Article Text

Download PDFPDF

Global prevalence of resistant hypertension: a meta-analysis of data from 3.2 million patients
  1. Jean Jacques Noubiap1,
  2. Jobert Richie Nansseu2,
  3. Ulrich Flore Nyaga3,
  4. Paule Sandra Sime3,
  5. Innocent Francis1,
  6. Jean Joel Bigna4,5
  1. 1 Department of Medicine, Groote Schuur Hospital and University of Cape Town, Cape Town, South Africa
  2. 2 Department of Public Health, Faculty of Medicine and Biomedical Sciences, University of Yaoundé I, Yaoundé, Cameroon
  3. 3 Department of Internal Medicine and Specialties, Faculty of Medicine and Biomedical Sciences, University of Yaoundé, Yaoundé, Cameroon
  4. 4 Faculty of Medicine, University of Paris Sud XI, Le Kremlin-Bicêtre, France
  5. 5 Department of Epidemiology and Public Health, Centre Pasteur of Cameroon, Yaoundé, Cameroon
  1. Correspondence to Dr Jean Joel Bigna, Department of Epidemiology and Public Health, Centre Pasteur of Cameroon, Yaoundé, Cameroon; bignarimjj{at}


Objective We conducted the first systematic review and meta-analysis to estimate the specific prevalence of apparent treatment-resistant, pseudo-resistant and true-resistant hypertension among treated patients with hypertension globally.

Methods We conducted a search in PubMed, EMBASE, Web of Science and Global Index Medicus to identify articles published from inception to 30 September 2017, and searched the reference list of retrieved articles. We used a random-effects model to estimate the prevalence of resistant hypertension across studies and heterogeneity was assessed via the χ² test on Cochran’s Q statistic.

Results We included 91 studies published between 1991 and 2017 reporting data of a pooled sample of 3 207 911 patients with hypertension on antihypertensive drugs globally. Most of the studies (n=64, 70%) only used office blood pressure (BP) measurement. In the general, population of treated patients with hypertension, the prevalence of true-resistant, apparent treatment-resistant and pseudo-resistant hypertension were 10.3% (95% CI 7.6% to 13.2%), 14.7% (95% CI 13.1% to 16.3%) and 10.3% (95% CI 6.0% to 15.5%). The prevalence of true-resistant hypertension was 22.9% (95% CI 19.1% to 27.0%), 56.0% (95% CI 52.7% to 59.3%) and 12.3% (95% CI 1.7% to 30.5%) in chronic kidney disease, renal transplant and elderly patients, respectively.

Conclusions This study shows a high prevalence of true-resistant hypertension. This prevalence is lower than that of apparent treatment-resistant hypertension, demonstrating the importance to exclude causes of pseudo-resistant hypertension including white-coat hypertension with the use of ambulatory BP measurement. The burden of resistant hypertension is highest in patients with chronic kidney disease. New treatments for resistant hypertension are highly needed, considering the disastrous complications of the disease.

  • resistant hypertension
  • white-coat hypertension
  • pseudo-resistant hypertension
  • meta-analysis
  • systemic review

Statistics from


Hypertension is the most common chronic disease and the leading risk factor for disability and premature deaths in the world.1 2 It affects about 1 billion adults, accounts for about 9% of global disability-adjusted life years and is associated with more than 9 million deaths annually.1 2 Blood pressure (BP) control in affected individuals significantly reduces the occurrence of complications such as stroke, coronary artery disease, progression of chronic kidney disease (CKD), heart failure and consequential disability and deaths.3 4 Unfortunately, a substantial proportion of hypertensive treated individuals fail to achieve long-term BP control.5–7 For instance, in sub-Saharan Africa, only as few as 7% of treated patients with hypertension have controlled BP.8 Several factors account for poor BP control, including absence of hypertension treatment guidelines or non-adherence to these guidelines by physicians, non-compliance to pharmacological treatment or life styles changes.6 Another increasingly recognised cause of uncontrolled BP is resistant hypertension (RH), which occurred in a proportion of patients who are fully adherent to a well-designed regimen including multiple antihypertensive drugs.6

