboptimal measurement of clinic BP.

Question #3c. What is the incremental gain in prediction of clinical outcomes from use of ambulatory devices beyond prediction from clinic BP alone?

Please see discussion regarding Question #3b.

Question #3d. What is the effect of treatment guided by ABP in comparison to treatment guided by clinic BP.

In both trials, ABP was used to titrate medications in a fashion that would lead to more aggressive use of medications in persons with elevated ABP and less aggressive medication use in persons with apparently low ABP. In the trial by Staessen, there was less use of medications in the ABP group compared to control group, while in the trial by Schrader medication use was similar, perhaps as a result of enrollment procedures. Specifically, in this trial, persons with WCH were excluded post-randomization in the ABP group but not the control group. Had these individuals with WCH been included in both groups, not just the control group, overall medication use might have been less in the ABP group.

During follow-up, BP related end-organ disease, as assessed by LV mass, was similar in the ABP and control groups in the trial by Staessen. In the trial by Schrader, clinical cardiovascular events and deaths were less common in the ABP group than the control group, despite similar mean levels of clinic BP in both groups. This pattern of findings occurred despite the fact that the ABP group in this trial was enriched with a relatively high risk group, sustained hypertensives, while the control group included 'white coat hypertensives'. The reduction in clinical cardiovascular events in the ABP group may have resulted a differential approach to persons with high ABP, specifically, those in the ABP group received upward titration of medications whereas those with high ABP remained undetected in the control group.

In summary, the availability of just two trials limits inferences about the utility of ABP to guide BP management. The dearth of studies might be related to several factors, including historical lack of reimbursement for ABP, difficulties in obtaining repeat ABP, and the perception that SMBP is a more suitable alternative to ABP for management. Still, it is noteworthy that there was no apparent excess in BP-related end organ damage in both trials and potentially even a reduction in clinical events, despite the fact that BP medications were sometimes titrated downward.

Question #4

Does the evidence for the above questions vary according to a patient's age, gender, income level, race/ethnicity, and clinical subgroups?

As discussed previously, the vast majority of studies included both men and women. However, few studies reported results separately by gender. Also, studies rarely documented enrollment African-Americans; accordingly, race-stratified data was extremely unusual. The remainder of this section documents reports of individual studies that provided subgroup findings. Except for the prevalence of WCH, it is impossible to draw distinct conclusions for separate subgroups.

Research Question 1

One study reported differences between SMBP and clinic BP by gender.²⁶ For both systolic and diastolic BP, clinic BP was greater than SMBP in women and men. Another two studies reported BP differences between ABP and clinic BP, separately by gender.^28,33 For both men and women, clinic BP exceeded daytime and 24 hour BP, but the differences appeared somewhat greater in women than men. The same pattern was evident for both systolic and diastolic BP.

The only apparent subgroup difference was the prevalence of WCH by gender. Specifically, in each study that presented WCH prevalence estimates by gender, the prevalence of WCH was higher in women compared to men.^{39,40,43,49,51,53}

Research Question 2

No observational study presented SMBP risk relationships separately by gender. In contrast, three trials that evaluated the effects of SMBP reported or commented on gender differences. In one trial, reductions in BP from the SMBP intervention were similar by gender,⁷⁰ while in two studies results were better in women compared to men.^71,73 One trial reported that the SMBP intervention significantly improved mean arterial pressure in blacks⁷⁰

Research Question 3

In one cross-sectional study,⁴³ correlations of left ventricular mass with BP appeared higher in women than in men. In the same study, left ventricular mass in sustained hypertensives was greater than that of individuals with WCH, for both men and women. In one prospective study,¹⁰⁴ non-dipping status was significantly associated with a greater risk of CVD morbidity and mortality in women but not in men.

Summary of Findings

• Key question 1. Comparison of clinic BP, SMBP, and ABP readings.
  

  • Question 1a. Distribution of BP differences.
    A total of 18 studies addressed the distribution of BP differences. BP levels measured outside the clinic setting differed from those obtained in the clinic. For both systolic and diastolic BP, clinic measurements exceeded SMBP, daytime ABP, nighttime ABP and 24 hour ABP. In the few studies that compared SMBP and ABP, daytime ABP and SMBP appeared similar, while nighttime ABP was consistently lower than SMBP. The literature was insufficient to determine whether these BP differences are reproducible.

  • Question 1b. Prevalence of WCH based on SMBP.
    A total of four studies addressed this issue. Hence, the literature was insufficient to determine the prevalence of WCH by SMBP.

  • Question 1c. Prevalence of WCH based on ABP.
    A total of 16 studies addressed this issue. Prevalence varied by WCH definition and study population. Overall, the prevalence was approximately 20 percent among patients with hypertension. Only two studies addressed the reproducibility of WCH. Hence, the literature was insufficient to determine whether WCH based on ABP is reproducible.

• Key question 2. The relationship of SMBP levels and WCH based on SMBP with target organ damage and clinical outcomes.
  

