does withdrawal of antihypertensive medication increase the risk of cardiovascular events?

Does Withdrawal of AntihypertensiveMedication Increase the Risk of

Cardiovascular Events?John B. Kostis, MD, Mark A. Espeland, PhD, Lawrence Appel, MD,

Karen C. Johnson, MD, MPH, June Pierce, AB, andJames L. Wofford, MD, for the Trial of Nonpharmacologic Interventions in the Elderly

(TONE) Cooperative Research Group

The Fifth Report of the Joint National Committee onDetection, Evaluation, and Treatment of High Blood Pres-sure recommends that attempts to discontinue antihy-pertensive drug therapy be considered after blood pres-sure (BP) has been controlled for 1 year. However,discontinuation of drug therapy could unmask underly-ing conditions and precipitate clinical cardiovascularevents. The Trial of Nonpharmacologic Interventions inthe Elderly (TONE) was a clinical trial of the efficacy ofweight loss and/or sodium reduction in controlling BPafter withdrawal of drug therapy in patients with aBP<145/85 mm Hg on 1 antihypertensive medication.Of 975 participants, 886 entered the drug withdrawalphase of the trial and 774 were successfully withdrawnfrom their medications. Thirty-three events (stroke, tran-sient ischemic attack, myocardial infarction, arrhythmia,

congestive heart failure, angina, other) occurred be-tween randomization and the onset of drug withdrawal(median time 3.6 months), 57 events occurred eitherduring or after drug withdrawal (14.0 months), and 36events occurred after resumption of antihypertensivetherapy (15.9 months). Event rates per 100 person-years were 5.5, 5.5, and 6.8 for the 3 time periods (p 50.84) in the nonoverweight group and 7.2, 5.2, and 5.6(p 5 0.08) in the overweight group. The study showsthat antihypertensive medication can be safely with-drawn in older persons without clinical evidence of car-diovascular disease who do not have diastolic pressure>150/90 mm Hg at withdrawal, providing that goodBP control can be maintained with nonpharmacologictherapy. Q1998 by Excerpta Medica, Inc.

(Am J Cardiol 1998;82:1501–1508)

The Fifth Report of the Joint National Committee onDetection, Evaluation, and Treatment of High

Blood Pressure recommends that after blood pressure(BP) has been controlled for 1 year, “step down” ordiscontinuation of antihypertensive drug therapy med-ication be considered, especially in the setting of life-style changes that reduce BP.1 However, antihyper-tensive medications may possess additional propertiesthat may treat or mask cardiovascular conditions suchas angina (e.g.,b blockers, calcium channel blockers)or heart failure (e.g., diuretics, angiotensin-convertingenzyme inhibitors).2–8 Thus, discontinuation of theseagents when not needed for control of BP after non-drug therapy may unmask underlying conditions orprecipitate clinical events such as angina, myocardialinfarction, or heart failure. In addition, less effectiveBP control during or after drug withdrawal may in-crease the risk of cardiovascular events. The Trial of

Nonpharmacologic Interventions in the Elderly(TONE) demonstrated the efficacy of sodium reduc-tion and/or weight loss as a means to control BP inolder-aged patients.9–11 The objectives of this studywere to report on the occurrence of cardiovascularevents before, during, and after drug withdrawal, andafter resumption of antihypertensive drug therapy inTONE and to identify predictors of events.

METHODSTONE was a 4-center, randomized, controlled clin-

ical trial designed to study the safety and effectivenessof nonpharmacologic therapy in maintaining satisfac-tory BP control in persons with hypertension aged 60to 80 years who were withdrawn from antihyperten-sive medication. The study used the 2 most promisingnonpharmacologic interventions to control BP in el-derly persons: weight loss and sodium reduction.Community dwelling persons who were currently on 1or 2 antihypertensive medications were recruited forparticipation in TONE at 4 clinical sites—Baltimore,Maryland (Johns Hopkins Health Institutions); Mem-phis, Tennessee (University of Tennessee at Mem-phis); New Brunswick, New Jersey (UMDNJ-RobertWood Johnson Medical School); and Winston-Salem,North Carolina (Bowman Gray School of Medicine)—from September 1992 through April 1994. Designcharacteristics, eligibility criteria, and the TONE re-cruitment experience have been described else-where.9,10 Major inclusion criteria were average base-line systolic BP ,145 mm Hg and average dias-

