ORIGINAL INVESTIGATION

Effects of Reduced Sodium Intake on Hypertension

Control in Older Individuals
Results From the Trial of Nonpharmacologic Interventions in the Elderly (TONE)
Lawrence J. Appel, MD, MPH; Mark A. Espeland, PhD; Linda Easter, MS, RD;
Alan C. Wilson, PhD; Steven Folmar, PhD; Clifton R. Lacy, MD

Background: Few trials have evaluated the effects of re-
duced sodium intake in older individuals, and no trial
has examined the effects in relevant subgroups such as
African Americans.
Patients and Methods: The effects of sodium reduc-
tion on blood pressure (BP) and hypertension control were
evaluated in 681 patients with hypertension, aged 60 to
80 years, randomly assigned to a reduced sodium inter-
vention or control group. Participants (47% women, 23%
African Americans) had systolic BP less than 145 mm Hg
and diastolic BP less than 85 mm Hg while taking 1 anti-
hypertensive medication. Three months after the start of
intervention, medication was withdrawn. The primary
end point was occurrence of an average systolic BP of 150
mm Hg or more, an average diastolic BP of 90 mm Hg
or more, the resumption of medication, or a cardiovas-
cular event during follow-up (mean, 27.8 months).
excretion was 40 mmol/d less in the reduced sodium
intervention group (P,.001); significant reductions in
sodium excretion occurred in subgroups defined by sex,
race, age, and obesity. Prior to medication withdrawal,
mean reductions in systolic and diastolic BPs from the
reduced sodium intervention, net of control, were 4.3 mm
Hg (P,.001) and 2.0 mm Hg (P =.001). During follow-
up, an end point occurred in 59% of reduced sodium
and 73% of control group participants (relative hazard
ratio=0.68, P,.001). In African Americans, the corre-
sponding relative hazard ratio was 0.56 (P = .005); re-
sults were similar in other subgroups. In dose-response
analyses, end points were progressively less frequent with
greater sodium reduction (P for trend=.002).
Conclusion: A reduced sodium intake is a broadly ef-
fective, nonpharmacologic therapy that can lower BP and
control hypertension in older individuals.

Results: Compared with control, mean urinary sodium Arch Intern Med. 2001;161:685-693

From the Welch Center for
Prevention, Epidemiology, and
Clinical Research, Johns
Hopkins Medical Institutions,
Baltimore, Md (Dr Appel);
Departments of Public Health
Sciences (Dr Espeland),
General Clinical Research
Center (Ms Easter), and
Anthropology (Dr Folmar),
Wake Forest University,
Winston-Salem, NC; and
Division of Cardiovascular
Diseases and Hypertension,
University of Medicine and
Dentistry of New Jersey–Robert
Wood Johnson Medical School,
New Brunswick (Drs Wilson
and Lacy).
H
YPERTENSION and its treat-
ment with medication are
extremely common in the
elderly. According to the
Third National Health and
Nutrition Examination Survey, con-
ducted between 1988 and 1991, the preva-
lence of hypertension, defined as a sys-
tolic blood pressure (BP) of 140 mm Hg or
more, a diastolic BP of 90 mm Hg or more,
or treatment with medication, exceeds 50%
in the civilian, noninstitutionalized popu-
lation aged 60 years and older.1 In certain
subgroups, hypertension is nearly ubiqui-
tous. For example, the prevalence of hy-
pertension among African American women
aged 60 to 69 years is 78%. Medication use
is also highly prevalent in the elderly, rang-
ing from 31% of Mexican American women
with hypertension 70 years and older to 70%
of African American women with hyper-
tension 70 years and older.
Sodium reduction is widely advo-
cated as a means to reduce BP and con-
trol hypertension in older persons,2,3 yet
empiric evidence is sparse. Little is
known about the ability and willingness
of older persons to reduce their sodium
intake, the effects of a reduced sodium
intake on BP and hypertension control in
this population, dose-response relation-
ships, and the effects in relevant sub-
groups, such as African Americans.
Applegate et al4 demonstrated that a mul-
tifactorial intervention, consisting of
sodium reduction, weight loss, and
increased physical activity, can reduce BP
in older persons. Otherwise, only a few
trials, each with small sample size, have
examined the impact of sodium reduc-
tion as a means to reduce BP5,6 and con-
trol hypertension7 in older persons. In
these trials, few, if any, participants were
African American.
Sodium reduction should be particu-
larly effective in older persons. First, be-
cause arterial compliance decreases with
age, any change in intravascular volume re-

