Author + information
- Received February 23, 2017
- Revision received May 26, 2017
- Accepted May 30, 2017
- Published online July 24, 2017.
- Dean S. Picone, BMedRes(Hons)a,
- Martin G. Schultz, PhDa,
- Petr Otahal, GDipScia,
- Svend Aakhus, MD, PhDb,
- Ahmed M. Al-Jumaily, PhDc,
- J. Andrew Black, MBBS(Hons)a,d,
- Willem J. Bos, MD, PhDe,
- John B. Chambers, MDf,
- Chen-Huan Chen, MDg,
- Hao-Min Cheng, MD, PhDg,
- Antoine Cremer, MDh,
- Justin E. Davies, PhDi,
- Nathan Dwyer, MBBS, PhDa,d,
- Brian A. Gould, MD, PhDj,
- Alun D. Hughes, MBBS, PhDk,
- Peter S. Lacy, PhDl,
- Esben Laugesen, MD, PhDm,
- Fuyou Liang, PhDn,
- Roman Melamed, MDo,
- Sandy Muecke, PhDp,
- Nobuyuki Ohte, MD, PhDq,
- Sho Okada, MD, PhDr,
- Stefano Omboni, MDs,
- Christian Ott, MDt,
- Xiaoqing Peng, MPharma,
- Telmo Pereira, PhDu,
- Giacomo Pucci, MDv,
- Ronak Rajani, MDf,
- Philip Roberts-Thomson, MBBSa,d,
- Niklas B. Rossen, MD, PhDm,
- Daisuke Sueta, MD, PhDw,
- Manish D. Sinha, PhDx,
- Roland E. Schmieder, MDt,
- Harold Smulyan, MDy,
- Velandai K. Srikanth, PhDa,z,aa,
- Ralph Stewart, MDbb,
- George A. Stouffer, MDcc,
- Kenji Takazawa, MD, PhDdd,
- Jiguang Wang, MD, PhDee,
- Berend E. Westerhof, PhDff,
- Franz Weber, MDgg,
- Thomas Weber, MDhh,
- Bryan Williams, MDl,
- Hirotsugu Yamada, MD, PhDii,
- Eiichiro Yamamoto, MD, PhDw and
- James E. Sharman, PhDa,∗ ()
- aMenzies Institute for Medical Research, University of Tasmania, Hobart, Tasmania, Australia
- bDepartment of Circulation and Medical Imaging, Faculty of Medicine, Norwegian University of Science and Technology, Trondheim, Norway
- cInstitute of Biomedical Technologies, Auckland University of Technology, Auckland, New Zealand
- dRoyal Hobart Hospital, Hobart, Tasmania, Australia
- eDepartment of Internal Medicine, St. Antonius Hospital, Nieuwegein, the Netherlands
- fCardiology Department, Guy’s and St. Thomas’ Hospitals, London, United Kingdom
- gDepartment of Medicine, National Yang-Ming University, Department of Medical Education, Taipei Veterans General Hospital, Taipei, Taiwan
- hDepartment of Cardiology/Hypertension, University Hospital of Bordeaux, Bordeaux, France
- iInternational Centre for Circulatory Health, Imperial College London, London, United Kingdom
- jBMI Hospital Blackheath, London, United Kingdom
- kInstitute of Cardiovascular Sciences, University College London, London, United Kingdom
- lInstitute of Cardiovascular Sciences University College London (UCL) and National Institute for Health Research (NIHR) UCL/UCL Hospitals Biomedical Research Centre, London, United Kingdom
- mDepartment of Endocrinology and Internal Medicine, Aarhus University Hospital, Aarhus, Denmark
- nSchool of Naval Architecture, Ocean and Civil Engineering, Shanghai Jiao Tong University, Shanghai, China
- oAbbott Northwestern Hospital, Allina Health, Minneapolis, Minnesota
- pDepartment of Critical Care Medicine, Flinders University, Adelaide, South Australia, Australia
- qDepartment of Cardio-Renal Medicine and Hypertension, Nagoya City University Graduate School of Medical Sciences, Nagoya, Japan
- rDepartment of Cardiovascular Medicine, Chiba University Graduate School of Medicine, Chiba, Japan
- sClinical Research Unit, Italian Institute of Telemedicine, Varese, Italy
- tDepartment of Nephrology and Hypertension, University Hospital Erlangen, Friedrich-Alexander University Erlangen-Nürnberg, Erlangen, Germany
- uPolytechnic Institute of Coimbra, Coimbra College of Health Technology, Department of Cardiopneumology, Lousã, Portugal
- vUnit of Internal Medicine at Terni University Hospital, Department of Medicine, University of Perugia, Perugia, Italy
- wDepartment of Cardiovascular Medicine, Graduate School of Medical Sciences, Kumamoto University, Kumamoto, Japan
- xDepartment of Clinical Pharmacology and Department of Paediatric Nephrology, Kings College London, Evelina London Children’s Hospital, Guy’s and St. Thomas’ NHS Foundation Trust, London, United Kingdom
- yDepartment of Medicine, State University of New York, Upstate Medical University, Cardiology Division, Syracuse, New York
- zCentral Clinical School, Faculty of Medicine, Monash University, Melbourne, Victoria, Australia
- aaDepartment of Medicine, Peninsula Health, Melbourne, Victoria, Australia
- bbGreen Lane Cardiovascular Service, Auckland City Hospital, University of Auckland, Auckland, New Zealand
