Objective: We conducted this systematic review to support the U.S. Preventive Services Task Force (USPSTF) in updating its recommendation on screening for high blood pressure (BP) in nonpregnant adults.

Data Sources: We searched relevant databases and literature sources from 2003 to June 17, 2013 to identify existing systematic reviews. For Key Questions (KQs) 1 and 5, we searched MEDLINE, PubMed, Cochrane Central Register for Controlled Clinical Trials, and the Cumulative Index to Nursing and Allied Health Literature from 2003 to February 24, 2014 to locate relevant studies. For KQs 2 and 3, we searched MEDLINE, PubMed, and the Cochrane Central Register for Controlled Clinical Trials from January 1, 1992 to February 24, 2014 for relevant studies. For KQ 4, we searched MEDLINE and PubMed from January 1, 1966 to February 24, 2014 to identify longitudinal cohort studies of rescreening.

Study Selection: We conducted a dual independent review of 19,309 abstracts and 1,171 full-text articles against a priori inclusion and exclusion criteria. Two investigators also independently critically appraised each included article using criteria defined by the USPSTF and supplemented with criteria from the Quality Assessment of Studies of Diagnostic Accuracy II, the Quality in Prognosis Studies tool, and the Newcastle-Ottawa Scale for diagnostic accuracy (KQs 2 and 3), prognostic (KQ 3), and observational (KQs 4 and 5) studies, respectively. We resolved discrepancies through discussion and consultation with a third reviewer, when necessary. We included only fair- or good-quality studies.

Data Analysis: For KQs 1 and 5, we qualitatively summarized results because of the small number of included studies. For KQ 2, we calculated the diagnostic accuracy of office-based BP measurement (OBPM) devices and protocols using the result from the most commonly recommended device (i.e., manual mercury sphygmomanometer) or protocol component (e.g., no caffeine) as the reference standard. We qualitatively summarized the results. For the prognosis component of KQ 3, we grouped outcomes into the categories of cardiovascular (CV), stroke, and cardiac events. We combined fatal and nonfatal events within these outcome categories. Risk was consistently expressed as a hazard ratio per increment in BP measurement across all included studies. Risk results for CV outcomes by BP measurement method at baseline were visualized in forest plots of hazard ratios. For diagnostic accuracy calculations, we used the BP measurement method that best predicted CV outcomes (i.e., ambulatory BP monitoring [ABPM]) as the reference standard. We qualitatively evaluated how patient or study characteristics influenced diagnostic accuracy. For KQ 4, we pooled incidence rates for the overall populations in included studies to generate a weighted mean incidence at various rescreening intervals, which were categorized into 1, 2, 3, 4, and 5 years. We qualitatively examined direct evidence from subgroup results reported within studies to address the influence of patient characteristics.

Results: One randomized, controlled trial (39 clusters; n=140,642) of a Canadian BP screening program that targeted adults age 65 years or older reported 3.02 fewer annual hospital admissions for cardiovascular disease per 1,000 persons in the intervention group compared with the no screening group. When the trial data were analyzed by number of unique persons with hospital admissions, there was a significant relative reduction only in the individual outcome of acute myocardial infarction (rate ratio, 0.89 [95% CI, 0.79 to 0.99]; p=0.03).

Few studies reported the necessary data to allow us to evaluate the diagnostic accuracy of specific BP measurement methods or protocols. In three studies, automated oscillometric office BP results showed a range of sensitivity (51%–68%) for elevated BP, defined by manual mercury sphygmomanometry, but more consistent specificity (97%–98%) and positive predictive value (PPV) (76%–84%). Three different diagnostic accuracy studies examined the impact of recommended protocols on OBPM. In one study, a single BP measurement had high sensitivity (0.95) but only moderate PPV (0.76) compared with the average of second and third BP measurements. Two small studies in normotensive subjects found that leg crossing elevated BP measurements within the normal range and caffeine ingestion falsely elevated BP measurements above the hypertensive threshold in 17% of normotensive participants.

