J Manipulative Physiol Ther 2002 (May); 25 (4): 221–239 ~ FULL TEXT
Gregory Plaugher, DC, Cynthia R. Long, PhD, Alyssa D. Silveus,
Herbert Wood, DC, Kapildeo Lotun, MD, J. Michael Menke, DC,
William C. Meeker, DC, MPH, Stephen H. Rowe, DC
Director of Research,
Life Chiropractic College West,
25001 Industrial Boulevard,
Hayward, CA 94545, USA
OBJECTIVE: To determine the feasibility of conducting a randomized clinical trial in the private practice setting examining short- and long-term effects of chiropractic adjustments for subjects with essential hypertension compared with a brief soft tissue massage, as well as a nontreatment control group.
DESIGN: Randomized controlled-comparison trial with 3 parallel groups.
SETTING: Private practice outpatient chiropractic clinic.
PATIENTS: Twenty-three subjects, aged 24 to 50 years with systolic or diastolic essential hypertension.
INTERVENTIONS: Two months of full-spine chiropractic care (ie, Gonstead) consisting primarily of specific-contact, short-lever-arm adjustments delivered at motion segments exhibiting signs of subluxation. The massage group had a brief effleurage procedure delivered at localized regions of the spine believed to be exhibiting signs of subluxation. The nontreatment control group rested alone for a period of approximately 5 minutes in an adjustment room.
MAIN OUTCOME MEASURES: Cost per enrolled subject, as well as systolic and diastolic blood pressure (BP) measured with a random-0 sphygmomanometer and patient reported health status (SF-36). Pilot study outcome measures also included an assessment of cooperation of subjects to randomization procedures and drop-out rates, recruitment effectiveness, analysis of temporal stability of BPs at the beginning of care, and the effects of inclusion/exclusion criteria on the subject pool.
RESULTS: Thirty subjects enrolled, yielding a cost of $161 per enrolled subject. One subject was later determined to be ineligible, and 6 others dropped out. In both the chiropractic and massage therapy groups, all subjects were classified as either overweight or obese; in the control group there were only 2 classified as such. SF-36 profiles for the groups were similar to that of a normal population. The mean change in diastolic BP was -4 (95% confidence interval [CI]: -8.6, 0.5) in the chiropractic care group, 0.5 (95% CI: -3.5, 4.5) in the brief massage treatment group, and -4.9 (95% CI: -9.7, -0.1) in the no treatment control group. At the end of the study period, this change was -6.3 (95% CI: 13.1, 0.4), -1.0 (95% CI: -7.5, 15.6), -7.2 (95% CI: -13.3, -1.1) in the 3 study groups. The mean improvements in the chiropractic care and no treatment control groups remained consistent over the follow-up period.
CONCLUSIONS: This pilot study elucidated several procedural issues that should be addressed before undertaking a full-scale clinical trial on the effects of chiropractic adjustments in patients with essential hypertension. A multidisciplinary approach to recruitment may need to be used in any future efforts because of the limited subject pool of patients who have hypertensive disease but are not taking medications for its control. Measures need to be used to assure comparable groups regarding prognostic variables such as weight. Studies such as these demonstrate the feasibility of conducting a full-scale 3-group randomized clinical trial in the private practice setting.
From the FULL TEXT Article
Our inquiry into this line of investigation is to determine whether specific chiropractic adjustments, applied at sites of subluxation have any effect on patients with concomitant hypertensive disease. The study represents the first phase of a systematic research program that will examine the efficacy of specific chiropractic spinal and sacroiliac adjustments delivered at sites of subluxation in patients with essential hypertension. Because BP is but 1 limited physiological parameter correlated with health and disease, a global assessment of the subject's quality of life was also included.
Initially, it is important that the basic effect, if any, of adjustments on elevated BP, be evaluated validly. For example, previous studies have attempted to analyze this effect under unrealistic treatment conditions in which subjects were given 1 adjustment or manipulation or a very short course of care. This study tracked patients over 2 months to analyze the cumulative effect of adjustive care. The accurate estimation of the effect of spinal adjustments on hypertension might serve as the basis for more in depth clinical and basic science studies concerning underlying physiologic mechanisms. Thus further research would attempt to measure concomitant physiological factors such as plasma catecholamines and aldosterone levels, and how these factors are influenced by chiropractic adjustments in the care of patients with essential hypertension. Noninvasive measurement of the autonomic nervous system is also possible with the orthostatic test, heart rate variability measurement, and cold pressor test. 