RH has a poor prognosis. As compared with those with controlled BP, patients with RH have higher risk of end-organ damage, including carotid intima–media thickening, retinopathy, left ventricular hypertrophy and heart failure, myocardial infarction, stroke, impaired renal function, and death.7 9 10 In addition to its clinical importance, RH carries a considerable public health, economic and societal burden due to treatment cost and associated disability and premature deaths.11

Reliable epidemiological data are crucial to support public health efforts to tackle the burden of RH and specifically to advocate for the development of new and effective treatment options. We present here the first systematic review and meta-analysis of the extant literature on the burden of RH which provides estimate of the prevalence apparent treatment RH, true RH and pseudo-RH in the general treated hypertensive population and in patients with other cardiovascular comorbidities.


Literature search

We performed a search of PubMed/MEDLINE, Excerpta Medica Database, Web of Science and Global Index Medicus to identify all relevant articles reporting data on the prevalence of RH in treated patients with hypertension from inception to 30 September 2017, without language restriction. Bibliographical search strategies are available in the online supplementary appendix. Furthermore, we manually searched the reference lists of all relevant articles and reviews to identify additional eligible studies.

Supplementary file 1

Selection of studies for inclusion in the review

We considered cross-sectional and cohort studies reporting on the prevalence of RH in adults (aged ≥18 years) treated for hypertension, or providing enough data to compute this estimate. Furthermore, eligible studies were those which used one of the following definitions of RH: (1) uncontrolled BP (≥140/90 mm Hg) despite antihypertensive regimen of ≥3 medications of different classes or (2) uncontrolled BP (≥140/90 mm Hg) despite antihypertensive regimen of ≥3 medications of different classes (including diuretics) or (3) uncontrolled BP (≥140/90 mm Hg) despite treatment with ≥4 antihypertensive agents of different classes irrespective of BP values.12 We categorised RH into three categories: (1) pseudo-RH: as found in patients who fulfil the above criteria for RH, but have white-coat hypertension as confirmed by 24-hour ambulatory BP measurement (ABPM) or homed BP measurement (HBPM), or have an identifiable cause of uncontrolled BP such as non-compliance to treatment; (2) true RH was present when pseudo-RH had been excluded and (3) apparent treatment RH was considered in patients for which pseudo-resistant RH had not been excluded. We excluded studies lacking primary data and/or explicit description of methods. For studies reported in two or more articles, we considered the most informative one.

Two assessors (JJN and JRN) independently screened the titles and abstracts of records retrieved from literature search, and then the full texts of studies found potentially eligible were obtained and further assessed for final inclusion (figure 1). Disagreements were solved through consensus.

Appraisal of the risk of bias of included studies

We used the tool developed by Hoy et al to evaluate the methodological quality of included studies.13 A score of 1 (yes) or 0 (no) was assigned for each item, and the scores across items were summed to generate an overall quality score that ranged from 0 to 10. According to the overall scores, studies were classified as having a low (>8), moderate (6–8) or high (≤5) risk of bias. Two assessors (JJB and JRN) independently assessed study methodological quality, with disagreements resolved by consensus.

Data extraction and management

The following information was collected for each study: first author’s name, year of publication, year of participants’ recruitment, country, setting (primary care, secondary care or specialised clinic), study design, timing of data collection (prospective or retrospective), follow-up time for cohort studies, method (office BP measurement (OBPM), HBPM or ABPM) and technique (auscultatory or oscillatory) of BP measurement, definition of RH, sample size, mean or median age and age range, proportion of male participants, any disease specific to the study population and the number of participants with true RH and number of those with apparent whenever available. Based on the country of recruitment, a continent and a World Bank category (high-income, middle-income or low-income country) was assigned to each study. This information was independently extracted by three investigators (IF, PSS and UFN) using a preconceived and standardised data extraction form. At the end of the extraction process, one investigator (JJN) cross-checked all the data for accuracy. Disagreements were resolved through consensus following a discussion.