We first evaluated the predictive value of home BP monitoring (HBPM) and ABPM methods for long-term CV events compared with OBPM. Eleven studies reported that daytime, nighttime, and 24-hour ABPM predicted stroke and other fatal and nonfatal CV events independently of OBPM. While the results of five studies suggest similar results for HBPM, too few studies are available to draw firm conclusions. Evidence from one study comparing HBPM with ABPM was insufficient to allow us to draw conclusions. Limited evidence suggested that cardiovascular disease outcomes for the patient subgroup with isolated clinic hypertension (elevated OBPM and normal ABPM) are more similar to those of normotensive subjects at baseline than those with sustained hypertension.

The proportion of participants with an elevated BP measurement who are normotensive upon confirmatory testing by ABPM (or HBPM) ranged from 5 to 65 percent across all studies. This population has false-positive results when screened by OBPM methods, or “isolated clinic hypertension.” Increasing baseline OBPM was associated with increasing PPV for ABPM-confirmed hypertension. As a result, the likelihood of misdiagnosis of hypertension based only on screening measurement is greater as measurements approach the threshold for a diagnosis of hypertension. We did not qualitatively detect any associations between reported race/ethnicity, sex, or smoking.

Estimates of the weighted mean incidence of hypertension at yearly intervals less than 6 years were derived from a small number of studies (except at 5 years) with highly variable results at each interval. The weighted mean incidence at 5 years of 14 percent, for example, actually ranged from 2 to 28 percent. In the small number of studies that used a separate confirmation step, a significant proportion of apparent incident hypertension cases were not confirmed. Thus, overall estimates at yearly intervals based on unconfirmed incident hypertension are likely to be falsely high. Variation in incidence estimates across studies also likely reflects differences in criteria for diagnosis, as well as differences in age, sex, baseline BP, and obesity status of the populations studied. Hypertension incidence increased as much as two- to four-fold between a younger (ages 18 to 40/45 years) and older (ages 40/45 to 60/65) age group, respectively. Within-study hypertension incidence consistently tripled when comparing participants with initial optimal versus normal BP, and was approximately doubled in those with initial normal versus high-normal BP. Incidence was generally higher in men than women, especially men in younger populations. While incidence was also two-fold higher in overweight persons and three-fold higher in obese persons compared with those of normal weight, it was not increased in smokers compared with nonsmokers or former smokers. African Americans had a consistently higher incidence of hypertension at rescreening than white participants.

Four trials found no significant differences in psychological distress or quality of life after patients were labeled as hypertensive or prehypertensive. One cohort study reported significantly increased absenteeism up to 4 years after labeling compared with the year before. Three cohort studies reported significant sleep disturbances associated with ABPM use and one study reported that a significant proportion of ABPM users experienced pain, skin irritation, and overall discomfort. Discomfort and restrictions in daily activities were more frequently reported with ABPM than HBPM in one study.

Limitations: Despite recent emphasis on the instability of single BP measurements and the need for multiple valid measurements to assess a patient’s actual elevated BP exposure, high-quality comparable diagnostic accuracy studies are not common. Given recent recognition of the impact of overdiagnosis in many diseases, the widespread availability of automated BP devices with variable performance, and the prevalence of essential hypertension in the United States, further research is needed to guide primary care clinicians and consumers.

Conclusions: ABPM (24-hour, daytime, or nighttime) is a better predictor of long-term CV outcomes than OBPM (usually manual sphygmomanometry) and should be considered the reference standard for evaluating noninvasive BP measurements. A small body of evidence suggests, but does not confirm, that HBPM can serve as a similar predictor of outcomes. Initial screening by office-based methods (manual sphygmomanometry or automated oscillometric devices) variably predicts hypertension as defined by ABPM, resulting in a significant population with isolated clinic hypertension. Limited evidence suggests that patients with isolated clinic hypertension have outcomes that are more similar to normotensive than hypertensive persons. Failure to confirm initial elevated OBPM results may result in misdiagnosis and overtreatment. Limited evidence suggests that repeated measurements and improved procedural control (e.g., by automation) may improve the diagnostic accuracy of OBPM when used to screen for high BP or confirm a diagnosis of hypertension. Studies of rescreening intervals at up to 6 years found a higher incidence of hypertension overall and at shorter intervals for persons with BP in the high-normal range, older adults, persons with an above normal BMI, and African Americans. These studies showed much lower incidence at longer rescreening intervals up to 6 years in persons without these risk factors.