Original Article
Systolic blood pressure variability is an important predictor of cardiovascular outcomes in elderly hypertensive patients Enayet K. Chowdhury a, Alice Owen a, Henry Krum a, Lindon M.H. Wing b, Mark R. Nelson c, and Christopher M. Reid a, on behalf of the Second Australian National Blood Pressure Study Management Committee
Objectives: In hypertensive persons aged 60 years or below, visit-to-visit SBP variability is directly associated with cardiovascular events, especially stroke. It is unclear whether such a relationship exists for older persons. We investigated whether there is a relationship between visit-to-visit SBP variability and cardiovascular events in an elderly population, and identified the factors associated with increased SBP variability. Methods: Information from 49 771 visits of 5880 patients aged at least 65 years being treated for hypertension in the Second Australian National Blood Pressure study was used. Patients were followed for 4.1 (median) years and had eight (median) doctor visits during the study. SBP variability was defined as within-individual SD of SBP across study follow-up visits. Results: Increased visit-to-visit SBP variability was found to be a strong predictor for future cardiovascular events in this elderly population. The hazard ratio (95% confidence interval) for any first fatal/nonfatal cardiovascular event for highest decile compared with lowest decile of SBP variability was 2.18 (1.52–3.13) after adjusting for sex, age, treatment including other baseline variables, and average on-treatment SBP. A similar effect was observed for stroke (hazard ratio 2.78, 1.28–6.05), myocardial infarction (hazard ratio 4.11, 1.87–9.06), and heart failure (hazard ratio 4.79, 1.82–12.62). Highest SBP variability was also a predictor of post-trial fatal cardiovascular events. Increased visit-to-visit SBP variability was related to age, pulse pressure, changing physicians, smoking, treatment allocation, and the use of multiple BPlowering drugs. Conclusion: These findings suggest that reducing visit-tovisit SBP variability might be an important objective in addition to conventional blood pressure-lowering in elderly hypertensive patients. Keywords: blood pressure, blood pressure variability, cardiovascular events, elderly, hypertension Abbreviations: ANBP2 Second Australian National Blood Pressure, study; SBPV, systolic blood pressure variability
INTRODUCTION
H
igh SBP is the major risk factor for cardiovascular events such as heart attack and stroke [1]. A reduction in SBP decreases the risk of cardiovascular events, especially stroke [2]. Arterial blood pressure (BP) is inherently variable and in the clinical setting this may be observed when examining multiple indirect clinical BP measurements for an individual at a single clinic visit or between clinic visits, by noninvasive ambulatory blood pressure monitoring (ABPM) or with home measurement of BP. A degree of intra-individual SBP variability is expected; however, there is growing evidence that increased SBP variability is related to adverse cardiovascular outcomes [3–5]. Recent findings by Rothwell et al. [6] have shown that BP variability, especially SBP, independently of mean BP, strongly predicts the risk of a stroke, particularly in those aged less than 60 years. Similar findings were reported in a study among postmenopausal women [3]. These compelling findings raise the question whether SBP variability, in addition to absolute BP, should be a target for reducing cardiovascular risk. It is also important to confirm whether there is an age-dependence in this relationship. To date, little has been reported on SBP variability in the elderly and its impact on clinical outcomes. It is not clear what factors are associated with SBP variability in this age group, although earlier studies have shown SBP variability is positively correlated with age [6–8]. The observations related to the prognostic outcome of SBP variability have to date been driven by studies, which mainly including
Journal of Hypertension 2014, 32:525–533 a Centre of Cardiovascular Research and Education in Therapeutics, Department of Epidemiology and Preventive Medicine, Monash University, Melbourne, bDepartment of Clinical Pharmacology, School of Medicine, Flinders University, Adelaide and c Menzies Research Institute Tasmania, University of Tasmania, Hobart, Australia
Correspondence to Professor Christopher M. Reid, School of Public Health & Preventive Medicine, Monash University, 6th Floor, Alfred Centre, 99 Commercial Rd, Melbourne, VIC 3004, Australia. Tel: +61 3 99030752; fax: +61 3 9903 0556; e-mail: [email protected] Received 18 February 2013 Revised 31 July 2013 Accepted 26 September 2013 J Hypertens 31:525–533 ß 2014 Wolters Kluwer Health | Lippincott Williams & Wilkins. DOI:10.1097/HJH.0000000000000028
Journal of Hypertension
www.jhypertension.com
525
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Chowdhury et al.
younger and middle-aged patients [3,6]. Limited evidence is available on the long-term effect of visit-to-visit SBP variability on cardiovascular outcome among elderly patients treated for hypertension. We used data from the Second Australian National Blood Pressure (ANBP2) study to investigate the relationship between visit-to-visit SBP variability and cardiovascular events in this cohort of elderly patients with treated hypertension. Our aim was also to identify the prognostic value of higher SBP variability on cardiovascular disease mortality using extended follow-up data from the ANBP2 study. In addition, we sought to identify factors associated with increased visit-to-visit SBP variability in this elderly cohort.
