Clinical Trial

Hypertension Improvement Project

Randomized Trial of Quality Improvement for Physicians and Lifestyle
Modification for Patients
Laura P. Svetkey, Kathryn I. Pollak, William S. Yancy, Jr, Rowena J. Dolor, Bryan C. Batch,
Greg Samsa, David B. Matchar, Pao-Hwa Lin

Abstract—Despite widely publicized hypertension treatment guidelines for physicians and lifestyle recommendations for
patients, blood pressure control rates remain low. In community-based primary care clinics, we performed a nested, 2⫻2
randomized, controlled trial of physician intervention versus control and/or patient intervention versus control.
Physician intervention included internet-based training, self-monitoring, and quarterly feedback reports. Patient
intervention included 20 weekly group sessions followed by 12 monthly telephone counseling contacts and focused on
weight loss, Dietary Approaches to Stop Hypertension dietary pattern, exercise, and reduced sodium intake. The primary
outcome was change in systolic blood pressure at 6 months. Eight primary care practices (32 physicians) were
randomized to physician intervention or control groups. Within those practices, 574 patients were randomized to patient
intervention or control groups. Patient mean age was 60 years, 61% were women, and 37% were black. Blood pressure
data were available for 91% of patients at 6 months. The main effect of physician intervention on systolic blood pressure
at 6 months, adjusted for baseline pressure, was 0.3 mm Hg (95% CI: ⫺1.5 to 2.2; P⫽0.72). The main effect of the
patient intervention was ⫺2.6 mm Hg (95% CI: ⫺4.4 to ⫺0.7; P⫽0.01). The interaction of the 2 interventions was
significant (P⫽0.03); the largest impact was observed with the combination of physician and patient intervention
(⫺9.7⫾12.7 mm Hg). Differences between treatment groups did not persist at 18 months. Combined physician and
patient interventions lowers blood pressure; future research should focus on enhancing effectiveness and sustainability
of these interventions. (Hypertension. 2009;54:1226-1233.)
Key Words: hypertension 䡲 blood pressure 䡲 behavioral intervention 䡲 quality improvement
Downloaded from http://ahajournals.org by on August 23, 2021

䡲 lifestyle 䡲 DASH dietary pattern

H ypertension is the most prevalent risk factor for cardio-

vascular and kidney diseases1 and accounts for ⬇35% of
atherosclerotic cardiovascular disease.2 Antihypertensive
therapy reduces the risk of stroke by ⬇35%, congestive heart
failure by 42%, and coronary heart disease by 28%.3 Preven-
tion and treatment guidelines for providers and patients are
readily available.4 Nonetheless, adherence to both provider
guidelines and lifestyle recommendations for patients is low,
resulting in continued low rates of blood pressure (BP)
control.5 We tested interventions to increase physician adher-
ence to national guidelines and patient adherence to lifestyle
recommendations for lowering BP.
Methods
The Hypertension Improvement Project (HIP) was approved by the
Duke Institutional Review Board. The design6 was a nested 2⫻2
randomized, controlled trial of a physician intervention, a patient
intervention, and both combined, compared with neither. Nesting
occurred both at the level of the practice and the level of the
physician. Primary care practices were randomly assigned to the
physician intervention (MD-I) or to the physician control (MD-C)
group. All of the participating physicians within a given practice had
the same randomization assignment. Within the participating prac-
tices, patients were individually randomized to the patient interven-
tion (Pt-I) or to usual care (Pt-C). Follow-up measurements were
performed at 6 and 18 months postrandomization. The primary
outcome was systolic BP change at 6 months. The study design is
displayed in Figure 1.
Enrollment and Randomization
Four matched pairs of community-based primary care practices in
central North Carolina were randomized between 2005 and 2007.
Practices were matched with regard to specialty (internal medicine or
family practice) and patient socioeconomic mix. One practice of

Received April 25, 2009; first decision May 14, 2009; revision accepted August 28, 2009.
From the Department of Medicine (L.P.S., W.S.Y., R.J.D., B.C.B., G.S., D.B.M., P.-H.L.), Duke Hypertension Center (L.P.S., B.C.B., P.-H.L.), Sarah
W. Stedman Nutrition and Metabolism Center (L.P.S., P.-H.L.), Department of Community and Family Medicine (K.I.P.), Cancer Prevention, Detection,
and Control Research Program (K.I.P., D.B.M.), Center for Health Policy Research (G.S.), and Department of Biostatistics and Bioinformatics (G.S.),
Duke University Medical Center, Durham, N.C.; Program in Health Services Research (D.B.M.), Duke-National University of Singapore Graduate
Medical School, Singapore.
This trial has been registered at www.clinicaltrials.gov (identifier NCT00201136).
Correspondence to Laura P. Svetkey, Stedman Building, Center for Living Campus, 3475 Erwin Rd, Suite 100, Durham, NC 27705. E-mail
svetk001@mc.duke.edu
© 2009 American Heart Association, Inc.
Hypertension is available at http://hyper.ahajournals.org DOI: 10.1161/HYPERTENSIONAHA.109.134874

1226
 Svetkey et al Hypertension Improvement Project: Main Results 1227

Randomize
primary care
practices
N=8

MD Control MD Intervention
N=4 N=4
practices/16 practices/16
MDs MDs

Patient Patient
recruitment recruitment
3045 Letters 2846 Letters
521 Patients 473 Patients

Screened but not Randomized Randomized Screened but not

enrolled N = 281 N = 293 enrolled
N = 270 N = 180
Figure 1. Study flow.