This study has addressed the feasibility of applying the procedures inherent in a randomized clinical trial to a practice-based setting. Previous studies involving hypertension have relied on student populations or used less rigorous, and less demanding, research designs. This study has examined the implementation of blinded assessment and a randomization process for subject allocation to treatment/care groups, design features essential to unbiased estimates of treatment effects.
The great majority of patients with hypertension (approximately 90%) have a disease of unknown cause. Idiopathic hypertension, termed essential or primary, is likely a multifactorial disease involving abnormalities of regulatory mechanisms normally associated with the control of systemic vascular resistance, blood volume, sodium excretion, and cardiac output.  Findings from biochemical, electrophysiological, pharmacologic, and hemodynamic studies support the existence of sympathetic activation in primary human hypertension. [11, 70, 72] Although stress, by itself, may not directly cause hypertensive disease, it can lead to repeated BP elevations, which may eventually lead to hypertension.  The role of the sympathetic nervous system in the development of hypertension is seen mostly in younger patients with borderline hypertension.  Genetic  or predisposing factors, susceptibility, and innate capability all play a role in this multifactorial disease.
Although many pharmacologic treatments have their basis in modulation of the autonomic nervous system, little is understood about the nervous system and its role in the induction of hypertensive disease.  Total plasma concentrations of catecholamines (eg, norepinephrine and epinephrine) in patients with hypertension appear to be higher than in individuals with normal BP. [72, 77–79] Cardiac sympathetic activity is increased in patients with hypertension.  Both parasympathetic and sympathetic autonomic dysfunction exist in children with essential hypertension. 
Experimental decreases in parasympathetic activity lower BP in patients with hypertension.  Decreases in parasympathetic activity are demonstrated in those patients with sustained essential hypertension. 
Plasma renin activity also appears to correlate with diastolic pressure and may serve as a predictor for the likelihood of success of various hypertension treatments.  The renin-angiotensin-aldosterone system regulates sodium balance and thus fluid volume and ultimately arterial pressure. Angiotensin (a vasoconstrictor agent) stimulates aldosterone secretion. Aldosterone's effect on the kidneys causes increased sodium retention, which then increases blood volume. It has been hypothesized that the interaction between angiotensin and the nervous system produces a differential activation of sympathetic outflow that spares the kidney.  It has also been proposed that the kidney malfunction in the renin-angiotensin-aldosterone system is preceded by sympathetic overactivity, which damages the peritubular capillaries resulting in salt-dependent hypertension.  One study attempted to determine the influence of chiropractic care on aldosterone levels. Wagnon et al  studied the effects of specific chiropractic full-spine adjustments (short-lever, manual-force, specific-contact procedure) on serum levels of aldosterone. In their group of patients with mild hypertension, a decrease in serum aldosterone levels were observed after chiropractic care. Although BP readings were lowered (exact data not specified) after each therapeutic intervention, lasting changes in BP were not obtained. The authors speculated that this could have been the result of the short time course of the study (ie, approximately 10 visits over a 1-month period).
To maintain homeostasis, the nervous system organization can be grouped into 3 general components: (1) an afferent system, (2) a control center, and (3) an efferent system.  We will examine these 3 areas, as well as the abnormalities associated with them leading to hypertension.
The afferent system
Afferent neurons with sensory endings in various receptor regions have central projections to the brain. Impulses from receptors travel via glossopharyngeal and vagal afferent nerve fibers (whose cell bodies are located in the petrosal and nodose ganglia, respectively) terminate in the brain in structures such as the nucleus tractus solitarius (NTS). The NTS also receives input by means of afferent nerve fibers from cranial nervesV, VII, and VIII, as well as from the supraoptic nuclei and paraventricular nuclei of the hypothalamus. The NTS neural activity is therefore modulated by inputs from sensory receptors. These receptors reside in the cardiovascular system such as from arterial baroreceptors, from cardiopulmonary located receptors and chemoreceptors, from somatic structures in skeletal muscle and skin and from receptors located in viscera such asfrom the liver and kidneys. These impulses monitor arterial pressure, blood gases, and blood flow to the various organs.