(REPRINTED) ARCH INTERN MED/ VOL 161, MAR 12, 2001 WWW.ARCHINTERNMED.COM
685
Downloaded from www.archinternmed.com on October 12, 2010
©2001 American Medical Association. All rights reserved.
 SUBJECTS AND METHODS
A detailed description of the design and methods of this
trial has been published. 11 The trial protocol was ap-
proved by institutional review boards at each participat-
ing center and by an external protocol review board ap-
pointed by the funding agencies: the National Institute on
Aging and the National Heart, Lung, and Blood Institute.
Each participant provided written informed consent.
STUDY POPULATION
The study population consisted of healthy persons (aged
60-80 years) with systolic BP of less than 145 mm Hg and
diastolic BP of less than 85 mm Hg (mean of 9 measure-
ments, ie, 3 BPs at each of 3 visits) while taking 1 antihy-
pertensive medication, ie, 1 type of medication whether or
not multiple doses were used. Individuals treated with 2
antihypertensive medications were also eligible, if they were
successfully weaned from 1 of these medications during the
screening phase. Major exclusion criteria were use of an-
tihypertensive medication for conditions other than hy-
pertension (eg, ischemic heart disease), myocardial infarc-
tion or stroke within 6 months, angina pectoris, congestive
heart failure, serum creatinine level of more than 176.8
µmol/L (.2 mg/dL), blood glucose level of more than 14.4
mmol/L (.260 mg/dL), and self-report of average alco-
holic beverage intake of more than 14 drinks per week.
A detailed description of recruitment procedures has
been published.12 In brief, each TONE clinical center imple-
mented site-specific strategies, which included (1) mass
mailings of brochures; (2) radio, television, and newspa-
per advertisements; (3) BP screenings; and (4) enrollment
of participants from previous studies. Enrollment began in
August 1992 and ended in June 1994.
DATA COLLECTION
Individuals provided a medical history, underwent a physi-
cal examination, and had routine laboratory tests to confirm
eligibility. TONE data collection visits included 2 screening
visits, a randomization visit, drug withdrawal visits (begin-
ning 90 days [±14 days] after the start of intervention), and
follow-up visits scheduled every 3 months. For screenees who
were taking 2 antihypertensive medications, medication step-
down visits took place after the second screening visit but
prior to the randomization visit. Closeout visits occurred be-
tween July 1995 and December 1995. The median duration
of follow-up was 29 months (maximum of 36 months). Dur-
ing follow-up, safety monitoring visits took place whenever
a participant’s mean systolic BP was 150 mm Hg or more or
diastolic BP was 90 mm Hg or more.
lated to sodium intake should result in a greater BP change
in older persons than in younger individuals.8 Second, be-
cause of the decline in kidney function associated with ag-
ing, older individuals may retain sodium to a greater ex-
tent than younger persons.9 Third, older individuals may
be more willing and able to reduce their sodium intake
than younger individuals who have yet to experience the
adverse health consequences of elevated BP.
(REPRINTED) ARCH INTERN MED/ VOL 161, MAR 12, 2001
686
Downloaded from www.archinternmed.com on October 12, 2010
©2001 American Medical Association. All rights reserved.
All BP measurements were obtained by trained and cer-
tified observers who were masked to intervention assign-
ment. Systolic BP was defined as the appearance of the first
Korotkoff sound, and diastolic BP as the point of disap-
pearance of the fifth Korotkoff sound. At each visit, 3 BP
measurements were obtained while the participant rested
quietly in the seated position. Random-zero sphygmoma-
nometers were used to minimize observer bias.
Twenty-four-hour urine collections were obtained
twice prior to randomization and once at the 9-, 18-, and
30-month follow-up visits and at closeout, if this visit did
not coincide with 1 of the follow-up visits. Twenty-four-
hour dietary recalls were obtained twice prior to random-
ization and again at the 9- and 12-month follow-up visits
and every 6 months thereafter by trained, certified techni-
cians who were masked to intervention assignment. Nu-
trient calculations were performed using the Minnesota Data
System software (food database version 6A; nutrient data-
base version 21), developed by the Nutrition Coordinat-
ing Center, University of Minnesota, Minneapolis.13 Data
from the 24-hour recalls were used to determine macro-
nutrient, micronutrient, and energy intake in randomized
groups at baseline and during follow-up.
INTERVENTION
Overweight participants were randomly assigned, in a 2 32
factorial design, to 1 of the following 4 groups: (1) com-
bined weight loss and reduced sodium, (2) reduced so-
dium alone, (3) weight loss alone, or (4) usual lifestyle (UL)
control group. Nonoverweight participants were ran-
domly assigned to reduced sodium alone or UL groups. Body
mass index values (calculated as weight in kilograms di-
vided by the square of height in meters) of 27.8 for men
and 27.3 for women were used as the thresholds to define
overweight.14 This article presents data on participants, both
overweight and nonoverweight, who were assigned to the
reduced sodium alone or UL group.
The intervention goal for reduced sodium groups was
to achieve and maintain a 24-hour dietary sodium intake
of 80 mmol/L or less of sodium, as measured by 24-hour
urine collections. This level of sodium intake is slightly be-
low the currently recommended upper limit of 100 mmol/L
of sodium per day for the prevention and treatment of hy-
pertension.3 Participants were expected to modify only those
aspects of their diet that led to a high sodium intake; a com-
prehensive change in diet was not expected. To achieve its
goal, TONE used intervention techniques derived from ex-
perience in clinical trials that achieved BP control through
interventions focusing on behavioral change.15-17 Social learn-
ing theory and behavioral approaches that enhance the un-
derstanding of behavior change and the ways to achieve it
were incorporated.
In view of these considerations, the Trial of Nonphar-
macologic Interventions in the Elderly (TONE) was a ran-
domized trial that tested whether reduced sodium intake
and/or weight loss can maintain satisfactory medication-
treated hypertension control in older persons with hyper-
tension. The main results of TONE have been pub-
lished.10 In brief, TONE demonstrated that a reduced sodium
intake and weight loss, alone or combined, could effec-
WWW.ARCHINTERNMED.COM
 In the reduced sodium group, each person had an in-
troductory individual session with the interventionist as-
signed to their group. This session occurred within 4 weeks
of randomization. Intervention groups, typically consist-
ing of 9 to 12 participants, were then formed in a timely
manner so that participants began the reduced sodium pro-
gram within 45 days after randomization. Each of the TONE
interventions consisted of a 4-month “intensive” phase with
weekly meetings, a 3-month “extended” phase with bi-
weekly meetings, and a maintenance phase. The interven-
tionist typically was a registered dietitian. The meetings were
conducted as group sessions with individual sessions held
at every fourth contact.
During the group and individual sessions, interven-
tionists provided information using both centrally and lo-
cally prepared materials, motivated participants to make
and sustain long-term lifestyle changes, and monitored in-
dividual and group progress at frequent intervals. In the
process, participants learned about sources of sodium, par-
ticularly those foods with a high salt content, and about
alternative foods, condiments, and spices. They also learned
how to adapt the reduced sodium lifestyle recommenda-
tions to their own individual situations, eg, how to select
appropriate foods at restaurants.
To enhance follow-up among participants assigned to
UL, meetings were held on a regular basis with speakers
who led discussions on topics unrelated to BP, cardiovas-
cular disease, or nutrition. To facilitate masking of the data
collectors, intervention visits were conducted at separate
times and places from data collection visits.
WITHDRAWAL OF ANTIHYPERTENSIVE
MEDICATION
Drug withdrawal began 90 days (±14 days) after the first
group intervention session. Participants randomized to
UL began drug withdrawal at a comparable time. The
drug withdrawal process, which was standardized across
the 4 clinical centers, used drug-specific tapering regi-
mens. As the medication was tapered, participants were
evaluated weekly. After discontinuation of the drug, par-
ticipants had 3 additional biweekly visits to confirm that
their systolic BP remained less than 150 mm Hg and
diastolic BP less than 90 mm Hg. The rationale for
beginning medication withdrawal 90 days after the start
of intervention, as opposed to immediately after ran-
domization, was to increase the likelihood of successful
drug withdrawal.
OUTCOME VARIABLES
In primary analyses, the outcome variable was a composite
end point defined by the need for, or actual resumption
tively control hypertension. The objectives of this article,
which focuses on only the sodium component of TONE,
are to examine the following issues: (1) the effects of a re-
duced sodium intake in subgroups defined by sex, ethnic-
ity, age, and obesity; (2) dose-response relationships; and
(3) the effects of a reduced sodium intake on subtypes of
the primary outcome variable, dietary intake of other nu-
trients, and the occurrence of adverse events.
(REPRINTED) ARCH INTERN MED/ VOL 161, MAR 12, 2001
687
Downloaded from www.archinternmed.com on October 12, 2010
©2001 American Medical Association. All rights reserved.
of, antihypertensive drug therapy. Specifically, a TONE par-
ticipant reached a trial end point when any of the follow-
ing occurred:
•High BP, as indicated by (1) elevated BP measured by TONE
BP technicians (at 1 visit, mean diastolic BP $110 mm Hg
or systolic BP $190 mm Hg; at 2 visits, mean diastolic BP
$100 mm Hg or systolic BP $170 mm Hg; or at 3 visits,
mean diastolic BP $90 mm Hg or systolic BP $150 mm Hg),
or (2) resumption of antihypertensive medication initi-
ated either by the participant or personal physician for el-
evated BP measured outside a clinical center.
•Resumption of antihypertensive medication initiated by
either a personal physician or participant for a symptom
or condition other than elevated BP or a cardiovascular
clinical event. Participants given medication for non-BP
and noncardiovascular conditions, such as migraine head-
aches and benign prostatic hypertrophy, were censored.
•A cardiovascular clinical event (myocardial infarction, an-
gina, congestive heart failure, hypertensive encephalopa-
thy, stroke, or procedure [bypass surgery, angioplasty,
endarterectomy]). These clinical events were included as
components of the composite end point to reduce the pos-
sibility of informative censoring.
Medical records were retrieved for all end points occur-
ring outside a TONE visit. An end point committee, masked
to intervention assignment, made final decisions concern-
ing the end point status of each participant.
To determine the impact of the interventions on BP,
change in BP was calculated as the difference between mean
BP prior to randomization and BP at the visit when medi-
cation withdrawal was first attempted. At both times, par-
ticipants were receiving drug therapy.
ANALYSES
Analyses were conducted on an intention-to-treat basis. In
primary analyses, the distributions of times until first occur-
rence of an end point were compared in participants as-
signed to the reduced sodium and UL groups. Times were
measured from the end of the drug withdrawal process until
the occurrence of the end point. Participants who met an end
point criterion prior to, or during, drug withdrawal were
treated as instantaneous failures. Each comparison was per-
formed using proportional hazards regression. Kaplan-
Meier curves were used to portray the distribution of times
until failure for each of the study cohorts. For continuous out-
comes, eg, BP and nutrient intake, differences in change (fol-
low-up minus baseline levels) between the reduced sodium
and UL groups were compared using unpaired t tests. In all
analyses, a 2-sided significance level of .05 was considered
statistically significant.
RESULTS
At baseline, mean (SD) age was 65.8 (4.6) years, and 79%
of participants were between 60 and 69 years of age. Of
the participants, 47% were women, 23% African Ameri-
cans, 34% college graduates, and 43% overweight. Mean
(SD) systolic and diastolic BPs at baseline, while partici-
WWW.ARCHINTERNMED.COM
 Table 1. Twenty-four-Hour Urinary Sodium Excretion Overall and by Subgroup*