- ccDivision of Cardiology, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina
- ddCenter for Health Surveillance and Preventive Medicine, Tokyo Medical University Hospital, Tokyo, Japan
- eeCentre for Epidemiological Studies and Clinical Trials, Shanghai Key Laboratory of Hypertension, The Shanghai Institute of Hypertension, Department of Hypertension, Ruijin Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, China
- ffDepartment of Pulmonary Diseases, VU University Medical Center, Amsterdam, the Netherlands
- ggDepartment of Nephrology, Center for Internal Medicine, University Clinic Essen, University Duisburg-Essen, Essen, Germany
- hhCardiology Department, Klinikum Wels-Grieskirchen, Wels, Austria
- iiDepartment of Cardiovascular Medicine, Tokushima University Hospital, Tokushima, Japan
- ↵∗Address for correspondence:
Prof. James E. Sharman, Menzies Institute for Medical Research, University of Tasmania, Private Bag 23, Hobart, Tasmania 7000, Australia.
Background Hypertension (HTN) is the single greatest cardiovascular risk factor worldwide. HTN management is usually guided by brachial cuff blood pressure (BP), but questions have been raised regarding accuracy.
Objectives This comprehensive analysis determined the accuracy of cuff BP and the consequent effect on BP classification compared with intra-arterial BP reference standards.
Methods Three individual participant data meta-analyses were conducted among studies (from the 1950s to 2016) that measured intra-arterial aortic BP, intra-arterial brachial BP, and cuff BP.
Results A total of 74 studies with 3,073 participants were included. Intra-arterial brachial systolic blood pressure (SBP) was higher than aortic values (8.0 mm Hg; 95% confidence interval [CI]: 5.9 to 10.1 mm Hg; p < 0.0001) and intra-arterial brachial diastolic BP was lower than aortic values (−1.0 mm Hg; 95% CI: −2.0 to −0.1 mm Hg; p = 0.038). Cuff BP underestimated intra-arterial brachial SBP (−5.7 mm Hg; 95% CI: −8.0 to −3.5 mm Hg; p < 0.0001) but overestimated intra-arterial diastolic BP (5.5 mm Hg; 95% CI: 3.5 to 7.5 mm Hg; p < 0.0001). Cuff and intra-arterial aortic SBP showed a small mean difference (0.3 mm Hg; 95% CI: −1.5 to 2.1 mm Hg; p = 0.77) but poor agreement (mean absolute difference 8.0 mm Hg; 95% CI: 7.1 to 8.9 mm Hg). Concordance between BP classification using the Seventh Report of the Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure cuff BP (normal, pre-HTN, and HTN stages 1 and 2) compared with intra-arterial brachial BP was 60%, 50%, 53%, and 80%, and using intra-arterial aortic BP was 79%, 57%, 52%, and 76%, respectively. Using revised intra-arterial thresholds based on cuff BP percentile rank, concordance between BP classification using cuff BP compared with intra-arterial brachial BP was 71%, 66%, 52%, and 76%, and using intra-arterial aortic BP was 74%, 61%, 56%, and 65%, respectively.
Conclusions Cuff BP has variable accuracy for measuring either brachial or aortic intra-arterial BP, and this adversely influences correct BP classification. These findings indicate that stronger accuracy standards for BP devices may improve cardiovascular risk management.
Cardiovascular disease is the number 1 cause of mortality worldwide, with elevated blood pressure (BP) as the single largest risk factor (1–3). Noninvasive brachial cuff BP is the principal method for hypertension (HTN) diagnosis and management; thus, accurate BP measurement is among the most important medical tests performed (4). Relatively small errors in cuff BP measurement can have major public health ramifications. An inaccuracy of 5 mm Hg is estimated to result in the misclassification of BP of 48 million people each year in the United States alone (21 million underestimated BP, 27 million overestimated BP) (5). BP underestimation leads to missed therapeutic potential and unnecessary elevation of cardiovascular risk (6). BP overestimation creates additional cost and exposure to the possible adverse effects of unnecessary treatment (5). The recognition of pre-hypertension as a nonbenign clinical presentation (7), and the benefit to some patient populations of achieving low BP targets (8), further emphasizes the need for accurate cuff BP across the range of BP classifications.