In the patients with hypertension, the arterial baroreceptors have been found to have a higher-pressure threshold and reduced sensitivity to pressure increases. This results in a decreased sensory input to the central nervous system and increased sympathetic activity. Mechanisms leading to baroreceptor disease may occur as a result of baroreceptor membrane defect, increased membrane sensitivity to cations, variations in neurotransmitter release (ie, angiotensin, leucine enkephalin, substance P) that suppress baroreflexes from other neuronal structures, and greater number of unmyelinated fibers from baroreceptors. Unmyelinated fibers have a higher threshold, irregular firing, and less input to the central nervous system. [87–97]
Cardiac baroreceptors are also important in that they may modulate sympathetic tone and neurohumoral drive to the circulation. [98–101] In particular, neurohumoral drive is inhibited when patients with borderline hypertension are placed in the supine position by causing cardiac distension and receptor activation. Finally, there are studies that point to renal afferents affecting sympathetic tone (rather than via water and sodium handling) resulting in renal and mesenteric vasoconstriction. [102–103]
The control center
Neural structures of sympathetic and parasympathetic neurons in the medulla and spinal cord as wells as other central nervous system structures such as the brain stem, the forebrain and the hypothalamus make appropriate responses to maintain homeostasis such as proper BP.
Animal studies indicate specific central nervous system sites, such as the structures surrounding the third ventricle, in the expression of experimental hypertension. These structures, including the paraventricular nucleus, have connections to the medulla, which controls sympathetic outflow to the peripheral nervous system.  Lesions of the anteroventral portion of the third ventricle in the anterior hypothalamus reverse, prevent or delay hypertension. [105–110] Lesions to the NTS in cats and dogs  result in chronic labile or sustained hypertension as a result of loss of arterial baroreceptor reflexes. Surgically induced lesions or the injection of 6-hydroxydopamine to the NTS results in labile or sustained hypertension. [112, 113] Stimulation studies of the paraventricular and supraoptic nuclei in the hypothalamus indicate that neurons of the NTS may be modulated by these structures and therefore play a role in hypertension. 
The efferent system
Neurons in the NTS that send axons to the hypothalamus and to the preganglionic vagal and sympathetic efferent neurons. Vagal and sympathetic efferent neurons in the medulla and spinal cord that innervate the heart and blood vessels through postganglionic fibers. More specifically, efferent fibers from the NTS project to 3 main groups of neurons. They are the vagal (nucleus ambiguus, dorsal motor vagal nucleus) and sympathetic preganglionic nuclei (intermediolateral nucleus in the spinal cord); other brainstem nuclei such as the nucleus ceruleus, the parabrachial nucleus and the reticular formation, and the hypothalamus and amygdala. Increased arterial pressure resulting from vasoconstrictor response to sympathetic overactivity is implicated in several models of hypertension. This may be the result of increased neuroepinephrine release or an augmented contractile response of hypertension.  More specifically, there may be membrane defects in muscle membranes contributing to increased arterial vasoconstriction. [114–116] With respect to prejunctional and postjunctional receptors, there are findings of an increased vasoconstrictor response due to reduced prejunctional alpha-2 receptors [117–119] and increased postjunctional alpha-1. At adrenergic terminals, sympathetic overactivity may facilitate increased norepinephrinerelease suppressing the sodium-potassium pump. [120, 121] Also, sympathetic overactivity may increase vascular resistance through the following mechanisms. First, released norepinephrine in, for example, splanchnic, renal, muscular, and cutaneous vascular beds may result in sustained arteriolar constriction. Second, increased sympathetic output to the heart results in increased heart rate, stroke volume, and therefore cardiac output. This in turn results in arterial hypertrophy or autoregulatory vasoconstriction and increased vascular resistance. Third, sympathetic outflow to the kidney causes sodium and water retention, as well as activation of the renin-angiotensin pathway. The result is arteriolar constriction, increased vascular stiffness resulting in increased vascular resistance. Fourth, sympathetic nervous system activation results in glomerular hyperfiltration, which contributes to or triggers the development of hypertension ; fifth, sympathetic overactivity results in arteriolar muscle membrane defect resulting in abnormal ion transport and an exaggerated vasoconstriction response, as well as the synthesis of contractile proteins in the vascular muscles. This would also result in enhanced vasoconstriction. [123–127]
Kuwahara et al  studied sympathetic activity in 49 patients with essential hypertension and 17 control subjects with normal BP. They concluded that in those patients with essential hypertension and cardiac hypertrophy, there was direct physiological evidence of impaired sympathetic innervation in the inferior and lateral regions of the left ventricle.