Between-Group Difference‡ P for Difference

Baseline, Within-Group in Effect Across
Group No. Mean (SD) Change, Mean (SD)† Mean (95% CI) P Subgroups
All participants
Reduced sodium 319 144 (53) −45 (55.8)
−40 (−48 to −32) ,.001 ...
Usual lifestyle 320 145 (55) −5 (50.0)
Men
Reduced sodium 162 162 (53) −59 (53.0)
−53 (−64 to −41) ,.001
Usual lifestyle 175 159 (55) −7 (53.6)
Women ,.001
Reduced sodium 157 125 (45) −30 (55)
−27 (−39 to −16) ,.001
Usual lifestyle 145 128 (48) −3 (45)
60-69 y age group
Reduced sodium 253 144 (54) −46 (57)
−38 (−48 to −29) ,.001
Usual lifestyle 250 151 (56) −8 (52)
70-80 y age group .96
Reduced sodium 66 142 (48) −41 (52)
−46 (−62 to −30) ,.001
Usual lifestyle 70 124 (39) 5 (42)
Nonoverweight
Reduced sodium 184 135 (49) −45 (50)
−44 (−54 to −35) ,.001
Usual lifestyle 183 136 (49) −1 (46)
Overweight .24
Reduced sodium 135 156 (55) −45 (63)
−34 (−48 to −20) ,.001
Usual lifestyle 137 157 (58) −10 (54)

*Urinary sodium excretion expressed as millimoles per 24 hours. CI indicates confidence interval; ellipses, not applicable.
†For each individual, average within-group change was the difference between the average of all 24-hour urine collections during follow-up minus the average
of 2 baseline 24-hour urine collections.
‡Within-group difference in reduced sodium group minus within-group difference in usual lifestyle group.