METHODS Study design and participants The ANBP2 study was a prospective, randomized, openlabel trial with blinded assessment of endpoints. The study was designed to compare the outcome of treatment with either a diuretic or an angiotensin-converting enzyme inhibitor (ACE-I)-based BP-lowering drug regimen in hypertensive patients aged 65–84 years at study entry. The study was conducted from 1995 through 2001 in 1594 family medical practices in five Australian states. A total of 6083 hypertensive patients were enrolled and randomized to one of the two different BP-lowering regimens. Details about the study design, and recruitment processes have been published previously [9,10]. Study entry required an average untreated baseline SBP of at least 160 mmHg or at least 140 mmHg if the average baseline DBP was at least 90 mmHg. Following randomization, patients were followed for cardiovascular or other morbidity until study termination, or if the patient died or withdrew from the study. Family practices were requested to treat patients with randomized therapy, and additional BP-lowering medication, if required, by a recommended regimen aimed to minimize cross-over of treatment. Target SBPs were a reduction of at least 20 mmHg to below 160 mmHg, and, if tolerated, below 140 mmHg. Target DBPs were a reduction of at least 10 mmHg to below 90 mmHg and, if tolerated, below 80 mmHg.
Follow-up and endpoints In the ANBP2 clinical trial, the patients were followed for a median of 4.1 years. The main results of the ANBP2 study have been published showing the effect of ACE-I and diuretic-based regimens on cardiovascular outcomes [10]. In the current investigation, we defined four endpoints for the clinical trial phase, either fatal or nonfatal, for the evaluation of the following: first stroke; first myocardial infarction; first heart failure; and any first cardiovascular event including coronary, cerebrovascular, and other cardiovascular events. All study endpoints were adjudicated by an independent endpoint committee whose members, unaware of treatment-group assignments, reviewed the cases and confirmed potential endpoints. Once the ANBP2 clinical trial was completed, the longterm survival information of the cohort was determined for 526
www.jhypertension.com
the extended phase by linkage to the Australian Institute of Health and Welfare National Death Index, and information received from the patient and their family practice by questionnaire. We used the International Statistical Classification of Diseases and Related Health Problems 10th Revision (ICD10) coding of the National Death Index to identify the cause of death.
Blood pressure measurement and variability After at least 5 min rest, BP was measured by either family physicians or nurses in the seated posture at least three times for each patient during each visit using a mercury sphygmomanometer with an occluding cuff of appropriate size for the patient’s upper arm. The average of the second and third BP measurements in each patient was recorded as the BP for that particular visit. We defined the SBP variability of an individual as the SD of SBP between visits. On-treatment visit-to-visit SBP variability was based on all available BP records for an individual following randomization. Patients with at least two postrandomization visits with SBP recorded were included in the study. We excluded the follow-up visits that took place in the first 6 months following randomization and in patients who experienced an event within the first 6 months. The reason for exclusion of the first 6-month data was that we assumed that during that period the patient’s BP might be in a less than steady-state associated with the phase of treatment initiation and dose adjustment. We also calculated other measures of visit-to-visit SBP variability, such as variability independent of mean, co-efficient of variation, and average real variability. The variability independent of mean and co-efficient of variation were highly correlated with SD (r ¼ 0.99 and r ¼ 0.98, respectively) and therefore, not used for assessing predictive effect of variability on cardiovascular outcome.