From the UMDNJ-Robert Wood Johnson Medical School, New Bruns-wick, New Jersey; Bowman Gray School of Medicine, Winston-Salem, North Carolina; The Johns Hopkins Medical Institutions, Balti-more, Maryland; and University of Tennessee, Memphis, Tennessee.This study was supported by Grants R01-HL48641, R01-HL48642,and K08-HL02642 from the National Heart, Lung, and Blood Instituteand Grants R01-AG09773, R01-AG09799, R01-AG09771, andR03-HL60197 from the National Institute on Aging, Bethesda, Mary-land. Manuscript received March 24, 1998; revised manuscript re-ceived and accepted July 20, 1998.

Address for reprints: John B. Kostis, MD, UMDNJ-Robert WoodJohnson Medical School, One Robert Wood Johnson Place, P.O. Box19, New Brunswick, New Jersey 08903-0019. E-mail: [email protected].

1501©1998 by Excerpta Medica, Inc. 0002-9149/98/$19.00All rights reserved. PII S0002-9149(98)00694-8

TABLE I Baseline Sociodemographic, Lifestyle, Clinical, and Family Medical History Characteristics of RandomizedParticipants in TONE

AttributeUsual Care(n 5 341)

Weight Loss(n 5 147)

Sodium Reduction(n 5 340)

Combined(n 5 147)

Total(n 5 975) p Value

Age (yr) 65.9 6 4.5 65.8 6 4.7 65.8 6 4.6 65.7 6 4.8 65.8 6 4.6 0.9960–69 78.5% 76.2% 79.4% 81.0% 78.9% 0.8070–79 21.5% 23.8% 20.6% 19.0% 21.1%

GenderMen 53.1% 49.0% 50.9% 56.5% 52.2% 0.57Women 46.9% 51.0% 49.1% 43.5% 47.8%

Ethnic backgroundAfrican-American 24.0% 26.5% 21.8% 23.8% 23.6%Caucasian 75.4% 73.5% 77.6% 76.2% 76.0% 0.80Other 0.6% 0.0% 0.6% 0.0% 0.4%

Marital statusNever married 2.9% 1.4% 3.8% 3.4% 3.1%Currently married 74.8% 68.0% 75.6% 69.4% 73.2% 0.42Divorced/separated 8.2% 10.2% 7.6% 8.8% 8.4%Widowed 14.1% 20.4% 12.9% 18.4% 15.3%

Cigarette smokingNever smoked 57.5% 56.5% 51.8% 51.7% 54.5%Former smoker 36.7% 40.8% 42.9% 41.5% 40.2% 0.45Current smoker 5.9% 2.7% 5.3% 6.8% 5.3%

Years since Htn DxOverweight (n 5 585) 13.6 6 9.5 13.8 6 11.6 13.9 6 10.7 13.4 6 10.5 13.7 6 10.5 0.98Nonoverweight (n 5 390) 13.5 6 10.0 13.5 6 10.3 13.5 6 10.2 0.99

Family medical historyHigh BP 72.1% 61.9% 69.4% 66.0% 68.7% 0.13Stroke 38.4% 36.7% 39.1% 33.3% 37.6% 0.66MI or angina 37.8% 31.3% 38.8% 44.2% 38.2% 0.16

Systolic BP (mm Hg)Overweight 127.8 6 9.6 129.1 6 9.1 128.3 6 9.4 128.4 6 9.4 128.4 6 9.4 0.66

,125 38.1% 30.6% 30.6% 33.3% 33.2%125–134 35.4% 36.0% 41.0% 37.4% 37.4% 0.731351 26.5% 33.3% 28.5% 29.2% 29.4%

Nonoverweight 127.4 6 9.3 128.4 6 9.4 127.9 6 9.3 0.33,125 37.1% 31.0% 34.1%125–134 34.5% 37.4% 35.9% 0.461351 28.4% 31.6% 30.0%