Several lines of evidence question the accuracy of cuff BP. First, many small studies indicate a possible bias for cuff BP to underestimate intra-arterial brachial systolic blood pressure (SBP) but overestimate intra-arterial brachial diastolic blood pressure (DBP), and thereby, underestimate intra-arterial pulse pressure (PP) (9–11). Second, cuff BP devices being tested for accuracy against other noninvasive measurements according to international validation protocols may perform to a “pass” standard even when clinically significant measurement errors occur among many patients (12). Third, there is large individual variability in intra-arterial BP between the aorta and brachial artery (9,13,14), and whether oscillometric or auscultatory cuff BP accurately measures either aortic or brachial BP has never been systematically determined. This question is important to resolve, given: 1) the possibility that aortic BP is more clinically relevant than brachial BP (13,15–17); and 2) the burgeoning of commercial devices purporting to measure aortic BP (18) to (theoretically) better assess cardiovascular risk (19). However, this is a controversial theory (20,21), with some investigators asserting that there is a lack of evidence to justify departing from standard cuff BP (20,22). Others suggest that brachial cuff BP may already accurately measure aortic BP, eliminating the need for specialist devices (23–25).
These issues create uncertainty as to whether cuff BP accurately measures intra-arterial BP, either at the brachial or aortic level. Better understanding of these issues is relevant to validation protocol standards for cuff BP devices, and could lead to improved clinical management of cardiovascular risk through more accurate BP measurement and classification. We completed 3 separate but inter-related systematic reviews and individual participant data meta-analyses to determine the accuracy of cuff BP measurement methods. We first aimed to determine the true level of intra-arterial BP agreement between the aorta and brachial artery (meta-analysis 1), and then whether cuff BP accurately measured either intra-arterial brachial BP (meta-analysis 2) or intra-arterial aortic BP (meta-analysis 3). Potential clinical consequences of cuff BP measurement error were determined by the concordance between cuff and intra-arterial BP for classifying HTN according to criteria of the Seventh Report of the Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure (JNC 7) (26).
Search technique and study eligibility
The search technique, study eligibility criteria, data collection, synthesis, and statistical analysis were conducted similarly across each meta-analysis, with minor differences reflecting the specific needs of each question. The Preferred Reporting Items for Systematic Review and Meta-Analyses of individual participant data were adhered to (Online Table 1) (27). Two reviewers (D.S.P, M.G.S) identified eligible studies by title, abstract, or full-text review and performed a separate data quality assessment. All of these activities were undertaken with each reviewer blinded to the other reviewer’s results. Discrepancies were resolved via consensus. Results from meta-analysis 1 are provided in the Online Appendix.
Four online databases (PubMed, Scopus, Embase, and Web of Knowledge) were systematically searched for eligible papers from database inception until May 9, 2016, with slight modifications for each meta-analysis (Online Table 2). Additional studies were found by searching the reference lists of identified studies and personal communication with authors. Unpublished data was accepted if sufficient methodology was provided (Online Appendix 1). Study eligibility was not restricted by subject age, language, or year of publication. We included studies that measured intra-arterial BP by high-fidelity micromanometer tip or fluid-filled catheters, as well as indwelling arterial needles and cannulas. For each meta-analysis, studies were only included if the BP measurements being compared were recorded within the immediate period of each other, rather than at different times (28), due to possible hemodynamic changes between measurement periods (29). Studies that measured BP at multiple arterial sites (e.g., brachial and radial) in the same study were included if authors were able to provide separated data. Studies that recorded data under nonbasal conditions involving hemodynamic shifts (e.g., exercise or administration of vasoactive drugs that altered BP during the recording procedure) were excluded. There was some minor variability of the inclusion and exclusion criteria that were specific to the goal of each meta-analysis. These included cuff BP methods of auscultation (mercury or aneroid), and oscillometric and automatic Korotkoff sound devices for meta-analyses 2 and 3. Studies were also excluded if the goal of the work was to determine the effect of different cuff sizes on the relationship between cuff and intra-arterial BP, because of the expectation of cuff BP measurement error (30). For meta-analyses 1 and 3, studies that measured aortic BP distal to the aortic arch were excluded because potential amplification of SBP along the aorta (31) could contribute to discordance of comparisons between BP measurements.
