Clinical trials have provided convincing evidence that blood pressure (BP) lowering treatment reduces the risk of cardiovascular disease (CVD) and total mortality. The objective of this study was to examine the association of hypertension treatment, control, and BP indexes with all-cause and cardiovascular mortality among US adults with hypertension.
Persons aged ≥18 years from the Third National Health and Nutrition Examination Survey (NHANES III) were identified as hypertensives based on a BP ≥140/90mmHg or current treatment for hypertension. Vital status in 2006 was ascertained by passive follow-up using the National Death Index. Cox regression models were used to assess correlates of survival.
At baseline, 52% of hypertensive adults reported currently taking prescription medicine for high BP and 38% of treated persons had BP controlled. Compared to treated controlled hypertensives, treated uncontrolled hypertensives had a 1.57-fold (95% confidence interval (CI) 1.28–1.91) and 1.74-fold (95%CI 1.36–2.22) risk of all-cause and cardiovascular mortality; untreated hypertensives had a 1.34-fold (95%CI 1.12–1.62) and 1.37-fold (95%CI 1.04–1.81) risk of all-cause and cardiovascular mortality, respectively. The association persisted after further excluding persons with pre-existing hypertension comorbidities. Mortality risk was linearly increased with systolic BP (SBP), pulse pressure (PP), and mean arterial pressure (MBP), whereas diastolic BP (DBP) was not a significant predictor of cardiovascular mortality overall. No significant associations were observed between drug classes and mortality risk.
This study indicates that uncontrolled and untreated hypertension was associated with increased risk of total and cardiovascular mortality among the general hypertensive population.
Hypertension is a well established independent risk factor for cardiovascular disease (CVD). Although substantial mortality from the long-term sequelae of hypertension can be prevented through existing antihypertensive drug therapy and blood pressure (BP) control, almost 22% of US adults are still unaware of their hypertension, and 32% of the hypertensive population are not on treatment.1 As mortality from hypertension has increased over the past 10 years, and >70 million Americans are living with hypertension currently,2 it is of great interest to examine the relationship between hypertension treatment and control with cardiovascular outcomes and total mortality in the general hypertensive population.
A number of effective antihypertensive drugs, including diuretics, β-blockers, calcium channel blockers, angiotensin-converting enzyme inhibitors, and angiotensin receptor blockers are currently available for hypertension therapy.3 Meta-analyses of the multiple randomized clinical trials have shown that lowering BP by any drugs compared with placebo reduces CVD morbidity and mortality; all classes of drugs reduce total and CVD mortality equally; and different classes of drugs provide differing degrees of protection from CVD morbidities.4,5,6,7 However morbidity and mortality trials have been the target of growing dissatisfaction and criticism in the past few years.8,9 Because most trials have been carried out on selected groups of high-risk hypertensive patients with obvious differences in their cardiovascular risk profile as opposed to the general hypertensive population, the documented effects of hypertension treatment and control in a particular clinical trial may have less general validity than what is often assumed. Thus, population-based observational studies offer an alternative approach to evaluate outcomes in hypertension treatment and control.
In the present study, we use the Third National Health and Nutrition Examination Survey (NHANES III) and mortality follow-up through 2006 to examine the association of hypertension treatment and control status with health outcomes in terms of all-cause and CVD mortality among US adults with hypertension. We also evaluate the relationship of systolic BP (SBP), diastolic BP (DBP), pulse pressure (PP), mean arterial pressure (MBP), and specific antihypertensive treatment regimens with the risk of total and CVD mortality.
Study design and subjects. The NHANES III was a nationally representative health survey conducted between1988 and 1994. The survey used a multistage, stratified probability cluster design to select a representative sample of the civilian noninstitutionalized US population. Detailed descriptions of the survey plan, sampling, operation, response, and analytic guidelines have been specified elsewhere.10 The NHANES III protocol was approved by the National Center for Health Statistics institutional review board and written informed consent for data collection was obtained from all survey participants.
The personal interviews and physical examination of NHANES III participants provided baseline data for this study. The NHANES III–linked mortality file through 2006 was created to permit longitudinal study of NHANES III participants. Mortality status was based on the result of a probabilistic matching between NHANES III participants and the US National Death Index. Details about the probabilistic matching methodology are available elsewhere.11
The overall NHANES III response rate for the interview and health examination was 78%. Among 17,250 participants (≥18 years) who had BP measurements, 5,128 hypertensive persons were identified on the basis of a mean SBP ≥140mmHg, or a mean DBP ≥90mmHg, or self-reported current use of a prescribed medicine for high BP. After further exclusion of pregnant women, and those missing data on prescription medication, loss to follow-up, or lack of information on cause of death, a total of 5,086 hypertensive subjects remained available to form the final analytical sample.
