Cardiovascular Events and Mortality in White Coat Hypertension

Abstract

Hypertension, the foremost preventable cause of disability and premature mortality worldwide (1), is diagnosed most commonly with in-office blood pressure (BP) measurements. However, recent guidelines strongly recommend out-of-office BP monitoring (including ambulatory BP monitoring [ABPM] and self or home BP monitoring [HBPM]) for the diagnosis and management of hypertension (24). Increased use of out-of-office BP monitoring in recent decades has led to the identification of several BP phenotypes with different prognostic implications regarding long-term cardiovascular risk (57). These BP phenotypes, which require a combination of in-office and out-of-office BP readings to ascertain, include sustained normotension (that is, normal in-office and out-of-office BP in persons not receiving antihypertensive treatment), controlled hypertension (normal in-office and out-of-office BP in persons receiving antihypertensive treatment), masked hypertension (normal in-office but elevated out-of-office BP), white coat hypertension (WCH) (elevated in-office but normal out-of-office BP, described as WCH in persons not receiving antihypertensive treatment and as white coat effect [WCE] or white coat uncontrolled hypertension in those receiving antihypertensive treatment), and uncontrolled hypertension (elevated in-office and out-of-office BP). Despite guideline recommendations, real-world practice has been slow to adopt out-of-office BP monitoring (8). The clinical inertia surrounding out-of-office BP monitoring seems to be driven by several provider-, patient-, and policy-related factors (9, 10). A major barrier is skepticism over the utility of screening for isolated office hypertension (that is, untreated WCH and treated WCE) due to unclear evidence (9). The burden and risks of WCH, in particular, differ across studies. In a systematic review for the U.S. Preventive Services Task Force, Piper and colleagues (11) reported that the prevalence of WCH ranged from 5% to 65% in studies using ABPM and 16% to 55% in those using HBPM. They also found that WCH carried a higher cardiovascular risk than normotension in several studies but that these findings were not consistent across studies (11). Furthermore, the authors noted that studies of treated WCE showed no increased risk for adverse cardiovascular outcomes. Likewise, previous meta-analyses demonstrated weak association of WCH with cardiovascular risk and weak or no association with all-cause mortality (12, 13). However, these meta-analyses did not adequately explore factors contributing to the inconsistent findings across studies. Moreover, several additional studies evaluating the association between WCH and adverse cardiovascular outcomes were subsequently published. In this meta-analysis, we aimed to thoroughly assess the association of untreated WCH and treated WCE with future cardiovascular events and all-cause mortality. This information might promote more widespread adoption of out-of-office BP monitoring as standard of care and may inform policy changes to provide greater reimbursement and support for out-of-office BP monitoring in routine practice. Methods Data Sources and Searches All steps of the review and meta-analysis were performed by using a predefined protocol (Supplement) completed on 5 July 2018 in accordance with MOOSE (Meta-analysis of Observational Studies in Epidemiology) guidelines (14). Publications were identified by searching PubMed and EMBASE from inception to 10 December 2018, without language restriction. Search algorithms incorporated hypertension, blood pressure, and several terms related to WCH, in-office BP, out-of-office BP monitoring, and cardiovascular outcomes (Supplement). Additional publications were identified by manual review of reference lists of relevant studies, reviews, and meta-analyses. Supplement. Supplemental Methods Study Selection Publications were eligible for inclusion if they were studies in adult humans that reported associations of WCH or WCE with nonfatal cardiovascular events (including incident coronary artery disease, myocardial infarction, angina, stroke, transient ischemic attack, peripheral artery disease, revascularization procedure, and hospitalization for congestive heart failure), fatal cardiovascular events, or all-cause mortality; had a mean follow-up of at least 3 years; and provided a reference group of persons with normotension or controlled hypertension. Two investigators independently screened abstracts and reviewed full texts to determine eligibility. Any discrepancies were resolved by a third reviewer. Data Extraction and Quality Assessment Two investigators independently extracted data from each eligible publication by using a standardized form (Supplement). Data extracted included cohort name; year of publication; country and location of the study; study design; inclusion and exclusion criteria; type and duration of out-of-office BP measurement; criteria for