Diastolic BP (mm Hg)Overweight 71.6 6 7.0 71.5 6 7.8 71.5 6 7.6 71.5 6 6.8 71.1 6 7.3 0.99

,65 14.3% 13.6% 16.7% 15.6% 15.0%65–74 49.7% 47.6% 45.1% 49.0% 47.9% 0.98751 36.0% 38.8% 38.2% 35.4% 37.1%

Nonoverweight 71.1 6 7.4 71.2 6 7.3 71.5 6 7.3 0.85,65 18.0% 20.4% 19.2%65–74 47.9% 39.3% 43.6% 0.72751 34.0% 40.3% 37.2%

Heart rate (per 30 sec)Overweight 35.2 6 5.0 34.7 6 5.0 35.8 6 5.1 34.4 6 5.0 35.0 6 5.0 0.09Nonoverweight 33.2 6 4.8 33.5 6 4.6 33.4 6 4.7 0.67

Body mass index (kg/m2)Overweight

Men (n 5 279) 30.2 6 1.6 30.4 6 1.7 30.1 6 1.8 30.8 6 1.6 30.4 6 1.7 0.07Women (n 5 306) 31.7 6 2.7 31.8 6 2.7 31.9 6 2.7 31.9 6 2.6 31.8 6 2.7 0.95

NonoverweightMen (n 5 230) 25.7 6 1.8 25.7 6 1.8 25.7 6 1.8 0.91Women (n 5 160) 25.1 6 1.7 25.0 6 1.7 25.0 6 1.7 0.56

Waist/hip ratioOverweight

Men 1.00 6 0.04 1.01 6 0.05 1.00 6 0.05 1.01 6 0.05 1.01 6 0.05 0.72Women 0.90 6 0.11 0.92 6 0.09 0.91 6 0.10 0.92 6 0.08 0.91 6 0.10 0.73

NonoverweightMen 0.98 6 0.06 0.97 6 0.08 0.98 6 0.07 0.70Women 0.88 6 0.11 0.88 6 0.11 0.88 6 0.11 0.90

Medical historyHeart attack 2.4% 4.1% 2.1% 2.0% 2.5% 0.59Angina 0.3% 0.7% 0.3% 0.7% 0.4% 0.60CHF 0.3% 0.0% 0.3% 0.0% 0.2% 0.83Stroke 2.4% 0.7% 2.1% 2.7% 2.1% 0.60Diabetes mellitis 3.5% 4.1% 5.3% 2.7% 4.1% 0.53

Values are expressed as mean 6 SD or number (%).Dx 5 diagnosis; Htn 5 hypertension; MI5 myocardial reinfarction.

1502 THE AMERICAN JOURNAL OF CARDIOLOGYT VOL. 82 DECEMBER 15, 1998

tolic BP ,85 mm Hg on 1 medication, willingness toparticipate in all aspects of the study; independence inactivities of daily living; capacity to alter diet andphysical activity in accordance with the requirementsof any TONE intervention as evaluated by TONEintervention staff; and approval from the participant’spersonal physician. Patients treated with 2 antihyper-tensive medications were eligible for recruitment aslong as they met BP eligibility criteria after step-downfrom 2 to 1 medication. For practical purposes, com-bination drugs consisting of a diuretic and nondiureticagent were considered to be a single medication. Ma-jor exclusion criteria were history of a heart attack orstroke within the preceding 6 months, current anginapectoris, congestive heart failure, insulin-dependentdiabetes mellitus, serious psychiatric or physical ill-ness, unexplained or involuntary weight loss$10 lbsduring the previous year, a body mass index (BMI;weight in kg divided by the square of height in meters),21 kg/m2 or .33 kg/m2 (men) or.37 kg/m2 (wom-en), high creatinine levels (.2.0 mg/dl), hyperkalemia(.5.5 mmol/L), hyperglycemia (nonfasting glucose.260 mg/dl), or anemia (hemoglobin,11 g/dl).