  • Question 2a. Cross-sectional associations of SMBP with target organ damage.
    Only one study addressed this issue. Hence, the literature was insufficient to determine the associations of absolute SMBP levels or WCH as determined by SMBP with left ventricular mass or proteinuria.

  • Question 2b. Associations of SMBP with clinical outcomes in prospective studies.
    Only one study addressed this issue. Hence, the literature was insufficient to determine whether absolute SMBP levels or WCH based on SMBP predicts subsequent CVD.

  • Question 2c. Comparison of risk prediction from SMBP and clinic BP.
    Only one study addressed this issue. The dearth of studies combined with the poor or uncertain quality of clinic BP measurements precluded an answer to this question.

  • Question 2d. Effect of treatment guided by SMBP.
    Twelve trials addressed this issue, but the evidence was inconsistent. In half of these trials, interventions that included SMBP led to reduced BP. Two trials used contemporary SMBP technology which can store and synthesize SMBP measurements and which can generate BP reports. In both of these trials, the SMBP intervention led to reduced BP.

• Key question 3. The relationship of ABP levels and WCH based on ABP with target organ damage and clinical outcomes.
  

  • Question 3a. Cross-sectional associations of ABP with target organ damage.
    A total of 25 studies addressed these issues. Left ventricular mass and albuminuria were positively associated with ABP.

  • Question 3b. Associations of ABP with clinical events in prospective studies.
    A total of 10 studies addressed this issue. In each study, at least one dimension of ABP predicted subsequent clinical events, primarily CVD. In two of these studies, WCH was associated with a reduced risk of CVD relative to the risk associated with sustained hypertension. No prospective study adequately compared the risk associated with WCH relative to the risk associated with non-hypertension. In four of five studies, a non-dipping or inverse dipping pattern predicted an increased risk of adverse events.

  • Question 3c. Comparison of risk prediction from ABP and clinic BP.
    A total of nine prospective studies addressed this issue, but only two studies assessed 'incremental' gain, that is, whether ABP provided additional information that was predictive of risk beyond that of clinic BP. However, the poor or uncertain quality of clinic BP measurements precluded a satisfactory comparison of risk prediction from ABP and clinic BP.

  • Question 3d. Effect of treatment guided by ABP.
    Only two trials addressed this issue. Hence, the literature was insufficient to determine the effects of treatment guided by ABP.

• Key question 4. Findings to research questions 1-3 in subgroups.
  The vast majority of studies included both men and women, but few studies reported results separately by gender. Few studies reported enrollment African-Americans, and race-stratified data were rarely presented. The only notable subgroup finding was a higher prevalence of WCH in women than men.

In summary, ABP levels and ABP patterns were associated with BP-related target organ damage in cross-sectional studies. Likewise, in prospective studies, higher ABP, sustained BP and a non-dipping ABP pattern were associated with an increased risk of subsequent CVD events. Few studies examined corresponding relationships for SMBP. The poor or uncertain quality of clinic BP measurements precluded satisfactory comparisons of risk prediction based on ABP or SMBP with risk prediction based on clinic BP. In aggregate, these findings provide some support for use of ABP monitoring in evaluating prognosis. However, evidence was insufficient to determine whether the risks associated with WCH are sufficiently low to consider withholding drug therapy in this large subgroup of hypertensive patients. For SMBP, available evidence from several trials suggested that use of SMBP can improve BP control; however, further trials are needed.

Limitations of Report

The potential scope of the project was beyond available resources. Hence, the EPC team made considerable efforts to focus on the most critical research questions, the most relevant populations, and the most important data collection items. In the process, certain research issues were not covered in this report, for example, the prevalence of non-dipping and its cross-sectional associations. By necessity, the EPC team focused on study populations that are now considered candidates for ABP and SMBP monitoring, that is, non-pregnant adults with hypertension.

The literature review was limited to articles published in English, thus increasing the potential for publication bias. The exclusion of articles not published in the English language reflects the practical realities of obtaining and reviewing non-English articles within the time frame and budget of this project.

The evaluation of diagnostic technologies is complex and often does not lend itself well to the traditional table-based format of an evidence report that synthesizes data from large numbers of basically similar studies, often clinical trials. Furthermore, technologies under evaluation rapidly change such that research is often dated by the time it is completed. In the case of SMBP, only two studies tested contemporary technologies that are capable of storing and transmitting data and generating reports. Finally, it is often unclear whether findings from studies of specific devices can be extrapolated to an entire class of devices.

Another set of issues pertain to the reference technology or 'gold standard' against which new technologies are compared. For this report, a critical issue was whether the standard should be clinic BP as recommended in guidelines or clinic BP as commonly (and sub-optimally) obtained in routine medical practice. In the end, most publications provided little information about clinic BP measurements; hence, it is doubtful that ABP and SMBP were compared to high quality clinic measurements. However, the uncertain or poor quality of clinic BP in these studies may actually parallel its routine use in medical practice.