diagnosis of WCH or WCE; number of study participants overall and with WCH or WCE; number of participants receiving antihypertensive treatment at baseline; number of participants with a history of diabetes, cardiovascular disease, or chronic kidney disease; number of participants who were current smokers and were male; mean age, body mass index, and duration of follow-up; covariates included in statistical adjustment; outcomes reported and outcome definitions; adjusted risk estimates, separated by antihypertensive treatment status (treated, untreated, or treated and untreated combined); and type of outcome (fatal and nonfatal cardiovascular event, fatal cardiovascular event, or all-cause mortality). Any discrepancies were resolved by a third reviewer. Study authors were contacted directly by the lead author if a publication met all inclusion criteria but did not report the outcomes in a way that could be extracted for meta-analysis (for example, if 95% CIs were not reported). Quality of the evidence was evaluated by 2 investigators using a modified QUADAS-2 (Quality Assessment of Diagnostic Accuracy Studies 2) tool to assess individual study bias for each outcome (Supplement). The QUADAS-2 tool assesses whether a study has low, high, or unclear risk of bias across 4 domains: patient selection, index test (modified to reflect the quality of the ABPM or HBPM assessment), reference standard (modified to reflect the quality of the in-office BP), and flow and timing (15). The modified tool incorporated quality of the statistical analyses, handling of confounding, and outcome assessment. Confounding was considered to be adequately addressed if adjustment was made for age, sex, previous cardiovascular events, antihypertensive medication, and at least 2 additional covariates among smoking status, lipid levels, diabetes mellitus, body mass index, kidney function, left ventricular hypertrophy, clinic BP, and alcohol use. Studies were determined to have a high risk of bias in the handling of confounding if the same covariates were used for analyzing cardiovascular events and all-cause mortality without otherwise accounting for potential differential confounding (for example, participant exclusion for important risk factors for all-cause mortality, such as cancer or high infectious risk). The primary analyses were restricted to studies determined to have a low risk of bias across at least 5 of 7 domains of the modified QUADAS-2. Data Synthesis and Analysis Meta-analyses were performed by calculating pooled log hazard ratios using random-effects inverse-variance models, with profile likelihood estimation (1618) and Bartlett correction (in analyses of more than 5 studies) (19) to address heterogeneity across the relatively small number of studies. All analyses incorporated multivariable-adjusted hazard ratios to quantify the association between WCH or WCE and each of the outcomes, with normotension or controlled hypertension as the reference group. The primary analyses were stratified by baseline antihypertensive treatment status reported in each study (WCH [untreated], WCE [treated], or combined). The primary outcomes evaluated were fatal and nonfatal cardiovascular events and all-cause mortality. Heterogeneity was assessed by Cochran Q test and quantified with the I 2 index (20) in analyses of 3 or more studies. Begg rank correlation (21) and Egger weighted linear regression (22) tests were planned to assess for small study effects (that is, publication bias). However, these tests do not perform well with fewer than 10 studies contributing to a given estimate, and consequently they were omitted. In instances with more than 1 publication from the same cohort, data from the most recent and applicable report were used for the primary analyses; other publications from that cohort were included in pertinent subgroup analyses if the data were not available from the most recent publication. Analyses were performed with the admetan and metabias packages in Stata, version 15.1 (StataCorp). Role of the Funding Source The funding source had no role in the study design or implementation. Results The search strategy identified 27 publications that were eligible for inclusion from 29 unique cohorts, involving 25786 persons with WCH or WCE and 38487 with normotension or controlled hypertension (Figure 1 and Supplement Table 1). Two studies were based in North America, 13 in Europe, 7 in Asia, and 5 across several regions. Fourteen studies reported funding from government, university, medical society, or research foundation grants; 3 reported only industry sponsorship; 4 reported a combination of industry and government or foundation funding; and 6 (encompassing 4 distinct cohorts) did not report any source of funding. Six studies were population based, 11 recruited participants from outpatient clinics, and 10 included patients who were referred for ABPM or to a specialized hypertension clinic. Eighteen studies assessed out-of-office BP with ABPM, 7 with HBPM, and 2 wit