Patient control Patient Patient control Patient

N = 141 intervention N = 148 intervention
N = 140 N = 145

6-month data collection

N = 132 (94%) N = 124 (89%) N = 137 (93%) N = 132 (91%)
Died = 2 Died = 0 Died = 0 Died = 2
Dropped out = 7 Dropped out = Dropped out = Dropped out =
16 11 11

18-month data collection

N = 122 (87%) N = 124 (89%) N = 134 (91%) N = 128 (88%)

Downloaded from http://ahajournals.org by on August 23, 2021

Additional drop- Additional drop- Additional drop- Add’l deaths = 1

outs = 10 outs = 0 outs = 3 Add’l drop-outs =
10

each pair was blindly assigned by the study statistician to interven-
tion or control status. Within each practice, all of the physicians were
invited to participate, with a goal of enrolling 4 physicians per clinic.
Each physician provided written informed consent. For logistical
reasons, we enrolled practices in waves or “cohorts” of 1 interven-
tion and 1 control clinic each.
We sent a recruitment letter from the physician to potentially
eligible patients. We sought to enroll 10 to 15 patients from each
physician. Patients were eligible if they were ⱖ25 years old and were
hypertensive on the basis of billing codes. Patients were excluded if
they had self-reported chronic kidney disease (CKD), a cardiovas-
cular disease event within the past 6 months, or were pregnant,
breastfeeding, or planning a pregnancy.
Potential study participants were prescreened by telephone and
then attended 2 screening visits at which eligibility was confirmed,
written informed consent was obtained, and baseline data were
collected. Randomization to Pt-C or Pt-I occurred in varying block
sizes using a computer-generated algorithm, stratified by cohort and
clinic. Randomization was performed by the study statistician; all of
the data collection staff remained blinded to the participant’s
treatment assignment. It was not feasible to blind patients or
providers to their own randomization assignment. Patients were
asked not to discuss their randomization assignment with their
provider, but strict blinding of the physicians was not feasible.
Interventions
MD-I lasted 18 months and consisted of 3 elements. First, 2 training
modules were provided on-line. The first module addressed the
Seventh Joint National Committee on Prevention, Detection, Evalu-
ation and Treatment of High Blood Pressure (JNC-7) guidelines,4
and the second addressed lifestyle modification for BP control. Each
module required ⬇45 minutes, included a quiz that gave immediate
feedback, and provided Continuing Medical Education credit
through Duke University. Participating physicians completed the
modules within 2 weeks of randomization and before the patient
intervention began. Second, each physician in MD-I received an
evaluation and treatment algorithm that summarized, on a color-
coded, pocket-sized laminated card, the major JNC-7 guidelines,
including lifestyle guidelines and a decision tree. Third, a quality
improvement (QI) procedure assessed clinical performance measures
and provided quarterly feedback to physicians on adherence to
JNC-7 guidelines. At each MD-I site, participating physicians
completed a clinical performance measure data form every time a
HIP patient (in either Pt-I or Pt-C) had a clinic visit. The form
recorded patient demographics, comorbidity, previous and current
BP measurements, and actions taken during the visit. In addition, 1
day each month, physicians completed a clinical performance mea-
sure form on all of the adult patients treated during that day, whether
or not the patient was a HIP participant, recording the same data as
for study participants but without patient identifiers.
These data were converted into personalized quarterly feedback
reports that indicated the following: (1) the proportion of hyperten-
sive patients in the practice with adequately controlled BP, the
change in that proportion over the course of the study, and compar-
ison with the other participating physicians; (2) the proportion of
patients with diabetes mellitus or CKD who were at JNC-7 goal BP;
(3) the proportion of patients prescribed specific classes of medica-
tion on the basis of JNC-7 guidelines; and (4) the proportion who
received lifestyle modification counseling on the basis of physician
self-report.
MD-C constituted “usual care.” There was no attempt to change or
monitor procedures already in place for QI and physician education
with regard to BP control. No performance data were collected from
these physicians, and no performance feedback was given.
 1228 Hypertension December 2009
Pt-I consisted of 20 weekly group sessions (n⫽10 to 15 patients
per group) over ⬇6 months. All of the intervention sessions occurred
at or near the patients’ primary care clinic.
The behavior goals of Pt-I included weight loss if overweight, the
Dietary Approaches to Stop Hypertension (DASH) dietary pattern,7
increased moderate-to-vigorous physical activity, reduced sodium
intake, and moderation of alcohol intake. In addition, the interven-
tion promoted adherence to antihypertensive medication regimen.
Pt-I was based on key theoretical constructs developed to guide
health behavior change efforts and on practical applications from
previous trials.8 The intervention was designed to promote frequent
self-monitoring, feedback, goal setting, and social support and used
motivational interviewing techniques.9
Pt-I was conducted by 2 experienced behavioral interventionists
who were trained and certified to deliver a group intervention
focusing on diet and exercise and to use motivational interviewing
techniques. They were assisted by community health advisors.10 Two
community health advisors assisted with group sessions at each
clinic and contacted participants who missed a session. After the
initial 6-month intensive intervention, community health advisors
contacted participants by telephone each month for 1 year to offer
brief lifestyle counseling.
Pt-C constituted usual care, composed of an individual visit with
an interventionist to receive advice and written materials on lifestyle
modification for BP control consistent with JNC-7 guidelines.
Measurements
Measurements were obtained from all of the randomized physicians
and all of the randomized patients at baseline and 6 and 18 months.
Physician Measurements
All of the physicians, regardless of treatment assignment, were asked