Data synthesis and analysis

Analyses were performed using ‘meta’ packages of R (V.3.5.0) (The R foundation for statistical computing, Vienna, Austria). For each study, the unadjusted prevalence estimates and SEs were calculated based on the information on crude numerators and denominators. To minimise the effect of studies with extremely small or extremely large prevalence on the overall estimate, we stabilised the variance of the study-specific prevalence estimates with the Freeman-Tukey double arcsine transformation before pooling the data using the DerSimonian-Laird random-effects meta-analysis model.14 The random-effects model was chosen in anticipation of substantial variations in RH prevalence estimates across the included studies. Data were pooled by type of population treated for hypertension. We assessed inter-rater agreement for inclusion and quality assessment using Cohen’s kappa (κ) coefficient.

We appraised heterogeneity between studies using Cochran’s Q statistic, H and the I2 statistics,15 16 which estimate the percentage of total variation across studies due to true between-study difference rather than chance, with I2 values of 25%, 50% and 75% representing low, medium and substantial heterogeneity, respectively. We explored sources of heterogeneity of the prevalence of apparent RH in general population treated for hypertension through subgroup and meta-regression analyses defined by mean/median age, proportion of males, sample size, GINI coefficient, WHO regions, United Nations Statistical Division (UNSD) of regions, setting, level of income and BP measurement. Comparisons between subgroups were performed using the Q-test based on the analysis of the variance. For categorical variables, the global p value was considered for inclusion in the multivariable model. We applied a manual backward selection procedure to identify factors independently associated with the variation of overall prevalence of apparent RH. We retained in the multivariable meta-regression analysis, variables associated at p<0.20. In the case of non-linear distribution, we log-transformed the covariate (year of publication, sample size and GINI coefficient) before performing the meta-regression analyses. Publication bias was evaluated with funnel plots supplemented by formal statistical assessment using Egger’s test.17 A p<0.10 was considered statistically significant to detect publication bias. In the case of publication bias, we adjusted the prevalence using trim-and-fill method, recomputing; therefore, the prevalence of each iteration until the funnel plot was symmetrical about the new prevalence.


The review process and characteristics of included studies

Initially, 2024 records were identified and 91 studies were retained in the meta-analysis (figure 1). The list of included studies is in the online supplementary appendix. Regarding methodological quality, 39 (42.9%) had low risk of bias, 48 (52.7%) had moderate risk of bias and 4 (4.4%) had high risk of bias. The inter-rater agreement for quality assessment was excellent (κ=0.96). In total, 3 207 911 participants were included. Year of publication varied between 1991 and 2017, proportion of males varied from 28.8% to 96.0% (and one study included only males and another one only females; n=66 studies), mean/median age varied from 33 to 88 years (range 18–103; n=66 studies) and the median GINI coefficient was 0.36 (1st–3rd quartiles, 0.33–0.41). Most of studies were from general population with hypertension, were from Europe, were conducted in high-income countries and measured BP in office (table 1). Individual characteristics of included studies are reported in the appendix (online supplementary table 2).

Table 1

Characteristics of included studies

Overall prevalence of RH

In the general population treated for hypertension, the prevalence of true, apparent and pseudo-RH were 10.3% (95% CI 7.6% to 13.2%) (figure 2), 14.7% (95% CI 13.1% to 16.3%) (figure 3) and 10.3% (95% CI 6.0% to 15.5%) (figure 2), respectively. For others specific populations treated for hypertension (table 2), the prevalence varied between 4.9% (obesity) and 56.0% (renal transplants) for true RH (online supplementary figure 1), between 11.5% (obesity) and 28.8% (CKD) for apparent RH, and between 3.1% (elderly) and 12.7% (renal transplants) for pseudo-RH. There was no difference between populations for the prevalence of apparent RH, while significant difference was found for true and pseudo-RH (table 2). There was a wide variation of the range of the 95% prediction interval (tables 2 and 3). The sensitivity analysis including only studies with low risk of bias yielded to prevalence estimates close to that of crude analysis for general population treated for hypertension (online supplementary table 3). There was substantial heterogeneity except for the pseudo-RH in elderly. Publication bias was detected for apparent (online supplementary figure 2) and pseudo-resistant (online supplementary figure 3) hypertension and not for true RH (online supplementary figure 4) in general hypertensive population treated for hypertension (table 2). The trim-and-fill adjusted prevalence of apparent and pseudo-RH in general hypertensive population were close to that of crude prevalence (table 2).