Measurements. BP was measured with participants aged ≥5 years in the sitting position after 5min of rest by a trained health interviewer at home and by a physician at the mobile examination center. Appropriate cuff sizes were used for participants based on manufacturer's recommendations. A mean SBP and a mean DBP for each individual from up to six measurements (three on each occasion) were used to define hypertension and classify BP categories. PP was calculated as the difference between a mean SBP and a mean DBP. MAP was calculated as a mean DBP plus one-third of PP value. BP indexes were expressed as each 10mmHg increment in SBP, PP, and MAP, or each 5mmHg increment in DBP. Based on BP questionnaire and BP examination data, we further defined three hypertension categories. Treated controlled hypertensives were subjects who reported currently taking prescribed medicine for hypertension and had SBP <140mmHg and DBP <90mmHg (SBP <130mmHg and DBP <80mmHg if diabetes or chronic kidney disease (CKD) coexisted3). Treated uncontrolled hypertensives were subjects who reported currently taking prescribed medicine for hypertension and had SBP ≥140mmHg or DBP ≥90mmHg (SBP ≥130mmHg or DBP ≥80mmHg if diabetes or CKD coexisted). The remaining hypertensive persons were classified as untreated hypertensives.
Detailed prescription medication use information was collected through the household interview. Antihypertensive agents reported by survey participants were identified and grouped into five drug classes including diuretics, β-blockers, calcium channel blockers, angiotensin-converting enzyme inhibitors, and other antihypertensive agents (α1-blockers, central α2 agonists, direct vasodilators, and other centrally acting drugs).3 Angiotensin receptor blockers were not available before the 1995. Persons taking an antihypertensive medication with only one active ingredient were considered as receiving monotherapy. Those taking more than one active ingredient (either in one combination pill or in more than one different single pills) were classified as receiving polytherapy.
The interview questionnaires also provided data on age, gender, race/ethnicity, cigarette smoking, leisure time physical activities, a history of medically diagnosed hypertension, high cholesterol, diabetes, stroke, heart attack, or heart failure. Height and weight were measured with a standard protocol and used to compute body mass index (kg/m2). Persons with a body mass index ≥30 were classified as obese. Serum cholesterol, creatinine and urine albumin, and creatinine were measured according to standard methods.12 Hypercholesterolemia was defined as serum total cholesterol ≥240mg/dl or a self-report of currently taking a prescription drug for lowering cholesterol. Serum creatinine values were corrected according to the recommended guideline.12 Glomerular filtration rate was estimated using the modification of diet in renal disease equation, 175 × standardized Scr−1.154 × age−0.203 × 1.210 (if black) × 0.742 (if female).13 CKD was defined as either an estimated glomerular filtration rate below 60ml/min/1.73m2 or urinary albumin >200mg/g creatinine.3
We merged NHANES III baseline data with the NHANES III–linked mortality file through 31 December 2006. Underlying causes of death were classified according to the International Classification of Diseases 10th Revision, and codes I00 to I78 were considered as CVD causes of death.14 Persons who survived were administratively censored at 31 December 2006. Persons who died of non-CVD causes were considered censored at the date of death for CVD mortality analysis. Follow-up time for each person was calculated as the difference between the NHANES III examination date and the last known date alive or censored from the NHANES III mortality study.
Statistical analyses. Statistical analyses were conducted using SAS (SAS Institute, Cary, NC) and SUDAAN (RTI, Research Triangle Park, NC).15 Sample weights were used to account for differential probabilities of selection and the complex sample design and to obtain estimates representative of the noninstitutionalized US hypertensive population. Variance estimates were computed using the Taylor series linearization approximation approach.16 Statistical differences between hypertension treatment and control status for baseline characteristics of the study cohort were tested univariately at the 0.05 level using a t statistic with a Bonferroni adjustment for three comparisons (α = 0.05/3). Cox proportional hazard models were used to assess the risk of all-cause and CVD mortality associated with hypertension treatment and control status using treated controlled hypertensives as the reference group because a previous study indicated that the relative risk of CVD mortality for treated controlled hypertensives (1.15, 95% confidence interval (CI) 0.73–1.82) was not different from those without hypertension in the US general population.17 Validity of the proportional hazard assumption for each variable was assessed by adding to the model a time-dependent corresponding variable.18 As the proportional hazard assumption was violated across age and gender, we fit models separately for age and gender in addition to the hypertensive samples. Examining the interaction of covariates with hypertension treatment and control categories in predicting all-cause and CVD death did not indicate any significant modifiers beside age and diabetes. To eliminate potential bias due to pre-existing illness, we further fit model by exclusion of hypertensive individuals with diabetes, CKD, heart attack, heart failure, or stroke at baseline.