Eligible participants were stratified by BMI:Women with BMI .27.3 kg/m2 and men with BMI.27.8 kg/m2 constituted the overweight cohort; oth-ers constituted the nonoverweight cohort. Overweightparticipants were randomized with equal probability

to 1 of 4 groups: usual care, dietarysodium reduction alone, weight lossalone, or combined sodium reductionand weight loss. Nonoverweight par-ticipants were randomized withequal probability to 1 of 2 groups:usual care or sodium reduction. Ac-tive interventions involved individ-ual and group counseling sessions,which began within 1 month of ran-domization.9–11 The goal of weightloss interventions was achievementand maintenance of$10-lb weightloss; the goal of the sodium reduc-tion interventions was achievementand maintenance of#80 mmol/24-hour intake of sodium. The medica-tion withdrawal process was stan-dardized across the 4 clinical centersand used drug-specific tapering regi-mens. Ninety days (6 2 weeks) afterdietary intervention was begun, par-ticipants began standardized drug-specific weaning protocols that in-volved weekly clinic visits. When nocomplications occurred, this protocolresulted in the participant being offdrugs within 1 to 5 weeks of the startof drug withdrawal. After discontin-uation of antihypertensive medica-tion, participants had 3 additional bi-weekly visits to confirm BP control(systolic BP,150 mm Hg; diastolicBP ,90 mm Hg). For some patients,

symptoms or increases in BP prolonged drug with-drawal, or resulted in termination of the attempteddrug withdrawal. At each drug withdrawal visit, par-ticipants were queried concerning symptoms and/orevents that had occurred since their last visit, and BPwas measured. Participants whose drug withdrawalprocess was interrupted or otherwise not completedwere told to revert to their original medication anddose. The choice of medication for participants whohad been successfully weaned but who subsequentlyresumed medication was left to the discretion of theirpersonal physicians.

After drug withdrawal, clinic visits occurred every3 months for 11.9 to 31.6 months (average 23.7) afterdrug withdrawal. Each visit included a brief medicalhistory and questionnaires designed to elicit responsesconcerning any intercurrent cardiovascular events.Event data described in this report were derived fromthese clinic visits, from drug withdrawal visits, andfrom unscheduled visits or reports of intercurrentevents. Each adverse event report and relevant medi-cal records were reviewed and adjudicated by clini-cians at the clinical site according to a standard pro-tocol.

For purposes of this analysis, follow-up time wasdivided into 3 spans: (1)on medication:from time ofrandomization until the first drug withdrawal visit (oruntil time of last contact, if the drug withdrawal visit

TABLE II Distribution of Cardiovascular Events by Period of Study and WeightCohort at Baseline

Type of EventOn Medication

Before DW

Off Medicationfrom Start of DW/Before End Point

MedicationResumedAfter DW

After End Point

NonoverweightStroke 0 0 2TIA 1 7 0Myocardial infarction 0 0 3Arrhythmia 1 3 0Congestive heart failure 1 0 0Angina 4 8 6Other* 4 6 7

Total 11 24 18

OverweightStroke 1 0 1TIA 1 2 2Myocardial infarction 1 3 2Arrhythmia 2 2 2Congestive heart failure 1 1 0Angina 6 18 5Other* 10 7 6

Total 22 33 18

*Nonobese: aortic aneurysm, coronary artery disease by cardiac cath (asymptomatic), carotid arterysurgery, central retinal vein occlusion, change on electrocardiography, chest pain, elevated BP/tachycardia, heart murmur, hypertension, carotid endarterectomy, pericarditis, pericarditis/myocarditis,repair abdominal aortic aneurysm, shortness of breath with tachycardia, sudden death/probablemyocardial infarction, syncope. Obese: abdominal aortic aneurysm, accelerated hypertension, coro-nary artery bypass graft surgery, coronary artery disease by cardiac cath, cardiomyopathy (unknownetiology), chest pain, dissecting aortic aneurysm, endocarditis, episodes of aphasia, hypertension,peripheral vascular disease, possible anginal equivalent, pulmonary embolus, syncope, tachycardia/hyperthyroidism.

DW 5 drug withdrawal.