pants were taking medication, were 128.0 (9.4) and 71.3
(7.3) mm Hg, respectively. On average, participants had
hypertension for 13 years and had been taking antihy-
pertensive medication for 12 years. At baseline, 32% of
participants were taking a diuretic, 27% a calcium chan-
nel blocker, 22% an angiotensin-converting enzyme in-
hibitor, 11% a b-blocker, and 8% another antihyperten-
sive agent. Mean (SD) urinary sodium excretion was 161
(54) mmol/d in men and 126 (47) mmol/d in women.
There was no evidence of a substantial imbalance be-
tween the reduced sodium and UL groups.
FOLLOW-UP
At the 9-month follow-up visit, attendance was 91% and
88% of expected in the reduced sodium and UL groups,
respectively; at the 18-month follow-up visit, correspond-
ing attendance was 85% and 83%, respectively. At the
9-month follow-up visit, 24-hour urine collections were
provided by 99% of reduced sodium and 97% of UL par-
ticipants attending these visits. Among participants at-
tending the 18-month follow-up visit, 97% of reduced so-
dium and 99% of UL participants provided 24-hour urine
collections. Closeout visits occurred from 15 to 36 months
after randomization. Attendance at the closeout visits was
90% and 93% in the reduced sodium and UL groups. Of
those attending a closeout visit, 97% of reduced sodium
and 98% of UL participants provided urine specimens. At
the end of follow-up, end point status was known in 98%
of reduced sodium and UL participants.
INTERVENTION RESULTS
The reduced sodium group achieved and maintained a
substantial reduction in sodium levels. At the 9-, 18-,
and 30-month follow-up visits, more than 40% of
reduced sodium participants had an absolute urinary
sodium excretion of 80 mmol/d or less in contrast to
less than 15% of UL participants. Table 1 displays
mean (SD) sodium urinary excretion at baseline,
within-group changes, and between-group differences
in sodium excretion. Overall, urinary sodium excretion
was reduced by 40 mmol/d in the reduced sodium
group, net of UL (P,.001). The extent of sodium
reduction was less in women than in men (27 vs 53
mmol/d; P,.001), in part as a result of baseline differ-
ences in sodium intake. In the 2 age groups (60-69 and
70-80 years), the reductions in sodium levels were simi-
lar. Sodium reduction tended to be greater in over-
weight persons than in nonoverweight persons (44 vs
34 mmol/d; P = .24). Because the sex distribution in
African Americans differed from that in non–African
Americans, we performed sex-stratified analyses
(Table 2). In these analyses, sodium reduction was
similar in African Americans and non–African Ameri-
cans; however, differences by sex persisted. The pattern
of findings from 24-hour dietary recalls was similar to
that of the 24-hour urinary excretion (data not
presented). Compared with the UL group, the reduced
sodium group lost an average of 1.1 kg (2.5 lb) during
follow-up (P,.001).

(REPRINTED) ARCH INTERN MED/ VOL 161, MAR 12, 2001 WWW.ARCHINTERNMED.COM
688
Downloaded from www.archinternmed.com on October 12, 2010
©2001 American Medical Association. All rights reserved.
 Table 2. Twenty-four-Hour Urinary Sodium Excretion Stratified by Sex and Race*

Between-Group Difference‡ P for Difference

Baseline, Within-Group in Effect Across
Group No. Mean (SD) Change, Mean (SD)† Mean (95% CI) P Subgroups
African American women
Reduced sodium 49 129 (49) −26 (64)
−25 (−47 to −3) .03
Usual lifestyle 56 123 (41) −1 (48)
Non–African American women .83
Reduced sodium 108 123 (43) −32 (51)
−28 (−41 to −15) ,.001
Usual lifestyle 89 131 (51) −4 (43)
African American men
Reduced sodium 20 158 (60) −55 (44)
−41 (−69 to −13) .007
Usual lifestyle 21 157 (47) −14 (48)
Non–African American men .46
Reduced sodium 142 163 (52) −60 (54)
−54 (−67 to −42) ,.001
Usual lifestyle 154 160 (56) −6 (54)

*Urinary sodium excretion expressed as millimoles per 24 hours. CI indicates confidence interval.
†For each individual, average within-group change was the difference between the average of all 24-hour urine collections during follow-up minus the average
of 2 baseline 24-hour urine collections.
‡Within-group difference in reduced sodium group minus within-group difference in usual lifestyle group.

Table 3. Systolic and Diastolic Blood Pressure (BP) Overall and by Subgroup*

Systolic BP, mm Hg Diastolic BP, mm Hg

Between-Group Between-Group
Within-Group Difference‡ Within-Group Difference‡
Baseline BP, BP Change, Baseline BP, BP Change,
Group No. Mean (SD) Mean (SD)† Mean (95% CI) P Mean (SD) Mean (SD)† Mean (95% CI) P
All participants
Reduced sodium 317 128.4 (9.3) −4.6 (11.3)† 71.4 (7.5) −2.2 (8.0)
−4.3 (−6.0 to −2.5) ,.001 −2.0 (−3.2 to −0.8) .001
Usual lifestyle 296 127.3 (9.4) −0.4 (10.5) 71.4 (7.2) −0.2 (7.0)
Men
Reduced sodium 162 129.1 (9.0) −5.6 (11.3) 72.7 (6.6) −2.7 (7.7)
−5.2 (−7.5 to −2.9) ,.001 −2.6 (−4.2 to −1.0) .002
Usual lifestyle 163 126.9 (9.7) −0.4 (9.5) 72.2 (7.0) −0.1 (7.0)
Women
Reduced sodium 155 127.7 (9.5) −3.7 (11.3) 70.0 (8.1) −1.6 (8.2)
−3.4 (−6.0 to −0.6) .02 −1.3 (−3.1 to 0.4) .14
Usual lifestyle 133 127.7 (8.9) −0.4 (11.7) 70.4 (7.2) −0.3 (7.0)
African American
Reduced sodium 66 125.6 (7.6) −3.8 (10.1) 71.6 (6.7) −2.7 (7.1)
−5.0 (−8.4 to −1.6) .005 −2.9 (−5.3 to −0.5) .02
Usual lifestyle 76 127.3 (8.4) 1.1 (10.7) 70.2 (7.5) 0.3 (7.4)
Non–African American
Reduced sodium 251 129.2 (9.5) −4.9 (11.6) 71.3 (7.7) −2.0 (8.2)
−4.0 (−5.9 to −2.0) ,.001 −1.7 (−3.0 to −0.3) .01
Usual lifestyle 220 127.2 (9.7) −0.9 (10.4) 71.8 (7.0) −0.4 (6.8)
60-69 year age group
Reduced sodium 251 128.2 (9.2) −5.2 (11.1) 72.1 (7.4) −2.3 (8.1)
−5.0 (−6.9 to −3.1) ,.001 −2.1 (−3.5 to −0.8) .002
Usual lifestyle 230 126.8 (9.6) −0.2 (10.3) 72.3 (6.4) −0.2 (7.0)
70-80 year age group
Reduced sodium 66 129.2 (9.4) −2.6 (11.8) 68.8 (7.2) −1.6 (7.5)
−1.5 (−5.4 to 2.4) .46 −1.4 (−3.9 to 1.0) .25
Usual lifestyle 66 128.7 (8.6) −1.1 (11.2) 68.2 (8.8) −0.2 (6.8)
Overweight
Reduced sodium 135 128.4 (9.3) −4.7 (10.6) 71.6 (7.6) −1.9 (8.7)
−4.9 (−7.2 to −2.6) ,.001 −1.7 (−3.9 to 0.6) .09
Usual lifestyle 123 126.9 (9.6) 0.2 (10.9) 71.7 (7.1) −0.2 (6.8)
Nonoverweight
Reduced sodium 182 128.5 (9.3) −4.6 (11.8) 71.2 (7.4) −2.4 (7.3)
−3.9 (−6.2 to −1.5) .001 −2.2 (−3.7 to 0.7) .004
Usual lifestyle 173 127.5 (9.2) −0.8 (10.3) 71.1 (7.3) −0.2 (7.1)

*CI indicates confidence interval.