At baseline, 52% of hypertensive persons reported taking prescribed medicine for hypertension, 38% of treated hypertensives achieved the therapeutic goals, and 77% of untreated hypertensives were undiagnosed or unaware of their hypertension (data not shown). Table 1 shows distributions of demographic characteristics, risk factors, comorbidities, and BP levels by hypertension treatment and control status. Compared with treated controlled hypertensives, treated uncontrolled hypertensives were older, had higher BP values and higher prevalence of comorbidities; untreated hypertensives were more likely to be men and Mexican American, had higher BP values but lower prevalence of hypercholesterolemia. The demographics, risk factors, comorbidities, and BP values were also different between the treated uncontrolled and untreated hypertensive groups.
During 58,130 person-years of follow-up (mean 11.4 years, maximum 18.1 years), there were a total of 2,521 deaths from all causes and 1,223 deaths from CVD, respectively. The patterns of hazard ratio variations using the fully adjusted model are similar to those using the partially adjusted model (Table 2). In the fully adjusted model, relative to treated controlled hypertension, treated uncontrolled hypertension was associated with a 57 and 74% increased risk of all-cause and CVD mortality, and untreated hypertension was associated with a 34 and 37% increased risk of all-cause and CVD mortality, respectively. The age and gender stratified analyses showed similar patterns, whereas the difference in CVD mortality between untreated and treated controlled hypertension did not reach statistical significant. However, after further exclusion of those with major comorbidities at baseline (diabetes, CKD, heart attack, heart failure, or stroke), untreated hypertensives had a 71% (95%CI 1.12–2.60) increased risk of CVD mortality compared with treated controlled hypertensives.
The individual contributions of SBP, DBP, PP, and MAP in predicting mortality are shown in Table 3. Both all-cause and CVD-related mortality risk progressively increased with each 10mmHg increment in SBP, PP, or MAP, in the overall analysis and in most subgroup analyses. These associations persisted even after the exclusion of pre-existing comorbidities. However, PP was not a significant predictor for the younger hypertensive population (fully adjusted model), and MAP was not a significant predictor of all-cause death among men. DBP was only a significant predictor of all-cause mortality for the younger hypertensive population and among women. Considering the joint effect of SBP and DBP, multivariable adjusted hazard ratios of CVD mortality were 3.83, 2.88, 2.39, and 1.98 for hypertensive subjects with either one of SBP/DBP values ≥160/100, 140–159/90–99, 130–139/85–89, and 120–129/80–84mmHg, respectively, compared to those with SBP <120mmHg and DBP <80mmHg (Ptrend < 0.01, data not shown).
The relationships of different antihypertensive drug classes with all-cause and CVD mortality are shown in Table 4. In the partially adjusted model, hypertensive persons receiving monotherapy with a calcium channel blocker had a significantly greater risk of all-cause mortality compared to those receiving monotherapy with a diuretic. However, the hazard ratio reduced to a nonsignificant 1.14 (95%CI 0.90–1.46) in the fully adjusted model. Among hypertensive persons receiving polytherapy, no statistically significant associations were observed between drug classes of any two-drug combinations and either all-cause or CVD-caused mortality. Furthermore, exclusion of hypertensive individuals with pre-existing illnesses did not show any significant associations either in the partially and the fully adjusted models (data not shown).
The primary findings from this large, nationally representative, population-based study indicate that both all-cause mortality and mortality from CVD were associated with hypertension treatment and control status. Specifically, relative to treated controlled hypertensives, treated uncontrolled hypertensives had a 57 and 74% greater risk of all-cause and CVD mortality and untreated hypertensives had a 34 and 37% greater risk of all-cause and CVD mortality, respectively. The association persisted and was significant after excluding subjects with hypertension comorbidities at baseline.
Large randomized clinical trials have provided strong evidence that BP-lowering treatment reduces the risk of overall mortality, stroke, coronary heart disease, heart failure, and CKD.3,4,5,6,7 However, some population-based studies have shown an increased risk of CVD mortality among treated hypertensive persons compared with untreated hypertensive persons.19,20 It would be more appropriate to use treated controlled hypertensive persons as the reference group to evaluate the risk of health outcomes associated with hypertension treatment and control status. Our findings emphasize the important role of BP control with antihypertensive treatment in reducing the risk of total and CVD mortality. Compared to clinical trials, the treated controlled rate was substantially lower in the US hypertensive population. Only 38% of treated persons had BP levels at therapeutic goals defined as SBP <140mmHg and DBP <90mmHg (SBP <130mmHg and DBP <80mmHg if diabetes or CKD coexisted) in the NHANES 1988–1994.
