key: cord-0431401-16bs30dz authors: Bourque, G.; Ilin, J. V.; Ruzicka, M.; Davis, A.; Hundemer, G.; Hiremath, S. title: The Prevalence of Non-Adherence in Patients with Resistant Hypertension: a Systematic Review and Meta-Analysis date: 2020-08-16 journal: nan DOI: 10.1101/2020.08.14.20175125 sha: 4d00e29014d962c0a984524982b9b905bf3832bf doc_id: 431401 cord_uid: 16bs30dz Background: Resistant hypertension is quite prevalent and a risk factor for cardiovascular events. Patients with suspected resistant hypertension undergo more screening intensity for secondary hypertension, despite some of them being non-adherent to prescribed pharmacotherapy. The prevalence of non-adherence in this setting varies from about 5 to 80% in the published literature. Apart from the wide range, the relation between method of assessment and prevalence is not well established. Our objective was to establish the overall prevalence of non-adherence in the apparent treatment resistant hypertension population, explore causes of heterogeneity, and evaluate the effect of the method of assessment on the estimate of non-adherence. Methods: We performed a systematic review and meta-analysis. MEDLINE, EMBASE Classic+EMBASE, Cochrane, CINAHL, and Web of Science were searched for relevant articles. Details about the method of adherence assessment were extracted from each included article and grouped into direct and indirect. Pooled analysis was performed using the random effects model and heterogeneity was explored with metaregression and subgroup analyses. Results: The literature search yielded 1428 studies, of which 36 were included. The pooled prevalence of non-adherence was 35% (95% confidence interval 25 to 46 %). For indirect methods of adherence assessment, it was 25% (95% CI 15 to 39 %), whereas for direct methods of assessment, it was 44% (95% CI 32 to 57 %). Metaregression suggested gender, age, and time of publication as potential factors contributing to the heterogeneity. Conclusions: Non-adherence to pharmacotherapy is quite common in resistant hypertension, with the prevalence varying with the methods of assessment. Resistant hypertension (RH) is defined as blood pressure (BP) remaining above target despite the concurrent use of three or more antihypertensive agents of different classes, with one of the classes preferably being a diuretic and all of the medications being prescribed at optimal doses 1,2,3 . RH is common, with an estimated prevalence of about 10 to 30 %, and known to be a risk factor for cardiovascular events 4 . Furthermore, patients with RH are often part of high-risk groups with multiple cardiovascular comorbidities as well as vulnerable or disadvantaged populations. These patients undergo higher screening intensity for secondary hypertension. However, not all patients with apparent treatment RH have true RH. Apart from white coat hypertension and suboptimal prescribed drug doses and combinations, some individuals with apparent treatment RH may be non-adherent to the prescribed pharmacotherapy 5, 6 . There are several ways of assessing medication non-adherence, which can be broadly divided into indirect and direct methods. Indirect methods include questionnaires, self-reports, pill counts, rates of prescription refills, medication event monitoring systems (MEMS), and patient diaries. Direct methods include BP response to directly observed therapy (DOT) and measurement of the levels of BP-lowering drugs in physiologic fluids such as blood and urine 5 . The long term prevalence of non-adherence in chronic diseases is reported at about 50% 7 . However, this varied from 3 to 86% in individual studies in apparent treatment RH patients from a recent systematic review 8 . Interestingly, the pooled prevalence in this review varied based on the method of adherence measurement, from a low of 13% (similar estimate from self-report and physician interview) and 19% (prescription refill) to a high of 45% (DOT) and 49% (physical test, i.e. blood or urine assay). Though they were not grouped in this fashion, the former are indirect measures and the latter are direct measures. Apart from potentially different accuracy, these methods may capture different facets of nonadherence which might require different management strategies. As an example, pill counts and pharmacy refill data might identify occasional forgetfulness and/or carelessness, which can sometimes be related to an inability to follow instructions, either because of cognitive or physical limitations 5, 7 . It can be managed using reminders, pill packs, and other interventions. In other settings, a patient may choose to alter the prescribed medication regimen, either by discontinuing medications entirely, skipping doses, or modifying doses or dosing intervals, however still continuing to refill prescriptions 6 . Underlying health beliefs and