For each eligible study, individual participant-level deidentified BP data were requested from authors. PP was calculated as SBP – DBP. Clinical information, including age, sex, anthropometry, medications, and disease status, was also requested if available. Data were standardized to be in the international system of units, except for pressure units. Individual data supplied by authors were checked for consistency with published aggregate data where available. If discrepancies were identified, clarification was sought from authors. If no response was received to data requests, or authors were not contactable, individual data were extracted from within published tables (Online Appendix 2) or from figure scatterplots using extraction software, when possible (32). Data obtained from scatterplots were only included in the meta-analyses when accuracy could be verified by comparison with published summary data or correlation coefficients (Online Table 3). A quality score was applied to each study to account for important study design attributes that may have affected data quality (Online Appendix 3, Online Tables 4 to 6). The University of Tasmania Health and Medical Human Research Ethics Committee approved this study (reference number: H0015048).
Magnitude of BP differences
The proportion of cuff BP measurements that were ≥5, ≥10, or ≥15 mm Hg different from intra-arterial BP were determined as a measure of accuracy (33).
To determine accuracy of cuff BP for BP classification, each individual’s cuff BP was classified according to JNC 7 criteria (normal BP <120/80 mm Hg; pre-HTN SBP 120 to 139 mm Hg or DBP 80 to 89 mm Hg; stage 1 HTN SBP 140 to 159 mm Hg or DBP 90 to 99 mm Hg; and stage 2 HTN SBP ≥160 mm Hg or DBP ≥100 mm Hg) (26), and then compared for concordance with the BP classification according to the measurement of BP by intra-arterial brachial and aortic BP. For example, for an individual with cuff BP classified as normal (<120/80 mm Hg), the corresponding intra-arterial BP for that individual was classified into the appropriate category (e.g., normal, pre-HTN, or stage 1 or 2 HTN), and found to be concordant if also falling into the same normal BP classification (<120/80 mm Hg). This approach enabled an assessment of the potential effect of cuff BP inaccuracy on clinical practice, but also involves a level of arbitrariness with BP cut points because there is a continuous relationship between BP and cardiovascular risk. Additional analyses were also undertaken in which the risk cut points for intra-arterial BP (both brachial and aortic) were drawn at equal percentile ranks to the traditional cuff BP cut points. Sensitivity and specificity of cuff BP for delineating HTN at a cut point of ≥140/90 mm Hg was also assessed.
BP and clinical characteristics are presented as mean and 95% confidence interval (95% CI) unless otherwise specified. BP differences were calculated as brachial artery BP minus aortic BP (meta-analysis 1) and cuff BP minus intra-arterial brachial or aortic BP (meta-analyses 2 and 3). Both 1- and 2-stage meta-analysis was used. The results generated from each method are considered equivalent in individual participant data meta-analysis (34). Two-stage meta-analyses were used to analyze mean BP differences, because this method allowed production of summary forest plots to illustrate the level of the BP difference across included studies. For this method, data were first analyzed study by study, and were then synthesized using random effects meta-analysis due to the observational nature of the data. Correlation coefficients from individual studies were used to calculate summary correlation coefficients on the relationship between BP measurements in each meta-analysis. This same method was used for sensitivity and specificity analyses for cuff BP delineating HTN based on the 140/90-mm Hg cut point. Linear mixed modeling (1-stage meta-analysis) was used to account for clustering of individuals within each study for mean absolute difference, BP classification analysis, percentile calculation for the revised intra-arterial BP thresholds, and potential predictors of BP differences. Mean absolute difference was calculated as the absolute value of the BP difference at the individual participant level. In meta-analysis 3, Laugesen et al. (35) and Rossen et al. (36) were pooled for analysis, because participants were from the same population and the measurement protocols used were identical, except for the type of cuff BP device.
Sensitivity analyses were among studies that received the maximum study quality score to assess whether results were influenced by study design factors and to separately assess published data sources compared with unpublished data sources. To determine the influence on results of meta-analyses 2 and 3, sensitivity analyses were conducted for single BP measurements compared with the average of multiple cuff BP measures, as well as the type of catheter used for intra-arterial BP measurement. A p value < 0.05 was considered statistically significant. Data were synthesized and analyzed using R version 3.1.2 (R Foundation for Statistical Computing, Vienna, Austria), primarily using the metafor and lme4 packages, and Stata version 14 (StataCorp, College Station, Texas) (2015, metandi module). Additional statistical methods are in Online Appendix 4.