SYSTEMIC HYPERTENSION/CV EVENTS AFTER STOPPING ANTIHYPERTENSIVE DRUGS 1503

did not occur), this interval includes the time betweenrandomization and onset of intervention and the inter-vention time of 90 days6 2 weeks; (2)off medica-tion: from first drug withdrawal visit until last studycontact or reinstatement of drug therapy (median 14.0months); and (3)medication resumed:from the timeof reinstitution of drug therapy until last study contact(median 15.9 months) for those who were successfullyweaned, but were subsequently placed back on anti-hypertensive therapy. The total numbers of events andthe distribution of the types of events were tallied inthese intervals. In these tallies, all participants con-tributed to the first time span; only participants whowere withdrawn from antihypertensive medication orattended at least 1 drug withdrawal visit contributed tothe second time span; and only persons who wereweaned but subsequently were represcribed medica-tion contributed to the third time span. Mean rates/100person-years based on time spent in each span werecalculated using Poisson regression,12 and comparedbetween the overweight and nonoverweight cohorts.Statistical comparisons of event rates across timespans were more difficult because the lengths of timepatients spent in time spans were influenced, in some

cases, by the occurrence or nonoc-currence of events. Because of this, anonparametric approach was adopt-ed: rerandomization tests13 based onFreeman-Tukey test statistics14 wereused to contrast event rates amongthe 3 time periods.

For participants who started drugwithdrawal, lifetable analyses wereprepared to describe the distributionof times from the first drug with-drawal visit until the first incidentcardiovascular event.15 Event-freeparticipants were censored in theseanalyses at the time of their laststudy contact. Times of events weremeasured from the time of successfuldrug weaning, i.e., when the partici-pant stopped pharmacologic therapy.Events that occurred during drugwithdrawal (including those occur-ring to persons who were not suc-cessfully withdrawn from medica-tions) were assigned the time 0. Toassess the importance of baselinefactors on the occurrence of cardio-vascular events, proportional-haz-ards regression was performed andthe relative hazards associated withdifferent subgroups were calculat-ed.16,17 A multivariate proportional-hazards model was developed inwhich predictive factors were en-tered and maintained if they had in-dependent predictability associatedwith a nominal p value,0.05.

RESULTSNine hundred seventy-five participants were ran-

domized to 1 of the following 6 study groups: non-overweight controls (n5 194); nonoverweight so-dium reduction (n5 196); overweight controls (n5147); overweight sodium reduction (n5 144); over-weight weight loss (n5 147); and overweight com-bined weight loss/sodium reduction (n5 147). Ofthese, 886 (90.9%) entered the drug withdrawal pro-gram from their antihypertensive medication, and 774(79.4%) of the entire cohort were successfully with-drawn from their antihypertensive medication. Table Ilists selected baseline characteristics including historyof cardiovascular disease at baseline. Baseline char-acteristics were similar across clinical centers. Thiscohort represents a rather healthy group of older hy-pertensive patients with low prevalence of coronaryartery disease (2.5%), congestive heart failure (0.2%),or stroke (2.1%) at baseline.

Table II describes the types of events that occurredduring these time spans in the nonoverweight andoverweight cohorts. Thirty-three events occurred be-fore the onset of drug withdrawal (median time 3.6months: 1 stroke, 2 transient ischemic attacks [TIAs],1 myocardial infarction, 3 arrhythmias, 2 congestive

TABLE III Incidence of Cardiovascular Events Before, During, and AfterWithdrawal from Antihypertensive Medications, by Weight Cohort at Baseline

Any CVD (includes stroke, TIA, MI, arrhythmia,CHF, angina, and other CVD)

On MedicationBefore DW*

Off Medicationfrom Start of DW

and Before End Point†

MedicationResumed

After DW AfterEnd Point‡

Nonoverweight

Number of persons 390 354 186Number of events 11 24 18Persons with 1 event (%) 5 (1.3%) 22 (6.2%) 13 (7.0%)Persons with 2 events (%) 3 (0.8%) 1 (0.3%) 1 (0.5%)Persons with 31 events (%) 0 (0.0%) 0 (0.0%) 1 (0.5%)Person-years of follow-up 189.1 408.6 244.8Event rates/100 person-years

of follow-up (95% CI)§5.5 (2.9, 9.1) 5.5 (3.7, 7.9) 6.8 (4.3, 10.1)