†For each individual, average within-group change was the difference between BP prior to drug withdrawal minus baseline BP.
‡Within-group difference in reduced sodium group minus within-group difference in usual lifestyle group.

EFFECTS OF INTERVENTIONS ON BP interval, 3.5 months). In Table 3, the reduced sodium

group experienced a mean reduction in systolic BP of
The effect of the reduced sodium intervention on BP was 4.3 mm Hg (P,.001) and in diastolic BP of 2.0 mm Hg
assessed by comparing the change in BP from baseline (P=.001), net of BP change in UL. In all subgroup analy-
with BP at the visit prior to medication withdrawal (mean ses, the mean reduction in systolic BP in the reduced so-

(REPRINTED) ARCH INTERN MED/ VOL 161, MAR 12, 2001 WWW.ARCHINTERNMED.COM
689
Downloaded from www.archinternmed.com on October 12, 2010
©2001 American Medical Association. All rights reserved.
 1.0 1.4
Usual Lifestyle
0.9 0.818 Reduced Sodium P for Trend = .002
1.2
0.8 ∗
∗

Relative Hazard Ratio

0.7
Free of End Point, %

0.614 1.0
0.6 ∗ 0.527
0.477
0.5 0.745 ∗ 0.8
∗ ∗ 0.408
∗ 0.357
0.4 0.471 ∗
0.397 ∗ ∗ 0.6
0.3 0.347 ∗
0.2 0.280 ∗
0.4
0.209
0.1

0.0 0.2
1st (+41) 2nd (+3) 3rd (–22) 4th (–51) 5th (–93)
0 (3) 6 (9) 12 (15) 18 (21) 24 (27) 30 (33) Quintile of Change in Urinary Sodium Excretion
Months After Drug Withdrawal (Months After Randomization) (Median Change, mmol/d)

Figure 1. Proportion of participants who remained free of a trial end point Figure 3. Relative hazard ratio (with 95% confidence interval) of a trial end
(elevated blood pressure, resumption of medication, and cardiovascular point by quintile of change in urinary sodium excretion (observational
events) during follow-up. analyses, n=639). The cut points for the quintiles are as follows: +19, −9,
−35, and −70 mmol/d.

1.0 portion without an end point was 36% in the reduced

0.926
∗ Usual Lifestyle sodium group and 21% in the UL group. The relative haz-
0.9 ∗
Reduced Sodium
0.8
ard ratios associated with assignment to reduced so-
0.712
∗
dium vs UL were 0.68 (95% confidence interval [CI], 0.56-
0.7
Free of End Point, %

0.900 0.611
0.6 ∗ ∗
0.5 0.596 ∗ ∗
0.505
0.4
0.3
0.2
0.1
0.0
0 (3) 6 (9) 12 (15)
Months After Drug Withdrawal (Months After Randomization)
Figure 2. Proportion of participants who remained free of elevated blood
pressure during follow-up. Cardiovascular events were censored.
dium group was greater than that of the UL group, achiev-
ing statistical significance in all but 1 stratum (the 70-80
year age group). Likewise, diastolic BP reductions were
consistently greater in the reduced sodium group than
in the UL group; however, a few between-group differ-
ences did not achieve statistical significance.
EFFECTS OF INTERVENTIONS
ON HYPERTENSION CONTROL
Of the 448 end points, 334 occurred as a result of el-
evated BP (of which 203 occurred from measurements
in the TONE clinic), 22 as a result of a clinic cardiovas-
cular event, and 92 as a result of participant and per-
sonal physician decisions for symptoms and conditions
other than elevated BP. The distributions of event sub-
type in the reduced sodium and UL groups were similar
(P=.45, x23 test).
Figure 1 displays the distribution of end point times
after completion of medication withdrawal by interven-
tion group. Beginning with completion of medication
withdrawal and continuing throughout follow-up, the pro-
portion who remained end point free in the reduced so-
dium group exceeded that of the UL group. After 30
months of follow-up (after drug withdrawal), the pro-
0.567 0.82; P,.001) and 0.69 (95% CI, 0.57-0.84; P,.001,
∗ 0.494 adjusting for weight change). In analyses restricted to end
0.432
points defined by elevated BP in any setting (Figure 2),
∗ ∗
0.449 ∗ the proportion without elevated BP was 43% in the re-
0.362 ∗ duced sodium group and 27% in the UL group at the end
0.270 of follow-up. The corresponding relative hazard ratio was
0.67 (95% CI, 0.54-0.83; P,.001). In analyses re-
stricted to end points defined by an elevated BP as mea-
18 (21) 24 (27) 30 (33)
sured in a TONE clinic, the relative hazard ratio re-
mained virtually identical, ie, 0.62 (95% CI, 0.47-0.82;
P,.001).
In dose-response analyses that assessed the risk of
an end point by quintiles of change in urinary sodium
excretion (Figure 3), the relative hazard ratio de-
creased with greater reductions in urinary sodium ex-
cretion (P for trend=.002). The risk of an end point was
unrelated to baseline dietary sodium intake or excretion
(data not presented).
In subgroups defined by sex, ethnicity, age group,
and weight (Table 4), results were similar to overall find-
ings. For instance, in African Americans, the relative haz-
ard ratio associated with assignment to reduced sodium
vs UL was 0.56 (95% CI, 0.37-0.84; P=.005). The rela-
tive hazard ratio by class of withdrawn medication was
0.62 (95% CI, 0.42-0.92; P=.02) for diuretic users, 0.61
(95% CI, 0.42-0.92; P =.02) for calcium channel blocker
users, 0.50 (95% CI, 0.40-0.94; P=.02) for angiotensin-
converting enzyme inhibitor users, 1.66 (95% CI, 0.79-
3.50; P=.18) for b-blocker users, and 1.53 (95% CI, 0.62-
3.80; P = .36) for users of other antihypertensive
medications (all adjusted for sex, ethnicity, age, weight,
and change in urine sodium excretion).
OTHER EFFECTS OF THE INTERVENTIONS
Except for a tendency toward less angina in the reduced
sodium group compared with the UL group (9 vs 17 in-
dividuals, P=.16), the occurrence of cardiovascular events
was similar in the 2 groups (Table 5). Headache oc-