The current BP control rates remain poor in the US hypertensive population especially in the older persons and those with coexisting medical conditions.21,22 Untreated hypertension presents another Public Health challenge as it affects a large proportion of the general hypertensive population. In our study, >75% of untreated hypertensive persons did not have pre-existing diabetes, CKD, heart attack, heart failure, or stroke. However, they are still at greater risk of all-cause and CVD mortality, mainly due to higher BP values compared with their treated controlled counterparts.
The relationship between BP and cardiovascular risk has long been recognized from large randomized clinical trials and epidemiological studies. However, there has been an ongoing debate regarding the relative importance of various components of BP in predicting CVD.23,24 Our findings support the prevailing belief that SBP is the most consistent and significant predictor of CVD death.25,26 The importance of other BP parameters in predicting CVD mortality may depend on the study population. For example, the Framingham heart study showed that PP was superior to SBP or DBP in predicting coronary heart disease risk in normotensive and untreated hypertensive middle-aged and elderly adults.27 However, PP and MAP are not routinely used in clinical practice. Taking into account both SBP and DBP levels, hypertensive persons with SBP/DBP ≥120/80mmHg had almost 2.0-fold increased risk of CVD death compare to those with SBP <120mmHg and DBP <80mmHg.
The relative impact of different drug treatment regimens on cardiovascular events and mortality is another controversial issue.28,29,30 Results from the Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial indicated that diuretics were equal or superior to other antihypertensive agents in preventing one or more major forms of CVD.31,32 In 2003, two meta-analyses of multiple randomized controlled trials both came to the conclusion that all classes of antihypertensive agents reduced all-cause and CVD-related mortality equally.4,5 The largest meta-analysis of randomized trials published in 2009 further confirms the benefit of cardiovascular protection with the five major classes of BP-lowering drugs.33 In contrast, the Women's Health Initiative Observation Study found that women receiving monotherapy with calcium channel blockers, compared to those receiving monotherapy with diuretics, were more likely to experience CVD mortality.34 Tardif et al. reported an angiotensin-converting enzyme inhibitor–based treatment strategy associated with reduced mortality from all-cause, CVD, and coronary heart disease compared with a calcium channel blocker–based strategy among hypertensives.35 The differences between observational studies and clinical trials of mortality in relation to antihypertensive drug classes might be partially due to different risk profiles of hypertensive persons at baseline. In addition, the difference might be due to the fact that in observational studies, current antihypertensive medication use is often the result of past successes or failures with other drug treatment regimens. The findings of our US hypertensive population data are comparable with the results of previously reported clinical trials and support the notion that control BP is more important than the type of drug selection.
The major strength of this study is the use of the NHANES data to provide population-based estimates of the association of hypertension treatment, control, and BP parameters with CVD mortality and premature death. The method of BP measurement in our study is fully consistent with the national standards. An average of six BP readings obtained on two separate occasions were used to define hypertension and classify BP categories. However, several limitations should be taken into consideration when interpreting our findings. First, the cross-sectional design of our observational study and the passive mortality data linkage approach limited our ability to update exposure and covariate status. BP readings here are just being taken as one point in time (the baseline), and may not be entirely representative of BP ranges that prevailed over the hypertensive person's lifetime. Also, the total length of duration of hypertension could not be controlled. Assuming it is likely that untreated, or uncontrolled hypertensive persons may become treated, or controlled during the follow-up period, the hazard ratio for mortality for untreated or uncontrolled hypertension is probably underestimated in the present study, and in general our hazard ratio estimates are biased toward the null. Second, possible bias arising from missing data in multivariable adjusted analyses can not be ruled out. Excluded hypertensive subjects were more likely to be non-Hispanic blacks, less likely to be former smokers, had a higher proportion of sedentary lifestyles, and a lower proportion of diabetes than those included in the fully adjusted model. However, these differences should have little impact on hazard ratio estimates considering the fact that excluded data are equally distributed in terms of age, gender, hypertension treatment and control status, BP levels, drug classes, monotherapy or polytherapy, all-cause mortality, and CVD deaths.
Although it is well established that hypertension is a strong, independent risk factor for stroke and coronary heart disease, the impact of hypertension treatment, control, and of specific BP indexes on the total and CVD mortality are less completely addressed in the general hypertensive population. Using the NHANES III and mortality follow-up data, our study found that compared to the treated controlled hypertensive population, both the treated uncontrolled and untreated hypertensive populations had an increased risk of both all-cause and CVD mortality, especially among those without comorbidities. SBP was the best independent predictor of all-cause and CVD mortality. This highlights the importance of achieving the SBP treatment goals as recommended by current national guidelines. We believe that our findings may have significant implications for both Public Health and clinical intervention programs to reduce excess mortality risk through hypertensive treatment and control among the high-risk hypertensive population.
The authors declared no conflict of interest.