to complete a self-administered questionnaire concerning demo-
graphics, education, training, characteristics of patients under their
care, and usual practice patterns.
Downloaded from http://ahajournals.org by on August 23, 2021
Patient Measurements
All of the study measurements were obtained during face-to-face
clinic visits by trained, certified study personnel who were blinded to
intervention assignment. For BP measurements, personnel were
trained and certified using methods used in previous BP trials.8
Duplicate measurements were obtained with a calibrated automated
appropriate-sized cuff after the participant had been seated quietly
for ⱖ5 minutes. At each time point (baseline and 6 and 18 months),
participants attended 2 study visits ⬇1 week apart. BP for that time
point was defined as the mean over the 2 study visits (ie, 4 BP
readings). For eligibility, hypertension was considered present on the
basis of billing codes. For all other purposes, hypertension was
defined as measured systolic (S)BP ⱖ140 mm Hg, diastolic (D)BP
ⱖ90 mm Hg, or taking antihypertensive medication. BP treatment
goals were defined on the basis of JNC-7 guidelines.4
Height was measured once to the nearest 0.1 cm, using a
calibrated, wall-mounted stadiometer. Weight was measured in
duplicate with the participant wearing light, indoor clothes without
shoes and using a high-quality, calibrated digital scale. Body mass
index was calculated as the Quetelet Index (kilograms per meter
squared). Dietary intake was assessed with the Block Food Fre-
quency Questionnaire.11
Physical activity was assessed by a calibrated, triaxial accelerom-
eter (RT3, Stayhealthy, Inc) worn for ⱖ10 hours per day for ⱖ4 days
over the course of a week, including 1 weekend day. Data are
expressed as total weekly minutes of moderate-to-vigorous physical
activity.12 Medications were self-reported and verified by study staff
by inspection of medication bottles. Laboratory measurements were
performed in a certified commercial laboratory (LabCorp), including
fasting glucose and lipid profile and 24-hour urinary excretion of
potassium, sodium, and phosphorus (to reflect intake of fruits and
vegetables [F/V], salt, and dairy products, respectively).
Dyslipidemia was defined as low-density lipoprotein cholesterol
level ⱖ160 mg/dL or taking lipid-lowering medications. Diabetes
mellitus was defined as fasting blood sugar ⬎125 mg/dL or taking
diabetes medication. The diagnosis of CKD was based on patient
self-report. Although self-report of CKD is likely to be underesti-
mated, self-report was used to identify patients with severe CKD,
which would preclude eating the high-potassium DASH dietary
pattern. For determining patient mix in MD-I feedback reports, these
diagnoses were based on physician report.
Outcomes
The primary outcome was change in SBP from baseline to 6 months.
Secondary outcomes included change in DBP at 6 months, BP
change at 18 months, the effect of treatment on weight loss, dietary
pattern, physical activity, fasting blood glucose and lipids, and the
proportion of patients with adequate BP control.
Power and Statistical Analysis
The original study design defined the primary outcome as the
proportion at goal BP, and, with a planned sample size of 500
patients, was powered to detect an effect size of 0.3. During the study
it became clear that the proportion of patients who were at goal BP
at baseline (assessed blinded to treatment group) was higher than
anticipated on the basis of national statistics, potentially inducing a
ceiling effect that could have left the study underpowered. Conse-
quently, with the permission of the trial’s data and safety monitoring
board, the primary outcome was changed to the continuous variable,
change in SBP, with change in proportion at goal relegated to a
secondary outcome. In addition, recruitment for the remaining study
participants targeted those with BP above goal at baseline. The
achieved sample size of 574 patients in 8 practices provided 80%
power to detect a difference of 4 mm Hg.
Comparisons across treatment groups were adjusted for baseline
value and cohort. There were no interim analyses. All of the
participants were analyzed according to their original intervention
assignment, but participants who did not complete the study mea-
surements were not included in analyses. We compared baseline
characteristics of completers and noncompleters using a ␹2 or t test
(with Satterthwaite adjustment if indicated), and we performed a
sensitivity analysis using last value carried forward. Intervention
effects were consistent across cohort (assessed by a treatment-by-
cohort interaction); accordingly, results are presented in aggregate.
Analysis of change in BP was performed using ANCOVA. The
2⫻2 factorial design allowed for tests of the main effect of MD-I, the
main effect of Pt-I, and the interaction between the 2 interventions.
For SBP and DBP, the “adjusted” (or “least-squares”) mean reflects
the mean change in BP within each group after accounting for any
differences in case mix between the groups. In addition, we consid-
ered the design to be equivalent to a 4-arm parallel trial and
evaluated differences across pairs of intervention arms (MD-C/Pt-C,
MD-I/Pt-C, MD-C/Pt-I, and MD-I/Pt-I). Similar analytic methods
were applied to secondary outcomes. The effect of treatment group
on the proportion “at goal” or “not at goal” was evaluated by ␹2
analysis.
Results
Eight primary care practices, composed of 32 physicians,
were randomized to MD-I or MD-C (Figure 1). There were
no significant differences between randomized groups with
regard to physician age, sex, race, specialty, years in practice,
or patient panels (Table 1). Approximately 31% reported that
they were “quite familiar” with JNC-7 guidelines. All of the
physicians in the MD-I group completed the 2 training
modules. Quarterly feedback reports were based on physi-
cians completing a mean of 36.4⫾16.4 forms per quarter on
patients who were not HIP participants and 8.2⫾5.7 forms
per quarter on patients who were HIP participants.
Within the 8 randomized practices, a total of 574 patients
were randomized to Pt-I or Pt-C. Approximately 56% of
screened patients were randomized (Figure 1), with most
 Svetkey et al Hypertension Improvement Project: Main Results 1229