(REPRINTED) ARCH INTERN MED/ VOL 161, MAR 12, 2001 WWW.ARCHINTERNMED.COM
690
Downloaded from www.archinternmed.com on October 12, 2010
©2001 American Medical Association. All rights reserved.
 Table 4. Relative Hazard Ratio for End Points Table 5. Number of Individuals Reporting Cardiovascular
Associated With Assignment to Reduced Sodium and Other Adverse Events (and Total Number
Intervention vs Usual Lifestyle Control Group of Events) During Follow-up by Randomized Group*

P for No. of Individuals

Intervention (No. of Events)
Relative Hazard Effect Within P for
Group Ratio (95% CI) Group Interaction† Reduced Usual
Type of Adverse Event Sodium Lifestyle P*
All participants 0.68 (0.56 to 0.82) ,.001 ...
Sex Cardiovascular event
Men 0.72 (0.56 to 0.94) .01 Stroke 1 (1) 2 (2) ..99
.41 Transient ischemic attack 7 (8) 7 (8) ..99
Women 0.64 (0.49 to 0.83) .001
Ethnicity Myocardial infarction 2 (2) 4 (4) .69
African American 0.56 (0.37 to 0.84) .005 Arrhythmia 6 (6) 3 (4) .34
.79 Congestive heart failure 2 (4) 1 (1) ..99
Other 0.72 (0.58 to 0.68) .002
Age group, y Angina 9 (10) 17 (19) .16
60-69 0.66 (0.54 to 0.82) ,.001 Other 12 (13) 19 (19) .27
.88 Any cardiovascular event 36 (44) 46 (57) .24
70-80 0.75 (0.50 to 1.14) .18
Weight status Other adverse events
Not overweight 0.75 (0.58 to 0.96) .02 Excessive weight loss 0 (0) 0 (0) ...
.80 Physical injury from exercise 1 (1) 0 (0) ..99
Overweight 0.60 (0.45 to 0.80) ,.001
Palpitations 7 (9) 11 (13) .47
*CI indicates confidence interval; ellipses, data not applicable. Nonischemic chest pain 20 (22) 17 (18) .62
†P for difference in intervention effect across subgroups. Dizziness 24 (29) 15 (17) .14
Edema 14 (15) 21 (22) .30
Excessive weight gain 2 (2) 3 (3) ..99
Headache 35 (36) 54 (63) .04
curred less frequently in the reduced sodium group com- Other 66 (86) 55 (66) .27
pared with the UL group (35 vs 54 individuals, P=.04); Any adverse event 169 (244) 176 (259) .70
otherwise, the occurrence of adverse symptoms was simi-
lar in the 2 groups. Compared with the UL group, in- *Fisher exact tests comparing numbers of individuals reporting events.
take of total energy, dietary fat, saturated fat, monoun-
saturated fat, iron, calcium, thiamin, and riboflavin by 3 months, then the BP reduction observed in TONE
declined in the reduced sodium group, while intake of is similar to that expected, ie, approximately half of the
potassium and magnesium increased (each P,.05; estimated 10– and 4–mm Hg declines in systolic and di-
Table 6). astolic BPs from a 100-mmol/d sodium reduction in per-
sons 60 to 69 years of age.18 TONE did not assess the im-
COMMENT pact of sodium reduction in persons with high BP.
However, it is likely that the extent of BP reduction would
This large randomized controlled trial demonstrated that be greater than that observed in this trial.
free-living, older people with hypertension can reduce their In the design, analysis, and presentation of main trial
sodium intake and that a reduced sodium intake can lower results, the primary outcome variable included clinical,
BP and the need for antihypertensive drug therapy. The ef- potentially BP-related events (eg, stroke and myocardial
fects were consistent in subgroups defined by sex, ethnic- infarction) as well as resumption of medication initi-
ity, and weight status. However, the effects of the inter- ated by either the participant or personal physician for
vention on BP and end points in the age group 70 to 80 reasons other than elevated BP. The rationale for this de-
years did not achieve statistical significance, perhaps as a cision was to minimize the potential for informative cen-
result of small sample size. In dose-response analyses, pro- soring that otherwise might occur in this trial of older
gressively greater reductions in sodium intake were asso- persons had the end point been restricted to only el-
ciated with a reduced risk of a trial end point. In aggre- evated BP. This decision tends to inflate the number of
gate, these data indicate that a modest reduction in sodium end points and might obscure differences between groups.
intake is a feasible and broadly effective nonpharmaco- It is illustrative that in subsidiary analyses, restricted to
logic therapy in older persons. the 334 events that were related only to elevated BP, the
The reduction in sodium intake observed in this trial main trial findings persist.
had a substantial impact on BP, despite the fact that mean In TONE, there was no evidence of an adverse im-
baseline BP, while participants were taking medication, pact of the reduced sodium intervention on the occur-
was within the nonhypertensive range. Specifically, from rence of cardiovascular events. In fact, there was a ten-
a mean baseline BP of 128/71 mm Hg, a reduced sodium dency toward fewer cardiovascular events in the reduced
intake lowered systolic and diastolic BPs by 4.3 and 2.0 sodium intervention group compared with the UL group,
mm Hg, respectively. The extent of sodium reduction at primarily as a result of fewer instances of angina. In terms
3 months after randomization (before drug withdrawal) of adverse symptoms reported by participants, there were
is not known, because the first 24-hour urine collection significantly fewer reports of headaches in the reduced
was obtained 9 months after randomization (about 6 sodium intervention group. No other adverse symptom
months after drug withdrawal). Nonetheless, if the av- achieved statistical significance. The reduced sodium in-
erage 40-mmol/d reduction in sodium intake occurred tervention did have an impact on several aspects of diet.