Statistical analysis Statistical analyses were performed based on the intentionto-treat principle and using Stata version 11.2 for Windows (StataCorp LP, College Station, Texas, USA). We first used a number of Cox proportional-hazard models to explore the relationship between visit-to-visit SBP variability and study endpoints (e.g. any first cardiovascular events, stroke, myocardial infarction or heart failure) from the ANBP2 clinical trial. For the purpose of analysis, we split the visit-to-visit SBP variability into deciles combining both treatment groups, and calculated hazard ratios in relation to the first decile (i.e. the decile with the lowest SBP variability). Our first Cox proportional model (model 1) was adjusted for patient’s age at baseline (above or below 75 years), sex, and in-study treatment regimen, as earlier published analyses from the ANBP2 study had shown these variables to have strong associations with one or more endpoints. Model 1 was also adjusted for variables which were different (P < 0.1) between top and lowest decile of SBP variability at baseline (Table 1), such as patient’s residential location, education status, history of diabetes, history of previous coronary heart disease, BMI, SBP at randomization, and smoking status. In addition, model 1 was adjusted for clustering of patients within family
Volume 32 Number 3 March 2014
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Blood pressure variability in elderly TABLE 1. Distribution of key patient baseline characteristics by on-treatment visit-to-visit SBP variability Bottom decile of SBP variability
Top decile of SBP variability
N ¼ 5880 48.3 71.9 4.9 30.2 23.1 43.6 33.3 6.9 45.2 73.3 6.5 27.1 4.2 168 13 17.4 7.7 4.4 7.1 62.5 11.8
N ¼ 589 53.0 71.2 4.5 24.3 21.2 43.5 35.3 5.9 44.8 75.7 5.6 27.7 4.6 164 12 13.9 6.3 4.4 6.3 63.2 11.5
N ¼ 588 45.8 73.2 5.2 39.5 25.7 44.1 30.2 8.9 46.3 71.2 7.2 26.8 4.6 173 13 19.4 9.0 5.6 10.9 59.3 13.9
22.3 13.0 32.4
21.8 13.6 31.9
24.8 11.4 33.3
29.1 38.5
29.0 39.1
27.4 39.3
Total Male (%) Age [mean SD (years)] 75 years (%) Education
Tobacco use Alcohol use BMI (kg/m2) SBP at randomization (mmHg) Remoteness Previous medical conditions
Laboratory values
Physical activities (in 2 weeks prior randomization)
Primary (%) Some high school (%) Completed high school/university (%) Current (%) Previous (%) Current (%) Previous (%) (Mean SD) Regional area (%) Coronary heart disease (%) Cerebrovascular disease (%) History of diabetes (%) Antihypertensive therapy (%) Increased plasma creatinine (M >120 umol/l; F >110umol/l) (%) Raised total cholesterol (>6.5 mmol/l) (%) Low HDL (
Systolic blood pressure variability is an important predictor of cardiovascular outcomes in elderly hypertensive patients Enayet K. Chowdhury a, Alice Owen a, Henry Krum a, Lindon M.H. Wing b, Mark R. Nelson c, and Christopher M. Reid a, on behalf of the Second Australian National Blood Pressure Study Management Committee
Objectives: In hypertensive persons aged 60 years or below, visit-to-visit SBP variability is directly associated with cardiovascular events, especially stroke. It is unclear whether such a relationship exists for older persons. We investigated whether there is a relationship between visit-to-visit SBP variability and cardiovascular events in an elderly population, and identified the factors associated with increased SBP variability. Methods: Information from 49 771 visits of 5880 patients aged at least 65 years being treated for hypertension in the Second Australian National Blood Pressure study was used. Patients were followed for 4.1 (median) years and had eight (median) doctor visits during the study. SBP variability was defined as within-individual SD of SBP across study follow-up visits. Results: Increased visit-to-visit SBP variability was found to be a strong predictor for future cardiovascular events in this elderly population. The hazard ratio (95% confidence interval) for any first fatal/nonfatal cardiovascular event for highest decile compared with lowest decile of SBP variability was 2.18 (1.52–3.13) after adjusting for sex, age, treatment including other baseline variables, and average on-treatment SBP. A similar effect was observed for stroke (hazard ratio 2.78, 1.28–6.05), myocardial infarction (hazard ratio 4.11, 1.87–9.06), and heart failure (hazard ratio 4.79, 1.82–12.62). Highest SBP variability was also a predictor of post-trial fatal cardiovascular events. Increased visit-to-visit SBP variability was related to age, pulse pressure, changing physicians, smoking, treatment allocation, and the use of multiple BPlowering drugs. Conclusion: These findings suggest that reducing visit-tovisit SBP variability might be an important objective in addition to conventional blood pressure-lowering in elderly hypertensive patients. Keywords: blood pressure, blood pressure variability, cardiovascular events, elderly, hypertension Abbreviations: ANBP2 Second Australian National Blood Pressure, study; SBPV, systolic blood pressure variability