Eligible studies and subject characteristics
A total of 75,071 studies were identified from the 3 meta-analysis searches. After review based on title and abstract, we performed a full-text review of 371 studies, and 152 of these were deemed eligible for inclusion in the meta-analyses. Individual participant data were not available from 7, 48, and 23 studies for meta-analyses 1, 2, and 3, respectively. This left 13, 22, and 39 studies for SBP analysis, respectively, whereas 12, 18, and 36 studies, respectively, were available for analysis relating to DBP and PP (Online Tables 7 to 10, Online References). Systematic review flow diagrams and study characteristics for all meta-analyses are detailed in Online Figures 1 to 6 and Online Tables 7 to 12. Data were extracted from published tables in 11 studies (Online Appendix 2), and from published figures in 6 studies (Online Table 3). Data was sourced from 18 countries (Australia, New Zealand, China, Japan, Singapore, United States, Canada, England, Scotland, France, Germany, Italy, Austria, Portugal, the Netherlands, Denmark, Norway, and Israel). Across the 3 meta-analyses, subjects were generally of middle-to-older age, predominately male, and overweight according to body mass index (Online Tables 13 to 15). When individual participant data were checked as per guidelines (27), no important issues, such as inconsistency with published aggregate data, arose. There were minor differences between the number of subjects in some published papers and the number of subjects used in the meta-analyses (see explanation in Online Appendix 5).
Meta-analyses on BP differences
In meta-analysis 2, brachial cuff BP methods significantly underestimated intra-arterial brachial SBP and PP, but significantly overestimated intra-arterial brachial DBP (p < 0.0001 for all) (Figures 1A to 1C). The mean absolute difference for SBP was 7.9 mm Hg (95% CI: 6.5 to 9.5 mm Hg). Intra-arterial brachial SBP was underestimated among studies that used either oscillometric or mercury sphygmomanometric techniques, although this was only of borderline significance for the latter (Online Table 16). However, both oscillometric and mercury sphygmomanometric cuff methods significantly overestimated intra-arterial brachial DBP, and therefore, also significantly underestimated intra-arterial brachial PP. Strong correlations were observed between brachial cuff and intra-arterial brachial SBP (r = 0.89; 95% CI: 0.86 to 0.93), DBP (r = 0.78; 95% CI: 0.72 to 0.85), and PP (r = 0.82; 95% CI: 0.76 to 0.88), with p < 0.0001 for all (Online Figure 10).
In meta-analysis 3, there was no significant difference between brachial cuff and intra-arterial aortic SBP (Figure 2A) (p = 0.77); however, this was due to a relative balance in the number of studies reporting either significant overestimation (7 studies) or significant underestimation (7 studies) of intra-arterial aortic SBP by cuff SBP. Indeed, the mean absolute difference was 8.0 mm Hg (95% CI: 7.1 to 8.9 mm Hg). Brachial cuff methods significantly overestimated intra-arterial aortic DBP, and thus significantly underestimated intra-arterial aortic PP (Figures 2B and 2C) (p < 0.0001 for both). Oscillometric and mercury sphygmomanometric cuff methods were not analyzed separately as per meta-analysis 2, because the mercury method was only used in 2 studies, totaling 21 individuals. There were strong relationships between brachial cuff and intra-arterial aortic SBP based on the pooled correlation coefficients (r = 0.88; 95% CI: 0.86 to 0.90), DBP (r = 0.75; 95% CI: 0.70 to 0.80), and PP (r = 0.81; 95% CI: 0.76 to 0.85), p < 0.0001 for all (Online Figure 11). In all meta-analyses, there was significant heterogeneity between studies for the SBP, DBP, and PP analyses (I2 > 86%; p < 0.0001 for all).