(REPRINTED) ARCH INTERN MED/ VOL 161, MAR 12, 2001 WWW.ARCHINTERNMED.COM
691
Downloaded from www.archinternmed.com on October 12, 2010
©2001 American Medical Association. All rights reserved.
 Table 6. Daily Nutrient Intake*

Reduced Sodium Usual Lifestyle Between-Group Difference‡

Nutrient Intake/24 h Baseline Within-Group Change† Baseline Within-Group Change† Mean (95% CI) P
Total energy, kcal§ 1810 ± 31.9 −115 ± 2.8 1646 ± 32 4 ± 28 −119 (−197 to −41) .003
Macronutrients, g
Total fat 62.4 ± 1.6 −7.9 ± 1.5 58.1 ± 1.6 −2.1 ± 1.5 −5.8 (−10.1 to −1.5) .008
Saturated fat 20.7 ± 0.6 −3.2 ± 0.5 18.9 ± 0.6 −0.8 ± 0.6 −2.4 (−4.0 to −0.8) .002
Monounsaturated fat 23.9 ± 0.6 −3.3 ± 0.6 22.6 ± 0.6 −1.2 ± 0.6 −2.2 (−4.0 to −0.4) .02
Polyunsaturated fat 13.1 ± 0.4 −1.4 ± 0.5 12.2 ± 0.4 −0.3 ± 0.5 −1.1 (−2.3 to 0.1) .10
Protein 19.8 ± 1.4 −2.8 ± 1.9 66.2 ± 1.4 −1.5 ± 1.4 −1.3 (−5.0 to 2.4) .51
Carbohydrates 242.2 ± 4.8 −1.9 ± 4.1 224.9 ± 4.8 −1.8 ± 4.1 −0.2 (−11.6 to 11.2) .98
Minerals, mg
Potassium 2733 ± 55 110 ± 65 2633 ± 55 −50 ± 49 160 (25 to 295) .02
Magnesium 284.8 ± 6.1 6.8 ± 5.4 277.1 ± 6.1 −16.7 ± 5.4 24 (8 to 39) .002
Iron 15.0 ± 0.4 −1.8 ± 0.4 13.9 ± 0.4 −0.1 ± 0.4 −2.8 (−3.8 to −1.8) ,.001
Calcium 732.0 ± 18.4 −71.8 ± 17.3 640.1 ± 18.2 −1.2 ± 17.4 −71 (−119 to −23) .004
Phosphorus 1124 ± 23 −35 ± 20 1032 ± 22 −8 ± 2 −27 (−84 to 30) .35
Zinc 10.0 ± 0.3 −0.5 ± 0.3 9.5 ± 0.3 −0.2 ± 0.3 −0.3 (−1.1 to 0.5) .50
Vitamins
A, µg 1141 ± 61 −2 ± 66 1133 ± 61 17 ± 67 −19 (−203 to 165) .84
B6, mg 1.81 ± 0.03 0.05 ± 0.05 1.77 ± 0.05 −0.03 ± 0.05 0.1 (−0.1 to 0.2) .26
B12, µg 4.11 ± 0.21 −0.20 ± 0.22 3.72 ± 0.21 0.19 ± 0.22 −0.4 (−1.0 to 0.2) .21
C, mg 126.9 ± 4.6 6.7 ± 4.4 119.3 ± 3.6 −1.6 ± 4.4 8.3 (−3.9 to 20.5) .18
D, µg 4.49 ± 0.19 −0.29 ± 0.20 4.48 ± 0.19 −0.32 ± 0.20 0.0 (−0.5 to 0.6) .92
E, mg 8.30 ± 0.26 −0.64 ± 0.27 8.19 ± 0.27 −0.35 ± 0.27 −0.3 (−1.0 to 0.5) .45
Folate, µg 294.8 ± 8.6 2.4 ± 8.3 290.9 ± 8.5 −9.5 ± 8.3 12 (−11 to 35) .31
Thiamin, mg 1.66 ± 0.03 −0.14 ± 0.03 1.54 ± 0.03 −0.02 ± 0.03 −0.12 (−0.22 to −0.02) .009
Niacin, mg 20.8 ± 0.5 −0.7 ± 0.5 20.3 ± 0.4 −0.2 ± 0.5 −0.4 (−1.8 to 1.0) .51
Riboflavin, mg 1.82 ± 0.26 −0.16 ± 0.04 1.66 ± 0.04 0.00 ± 0.04 −0.2 (−0.3 to −0.1) .002

*Data are given as mean ± SE baseline, mean ± SE within-group change, and mean (95% confidence interval [CI]) between-group differences. Nutrient intake
from supplements not included; in each randomized group, data are available for 315 to 341 participants.
†For each nutrient, average within-group change was the difference between the average of all 24-hour dietary recalls during follow-up minus the average of 2
baseline 24-hour dietary recalls.
‡Within-group difference in reduced sodium group minus within-group difference in usual lifestyle group.
§To convert kilocalories to kilojoules, multiply by 4.184.

For most nutrients, eg, total energy, fat, and potassium,
the changes were favorable. In a few instances, such as a
reduced intake of calcium, the changes were potentially
deleterious. Whether this reduction in calcium intake is
clinically relevant is unclear, because a reduced intake
of sodium has opposite effects on bone demineraliza-
tion.19 Hence, the net impact on bone mineral density is
uncertain. Overall, TONE results reaffirm the safety of
moderate dietary sodium reduction.20
Findings from this trial have important public health
and clinical implications. Because hypertension is com-
mon in the elderly and because sodium reduction can sub-
stantially reduce BP, population-based and individual-
ized efforts to reduce sodium intake are appropriate. The
elderly often live and/or dine in common settings (eg, com-
munity centers, senior centers, or nursing homes). As
such, institutional changes in food preparation can pro-
vide an efficient means to reduce BP in broad popula-
tions. While TONE participants were able to achieve mod-
est reductions in sodium intake through careful selection
of food products, the availability of low-sodium foods was
limited. To facilitate easy access to such items, food manu-
facturers should minimize the addition of sodium and
should use alternative seasonings for flavor.
Successful sodium reduction will require individu-
alized counseling as a routine component of hyperten-
sion management. Unfortunately, Medicare, the pri-
mary insurer of older persons in the United States, does
not cover nutrition therapy for most outpatient condi-
tions such as hypertension. The TONE trial provides con-
vincing evidence that individualized counseling can re-
duce BP and control hypertension. Such evidence supports
current efforts to expand Medicare coverage of nutri-
tion services.21
From a clinical perspective, TONE results indicate
that sodium reduction can control hypertension in a size-
able proportion of medication-controlled patients with
hypertension. Before clinicians attempt medication with-
drawal, candidate patients must be committed to reduc-
ing their sodium intake, ideally in the setting of a super-
vised counseling program. Furthermore, regular BP
monitoring is warranted because many individuals will
require resumption of drug therapy. In another article
from the TONE study, patients with recently diagnosed
and well-controlled hypertension were most likely to be
successful at medication withdrawal.22 Still, physicians
may decide to promote sodium reduction without medi-
cation withdrawal; in this setting, a reduced sodium in-
take can substantially lower BP and presumably de-
crease the risk of atherosclerotic cardiovascular events.
In summary, a reduced sodium intake is a broadly
effective, nonpharmacologic therapy that lowers BP and
controls hypertension in older individuals. Our results,
in combination with the high prevalence of hyperten-