Crawford et al  have reviewed the basis for chiropractic adjustments of subluxations in the management of patients with hypertension. They have primarily focused on the role of adjustment/manipulation in altering abnormalities of sympathetic tone. The facilitated spinal segment has been described by Korr  as a possible precursor to a variety of pathologic visceral states, including hypertension. Adjustments may be able to modulate general sympathetic nervous system activity if directed at segments exhibiting signs of subluxation.  Harris and Wagnon  studied the effects of specific chiropractic adjustments at sites of subluxation on distal skin temperature in 196 patients attending a chiropractic college teaching clinic. Certain regions of the spine had a nonsympathetic or vasodilatory effect on the cutaneous blood vessels, whereas other areas caused sympathetic changes (ie, vasoconstriction). That adjustments applied to different vertebrae of the spine may have vastly different effects in terms of BP is further supported by Tran and Kirby. [131–132] They have documented BP increases in healthy individuals with normal BP after adjustments.
Wagnon and Rupert et al  determined, in a retrospective analysis of 467 patients with hypertension and 467 patients with normal BP, that certain locations of spinal lesions had greater prevalence in patients with hypertension. Both osteopathic [134–135] and chiropractic clinicians [43, 47, 136] have confirmed the seeming predilection for dysfunction in specific spinal regions in hypertensive patients. Crawford et al,  in their review of the literature, cite primarily 3 spinal regions that appear to be related to hypertension. They are the upper thoracic region, particularly the T2-T3 motion segment, the upper cervical area (C0-C2), and the lower thoracic region (ie, T11-T12).
Johnston et al  studied changes in the presence of a segmental dysfunction pattern (ie, subluxation) longitudinally over four and eight months. The C6-T2-T6 pattern occurred with more frequency in hypertensives compared to normal subjects. Sixty-one subjects with the C6-T2-T6 pattern were followed up from 3 to 10 years.138 The C6-T2-T6 pattern persisted in 16 of 16 subjects with grade 2 or greater hypertension and 4 of 9 individuals with normal BP who had shown the pattern initially.
Brainstem compression at the ventrolateral or lateral medulla oblongata can be produced by pulsatile compression from the vertebral or posterior inferior cerebellar artery. This disorder has been linked to subjects with essential hypertension. [139–143]
Jannetta et al  reported in 1985 on a group of 53 patients with hypertension. Forty-two underwent operation for decompression of the medulla. Relief of the hypertension was seen in 32 patients and improvement in 4.
Levy et al  report on 12 patients with medically intractable (refractory) hypertension with a median follow-up period of 4.1 years after surgery for relief of the compression at the medulla. Ten of the 12 patients experienced reductions in systolic pressure greater than 20 mm Hg. Of these 10, results lasted for 6 months in 2 patients. Seven of 8 patients experienced improvement in BP lability and autonomic dysreflexia. Five patients showed sustained improvements.
Naraghi et al  report on pathomorphologic findings from the autopsies of 24 patients with essential hypertension. All of the patients exhibited a definitive neurovascular compression of the left root entry zone of cranial nerves IX and X. In the control subjects (n = 21) and patients with renal hypertension (n = 10), no findings were noted. Naraghi et al  studied this phenomenon with magnetic resonance imaging (MRI) in 24 patients with essential hypertension and in 14 control subjects. Twenty patients with essential hypertension had magnetic tomographic evidence of left-sided neurovascular compression at the ventrolateral medulla.
Morimoto et al  report on a case series of 21 patients with essential hypertension examined with MRI. Ten patients with secondary hypertension and 18 individuals with normal BP served as control subjects. Neurovascular compression of the medulla was seen in 75% of the patients with essential hypertension. This was detected in 1 patient (10%) in the group with secondary hypertension and 2 of the 18 (11%) of the subjects with normal BP. According to Colon et al,  who do not deny the existence of the disorder, certain imaging practices such as thin-slice MRI may overestimate its prevalence. In 1 recent study it was detectable with MRI in many symptom-free (ie, with normal BP) individuals. 