certain demographic factors and comorbid conditions may be associated with this, which would evade detection by indirect measures 9,10 . It requires more intensive measures (such as therapeutic drug monitoring (TDM) or DOT) to diagnose 5, 6, 11 . Interventions to address this form of non-adherence are also not well studied. In the present study, we report on the overall prevalence of non-adherence in the apparent treatment RH population and what are the relative contributions of non-adherence based on different methods of assessment, with an emphasis on attempting to explain the heterogeneity of the data. This systematic review and meta-analysis has been conducted and is reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement 12 . It is registered with the International Prospective Register of Systematic Reviews database (registration number: CRD42020137944). The study protocol has also been published separately 13 . The databases MEDLINE (Ovid Interface, 1946 through April 02, 2019), EMBASE Classic+EMBASE (1947 through April 02, 2019), Cochrane (Cochrane Database of Systematic Reviews and Cochrane Central Register of Controlled Trials), Cumulative Index of Nursing and Allied Health Literature (CINAHL), and Web of Science were searched for relevant articles for this systematic review and meta-analysis. The search was designed by an information specialist (AD) and the search terms were adapted for the different databases. The MEDLINE search strategy is included in the supplementary appendix, table 1. Additional relevant records were identified through review of references of selected articles and from suggestions from experts in the field. We included observational studies, including cross sectional, retrospective, and prospective studies, as well as randomized controlled trials (RCTs) done on adult human participants aged 18 years or older with a diagnosis of RH, either uncontrolled on three or more drugs, or controlled with four or more drugs 1,2,3 . We only included studies where adherence to blood pressure lowering medications was measured using a test of adherence, either direct (such as TDM or DOT) or indirect (for example pill counts or pharmacy refill data). Only studies published in the English language were included. Studies published in other languages were included if a full-text version was available in English. Titles and abstracts of studies identified through the various database searches were uploaded to Covidence, an Internet based software program that facilitates collaboration among reviewers during the study selection process 14 . Two reviewers (GB and JVI) independently screened the titles and abstracts retrieved after the literature search to evaluate whether they met the pre-defined inclusion criteria. Conflicts between reviewers were resolved through discussion between the two reviewers until a consensus was reached. Full-text articles for the studies meeting inclusion criteria were retrieved and screened by the same two reviewers in order to select studies to be included in the systematic review. The reasons for excluding trials were recorded, both after title and abstract screening and after full-text screening. Reviewers were not blinded to the authors or journals when screening articles. We extracted key elements of study design, definitions of RH used, demographic characteristics, comorbidities, and data on adherence from the included studies. For adherence, details about the method of assessment were extracted and grouped into direct (eg urine or plasma assay or TDM, DOT) and indirect (eg pill counts, questionnaires, adherence scales). For studies that reported more than one method of assessment, the highest value was used for the pooled analysis. Additionally, for studies that reported both direct and indirect methods, these were compared in a separate meta-analysis. All data was extracted in duplicate by two reviewers (GB and JVI). The study quality and the presence of potential bias within individual studies included in this systematic review were evaluated at both the outcome and study levels. The methodological quality of eligible full-text articles was independently assessed by two reviewers (GB and JVI) using the Newcastle-Ottawa scale 15 . The Newcastle-Ottawa scale includes the following domains: selection, comparability (not applicable to the studies included in this systematic review and meta-analysis), exposure (not applicable to the studies included in this systematic review and meta-analysis), and outcome. Disagreements were resolved through discussion until consensus was reached. A pooled estimate of the prevalence of non-adherence was generated. The summary prevalence was estimated using the random effects modeling as described by DerSimonian-Laird 16 . We chose the random effects method because of its conservative summary estimate