(REPRINTED) ARCH INTERN MED/ VOL 161, MAR 12, 2001 WWW.ARCHINTERNMED.COM
692
Downloaded from www.archinternmed.com on October 12, 2010
©2001 American Medical Association. All rights reserved.
sion and its treatment with medication in the elderly, ar-
gue for substantial efforts to reduce sodium intake in older
persons.
Accepted for publication October 3, 2000.
This work was supported by grants HL02642, HL43641,
HL48642, and HL60197 from the National Heart, Lung, and
Blood Institute; AG09799, AG09771, and AG09773 from
the National Institute on Aging; and RR00722 from the Na-
tional Center for Research Resources of the National Insti-
tutes of Health.
We thank the trial participants and the entire TONE
Collaborative Research Group.
Corresponding author and reprints: Lawrence J. Ap-
pel, MD, MPH, Johns Hopkins University, 2024 E Monu-
ment St, Suite 2-645, Baltimore, MD 21205-2223 (e-mail:
lappel@jhmi.edu).
REFERENCES
1. Burt VL, Whelton P, Roccella EJ, et al. Prevalence of hypertension in the US adult
population: results from the Third National Health and Nutrition Examination Sur-
vey, 1988-1991. Hypertension. 1995;25:305-313.
2. National High Blood Pressure Education Program Working Group. National High
Blood Pressure Education Program Working Group Report on Hypertension in
the Elderly. Hypertension. 1994;23:275-285.
3. Joint National Committee on Prevention, Detection, Evaluation, and Treatment
of High Blood Pressure. The sixth report of the Joint National Committee on Pre-
vention, Detection, Evaluation, and Treatment of High Blood Pressure. Arch In-
tern Med. 1997;157:2413-2446.
4. Applegate WB, Miller ST, Elam JT, et al. Nonpharmacologic intervention to re-
duce blood pressure in older patients with mild hypertension. Arch Intern Med.
1992;152:1162-1166.
5. Fotherby MD, Potter JF. Effects of moderate sodium restriction on clinic and twenty-
four-hour ambulatory blood pressure in elderly hypertensive subjects. J Hyper-
tens. 1993;11:657-663.
6. Cappuccio FP, Markandu ND, Carney C, Sagnella GA, MacGregor GA. Double-
blind randomised trial of modest salt restriction in older people. Lancet. 1997;
350:850-854.
(REPRINTED) ARCH INTERN MED/ VOL 161, MAR 12, 2001
693
Downloaded from www.archinternmed.com on October 12, 2010
©2001 American Medical Association. All rights reserved.
7. Morgan T, Anderson A. Sodium restriction can delay the return of hypertension
in patients previously well-controlled on drug therapy. Can J Physiol Pharma-
col. 1987;65:1752-1755.
8. Kelly R, Hayward C, Avolio A, O’Rourke M. Noninvasive determination of age-
related changes in the human arterial pulse. Circulation. 1989;80:1652-1659.
9. Luft FC, Weinberger MH, Fineberg NS, Miller JZ, Grim CE. Effects of age on re-
nal sodium homeostasis and its relevance to sodium sensitivity. Am J Med. 1987;
82:9-15.
10. Whelton PK, Appel LJ, Espeland MA, et al. Efficacy of sodium reduction and weight
loss in the treatment of hypertension in older persons: main results of the ran-
domized, controlled Trial of Nonpharmacologic Interventions in the Elderly (TONE).
JAMA. 1998;279:839-846.
11. Appel LJ, Espeland MA, Whelton PK, et al. Trial of Nonpharmacologic Interven-
tions in the Elderly (TONE): design and rationale of a blood pressure control trial.
Ann Epidemiol. 1995;5:119-129.
12. Whelton PK, Bahnson J, Appel LJ, et al. Recruitment in the Trial of Nonpharma-
cologic Interventions in the Elderly (TONE). J Am Coll Surg. 1997;45:185-193.
13. Schakel SF, Sievart YA, Buzzard IM. Sources of data for developing and main-
taining a nutrient database. J Am Diet Assoc. 1988;10:1268-1271.
14. National Institutes of Health. Health implications of obesity: consensus confer-
ence statement. Ann Intern Med. 1985;103(6, pt 2):1073-1077.
15. Langford HG, Blaufox D, Oberman A, et al. Dietary therapy slows the return of
hypertension after stopping prolonged medication. JAMA. 1985;253:657-664.
16. Stamler R, Stamler J, Gosch FC, et al. Primary prevention of hypertension by
nutritional-hygienic means: final report of a randomized, controlled trial. JAMA.
1989;262:1801-1807.
17. The Trials of Hypertension Prevention Collaborative Research Group. Effects of
weight loss and sodium reduction intervention on blood pressure and hyperten-
sion incidence in overweight people with high-normal blood pressure: the Trials
of Hypertension Prevention, phase 2. Arch Intern Med. 1997;157:657-667.
18. Law MR, Frost CD, Wald NJ. By how much does salt reduction lower blood pres-
sure? I: analysis of observational data among populations. BMJ. 1991;302:811-
815.
19. Devine A, Criddle RA, Dick IM, Kerr DA, Prince RL. A longitudinal study of the
effect of sodium and calcium intakes on regional bone density in postmeno-
pausal women. Am J Clin Nutr. 1995;62:740-745.
20. Kumanyika SK, Cutler JA. Dietary sodium reduction: is there cause for concern?
J Am Coll Nutr. 1997;16:192-203.
21. Institute of Medicine. The Role of Nutrition in Maintaining Health in the Nation’s
Elderly: Evaluating Coverage of Nutrition Services for the Medicare Population.
Washington, DC: National Academy Press; 2000.
22. Espeland MA, Whelton PK, Kostos JB, et al. Predictors and mediators of suc-
cessful long-term withdrawal from antihypertensive medications. Arch Fam Med.
1999;8:228-236.
WWW.ARCHINTERNMED.COM