Subluxation or sprain  of the cervical spine is a relatively common injury to this region. Especially if one considers a wide continuum of subluxation or sprain from a mild plastic deformation because of postural habits, up to a grade III or complete rupture of the ligamentous elements. There is evidence that subluxation of cervical vertebrae can produce plastic stretch and other injuries to the vertebral arteries. [148–149] Severe subluxation and fracture-subluxations can also produce acute (sometimes life-threatening) injuries of the vertebral arteries. [150–152] They occurred in 46% of patients in one study. 
Does some smaller effect exist on the continuum of sprain injury to the neck or its sequelae to osteoarthritis, where repetitive asymmetrical movements produce a stretching effect on the vertebral arteries? Does this combine with disc degeneration to cause the longer vertebral artery (relative to a shorter spinal canal) to loop and press on the ventrolateral medulla oblongata activating the sympathetic nervous system? Is subluxation and subluxation-degeneration or sprain injury a marker for abnormal looping of the vertebral artery? Further studies are required to answer these questions.
A pilot study is essential to planning an adequate full-scale trial since a number of critical issues need to be resolved before time and significant funding is committed. This study was conducted in a chiropractic private practice setting and in a partial way thus examined the willingness and compliance of patients in the care of a condition/concern not strongly associated with chiropractic at present.
Previous studies have only evaluated short-term treatment protocols. In contrast, this study examined an extended period of treatment (ie, 2 months of intervention) to evaluate both short- and long-term effects.
The pilot study answered several specific questions. Will subjects allow themselves to be randomly assigned to treatment groups considering that one is not chiropractic (ie, massage) and the other suspends treatment for at least 2 months? Although most subjects agreed to be randomized, 2 of the subjects essentially rejected their study group assignment early on because of their desire for immediate chiropractic care. The time demand placed on subjects in comparison to their remuneration needs to be considered. There appeared to be higher dropout rates in the 2 control groups, which should be accounted for in any full-scale implementation. Paying subjects more for trials where they might participate as a no-treatment control for 4 to 6 months seems reasonable.
Do subject recruitment strategies (newspaper advertisements; radio spots; patient and doctor referrals) generate a sufficient flow of subjects to allow the timely completion of a full-scale trial? The use of radio advertisements appeared to influence subject recruitment the most. No referrals were received from other practitioners such as chiropractic or medical doctors. Future efforts should attempt to augment the influence of professionals in other disciplines (ie, medicine) in obtaining a sufficient subject pool. Subjects taking antihypertensive medications would have their drugs reduced and eventually eliminated before group randomization, and this would necessitate a medical professional in a trial with this design aspect. The potential pool of subjects would have significantly increased if we had chosen to admit subjects on high BP medications and this should be considered in large-scale clinical trials. Medical doctors would have to co-manage the patient by withdrawing medication from them during the study or run-in period. The radio spots made note that patients not taking medications were needed for the trial. The other major factor that limited patients was the smoking restriction. The age restriction in this study was also probably unnecessary.
Even though random assignment to study group was used in this pilot study, all subjects in both the chiropractic care and brief massage treatment groups were classified as overweight or obese, whereas only 2 subjects in the control group were classified as such. Unfortunately, such an imbalance on a possibly important prognostic factor can happen in small studies even with random assignment in place. In planning for a full-scale trial, it may be important to have a balance of overweight and normal weight subjects among the study groups. This could be achieved through the use of stratification on this factor; however, given the small number of normal weight subjects recruited in this study, we may be unable to fill the normal weight stratum. Rather than risk this, we propose to determine study group allocation through a dynamic randomization scheme. Adaptive computer-generated randomization will be used to minimize group differences on overweight status, as well as other possibly prognostic variables. [153, 154]
The pilot study generated estimates of the effect size and its variability, and these estimates will be used in the power analysis calculations to determine sample size requirements for a full-scale clinical trial. A major application in the conduct of a pilot study in this area is to derive these estimates, because little, if any, previous data exist for this purpose.
A 1-week run-in period may be necessary before randomization to exclude subjects who may only be exhibiting white-coat hypertension. A substantial drop in pressures occurred in the first week with the no-treatment/control group, perhaps reflecting regression to the mean.
Although exercise, dietary, and drug questionnaires were administered, alcohol use was not specifically queried. Alcohol consumption, especially heavy drinking, has a substantially additive impact to hypertension.  This factor, although likely controlled for in a large-scale randomized study would be an important addition to the questionnaire we used.