and because it incorporates between-and within-study variance. To assess heterogeneity of the event frequencies across studies, we used the Cochran's Q-statistic test and the I 2 statistic. All analyses were conducted using the Comprehensive Metaanalysis V2 software (Version 2.2, Biostat, Englewood NJ). Subgroup analyses were used to explore possible sources of heterogeneity, based on type of test used to measure adherence (grouped as direct versus indirect), study design, and definition of RH. We conducted univariate metaregression to assess moderator variables which are continuous in nature. The subgroup analyses and metaregression were also assessed as a method of resolving any statistical heterogeneity. Sensitivity analyses were conducted by excluding one study at a time and observing change in pooled estimate (with a > 10% change being considered significant). Publication bias was assessed by visual examination of the funnel plot and the Duval and Tweedie's trim and fill method 17,18 . . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted August 16, 2020. . https://doi.org/10.1101/2020.08.14.20175125 doi: medRxiv preprint The figure 1 shows a flow diagram of the systematic review process. The initial database search identified 2770 articles and 13 additional records were identified through other sources. After removal of duplicates, 1428 titles and abstracts were screened. 1218 articles were excluded at this stage and therefore 210 full-text articles were assessed for eligibility. 174 of these were excluded, with the most common reasons for exclusion being absence of RH as defined above and ineligible study design, such as case reports, reviews, or editorials. Therefore, 36 full-text articles were eligible for inclusion in this systematic review and metaanalysis. A summary of the included articles is included in Legend for Figures and Tables Table 1 Study Characteristics Table 2 Population Characteristics CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted August 16, 2020. . Table 1 and Table 2 , which describe the study and population characteristics. 36 studies were included in this systematic review. Of these, 21 were conducted in Europe, 11 in North America, and 4 in South America. The majority of the included studies were observational, with 15 prospective cohort studies, 11 cross-sectional studies, and 6 retrospective cohort studies. 3 of the included studies were RCTs and 1 was a pre-post intervention study. The total number of participants included was 69,912. One study 19 accounted for the vast majority of these. The median sample size was 83, with a range from 18 to 60,327. The proportion of patients with comorbidities, such as diabetes, chronic kidney disease (CKD), congestive heart failure (CHF), and cerebrovascular disease (CVD), was inconsistently reported in the individual studies. The definition of RH as well as the thresholds used to define non-adherence to prescribed pharmacotherapy varied between studies. 32 studies used office BP for defining RH and only 3 used ambulatory blood pressure monitoring (ABPM). One study used both methods. Antihypertensive regimens were not consistently reported and it often was not specified whether patients were on an optimized regimen. The most common threshold for diagnosis of RH was systolic blood pressure (SBP) ≥ 140 and/or diastolic blood pressure (DBP) ≥ 90. Patients with secondary forms of hypertension as well as white-coat hypertension were not consistently excluded, as is apparent from the low use of ABPM for diagnosis of RH. 18 studies assessed non-adherence using direct methods, 12 with indirect methods, and 6 with both direct and indirect methods, either concurrently or sequentially (see Table 1 ). DOT was used in 5 studies, plasma and/or urine assays in 19, MEMS in 2, non-standardized questionnaires in 4, standardized questionnaires such as the Morisky Medication Adherence Scale (MMAS) and the Medication Adherence Report Scale (MARS) in 8, prescription refill in 2, Proportion of Days Covered (PDC) in 3, and pill counts in 2 (see Table 1 ). The pooled prevalence of non-adherence to antihypertensive pharmacotherapy was 35% (95% confidence interval [CI] 25 to 46%) as shown in figure 2 . The prevalence of non-adherence varied across studies, with values ranging from 3.3% 20 to 86.1% 21 , leading to a high degree of statistical heterogeneity (I 2 = 99, p < 0.001). . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted August 16, 2020. . Using indirect methods of adherence assessment, the pooled prevalence of nonadherence was 25% (95% CI 15 -39%, I 2 = 99), whereas with direct methods of assessment, it was 44% (95% CI 32 -57%, I 2 = 89). The prevalence of nonadherence was higher in the subgroup of studies where ABPM was required for definition of RH (61%, 95% CI 27 -86%, I 2 = 89), compared to the subgroup of studies which defined RH based on office BP (32%, 95% CI 22 -44%, I 2 =99). Lastly, there was no difference in the prevalence of adherence in retrospective studies (38%, 95% CI 16 -66%, I 2 = 94) or prospective studies (34%, 95% CI 24 -47%, I 2 = 99). Excluding one study did not change the overall pooled estimate by > 10% (range 34 -36%, compared to 35% with all studies). The subgroup analyses did not resolve the heterogeneity, which remained high. With univariate metaregression (supplementary appendix, figures 1 to 5), there was an association of higher prevalence of non-adherence with younger age (slope coefficient 0.20, p < 0.001), higher proportion of men (slope coefficient 0.08, p < 0.001), and with more recent studies (slope coefficient 0.04, p < 0.001). An association of larger sample size with lower prevalence of non-adherence was also seen (slope coefficient 0.003, p < 0.001), which was driven by one large study with a sample size of 60,327 and a non-adherence prevalence of 11%. Exclusion of this study resulted in the association no longer being significant (slope coefficient 0.0001, p = 0.40). Six studies reported both direct and indirect methods of assessment 22,23,24,25,26,27 . In a pooled analysis, the odds of detecting non-adherence were higher with direct methods (summary odds ratio 1.95, 95% CI 1.31 to 1.95, p = 0.001), though with significant statistical heterogeneity (I 2 = 98). The overall quality of the studies was fair, with a median of 4 stars (range 3 to 5). A large proportion of studies rated high on the representativeness subsection of the selection domain of the Newcastle-Ottawa scale. However, they rated poorly on the subsection demonstration that the outcome of interest, in this case nonadherence to antihypertensive medications, was not present at start of study. They rated mostly high on the assessment of outcome, length of follow-up, and adequacy of follow-up subsections of the outcome domain of the scale (see supplementary appendix, table 2). There was no evidence of publication bias from visual examination of the funnel plot and no additional studies were imputed using the Duval and Tweedie's trim and fill method (see supplementary appendix, figure 6 ). In this systematic review, we report a high degree of non-adherence prevalence in patients with apparent treatment RH, at 35%. However, there was significant heterogeneity identified among the studies, with a range of non-adherence from . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted August 16, 2020. . https://doi.org/10.1101/2020.08.14.20175125 doi: medRxiv preprint 3% to 86%. Studies using direct methods reported a higher degree of nonadherence, at 44%, compared to indirect methods, at 25%. The study level metaregression reveals younger age, male sex, and more recent publication date as being associated with higher rates of non-adherence. Though non-adherence has always been part of the evaluation for RH, the extent of non-adherence, approximately at 1 in 3 patients in this systematic review, is so high, that an accurate assessment of adherence is essential before the patient is labeled as RH. Instead of investigations for secondary hypertension, consuming valuable healthcare resources, more attention needs to be focused on assessment and management of non-adherence in these patients. Additionally, non-adherence in most clinical settings is assessed by patient questionnaire, pharmacy filling data, or pill counts. As we demonstrate, these indirect methods of adherence provide a lower estimate compared to direct methods of assessment of non-adherence, such as TDM or DOT. Unfortunately, both TDM and DOT are mostly research tools and not available or funded in routine clinical practice. We hope this systematic review elevates their status for more routine use, especially in the setting of RH, where the high prevalence of non-adherence makes these tools more valuable and possibly cost effective. A previous systematic review 8 also reports on a high degree of non-adherence in this population, though with 24 studies, compared to 36 in the present study. Additionally, we believe grouping assessment methods into direct and indirect methods provides a more accurate picture of the non-adherence phenotype. Additionally, in the present study, the exploration of heterogeneity provides more insight into the fact that patients of a younger age and men might have a higher non-adherence phenotype. We also report the trend for higher rates of nonadherence in the recently published studies. It is not quite clear whether this is because of ascertainment bias or more accurate assessment of non-adherence. Regardless, pill counts and pharmacy filling data may be necessary, but are far from sufficient for assessment of adherence. The dichotomy of direct and indirect assessment of adherence does not necessarily