Table 1. Baseline Characteristics of Physicians

Characteristic Overall Intervention Control P*
N 32 16 16
Age, mean (SD), y 48 (10) 47 (13) 49 (7) 0.61
Women, n (%) 11 (34) 3 (19) 8 (50) 0.06
Black, n (%) 5 (16) 2 (12) 3 (19) 0.63
Family medicine (all others internal medicine), n (%) 17 (53) 8 (50) 9 (56) 0.72
Years postphysician degree, mean (SD), n 21 (10) 20 (12) 22 (8) 0.64
Clinic patients per day, mean (SD), n 21 (3) 22 (3) 20 (3) 0.13
Patients with hypertension per day, mean (SD), % 29 (16) 30 (14) 28 (18) 0.79
⬙Quite familiar⬙ with JNC-7 guidelines, n (%) 10 (31) 5 (31) 5 (31) 0.99
*P is for intervention vs control.

exclusions attributed to patients declining to participate. The
mean age of patients was 60.5 years (range: 28.0 to 94.0
years), 61% were women, 37% were black, and 1% were
Hispanic/Latino (Table 2). Most participants completed high
school and reported that their income was “adequate” (93%
and 85%, respectively), without differences across treatment
groups. Body mass index ranged from 20.5 to 47.9 kg/m2,
but, on average, participants were obese (body mass index:
ⱖ30 kg/m2). Three percent self-reported CKD. Billing code
diagnosis of hypertension was confirmed in 97% of partici-
pants, who were taking a mean of 2 antihypertensive medi-
cations. Mean baseline BP was 133.1/74.1 mm Hg. At base-
line, ⬇60% of study participants were at goal BP.
Outcome data are presented for patients with both baseline
Downloaded from http://ahajournals.org by on August 23, 2021
had higher baseline urinary potassium excretion (60.0 versus
53.1 mmol/24 hours; P⫽0.05), and had lower baseline SBP
(132.7/73.9 versus 137.5/76.2 mm Hg; P⫽0.05 for SBP,
P⫽0.17 for DBP).
With neither intervention (MD-C/Pt-C), SBP fell by a
mean of 6.7⫾12.8 mm Hg at 6 months. With the physician
intervention alone (MD-I/Pt-C), SBP fell by 5.3⫾12.1 mm Hg,
and with the patient intervention alone (MD-C/Pt-I), SBP fell
by 7.1⫾12.1 mm Hg (Figure 2A). With the combination of
physician and patient interventions (MD-I/Pt-I), SBP fell by
9.7⫾12.7 mm Hg (P⫽0.0072 compared with all other
groups). In the main effects model, at 6 months there was no
significant effect of MD-I (0.3 mm Hg; 95% CI: ⫺1.5 to 2.2;
P⫽0.76). In contrast, the main effect of Pt-I was a net

and follow-up data for each variable. Follow-up data were
available for 91.0% of randomized participants at 6 months
and 88.5% at 18 months, without difference by treatment
group (Figure 1). At baseline, there were no significant
differences between completers and noncompleters with re-
gard to dietary intake, excretion of sodium, and weight.
Completers were more physically active at baseline than
noncompleters (36.5 versus 19.1 minutes per week; P⫽0.02),
reduction of 2.6 mm Hg (95% CI: ⫺4.4 to ⫺0.7; P⫽0.01). In
addition, there was a significant interaction between MD-I
and Pt-I (P⫽0.03), suggesting that the effect of Pt-I was
enhanced by coincident exposure to MD-I. Similar results
were seen for diastolic BP (Figure 2B). Figure 2 also
demonstrates BP changes from baseline to 18 months. Al-
though changes at 18 months were similar to changes at 6
months, they were no longer significant: at 18 months, the