Because the administration of the study involved one room dedicated to a blinded assessor and an approximately 1 m2 floor space at the front desk for subjects' records and schedules, the physical impact of the study on the regular private practice was minimized. A chair was available for the project coordinator nearby.
We initially had attempted to use the private practitioner's office staff when scheduling appointments for the research subjects. This proved to be too cumbersome for the front desk because the normal patient flow of the office was already quite busy. There were also certain research requirements for the study participants, which would have involved additional training of the office's staff. Because of these developments, in the first week of the study, we changed to a scheduling procedure separate from the front desk. The project coordinator then checked with the front desk appointment book to determine when to schedule a research subject. During the initial phase of the study, when recruitment was at its maximum, this process involved nearly 30 subjects. The volume impact was the equivalent of adding approximately 30 new patients immediately to the doctor's practice.
This study demonstrates the feasibility of conducting a full-scale 3-group randomized clinical trial of chiropractic care in patients with subluxations and hypertension in a solo private practice setting. In 1989 Keating and Smallie  proposed a similar practice-based approach to clinical trials as a means of augmenting the productivity of clinical trials beyond what currently can be accomplished in college-based environments. However, the current study was more a lesson of bringing the clinical trial to the practice environment, rather than the model of setting up a clinic whose primary purpose is research.
Trial risk management
The risks associated with participating in a trial of chiropractic care in patients with hypertension were thought to be minimal at the onset. No adverse events or complications occurred in any of the study participants, nor has the care of patients with hypertension resulted in the reporting of adverse effects in case reports. Potential risks from this study primarily revolve around the care administered. There are no known risks for a patient receiving massage of the lumbar and thoracic spinal musculature. There are small risks associated with a patient receiving lumbar manipulation. Complications (ie, cauda equina syndrome) have been noted in less than 1 out of every 100,000,000 manipulations.157 Risks of cerebrovascular accidents after cervical manipulations have been recorded.  These risks are primarily associated with rotational manipulations of the cervical spine, with the risk being approximately 1.46 per 1 million manipulations.  The adjustments applied to the cervical spine in this pilot study did not involve rotation (ie, >30 degrees Y-axis rotation). This would have theoretically lessened the risks for cerebrovascular accidents, although the influence of rotational techniques is still debated in the literature. [158, 160]
All patients undergoing spinal adjustments received a comprehensive radiologic assessment of the regions of the spine to be adjusted. This may have lessened the chances of a subject receiving an adjustment in an inappropriate area and also disclosed contraindications to manipulation, such as areas with significant vertebral anomaly or areas of bone destruction resulting from fracture or metastatic disease. The radiographic assessments used some of the latest techniques of radiation safety, including a high screen/film speed, a long film focal distance, as well as breast and gonadal shielding.
There may be risks associated with not undergoing drug treatment for high BP. The risk, in the opinion of the investigators was quite small for a study of such a short duration. Other studies have used extended waiting or baseline periods. [59–60] Including in the study only those subjects with mild to moderate hypertensive disease minimized these risks. Patients were informed through the informed consent process that they might be randomized to 1 of 3 groups: no-care control; brief massage (sham); or chiropractic care.
No subjects withdrew or were removed from the study because of an adverse effect, nor did any subject report an adverse effect. The risks associated with participation in this type of study are extremely small and must be balanced against the potential for significant health benefits. If the care demonstrates a scientifically verified alternative to pharmacological treatment of hypertension, the benefits to patients with this disease would be quite substantial. Potential patient benefits may include reduced costs, decreased drug-related side effects, improved quality of life, and more effective health care.
This pilot study elucidated several procedural issues that should be addressed before undertaking a full-scale clinical trial on the effects of chiropractic adjustments at sites of subluxation in patients with essential hypertension. Among the most important of these is using measures to assure comparable groups regarding prognostic variables such as weight. A multidisciplinary approach to recruitment may need to be used in any future efforts due to the limited subject pool of patients who have hypertensive disease but are not taking medications for its control. A 1-week run-in period may be necessary before randomization, to exclude subjects who may only be exhibiting white coat hypertension. Studies such as these demonstrate the feasibility of conducting a full-scale 3-group randomized clinical trial of patients with hypertension in the practice-based setting.