BP classification based on cuff BP compared with intra-arterial BP
Among individuals with BP classified as either pre-HTN or stage 1 HTN, only 50% to 60% of brachial cuff BP measures were concordant with intra-arterial BP measures. Underestimation of BP classification was the predominant issue for brachial cuff comparisons with intra-arterial brachial BP, whereas intra-arterial aortic BP classifications were similarly overestimated and underestimated. However, there was reasonable concordance between brachial cuff and intra-arterial BP (brachial or aortic) values measured among individuals with stage 2 HTN (≥160/100 mm Hg) according to intra-arterial BP. There was also reasonable concordance between cuff and intra-arterial aortic BP for BP classification in the normal range (<120/80 mm Hg) (Table 1). There were similar findings when BP classification was only based on SBP thresholds (Online Table 17). When revised percentile rank intra-arterial BP thresholds were used, there was an improvement in concordance compared with the traditional threshold analysis in some BP categories (for example, in meta-analysis 2, normal and pre-HTN categories changed from 60% to 71% and from 50% to 66%). However, concordance remained similar or was reduced among other categories (Table 2). The revised thresholds shifted the systematic underestimation of risk using cuff BP compared with intra-arterial brachial BP among the categories of pre-HTN and stage 1 HTN to a more even distribution of overestimation and underestimation of the correct BP classification category. For example, in the category of cuff BP pre-HTN, the percentage of intra-arterial brachial BP cases that were in the stage 1 HTN category was reduced from 36% to 17% (cuff underestimation). However, in the category of cuff BP pre-HTN, the percentage of intra-arterial brachial BP in the normal category increased from 9% to 13% (cuff overestimation). Similarly, in the category of cuff BP stage 1 HTN, the percentage of intra-arterial brachial BP cases that were in either stage 2 HTN or pre-HTN categories changed from 32% to 20% (cuff underestimation) and from 13% to 26% (cuff overestimation), respectively. With respect to delineating HTN at the traditional cut-off of 140/90 mm Hg, in meta-analysis 2 the sensitivity was 78.5% (95% CI: 66.8% to 87.0%), and specificity was 95.2% (95% CI: 86.5% to 98.4%). In meta-analysis 3, sensitivity was 81.7% (95% CI: 74.9% to 87.0%) and specificity was 88.5% (95% CI: 83.4% to 92.2%).
Magnitude of difference between cuff and intra-arterial BP
Brachial cuff BP readings were ≥5, ≥10 or ≥15 mm Hg different from intra-arterial brachial SBP in 465 (67%), 275 (41%), and 173 (26%) of subjects of meta-analyses 2 (Figure 3A). Similarly, when compared with intra-arterial aortic BP, brachial cuff SBP was ≥5, ≥10 or ≥15 mm Hg different in 1,236 (67%), 748 (40%), and 411 (22%) of subjects of meta-analyses 3 (Figure 3B). Results were similar for DBP differences, although there was better agreement for DBP differences ≥15 mm Hg (Online Figure 12).
Clinical and demographic correlates
Older age and higher body mass index were related in univariable analysis to less underestimation of intra-arterial brachial and aortic SBP and PP by brachial cuff SBP and PP (Online Tables 18 and 19). In multivariable analysis, age and body mass index both remained significantly related to the difference in PP, but age was not significantly related to the difference between brachial cuff and intra-arterial brachial SBP, whereas body mass index was not significantly related to the difference between brachial cuff and intra-arterial aortic SBP. There were no consistent associations observed for brachial cuff DBP versus intra-arterial DBP.
There were significantly more males in the maximum-rated studies in meta-analyses 2 and 3. There were no other significant differences between the maximum-rated and non–maximum-rated studies (p > 0.05 all) (Online Tables 20 to 22). There were no significant differences in BP values for published versus unpublished data (p > 0.05) (Online Tables 23 to 25). In meta-analyses 2 and 3, there were no significant differences when data was analyzed based on single-cuff BP measures versus the average of multiple-cuff BP measures. Furthermore, BP classification analysis was consistent irrespective of the number of cuff measures. Correlations between cuff and intra-arterial BP were also similar irrespective of the number of cuff BP measures. Differences between cuff and intra-arterial BP were not significantly influenced by the type of catheter used for intra-arterial BP measurement (data not shown).
With HTN as the single major risk factor for global disease burden (1), the accuracy of clinic BP methods is critical. Our study had several key findings. First, we confirmed the expectation that intra-arterial brachial SBP was higher than intra-arterial aortic SBP, and also that there was little difference in DBP between the central and peripheral arterial sites. However, there was extreme individual variability in the magnitude of central-to-peripheral differences for both SBP and DBP. Second, we found that cuff BP underestimated intra-arterial brachial SBP (and PP), but overestimated intra-arterial brachial DBP irrespective of BP technique (e.g., oscillometric or auscultation using mercury methods). This is confirmation of perceived dogma relating to oscillometric devices, but as far as we know is the first comprehensive analysis of all cuff BP methods to be reported. Third, when cuff SBP was compared with intra-arterial aortic SBP, there was a small mean difference but poor agreement between measures at the individual level, whereas cuff DBP overestimated and cuff PP underestimated intra-arterial aortic values. Finally, the observed variability in cuff BP accuracy adversely influenced correct classification of BP (compared against intra-arterial classification) across all JNC 7 categories, with particular discordance in the range from pre-HTN to stage 1 HTN. These data are summarized in the Central Illustration and indicate the need to improve accuracy standards of cuff BP devices.