Overweight

Number of persons 585 531 248Number of events 22 33 18Persons with 1 event (%) 16 (2.7%) 31 (5.8%) 12 (4.8%)Persons with 2 events (%) 1 (0.2%) 1 (0.2%) 3 (1.2%)Persons with 31 events (%) 1 (0.2%) 0 (0.0%) 0 (0.0%)Person-years of follow-up 282.3 606.9 305.0Event rates/100 person-years

of follow-up (95% CI)7.2 (4.7, 10.3) 5.2 (3.6, 7.0) 5.6 (3.5, 8.3)

*All randomized participants: from time of randomization to first drug withdrawal visit (or, if no drugwithdrawal visit, until time of last contact).

†Only participants who had at least 1 drug withdrawal visit: from time of first drug withdrawal visit untillast study contact or end point (i.e., resumption of drug therapy).

‡Only participants who were successfully weaned from drug therapy and subsequently became endpoints (i.e., resumed drug therapy): from time of end point until last study contact.

§Computed using Poisson regression.Comparisons of rates between overweight and nonoverweight cohorts: On medication (p 5 0.43), off

medication (p 5 0.77), medication resumed (p 5 0.51).CHF 5 congestive heart failure; CI 5 confidence interval; CVD 5 cardiovascular disease; MI 5

myocardial infarction; other abbreviation as in Table I.


heart failures, 10 anginas, and 14 other events speci-fied in Table II) compared with 57 during and afterdrug withdrawal (14.0 months: 9 TIAs, 3 myocardialinfarctions, 5 arrhythmias, 1 congestive heart failure,25 anginas, and 14 other events) and 36 after resump-tion of antihypertensive therapy (15.9 months: 3

strokes, 2 TIAs, 5 myocardial infarc-tions, 2 arrhythmias, 11 angina, and13 other events). The combined TIA/angina event rates/100 person-yearswere slightly higher while partici-pants were off drugs (3.0 off drugs vs2.4 on drugs in the overweight groupand 3.5 vs 2.6 in the nonoverweightgroup, p.0.05).

Mean event rates for the 3 timeperiods are listed in Table III. Thesewere similar for the nonoverweightgroup: 5.5, 5.5, and 6.8/100 person-years of follow-up (p5 0.84, basedon a rerandomization test). Amongoverweight participants, events wereslightly more frequent before medi-cation withdrawal than during theother 2 time periods: 7.2 versus 5.2and 5.6/100 person-years of follow-up; however, this difference did notreach statistical significance (p50.08). Based on Poisson regression,during none of the 3 time spans wasthe difference between the eventrates of the overweight versus non-overweight cohorts statistically sig-nificant (p.0.40). Table III also liststhe numbers of persons who experi-enced 1, 2, or.2 events during eachof the time periods. During the rela-tively short period of time beforedrug withdrawal, 26 of 975 partici-pants (2.76 0.5%) experienced$1cardiovascular disease event. Duringdrug withdrawal and the typicallylonger period while off medication,this proportion was 55 of 885 partic-ipants (6.26 0.8%); after medica-tion was resumed the proportion was30 of 434 participants (6.96 1.2%).

Table IV describes the resultsfrom univariable proportional-haz-ards regression models fitted to eventtimes. There were relatively smalldifferences in the event rates amongsubgroups defined by obesity, gen-der, ethnicity, number of antihyper-tensive drugs at first screening con-tact, class of antihypertensive medi-cation, systolic pressure at time ofdrug withdrawal, family history ofheart disease, smoking or interven-tion assignment (p.0.20). Postwith-drawal rates of angina were higheramong participants who were with-

drawn fromb blockers and calcium channel blockers.However, the trend of this post hoc comparison doesnot reach statistical significance (p5 0.27): relativehazards ratio 2.106 1.71 forb blockers, 1.786 1.71for calcium channel blockers, 1.01 for angiotensin-converting enzyme I, 0.78 for diuretics, and 1.00 for