Limitations of Literature

The ABP and SMBP literature is vast, heterogeneous and poorly indexed. These aspects of the literature created enormous logistic challenges at each point in the process, including the review of 4,852 abstracts, review of 596 articles, the design of appropriate data collection instruments, the abstraction of data, and the construction of evidence tables. In several instances, summary statistics had to be recalculated in order to present data in a common format. Because of heterogeneity in study design and data presentation, results from prospective observational studies and clinical trials were entered directly into separate databases or spreadsheets and into open fields rather than as fixed pre-coded fields.

The quality of publications and presentation of data were often suboptimal. In many instances, core methods and basic descriptive information were presented in an unusual fashion that complicated data abstraction. Likewise, statistical analyses were often suboptimal. In the end, several studies that addressed our research questions could not be included because data were not presented in an abstractable format.

Most studies were single center studies, often with small sample size and without government support. Despite the vital importance of accurate BP measurement, governments have sponsored relatively little research that compares the utility of different techniques.

In most papers, the methods sections provided an incomplete description of clinic measurements. Often the type and training of the manual observer, the type of device, the number of measurement days, the number of BP readings per day, and the use (or non-use) of standard measurement techniques was not reported. When standard BP technique was reported, the measurement was often the average of a few readings, sometimes just one or two from a single visit. Training of manual observers was rarely mentioned. Despite this limitation, it should be recognized that the poor and uncertain quality of clinic measurements likely reflects actual clinical practice, in which high quality clinic BP measurements may never be routinely obtained. In contrast, ABP measurement technique in clinic practice is likely to be similar to that of the research setting.

Other limitations of the literature were evident, including the following:

• Of the available prospective observational studies, most were comparatively small. ABP and SMBP have not been used in the major observational studies that documented the relation between BP and CVD risk.

• Few studies assessed the relation between SMBP and either prevalent BP-related target organ damage (cross-sectional studies) or clinical outcomes (longitudinal studies).

• Few trials assessed the utility of ABP to guide BP therapy.

• Few studies assessed the reproducibility of the diagnosis of WCH or the reproducibility of differences between clinic BP and either ABP or SMBP.

• In the trials that evaluated the utility of SMBP measurements, it is unclear how SMBP data were used to guide BP therapy.

• Few studies have compared SMBP and ABP as predictors of outcomes or as tools to guide BP management.

• Definitions of ABP variables, such as WCH, were exceedingly variable.

• Few studies tested for incremental gain from use of ABP, that is, the gain from concomitant use of ABP with clinic BP beyond that of clinic BP alone. The appropriate analytic model would be simultaneous inclusion of both ABP and clinic BP in regression models rather than stepwise analyses. This proposed analytic strategy would actually parallel the intended use of ABP in clinic practice because ABP would likely be used with clinic BP, not by itself. Specifically, the decision to use ABP and the interpretation of subsequent data is contingent upon clinic BP readings.

• Adjustment procedures were often inadequate leading to the potential for residual confounding

Use of Evidence Report

This report synthesizes evidence that should facilitate clinical decision making and inform policy makers about the utility of BP measurements outside of the clinic setting. The importance of this report is heightened by concurrent concerns and uncertainties over standard clinic measurements. The EPC team intends to disseminate this report through several venues. The full report will be available through AHRQ's Publications Clearinghouse and its Web Site. Condensed versions of key components will be submitted for publication in peer-reviewed publications that are widely read by physicians and other health care providers who manage patients with hypertension. The NHBPEP will also assist in dissemination of this report through its ongoing activities and meetings. Key findings will also be presented at national meetings of major professional organizations, including the American Society of Hypertension and the American Heart Association. The EPC team anticipates that this report will be used by policy makers who are presently evaluating alternative strategies to measure BP and considering an appropriate research agenda. This report might also stimulate development and dissemination of guidelines for better reporting of ABP and SMBP studies.

ECP BP: Peer Reviewers

In addition to members of the technical advisory group, the partner and individuals within the AHRQ, feedback was received from individuals from the following organizations.

American Academy of Family Physicians
American Academy of Neurology
Association for the Advancement of Medical Instrumentation
American Association of Health Plans
American College of Cardiology
American College of Physicians-American Society of Internal Medicine
American Society of Hypertension
National High Blood Pressure Education Program Coordinating Committee

Journals Hand Searched

All journals searched January 2001 to May 2001, unless otherwise noted.

Journal Title

American Journal of Hypertension

Annals of Internal Medicine

Archives of Internal Medicine

Blood Pressure Monitoring

Blood Pressure

Blood Pressure Supplementum

British Medical Journal

Circulation

Hypertension

Journal of American Medical Association

Journal of Clinical Hypertension^*

Journal of Hypertension

Journal of Hypertension Supplementum

Journal of Human Hypertension

Lancet

New England Journal of Medicine

Search Strategies