A key problem in addressing the global burden of disease related to high BP is improving the diagnosis and characterization of the hypertensive phenotype (37). A fundamental problem with BP accuracy was identified in our study that affects most (but not all) cuff BP devices. Despite strong correlations between cuff BP and intra-arterial BP, 16 of 22 examined cuff BP devices significantly underestimated intra-arterial brachial SBP (Figure 1A) and 15 of 18 significantly underestimated PP (Figure 1C). The mean difference in the magnitude of the underestimation often exceeded 10 mm Hg. Translating these error margins to the traditional classification of BP based on intra-arterial SBP readings, cuff BP correctly identified pre-HTN and stage 1 HTN in only about one-half of the participants, whether based on intra-arterial brachial or aortic SBP (Table 1). Concordance with revised intra-arterial brachial BP thresholds (based on cuff BP percentile rank) was improved from 50% to 66% in the pre-HTN range (Tables 1 and 2). This analysis also resulted in reduced systematic underestimation of risk using cuff BP among the categories of pre-HTN and stage 1 HTN. Instead, a relatively even distribution was observed toward both overestimation and underestimation of correct classification of intra-arterial BP (Table 2). The true implications of these findings with respect to identification of risk related to BP in clinical practice will require future studies.
It could be argued that our findings are not a major clinical problem, because HTN thresholds have been derived from well conducted clinical trial data using the same (or similar) cuff BP methods to that analyzed in this current work. Thus, whether cuff BP is measuring the intra-arterial BP could be largely irrelevant if risk can still be gauged relative to the BP methods employed in the clinical trials. This contention would be valid if there were consistent systematic error(s), but in fact there was wide interdevice variability with respect to SBP, DBP, and PP accuracy. To clarify the issue, separate analysis on the accuracy of BP devices used in all the seminal clinical trials would be required. In any case, a reasonable degree of confidence that cuff BP is representative of intra-arterial brachial or aortic SBP is associated with readings <120/80 mm Hg or ≥160/100 mm Hg (Tables 1 and 2).
Cuff BP validation standards
Guidance on validation protocols for cuff BP devices is provided by several scientific bodies (33,38–42); however, there are many procedural differences between guidelines on features such as sample size, acceptable margin of error, and pass criteria (43). When comparing BP device performance with the reference standard (which can be intra-arterial BP or, most often, mercury sphygmomanometry), differences of 0 to 5 mm Hg are considered to be “very accurate,” whereas differences >15 mm Hg are “very inaccurate” (40). Although there are many ways to determine “pass” criteria for BP devices, the British Hypertension Society provide the highest grade pass (A) if 60% of differences fall within 5 mm Hg and only 5% of differences fall outside of 15 mm Hg (33). The analysis we have conducted cannot be directly compared with results of validation studies assessing the performance of individual BP devices. However, it is of note that only 33% of cuff SBP readings fell within 0 to 5 mm Hg, and >20% were >15 mm Hg from intra-arterial SBP (Figure 3). That would equate to a grade D (fail) device performance. From the available data, weak associations among age, body mass index, and cuff BP differences were observed in meta-analyses 2 and 3, but we were unable to determine clear-cut reasons for the disparity between cuff and intra-arterial BP.
A novel finding with respect to the use of mercury sphygmomanometry as a reference standard in BP validation protocols is that this method demonstrated sizable imprecision. Compared with intra-arterial brachial BP, the mercury method performed better than oscillometric BP with respect to the level of SBP underestimation, but significant overestimation of DBP and underestimation of PP was still observed (Online Table 16). There was insufficient data on mercury BP to compare this method with oscillometric BP for accuracy compared with intra-arterial aortic BP. Overall, the analyses cast some doubt on the robustness of mercury sphygmomanometry as the standard against which BP device performance is gauged (possibly due to influences of operator error), albeit acknowledging that it is the best noninvasive option currently available. Intra-arterial BP measured under rigorous criteria has the strongest level of BP accuracy and may be a better choice as the comparator for BP device validation. But, it is less practical, and it is not ethical to use among some populations. In any case, our observation of significant differences (and marked variability) between intra-arterial aortic and brachial BP clearly shows that it is not acceptable to assume peripheral BP is representative of central BP. This finding is applicable to BP device validation protocols in which cuff BP is compared against intra-arterial BP at the radial (44), brachial (10), or aortic (45) level. Improvement of BP device accuracy standards is desirable (29).