mean that one is more accurate than the other. Indirect assessment of non-adherence with pill counts and pharmacy filling data may provide more insight into the occasional forgetfulness or missed doses in a patient who is otherwise engaged with the medication regimen. Direct methods of nonadherence provide pharmacokinetic and pharmacodynamic data, however in a cross-sectional manner. We would posit that these two broad techniques are complementary in the complete diagnosis of non-adherence. The assessment of non-adherence is merely the first step in adequate blood pressure control. Management of non-adherence requires further exploration of the reasons of non-adherence and taking steps accordingly. Much of the focus has been on patient aids, such as reminders and smart pill bottles, and these would be useful for the unintentional non-adherence phenotype. For patients who . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted August 16, 2020. . are otherwise disengaged in the medication and therapy plan, further exploration of their motives and beliefs is necessary. The diagnosis of non-adherence is but the initial step in this process. Possible limitations of this study include the finding of severe heterogeneity, which has not been completely resolved by the analytic plan, the limitations of the literature search), and the potential lack of granular data in the published studies for a patient-level meta-analysis. However, there was no publication bias seen in the analysis and an information specialist was used for the literature search, attenuating these issues. In patients with apparent treatment RH, standard indirect measures of assessing adherence can miss a substantial proportion of patients who are non-adherent to prescribed pharmacotherapy. This systematic review and meta-analysis also provides a basis for future research on strategies to better address the different factors that contribute to medication non-adherence in this setting. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted August 16, 2020. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted August 16, 2020. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted August 16, 2020. . Tables Table 1 Study Characteristics Table 2 CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted August 16, 2020. . https://doi.org/10.1101/2020.08.14.20175125 doi: medRxiv preprint Resistant Hypertension: Diagnosis, Evaluation, and Treatment. A Scientific Statement from the American Heart Association Professional Education Committee of the Council for High Blood Pressure Research ESH/ESC Guidelines for the Management of Arterial Hypertension: the Task Force for the Management of Arterial Hypertension of the European Society of Hypertension (ESH) and of the European Society of Cardiology (ESC) Evidence Review and Guidelines for the Management of Resistant Hypertension Trends for Prevalence and Incidence of Resistant Hypertension: Population Based Cohort Study in the UK 1995-2015 Can Drugs Work in Patients who do Not Take Them? The Problem of Non-Adherence in Resistant Hypertension Renal Sympathetic Denervation in Patients with Treatment-Resistant Hypertension after Witnessed Intake of Medication before Qualifying Ambulatory Blood Pressure Blood Pressure Changes after Catheter-Based Renal Denervation are Related to Reductions in Total Peripheral Resistance Blood Pressure Reductions following Catheter-Based Renal Denervation are not Related to Improvements in Adherence to Antihypertensive Drugs Measured by Urine/Plasma Toxicological Analysis Resistant Hypertension Revisited: a Comparison of Two University-Based Cohorts Characteristics, Drug Combinations and Dosages of Primary Care Patients with Uncontrolled Ambulatory Blood Pressure and High Medication Adherence Non-Adherence to Antihypertensive Medication is Very Common among Resistant Hypertensives: Results of a Directly Observed Therapy Clinic Prevalence of Treatment-Resistant Hypertension after Considering Pseudo-Resistance and Morbidity: a Cross-Sectional Study in Irish Primary Care Tertiary Work-Up of Apparent Treatment-Resistant Hypertension Prevalence and Correlates of Low Medication Adherence in Apparent Treatment Resistant Hypertension Resistant Hypertension? Assessment of Adherence by Toxicological Urine Analysis Heart Rate is a Useful Marker of Adherence to Beta-Blocker Treatment in Hypertension Prevalence of Resistant Hypertension in Non-Elderly Adults: Prospective Study in a Clinical Setting Detecting Non-Adherence by Urine Analysis in Patients with Uncontrolled Hypertension: Rates, Reasons and Reactions Importance of Thorough Investigation of Resistant Hypertension before Renal Denervation: Should Compliance to Treatment be Evaluated Systematically? Adherence to Antihypertensive Medication in Treatment-Resistant Hypertension Undergoing Renal Denervation