INTRODUCTION
H
igh SBP is the major risk factor for cardiovascular events such as heart attack and stroke [1]. A reduction in SBP decreases the risk of cardiovascular events, especially stroke [2]. Arterial blood pressure (BP) is inherently variable and in the clinical setting this may be observed when examining multiple indirect clinical BP measurements for an individual at a single clinic visit or between clinic visits, by noninvasive ambulatory blood pressure monitoring (ABPM) or with home measurement of BP. A degree of intra-individual SBP variability is expected; however, there is growing evidence that increased SBP variability is related to adverse cardiovascular outcomes [3–5]. Recent findings by Rothwell et al. [6] have shown that BP variability, especially SBP, independently of mean BP, strongly predicts the risk of a stroke, particularly in those aged less than 60 years. Similar findings were reported in a study among postmenopausal women [3]. These compelling findings raise the question whether SBP variability, in addition to absolute BP, should be a target for reducing cardiovascular risk. It is also important to confirm whether there is an age-dependence in this relationship. To date, little has been reported on SBP variability in the elderly and its impact on clinical outcomes. It is not clear what factors are associated with SBP variability in this age group, although earlier studies have shown SBP variability is positively correlated with age [6–8]. The observations related to the prognostic outcome of SBP variability have to date been driven by studies, which mainly including
Journal of Hypertension 2014, 32:525–533 a Centre of Cardiovascular Research and Education in Therapeutics, Department of Epidemiology and Preventive Medicine, Monash University, Melbourne, bDepartment of Clinical Pharmacology, School of Medicine, Flinders University, Adelaide and c Menzies Research Institute Tasmania, University of Tasmania, Hobart, Australia
Correspondence to Professor Christopher M. Reid, School of Public Health & Preventive Medicine, Monash University, 6th Floor, Alfred Centre, 99 Commercial Rd, Melbourne, VIC 3004, Australia. Tel: +61 3 99030752; fax: +61 3 9903 0556; e-mail: [email protected] Received 18 February 2013 Revised 31 July 2013 Accepted 26 September 2013 J Hypertens 31:525–533 ß 2014 Wolters Kluwer Health | Lippincott Williams & Wilkins. DOI:10.1097/HJH.0000000000000028
Journal of Hypertension
www.jhypertension.com
525
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Chowdhury et al.
younger and middle-aged patients [3,6]. Limited evidence is available on the long-term effect of visit-to-visit SBP variability on cardiovascular outcome among elderly patients treated for hypertension. We used data from the Second Australian National Blood Pressure (ANBP2) study to investigate the relationship between visit-to-visit SBP variability and cardiovascular events in this cohort of elderly patients with treated hypertension. Our aim was also to identify the prognostic value of higher SBP variability on cardiovascular disease mortality using extended follow-up data from the ANBP2 study. In addition, we sought to identify factors associated with increased visit-to-visit SBP variability in this elderly cohort.
METHODS Study design and participants The ANBP2 study was a prospective, randomized, openlabel trial with blinded assessment of endpoints. The study was designed to compare the outcome of treatment with either a diuretic or an angiotensin-converting enzyme inhibitor (ACE-I)-based BP-lowering drug regimen in hypertensive patients aged 65–84 years at study entry. The study was conducted from 1995 through 2001 in 1594 family medical practices in five Australian states. A total of 6083 hypertensive patients were enrolled and randomized to one of the two different BP-lowering regimens. Details about the study design, and recruitment processes have been published previously [9,10]. Study entry required an average untreated baseline SBP of at least 160 mmHg or at least 140 mmHg if the average baseline DBP was at least 90 mmHg. Following randomization, patients were followed for cardiovascular or other morbidity until study termination, or if the patient died or withdrew from the study. Family practices were requested to treat patients with randomized therapy, and additional BP-lowering medication, if required, by a recommended regimen aimed to minimize cross-over of treatment. Target SBPs were a reduction of at least 20 mmHg to below 160 mmHg, and, if tolerated, below 140 mmHg. Target DBPs were a reduction of at least 10 mmHg to below 90 mmHg and, if tolerated, below 80 mmHg.