Study strengths and limitations
Individual-level data were acquired from a wide variety of studies employing high-quality techniques and spanning several decades of investigations, altogether comprising relatively large sample sizes for each meta-analysis. However, this also probably contributed to the observed statistical heterogeneity, indicating excess variation among experimental protocols and a degree of uncertainty regarding effect estimates. Although intra-arterial BP is the reference standard measurement of BP (46,47), inaccurate BP is possible due to numerous sources of error: 1) if operators do not follow appropriate techniques (e.g., catheter handling and dynamic response) (48); 2) variability in BP between the recording of cuff and intra-arterial measurements; 3) if measures being compared are recorded sequentially rather than simultaneously; or 4) if measures are being compared within contralateral rather than ipsilateral arterial sites. Reassuringly, sensitivity analyses showed no significant difference between the studies that received the maximum quality rating for experimental design taking into consideration these sources of error versus those that did not. Availability of repeated data would have helped address this issue further, but this was unavailable in most studies. Finally, the study populations were generally typical of patients presenting with clinical indications for coronary artery catheterization, and therefore, there was bias toward overweight, middle- to older-age men, and the findings cannot be widely generalized.
Cuff BP is the cornerstone measurement in HTN management. The most important finding of the present study was the inaccuracy of cuff BP when compared with intra-arterial brachial BP and aortic BP. These deviations substantially influenced BP classification according to clinical guideline criteria. The inadequacies of cuff BP identified within this work could be improved with better noninvasive cuff BP methods to estimate brachial or aortic BP. This should then lead to enhanced clinical diagnosis and management of HTN.
COMPETENCY IN MEDICAL KNOWLEDGE: Measurement of BP with pneumatic cuff devices is subject to considerable variability that affects correlations with direct intra-arterial brachial and aortic pressure measurements. When compared with intra-arterial pressures, brachial cuff sphygmomanometry generally underestimates systolic and overestimates diastolic BP.
TRANSLATIONAL OUTLOOK: New methods of noninvasive BP measurement should undergo robust validation to ensure accuracy before they are employed in patient care or population health studies.
For an expanded Methods section, as well as supplemental figures and tables, please see the online version of this article.
Microlife Co., Ltd., and National Yang-Ming University have signed a contract for transfer of the noninvasive central blood pressure technique. The contract of technology transfer includes research funding for conducting the validation study. Dr. Chen has served as a speaker or a member of a speakers bureau for AstraZeneca, Pfizer, Bayer AG, Bristol-Myers Squibb, Boehringer Ingelheim, Daiichi-Sankyo, Novartis Pharmaceuticals, Servier, Merck & Co., Sanofi, and Takeda Pharmaceuticals International. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- blood pressure
- diastolic blood pressure
- JNC 7
- Seventh Report of the Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure
- pulse pressure
- systolic blood pressure
- Received February 23, 2017.
- Revision received May 26, 2017.
- Accepted May 30, 2017.
- 2017 American College of Cardiology Foundation
- Forouzanfar M.H.,
- Alexander L.,
- Anderson H.R.,
- et al.
- Pickering T.G.,
- Hall J.E.,
- Appel L.J.,
- et al.
- Liszka H.A.,
- Mainous A.G. 3rd.,
- King D.E.,
- Everett C.J.,
- Egan B.M.
- Hunyor S.N.,
- Flynn J.M.,
- Cochineas C.
- Kelly R.P.,
- Gibbs H.H.,
- O'Rourke M.F.,
- et al.
- Kollias A.,
- Lagou S.,
- Zeniodi M.E.,
- Boubouchairopoulou N.,
- Stergiou G.S.
- Vlachopoulos C.,
- Aznaouridis K.,
- O'Rourke M.F.,
- Safar M.E.,
- Baou K.,
- Stefanadis C.
- Cheng H.M.,
- Chuang S.Y.,
- Sung S.H.,
- et al.
- Laugesen E.,
- Knudsen S.T.,
- Hansen K.W.,
- et al.
- Sharman J.E.,
- Avolio A.P.,
- Baulmann J.,
- et al.
- Tummers B.
- O'Brien E.,
- Petrie J.,
- Littler W.,
- et al.
- Laugesen E.,
- Rossen N.B.,
- Peters C.D.,
- et al.
- Rossen N.B.,
- Laugesen E.,
- Peters C.D.,
- et al.
- Olsen M.H.,
- Angell S.Y.,
- Asma S.,
- et al.
- ↵(2009) American National Standard Non-Invasive Sphygmomanometers—Part 2: Clinical Validation of Automated Measurement Type. ANSI/AAMI/ISO 81060-2:2009 (AAMI, Arlington, Virginia).
- Beime B.,
- Deutsch C.,
- Gomez T.,
- Zwingers T.,
- Mengden T.,
- Bramlage P.
- O'Callaghan W.G.,
- Fitzgerald D.J.,
- O'Malley K.,
- O'Brien E.
- Perloff D.,
- Grim C.,
- Flack J.,
- et al.