Table 2. Baseline Characteristics of Patients, Overall and by Treatment Group

Characteristic Overall MD-C/Pt-C MD-I/Pt-C MD-C/Pt-I MD-I/Pt-I P*
N (total) 574 141 148 140 145
Age, mean (SD), y 60.5 (11.4) 60.7 (12.2) 61.6 (10.2) 59.0 (12.3) 60.7 (11.0) 0.27
Women 61 65 58 66 55 0.17
Black 37 42 31 44 33 0.07
Hispanic 1 1 1 0 1 0.59
Current smoker 9 11 9 8 8 0.85
Diabetes mellitus 30 33 31 26 29 0.69
Body mass index, mean (SD), kg/m2 32.5 (5.5) 32.9 (5.7) 32.7 (5.4) 31.8 (5.5) 32.6 (5.2) 0.32
Hypertension 97 99 97 98 95 0.34
No. of BP medications, mean (SD) 2.0 (1.1) 2.1 (1.2) 2.0 (1.1) 2.0 (1.1) 1.9 (1.2) 0.58
SBP, mm Hg, mean (SD) 133.1 (16.1) 131.6 (14.6) 134.6 (15.7) 132.1 (17.6) 133.8 (16.3) 0.34
DBP, mm Hg, mean (SD) 74.1 (11.3) 73.3 (10.5) 74.3 (11.0) 73.3 (12.6) 75.3 (11.1) 0.39
Dyslipidemia 48 49 54 44 44 0.28
Previous cardiovascular disease event 16 16 14 14 21 0.39
Data are percentage except where otherwise noted.
*P is for comparison across treatment groups.
 1230 Hypertension
A SBP
0
-2
Change in SBP, mmHg
-4
-5.3
-6 -6.7
-7.5 -7.5
-8
-10
-12
MD-C/Pt-C MD-I/Pt-C
* p < .05 compared to MD-I/Pt-C; † p = .0006 compared to MD-I/Pt-C
B DBP
0
-1
Change in DBP, mmHg
-2
-3 -3.2
-3.6
-4
-5
-4.9
-4.6
-6
-7
MD-C/Pt-C MD-I/Pt-C
* p < .05 compared to MD-I/Pt-C.
Figure 2. Change in BP at 6 and 18 months (A, SBP; B, DBP).
main effect for the physician intervention was 0.3 (P⫽0.81),
the main effect for the patient intervention was ⫺0.2
(P⫽0.89), and the interaction was no longer significant.
There were no statistically significant interactions for effects
by race, sex, or age at either time point. Participants who were
Downloaded from http://ahajournals.org by on August 23, 2021
above goal BP at baseline had a 4.0 mm Hg greater reduction
in SBP than patients in whom BP was already at goal at
baseline (P⬍0.0001; 95% CI: 2.6 to 5.4). This finding
suggests that the HIP interventions will be most effective in
those who need it most, which may have implications for
targeting implementation of this kind of intervention in
clinical settings. These findings were confirmed in a separate
nonparametric analysis, in a model that included physician as
a covariate, and in a sensitivity analysis in which missing
values were replaced by the last value carried forward.
Figure 3 shows the percentage of patients with BP at the
JNC-7-defined goal. The proportion of patients with BP at
goal increased in the first 6 months postrandomization, with
significant differences across treatment groups (P⫽0.03),
seemingly because of higher rates in those exposed to the
patient intervention (MD-C/Pt-I and MD-I/Pt-I), but differ-
ences between treatment groups did not persist at 18 months.
Table 3 shows that, at baseline, treatment groups were
comparable with regard to physical activity, dietary intake,
and weight. In general, the effects on behavior of MD-I and
Pt-I were independent of each other; the only significant
interaction was for the effect on intake of F/V (P for
interaction⫽0.05). In the main effects analysis, there was no
significant effect of either intervention on MV-PA. At 6
months, MD-I resulted in a significant increase in F/V intake
and decreased total fat intake. Pt-I increased intake of F/V
and dairy products and decreased intake of both total and
saturated fat. Each of these dietary changes is consistent with
the DASH dietary pattern. These changes were not reflected
December 2009
90
85
80
75
*
Percent at goal
-7.1 -6.8
70
† -8.6 6 mo
65
-9.7 18 mo
60
MD-C/Pt-I MD-I/Pt-I 55
MD-C/Pt-C
50 MD-I/Pt-C
MD-C/Pt-I
45 MD-I/Pt-I
40
Baseline 6 Mo* 18 Mo
* p = .03 for differences across treatment groups
-3.4 -3.4
based on Chi-square statistic.
Figure 3. Percentage of patients at goal BP.
*
-5.4 -5.3
6 mo
18 mo
in the urinary excretion data (data not shown). Pt-I, but not
MD-I, led to a significant reduction in weight (⫺6.1 and ⫹0.6
MD-C/Pt-I MD-I/Pt-I
lb, respectively; P⬍0.0001 for Pt-I main effect). At 18
months, the effect of MD-I on F/V intake was no longer
significant, but the effect on total fat intake persisted. The
effects of Pt-I on F/V, total fat, and saturated fat intake
persisted, but effects on dairy intake and weight did not.
There were no significant changes in the number of antihy-
pertensive medications, but we did not measure changes in
dose.
Discussion
The HIP trial demonstrated that, in the setting of primary care
clinics, an intensive behavioral lifestyle intervention signifi-
cantly reduced BP at 6 months, with evidence that this effect
occurred because of adoption of a healthy dietary pattern and
weight loss. A QI-type intervention for the primary care
physicians treating these patients apparently did not lead to
more effective BP management. However, a key finding of
the trial was that the effect of the patient intervention was
significantly enhanced by simultaneous exposure of the
primary care provider to the QI-type intervention.
MD-I may have enhanced the effect of Pt-I through several
mechanisms. In previous research, providers reported a lack
of confidence in addressing behavior change and patient
lifestyle,13 and the training on lifestyle counseling included in
the physician intervention may have increased their confi-
dence and enhanced their counseling, reinforcing the lifestyle
advice that patients were receiving in the patient intervention.
Such reinforcement alone may not be sufficient to change
patient behavior, but coupling it with intensive support for
behavior change may have encouraged patients to adopt
BP-lowering behaviors. In addition, the combined interven-
tion may have changed the doctor-patient interaction in such
a way that the physician was more likely to intensify
antihypertensive treatment. Although there was no clear
effect of the MD intervention on the number of antihyperten-
sive medications prescribed, we were unable to assess
changes in dose, which perhaps would more sensitively
reflect intensification of therapy.
 Svetkey et al Hypertension Improvement Project: Main Results 1231

Table 3. Effect of Intervention on Physical Activity, Dietary Intake, and Weight

Outcome MD-C/Pt-C MD-I/Pt-C MD-C/Pt-I MD-I/Pt-I P for MD-I Main Effect P for Pt-I Main Effect
Physical activity by
accelerometry
(moderate-to-vigorous
physical activity, min/wk)
Baseline 43.9 (122.5) 36.4 (127.1) 37.9 (89.1) 28.8 (106.7)
Change (6 mo) ⫺15.7 (122.0) 18.5 (297.8) 6.2 (103.2) 28.4 (134.9) 0.15 0.49
Change (18 mo) ⫺13.0 (145.7) 5.0 (95.1) ⫺21.5 (138.8) ⫺0.7 (112.3) 0.07 0.10
Dietary pattern
F/V, servings per day
Baseline 1.28 (0.90) 1.23 (0.88) 1.33 (0.99) 1.42 (1.13)
Change (6 mo) 0.04 (0.81) 0.09 (0.72) 0.59 (1.27) 0.92 (1.34) 0.02 ⬍0.0001
Change (18 mo) 0.01 (0.90) ⫺0.03 (0.87) 0.41 (1.13) 0.55 (1.13) 0.53 ⬍0.0001
Dairy, servings per day
Baseline 0.94 (0.90) 0.89 (0.89) 0.94 (0.89) 0.95 (0.81)
Change (6 mo) ⫺0.01 (0.72) 0.01 (0.56) 0.06 (0.67) 0.21 (0.75) 0.17 0.01
Change (18 mo) ⫺0.03 (0.80) ⫺0.00 (0.61) 0.08 (0.51) ⫺0.01 (0.79) 0.57 0.56
Total fat, % kcal
Baseline 38.1 (8.0) 39.8 (8.3) 38.4 (8.1) 37.3 (6.8)
Change (6 mo) 0.6 (7.0) ⫺1.2 (6.7) ⫺2.3 (8.5) ⫺4.3 (7.1) 0.002 ⬍0.0001
Change (18 mo) 0.8 (8.0) ⫺1.1 (6.6) ⫺1.7 (7.7) ⫺2.8 (7.8) 0.02 ⬍0.0001
Saturated fat, % kcal
Baseline 10.6 (2.5) 11.0 (2.3) 10.9 (2.7) 10.5 (2.4)
Change (6 mo) 0.2 (2.3) ⫺0.2 (2.0) ⫺1.0 (2.7) ⫺1.3 (2.1) 0.07 ⬍0.0001
Change (18 mo) 0.1 (2.3) ⫺0.2 (2.0) ⫺0.9 (2.1) ⫺1.0 (2.2) 0.28 ⬍0.0001
Downloaded from http://ahajournals.org by on August 23, 2021