TABLE IV Results from Proportional-Hazards Regression to Identify Predictors ofCardiovascular Events: Results from Proportional Hazards Regression of TimesUntil First Event Among All Persons Entering Drug Withdrawal

No. ofPatients

Relative-HazardsRatios (6SE) p Value

Predictors 354* 1.16 6 0.26† 0.51532† 1.00†

Clinic Site1 277 1.06 6 0.382 211 2.08 6 0.69 0.023 219 0.92 6 0.354 179 1.00

Age (yr)60–64 398 0.45 6 0.1265–69 299 0.37 6 0.11 0.00170–79 189 1.00

GenderMen 469 1.14 6 0.26 0.55Women 417 1.00

EthnicityAfrican-American 211 0.72 6 0.21 0.25Other 675 1.00

Number of BP drugs during screening1 751 1.11 6 0.36 0.752 135 1.00

Class of antihypertensive medicationACE inhibitor 193 1.18 6 0.67b blocker 95 1.41 6 0.86Ca channel blocker 242 2.04 6 1.09 0.35Diuretic 291 1.35 6 0.73Other 65 1.00

Systolic BP at first drug withdrawal visit (mm Hg),145 823 1.10 6 0.511451 62 1.00 0.83

Diastolic BP at first drug withdrawal visit (mm Hg),85 840 0.54 6 0.21851 45 1.00 0.12

Years diagnosed with hypertension0–4 610 0.47 6 0.185–9 231 1.05 6 0.28 0.07101 45 1.00

Years prescribed antihypertensive drugs0–4 201 0.55 6 0.185–9 481 1.12 6 0.29 0.10101 204 1.00

Family history of heart diseaseNo 549 0.88 6 0.20 0.59Yes 337 1.00

CVD symptoms at baselineNo 619 0.67 6 0.15 0.08Yes 267 1.00

Smoking statusCurrent 48 1.14 6 0.54Former 353 0.87 6 0.21 0.77Never 485 1.00

Intervention assignmentSodium/weight 137 0.95 6 0.32Sodium only 316 0.71 6 0.20 0.61Weight only 137 0.96 6 0.32Usual care 296 1.00

*Nonoverweight.†Overweight.ACE 5 angiotensin-converting enzyme; Ca 5 calcium; other abbreviations as in Table III.


other (p5 0.27). Participants who were aged 60 to 64and 65 to 69 years had less than half the hazard forevents as participants who were aged 70 to 79 years(p 5 0.001). For participants aged 70 to 79 years, theevent rate/100 person-years without medication was11.6 in the nonoverweight group and 11.3 for theoverweight group compared with 12.8 and 10.9, re-spectively, before withdrawal. Significant clinic-spe-cific event rates also existed (p5 0.02). Participantswho had been diagnosed and pharmacologicallytreated for hypertension for,5 years had about halfthe hazard for events of those diagnosed or treatedlonger (p5 0.07 and 0.10, respectively). Participantswho reported symptoms of cardiovascular disease be-fore the start of the study were marginally more likelyto experience events (p5 0.08). Those with diastolicBP ,85 mm Hg also appeared to be at less risk forevents (p5 0.12). A significant relation between BPcontrol and events was not observed (p5 0.17 forsystolic and p5 0.53 for diastolic BP). Patients whoseBP increased above the TONE safety limits (150/90mm Hg) during drug withdrawal were excluded fromanalyses of postwithdrawal effects. Those who werelost to follow-up contributed to events and person-years of follow-up in analyses shown in Table III.They were censored at the end of their observedfollow-up in life-table analyses (Table IV). Occur-

rence of a cardiovascular disease event during at-tempted drug withdrawal led to suspension of theweaning process. This resulted in higher event ratesamong persons for whom drugs were not successfullywithdrawn.

Figures 1 and 2 present univariable life-table plotsfor the distribution of event times for participantsgrouped by age and years diagnosed with hyperten-sion. As can be seen from these figures, the incidenceof (first) events was spread fairly uniformly acrossfollow-up.