Table 2

Summary statistics and comparison by population of apparent, true and pseudo-resistant hypertension

Table 3

Summary statistics and comparison of subgroups of apparent resistant hypertension in general population with hypertension

Figure 2

Meta-analysis results for prevalence of true resistant hypertension and pseudo-resistant hypertension.

Figure 3

Meta-analysis results for prevalence of apparent resistant hypertension.

Subgroup meta-analyses of the prevalence of apparent RH

The prevalence of apparent RH did not vary significantly across WHO regions. The highest prevalence was found in the Western Pacific (19.2%, 95% CI 14.0% to 25.1%) and the lowest in the African region (10.1%, 95% CI 4.6% to 17.5%) (table 3, online supplementary figure 5). According to UNSD regions, the prevalence of apparent RH was similar in the Americas, Asia and Europe (~14%) (table 3, online supplementary figure 6). This prevalence did not differ significantly pertaining to country income level (table 3, online supplementary figure 7) and according to studies’ settings (table 3, online supplementary figure 8).

Sources of heterogeneity in the prevalence of apparent RH in the general population treated for hypertension

In the univariable and multivariable meta-regression analyses, the prevalence was higher when using both ambulatory and OBPM. Variables (setting and BP measurement) included in the final model explained 43.4% of the 99.9% residual of the prevalence of apparent RH (online supplementary table 4).


This systematic review and meta-analysis compiled data from 3 207 911 treated patients with hypertension and revealed a high prevalence of true, apparent and pseudo-RH at 10.3%, 14.7% and 10.3%, respectively. Although there was no difference in prevalence estimates of apparent RH according to various regions of the world, setting and income, we found a significant difference with regard to the method used to measure BP. OBPM remained the method mostly used to follow up patients with hypertension, by contrast to ABPM or HBPM.

The prevalence of apparent RH found in this review (14.7%) confirms the finding of a previous meta-analysis which included 20 observational studies compiling data from 9 61 035 treated patients with hypertension and reporting a prevalence of apparent RH at 13.7%.18 Our results are also in compliance with other reports from European or American countries (between 8.4% and 17.7%),19 20 or even from Africa (12.1%).21 Unsurprisingly, the prevalence of RH was highest for patients with CKD or renal transplant patients (29%–56%), aligning to previous observations indicating that the prevalence of RH is almost three times higher among patients with CKD than the general hypertensive population.22 Therefore, more attention should be put on patients with CKD treated for hypertension, considering that CKD on one hand and RH on the other, are both associated with an increased risk of cardiovascular morbidity and mortality.

The prevalence of apparent RH seemed not to differ significantly across the various regions of the world, according to the setting or income, even though the income was assessed at a macroeconomic level rather than at an individual level. This suggests that the interventions to curb the burden of RH can be effectively implemented in all regions of the world and irrespective of the setting and income. Among these interventions, recent clinical trials have demonstrated the benefit of mineralocorticoid receptor antagonists in the treatment of RH.23 Indeed, spironolactone has been shown to be the better add-on therapy as compared with some α-blockers (clonidine, doxazosin) and β-blockers (bisoprolol) in patients with RH.23 On the other hand, non-pharmacological therapies have been proposed as effective in the context of RH, notably renal denervation and baroreceptor stimulation, though highly sophisticated and costly. Hence, their cost-effectiveness needs to be evaluated around the various regions of the world for them to be vulgarised in routine clinical practice.