Urinary sodium, mmol/24 h

Baseline 174.7 (77.0) 175.2 (82.9) 150.9 (68.0) 170.3 (76.2)
Change (6 mo) ⫺22.8 (71.2) ⫺23.6 (75.2) ⫺13.1 (62.2) ⫺31.4 (79.7) 0.62 0.14
Change (18 mo) ⫺8.3 (84.1) ⫺1.4 (69.9) ⫺24.0 (85.2) ⫺28.0 (76.6) 0.03 0.32
Weight, lb
Baseline 202.1 (39.6) 199.0 (38.4) 192.3 (38.0) 202.5 (37.5)
Change (6 mo) ⫺0.3 (6.3) 0.1 (11.2) ⫺5.2 (10.7) ⫺7.0 (10.7) 0.44 ⬍0.0001
Change (18 mo) ⫺2.1 (12.0) ⫺0.4 (13.4) ⫺2.6 (12.2) ⫺3.8 (9.8) 0.83 0.06
Values represent means (SD). N in each category ranged from 74% (urine) to 91% (weight) of randomized participants at 6 months and from 71% (urine) to 89%
(weight) at 18 months.

Our findings are consistent with a systematic review of QI
interventions for BP control.14 In evaluating 44 articles,
Walsh et al14 found that QI strategies that target the provider
have limited effect compared with strategies that target the
patient. The median reduction in BP associated with QI
interventions that provided monitoring and feedback for
providers was 1.5/0.6 mm Hg compared with 3.3/2.8 mm Hg
for interventions that promoted patient self-management. In
this systematic review, there was no opportunity to evaluate
combinations of physician and patient strategies. However,
subsequently, Roumie et al15 randomized 205 primary care
providers to interventions that were similar but less intense
than HIP: provider education consisting of a letter to the MD,
provider education plus an electronic alert system that added
a single reminder for each patient, or provider education and
alert plus patient education that added mailed, written advice
concerning adherence to medication and lifestyle changes.
This study was conducted largely in academic medical center
clinics, in a study population that was older than HIP
participants (mean age: 65 versus 60 years), 97% male
(compared with 39% in HIP), and with higher baseline BP
(157/82 versus 133/74 mm Hg). Despite these differences,
results were similar: at 6 months, the group receiving pro-
vider education and alert plus patient education was 33%
more likely to have SBP ⬍140 mm Hg than the group
receiving provider education only (P⫽0.013). Provider edu-
cation plus alerts without patient education did not improve
the SBP control rate.
Our provider intervention was unique in its emphasis on
lifestyle counseling. There are numerous challenges to pro-
viding lifestyle counseling in the context of a primary care
visit, including time, provider confidence in his or her own or
the patient’s ability to change behavior, and lack of reim-
bursement for these activities. We expected that the physician
training in the HIP provider intervention would increase
lifestyle counseling, but the extent to which counseling
 1232 Hypertension December 2009
occurred was based on physician self-report. It is possible that
physicians in the intervention group were more likely to
report counseling whether or not they were actually providing
it, knowing that was expected of them. A discrepancy
between reported and actual counseling may help explain the
lack of effect of MD-I on BP.
Unlike the trial by Roumie et al,15 we tested the patient
intervention alone. The HIP Pt-I is similar to what has been
effective in previous studies16 –18 but is unique in its applica-
tion in this study in the community practice setting. In
addition, the HIP Pt-I focused heavily on adoption of the
DASH dietary pattern, which has been shown to lower BP7
and low-density lipoprotein cholesterol19 and has been asso-
ciated with successful weight loss.16,20 In HIP, Pt-I signifi-
cantly improved dietary pattern consistent with the DASH
dietary pattern. The DASH dietary pattern is not specifically
a low-salt diet, but the BP effect of DASH is increased by
simultaneous reduction in sodium intake.21 Reducing sodium
intake was a goal of the HIP Pt-I, but urinary excretion data
did not suggest an effect. The lack of significant decrease in
sodium excretion could reflect the limitations of a single
24-hour urine collection for assessing intake22 or an actual
absence of change. Nonetheless, clinical trial and meta-anal-
ysis would suggest that the impact on BP of weight loss and
DASH is greater than the impact of currently recommended
sodium reduction.23,24 Thus, it may be advantageous that the
HIP Pt-I had its predominant impact on DASH adherence and
weight loss. In general, effects on BP were associated with
changes in behavior. For example, patients who lost weight
Downloaded from http://ahajournals.org by on August 23, 2021
had a 2.1 mm Hg greater improvement in SBP (P⫽0.0013;
95% CI: 0.8 to 3.3) than patients who did not lose weight.
The main effect of the HIP Pt-I on behavior and BP was
somewhat less than in previous similar studies. For example,
the PREMIER Study exposed a similar patient population to
a similar intervention. In the 2 active treatment groups in the
PREMIER Study, BP fell by ⬇4/3 mm Hg (net of control)16
compared with a reduction of 2.6/1.0 mm Hg in HIP (Pt-I
main effect). This difference could well be within the vari-
ability of the effect estimate, or it could be because of
increased effects in a younger population (mean age: 50 years
in PREMIER versus 60 in HIP). However, PREMIER par-
ticipants were otherwise similar to HIP participants (gener-
ally healthy, 36% men, 36% black, and BP 134/84 mm Hg at
baseline). Therefore, the lesser effect on BP is more likely to
reflect mild dilution of the intervention’s effect as the study
design moves from academic medical centers (PREMIER)