Follow-up and endpoints In the ANBP2 clinical trial, the patients were followed for a median of 4.1 years. The main results of the ANBP2 study have been published showing the effect of ACE-I and diuretic-based regimens on cardiovascular outcomes [10]. In the current investigation, we defined four endpoints for the clinical trial phase, either fatal or nonfatal, for the evaluation of the following: first stroke; first myocardial infarction; first heart failure; and any first cardiovascular event including coronary, cerebrovascular, and other cardiovascular events. All study endpoints were adjudicated by an independent endpoint committee whose members, unaware of treatment-group assignments, reviewed the cases and confirmed potential endpoints. Once the ANBP2 clinical trial was completed, the longterm survival information of the cohort was determined for 526
www.jhypertension.com
the extended phase by linkage to the Australian Institute of Health and Welfare National Death Index, and information received from the patient and their family practice by questionnaire. We used the International Statistical Classification of Diseases and Related Health Problems 10th Revision (ICD10) coding of the National Death Index to identify the cause of death.
Blood pressure measurement and variability After at least 5 min rest, BP was measured by either family physicians or nurses in the seated posture at least three times for each patient during each visit using a mercury sphygmomanometer with an occluding cuff of appropriate size for the patient’s upper arm. The average of the second and third BP measurements in each patient was recorded as the BP for that particular visit. We defined the SBP variability of an individual as the SD of SBP between visits. On-treatment visit-to-visit SBP variability was based on all available BP records for an individual following randomization. Patients with at least two postrandomization visits with SBP recorded were included in the study. We excluded the follow-up visits that took place in the first 6 months following randomization and in patients who experienced an event within the first 6 months. The reason for exclusion of the first 6-month data was that we assumed that during that period the patient’s BP might be in a less than steady-state associated with the phase of treatment initiation and dose adjustment. We also calculated other measures of visit-to-visit SBP variability, such as variability independent of mean, co-efficient of variation, and average real variability. The variability independent of mean and co-efficient of variation were highly correlated with SD (r ¼ 0.99 and r ¼ 0.98, respectively) and therefore, not used for assessing predictive effect of variability on cardiovascular outcome.
Statistical analysis Statistical analyses were performed based on the intentionto-treat principle and using Stata version 11.2 for Windows (StataCorp LP, College Station, Texas, USA). We first used a number of Cox proportional-hazard models to explore the relationship between visit-to-visit SBP variability and study endpoints (e.g. any first cardiovascular events, stroke, myocardial infarction or heart failure) from the ANBP2 clinical trial. For the purpose of analysis, we split the visit-to-visit SBP variability into deciles combining both treatment groups, and calculated hazard ratios in relation to the first decile (i.e. the decile with the lowest SBP variability). Our first Cox proportional model (model 1) was adjusted for patient’s age at baseline (above or below 75 years), sex, and in-study treatment regimen, as earlier published analyses from the ANBP2 study had shown these variables to have strong associations with one or more endpoints. Model 1 was also adjusted for variables which were different (P < 0.1) between top and lowest decile of SBP variability at baseline (Table 1), such as patient’s residential location, education status, history of diabetes, history of previous coronary heart disease, BMI, SBP at randomization, and smoking status. In addition, model 1 was adjusted for clustering of patients within family
Volume 32 Number 3 March 2014
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Blood pressure variability in elderly TABLE 1. Distribution of key patient baseline characteristics by on-treatment visit-to-visit SBP variability Bottom decile of SBP variability
Top decile of SBP variability
N ¼ 5880 48.3 71.9 4.9 30.2 23.1 43.6 33.3 6.9 45.2 73.3 6.5 27.1 4.2 168 13 17.4 7.7 4.4 7.1 62.5 11.8
N ¼ 589 53.0 71.2 4.5 24.3 21.2 43.5 35.3 5.9 44.8 75.7 5.6 27.7 4.6 164 12 13.9 6.3 4.4 6.3 63.2 11.5
N ¼ 588 45.8 73.2 5.2 39.5 25.7 44.1 30.2 8.9 46.3 71.2 7.2 26.8 4.6 173 13 19.4 9.0 5.6 10.9 59.3 13.9
22.3 13.0 32.4
21.8 13.6 31.9
24.8 11.4 33.3
29.1 38.5
29.0 39.1
27.4 39.3
Total Male (%) Age [mean SD (years)] 75 years (%) Education
Tobacco use Alcohol use BMI (kg/m2) SBP at randomization (mmHg) Remoteness Previous medical conditions
Laboratory values
Physical activities (in 2 weeks prior randomization)
Primary (%) Some high school (%) Completed high school/university (%) Current (%) Previous (%) Current (%) Previous (%) (Mean SD) Regional area (%) Coronary heart disease (%) Cerebrovascular disease (%) History of diabetes (%) Antihypertensive therapy (%) Increased plasma creatinine (M >120 umol/l; F >110umol/l) (%) Raised total cholesterol (>6.5 mmol/l) (%) Low HDL (