Singularly, this is the first review presenting an estimate of the prevalence of true RH and pseudo-RH, both at 10.3%. This elevated prevalence of pseudo-RH highlights the important number of patients with hypertension wrongly classified as having RH. Pseudo-RH is largely explained by improper BP measurement technique, suboptimal antihypertensive regimen, clinical inertia, non-compliance or poor compliance to treatment and white-coat effect.23 Non-compliance or poor compliance to hypertensive medication has been incriminated as playing a major role in pseudo-RH as it seems very common among patients with hypertension, reaching rates of 50% or more.21 24 Non-adherence to treatment is multifactorial: among other causes, it can be due to side effects of polymedication, unavailability and affordability of some medications especially in resource-poor settings25 26 or lack of education of patients about their condition and the treatment. To improve medication compliance, it is therefore important to simplify BP treatment regimens, to educate patients in order to empower patients-centred care and self-management, and to improve access to medications in order to avoid shortages in those with low income. Moreover, the prevalence of true-RH reported in this review is more than two times higher than previous estimates at 4%–5%.19 This finding puts in light the fast increasing number of patients with hypertension for whom early identification and aggressive treatment of RH would benefit the most, provided the increased cardiovascular morbidity and mortality risk linked with RH.19

White-coat effect is another important factor to take into consideration when dealing with RH, which can only be ruled out by using ABPM designated as the gold standard to diagnose white-coat and masked hypertension.27 Our study reveals, for instance, the dire lack of use of ABPM in routine clinical practice. In fact, ABPM had been used in 19.7% of studies by contrast to OBPM used in 72.1% of studies. Although being popular, simple and convenient to use, OBPM does not permit to monitor BP over the 24 hours of the day, especially at night, and to predict cardiovascular events.28 Therefore, other methods were developed among which ABPM and HBPM. If it was demonstrated that ABPM is superior to HBPM in diagnosing white-coat hypertension,27 it was equally shown that both techniques have similar performances in predicting cardiovascular events and evaluating response to treatment.28 Compared with ABPM, HBPM is less costly, more available and convenient to use; it presents a greater potential towards achieving optimal BP control and treatment compliance.29 Accordingly, it has been claimed that HBPM could constitute a reliable surrogate to ABPM, particularly when the latter technique is uncomfortable for the patient, unavailable or unaffordable.29 30 There is therefore a clear need to step up HBPM in routine clinical practice in environments where ABPM may appear cumbersome. BP monitors should be made available everywhere and at affordable prices.

However, our review must be interpreted in consideration of certain limitations. First and common to most reviews of this type, we found a high statistical heterogeneity between studies, which could not be explained by the study sample size, world region, setting or income; the major source of heterogeneity came from the different methods used to assess BP control and different definitions, though accepted, of RH. Second, the various regions of the world were not proportionally represented in the review, which may hinder the translatability of our results at a global scale. Third, the large majority of studies were hospital based and used OBPM to measure and follow up the BP; consequently, our results may present an underestimated burden of RH in the general population of treated patients with hypertension. Nonetheless and to the very best of our knowledge, this is the first systematic review and meta-analysis which has given a clear estimate of the burden of apparent, true RH and pseudo-RH. We compiled data from more than 3 million hypertensive people, more than three times the total population from a previous systematic review and meta-analysis on the prevalence of RH.18 We searched the major electronic databases and used rigorous methodological and statistical procedures to generate our prevalence estimates. The prevalence of sensitivity analysis including only studies with low risk of bias was close to that of crude analysis indicating that we can be confident in our principal findings.


This systematic review and meta-analysis revealed a high prevalence of true RH. It is high time new techniques and strategies be vulgarised to address efficiently this increased burden of RH globally. Moreover, almost 1 in 10 patients with hypertension may present pseudo-RH, which could be better tacked if recommended guidelines are applied when following up patients with hypertension. Specifically, ABPM should be introduced in routine clinical practice wherever possible or HBPM in alternative, rather than OBPM. Patients should be followed up carefully and treatment regimen simplified as much as possible, in order to reduce the rate of non-compliance.

Abstract translation

This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.



  • UFN, PSS and IF contributed equally.

  • Contributors JJN, JJB and JRN conceived the study. JJN and JJB did the literature search. JJN and JRN selected the studies. UFN, IF, PSS and JJN extracted data from the included studies. JJB and JJN synthesised the data. JJN, JJB and JRN wrote the first draft of the paper. JJN, JJB, JRN, UFN, IF and PSS critically revised successive drafts of the paper and approved its final version. JJN is the guarantor of the review.

  • Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.

  • Competing interests None declared.

  • Patient consent Not required.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Linked Articles