closer to an effectiveness trial conducted in community-based
practices (HIP). Indeed the results of HIP may be more
generalizable than other trials, because participants were
recruited from community practices, received the intervention
locally, and were diverse with respect to race, sex, and age.
As noted in the PREMIER Study and other trials, inter-
vention effects on BP did not persist to the end of the study
(18 months after randomization or 12 months after the
intensive behavioral intervention for patients). We originally
speculated that the 6-month outcomes would reflect the
maximum impact of the patient intervention but might be too
early to reflect maximum impact of the MD intervention. The
18-month outcomes presumably assess the durability of Pt-I,
as well as the cumulative (perhaps maximal) effect of MD-I.
We noted that there was some persistent effect on behavior
(ie, improved dietary pattern and some weight loss), but there
was no significant effect of either Pt-I or MD-I on BP at 18
months. The effect in the MD-I/Pt-I group at 18 months,
however, was comparable to that found at 6 months.
The HIP study has 3 potential limitations. First, the
analysis of primary outcome is based on those who completed
follow-up. No imputation procedure was used for missing
data, because no method was considered satisfactory, given
that BP and other measurements were collected at baseline
and then at the time of primary outcome assessment, with no
intermediate measurements. Although a completers analysis
is potentially subject to bias, the follow-up rate exceeded 91%
at 6 months and 88% at 18 months, and a sensitivity analysis
assuming no change in BP in noncompleters yielded similar
results.
Second, because BP control rates were unexpectedly high
at baseline, the primary outcome was changed from “propor-
tion at goal BP” to “change in SBP.” However, the new
primary outcome initially had been designated as an impor-
tant secondary outcome, the decision to change the primary
outcome was made blinded to treatment group effects and
was reviewed and approved by the data and safety monitoring
board, and the results are consistent.
Finally, the study population represents a relatively healthy
cohort with high rates of BP control at baseline and limited
comorbidity. In addition, only ⬇10% of potentially eligible
patients were randomized. These factors suggest potential
limitations to both the generalizability of the results and the
implementation of the interventions.
Perspectives
The HIP behavioral intervention improved dietary pattern and
lowered BP over 6 months, but the intervention was inten-
sive, the effect size relatively small, and the effect did not
persist. Implementation of a similar lifestyle intervention
program would require further development to make it
affordable, scalable within health systems, and able to pro-
duce sustained improvements in behavior and BP. The same
is true for the HIP MD intervention. Improving the MD
intervention might require integrated tools for easily assess-
ing patient behavior and providing brief targeted advice
within the context of the modern primary care practice.
Nonetheless, given the potential impact of lifestyle modifi-
cation on BP and the apparent role that doctors play in
encouraging healthy behaviors, future development and test-
ing of both patient and provider interventions should be a
high priority.
Acknowledgments
The investigators gratefully acknowledge the valuable contributions
of the study participants and the following individuals: research staff
members Kathleen Aicher, Blondeaner Brown, Gwendolyn Davis,
LaVerne Johnson-Pruden, Martis King, Tonya Milligan, Rhonda
Mooney, LaChanda Reams, Patrice Reams, and Sonia P. Steele;
community health advisors Elaine Armstrong, Rachel Blackwell,
Frances Cagle, Mary Crispell, Maurice Darden, Carolyn Dean, Jesse
Edwards, Ralph Glover, Shirley Griles, Sarah Johnson, Jeaneen Lile,
Melissa O’Connor, and Laura Wickwar; Duke Medical Media (Ann
Bushyhead) and the Duke Continuing Medical Education office;
 Svetkey et al
Data and Safety Monitoring Board members Jean Spaulding, MD
(chair); L. Kristen Newby, MD; Michael Pignone MD, MD; Steven
C. Grambow, PhD; Wendy Demark-Wahnefried, PhD; physicians
and staff at the participating clinics: Durham Medical Center,
Durham, N.C.; Harps Mill Internal Medicine, Raleigh, N.C.; Hen-
derson Family Medicine Clinic, Henderson, N.C.; Hillsborough
Family Practice, Hillsborough, N.C.; Metropolitan Durham Medical
Group, Durham, N.C.; Oxford Family Physicians, Oxford, N.C.;
Roxboro Medical Associates, Roxboro, N.C.; Triangle Family Prac-
tice, Durham, N.C.
Sources of Funding
This work was supported by National Institutes of Health grant
R01-HL75373.
Disclosures
None.
References
1. Working Group Report on Primary Prevention of Hypertension. National
High Blood Pressure Education Program. National Institutes of Health,
National Heart, Lung, and Blood Institutes Document. Bethesda, MD:
National Institutes of Health; 2000.
2. Kannel WB. Blood pressure as a cardiovascular risk factor: prevention
and treatment. JAMA. 1996;275:1571–1576.
3. Psaty B, Smith NL, Siscovick DS, Koepsell TD, Weiss NS, Heckbert SR,
Lemaitre RN, Wagner EH, Furberg CD. Health outcomes associated with
anti-hypertensive therapies used as first line agents: a systematic review
and meta-analysis. JAMA. 1997;277:739 –745.
4. Chobanian AV, Bakris GL, Black HR, Cushman WC, Green LA, Izzo JL
Jr, Jones DW, Materson BJ, Oparil S, Wright JT Jr, Roccella EJ. The
seventh report of the Joint National Committee on Prevention, Detection,
Evaluation, and Treatment of High Blood Pressure: the JNC 7 report.
Hypertension. 2003;42:1206 –1252.
5. Singer GM, Izhar M, Black HR. Guidelines for hypertension: are quality-
Downloaded from http://ahajournals.org by on August 23, 2021
assurance measures on target? Hypertension. 2004;43:198 –202.
6. Dolor RJ, Yancy WS Jr, Owen WF, Matchar DB, Samsa GP, Pollak KI,
Lin PH, Ard JD, Prempeh M, McGuire HL, Batch BC, Fan W, Svetkey
LP. Hypertension Improvement Project (HIP): study protocol and imple-
mentation challenges. Trials. 2009;10:13.
7. Appel L, Moore TJ, Obarzanek E, Vollmer WM, Svetkey LP, Sacks FM,
Bray GA, Vogt TM, Cutler JA, Windhauser MM, Lin P-H, Karanja N; for
the DASH Collaborative Research Group. A clinical trial of the effects of
dietary patterns on blood pressure. N Engl J Med. 1997;336:1117–1124.
8. Svetkey LP, Harsha DW, Vollmer WM, Stevens VJ, Obarzanek E, Elmer
PJ, Lin PH, Champagne C, Simons-Morton DG, Aickin M, Proschan MA,
Appel LJ. Premier: a clinical trial of comprehensive lifestyle modification
for blood pressure control–rationale, design and baseline characteristics.
Ann Epidemiol. 2003;13:462– 471.
9. Miller WR, Rollnick S. Motivational Interviewing: Preparing People for
Change. 2nd ed. New York, NY: Guilford Press; 2002.
10. Brownstein JN, Chowdhury FM, Norris SL, Horsley T, Jack L Jr, Zhang
X, Satterfield D. Effectiveness of community health workers in the care
of people with hypertension. Am J Prev Med. 2007;32:435– 447.
11. Block G, Thompson FE, Hartman AM, Larkin FA, Guire KE. Com-
parison of two dietary questionnaires validated against multiple dietary
Hypertension Improvement Project: Main Results 1233
records collected during a 1-year period. J Am Diet Assoc. 1992;92:
686 – 693.
12. Chen C, Jerome GJ, Laferriere D, Young DR, Vollmer WM. Procedures
used to standardize data collected by RT3 triaxial accelerometers in a
large-scale weight-loss trial. J Phys Act Health. 2009;6:354 –359.
13. Cook S, Drum ML, Kirchhoff AC, Jin L, Levie J, Harrison JF, Lippold
SA, Schaefer CT, Chin MH. Providers’ assessment of barriers to effective
management of hypertension and hyperlipidemia in community health
centers. J Health Care Poor Underserved. 2006;17:70 – 85.
14. Walsh JM, McDonald KM, Shojania KG, Sundaram V, Nayak S, Lewis
R, Owens DK, Goldstein MK. Quality improvement strategies for hyper-
tension management: a systematic review. Med Care. 2006;44:646 – 657.
15. Roumie CL, Elasy TA, Greevy R, Griffin MR, Liu X, Stone WJ, Wallston
KA, Dittus RS, Alvarez V, Cobb J, Speroff T. Improving blood pressure
control through provider education, provider alerts, and patient education:
a cluster randomized trial. Ann Intern Med. 2006;145:165–175.
16. Appel LJ, Champagne CM, Harsha DW, Cooper LS, Obarzanek E, Elmer
PJ, Stevens VJ, Vollmer WM, Lin PH, Svetkey LP, Stedman SW, Young
DR; for the Writing Group of the PCRG. Effects of comprehensive
lifestyle modification on blood pressure control: main results of the
PREMIER clinical trial. JAMA. 2003;289:2083–2093.
17. Whelton PK, Appel LJ, Espeland MA, Applegate WB, Ettinger WH Jr,
Kostis JB, Kumanyika S, Lacy CR, Johnson KC, Folmar S, Cutler JA; for
the TONE Collaborative Research Group. Sodium reduction and weight
loss in the treatment of hypertension in older persons: a randomized
controlled trial of nonpharmacologic interventions in the elderly (TONE).
JAMA. 1998;279:839 – 846.
18. Trials of Hypertension Prevention Collaborative Research Group. Effects
of weight loss and sodium reduction intervention on blood pressure and
hypertension incidence in overweight people with high-normal blood
pressure: the Trials of Hypertension Prevention, phase II. Arch Intern
Med. 1997;157:657– 667.
19. Obarzanek E, Sacks FM, Vollmer WM, Bray GA, Miller ER III, Lin PH,
Karanja NM, Most-Windhauser MM, Moore TJ, Swain JF, Bales CW,
Proschan MA; for the DASH Research Group. Effects on blood lipids of
a blood pressure-lowering diet: the Dietary Approaches to Stop Hyper-
tension (DASH) Trial. Am J Clin Nutr. 2001;74:80 – 89.
20. Svetkey LP, Stevens VJ, Brantley PJ, Appel LJ, Hollis JF, Loria CM,
Vollmer WM, Gullion CM, Funk K, Smith P, Samuel-Hodge C, Myers V,
Lien LF, Laferriere D, Kennedy B, Jerome GJ, Heinith F, Harsha DW,
Evans P, Erlinger TP, Dalcin AT, Coughlin J, Charleston J, Champagne
CM, Bauck A, Ard JD, Aicher K; for the Weight Loss Maintenance
Collaborative Research Group. Comparison of strategies for sustaining
weight loss: the weight loss maintenance randomized controlled trial.
JAMA. 2008;299:1139 –1148.
21. Sacks FM, Svetkey LP, Vollmer WM, Appel LJ, Bray GA, Harsha D,
Obarzanek E, Conlin PR, Miller ER III, Simons-Morton DG, Karanja N,
Lin PH; for the DASH-Sodium Collaborative Research Group. Effects on
blood pressure of reduced dietary sodium and the Dietary Approaches to
Stop Hypertension (DASH) diet. N Engl J Med. 2001;344:3–10.
22. Liu K, Stamler J. Assessment of sodium intake in epidemiological studies
on blood pressure. Ann Clin Res. 1984;16(suppl 43):49 –54.
23. Neter JE, Stam BE, Kok FJ, Grobbee DE, Geleijnse JM. Influence of
weight reduction on blood pressure: a meta-analysis of randomized con-
trolled trials. Hypertension. 2003;42:878 – 884.
24. Graudal NA, Galloe AM, Garred P. Effects of sodium restriction on blood
pressure, renin, aldosterone, catecholamines, cholesterols, and triglycer-
ide: a meta-analysis. JAMA. 1998;279:1383–1391.