Multivariable analysis was performed to identifyimportant sets of independent predictors of events.Three predictors were identified, each associated witha nominal significance level of p,0.05 after control-ling for the other 2 factors: age (p5 0.0008), diastolicBP (p 5 0.04), and clinic site (p5 0.01). Based onthis model, older persons and those with diastolic BP.85 mm Hg had a higher adjusted risk of events.Years after diagnosis (p5 0.08) and treatment (p50.09) for hypertension, history of cardiovascular dis-ease symptoms (p5 0.09), and medication class (p50.19) appeared to be the most important of the remain-ing predictors. There was a slight trend for patientswho were on calcium channel blockers to develop acardiovascular disease event during or after drug with-drawal compared with other agents (p5 0.19).

FIGURE 1. Distribution of times from start of drug withdrawal until the first cardiovascular event for participants grouped by age atbaseline.


DISCUSSIONTONE has demonstrated the efficacy of weight loss

and sodium reduction in the treatment of hyperten-sion.11 At 30 months after attempted drug withdrawal,37.8% of those assigned to 1 of the 2 sodium inter-ventions (alone or combined with weight loss) contin-ued to demonstrate successful BP control, comparedwith 24.4% of those not assigned to sodium interven-tion (usual care or weight loss only). Among theoverweight participants, 39.2% of those assigned toweight loss (alone or combined with sodium reduc-tion) continued to demonstrate successful BP controlcompared with 26.2% of those not assigned to weightloss. Successful BP control at 30 months was achievedby 43.6% of the overweight participants assigned toreceive the combined sodium/weight intervention. Inthis rather healthy hypertensive cohort, there was noapparent increase in cardiovascular events during orafter withdrawal of antihypertensive medication. Thiswas true regardless of whether all cardiovascularevents were considered or the expected events foreach medication class were analyzed (i.e., angina forb blockers or calcium channel blockers, heart failurefor diuretics or angiotensin-converting enzyme inhib-itors, etc., data not shown). This indicates that whenBP is controlled by non-drug therapy, antihyperten-sive drug withdrawal may be done safely (either forclinical or research reasons), provided that there is no

active cardiovascular disease and provided that BP isclosely monitored during and after drug withdrawal.The absence of a higher event rate during and afterdrug withdrawal in this study is probably due to goodBP control with nonpharmacologic therapy (averageBP was 127.56 11.8/71.36 8.7 mm Hg at baseline,was lower by 3.56 0.5/1.96 0.3 mm Hg before drugwithdrawal, higher by 6.16 0.4/2.76 0.3 mm Hg offdrug therapy, and higher by 1.56 0.5/20.16 0.3 mmHg after resumption of drug therapy), to the lowpercentage of patients with history of myocardial in-farction (2.5%) or stroke (2.1%), and to the exclusionof patients with current angina, congestive heart fail-ure, insulin-dependent diabetes, and those with hyper-tension requiring.2 medications for control. Thesedata, therefore, document the safety of discontinuingantihypertensive drug therapy in older persons underthese conditions and support the “step down” strategyof the fifth report of the Joint National Committee.1

Special attention should be paid to higher risk subsetssuch as those with higher diastolic BP ($85 mm Hg,5% of the population in this study) at first withdrawalvisit and aged$70 years (21% of the population). Insuch persons, clinical cardiovascular events weremore common. However, there was no evidence thatmedication withdrawal contributed to the occurrenceof events. Age and BP are well known risk factors forevents among hypertensive patients.18,19It is also pos-

FIGURE 2. Distribution of times from start of drug withdrawal until the first cardiovascular event for participants grouped by yearsdiagnosed with hypertension (<5 vs >5 years).


sible that the higher rate of events in 1 clinical centeris due to a more careful ascertainment of events at thatclinical center.

Limitations of this study are the lack of random-ization to drug withdrawal versus no drug withdrawalafter non-drug therapy and the rather small number ofevents. The primary objective of TONE was to deter-mine the effectiveness of non-drug therapy in control-ling BP rather than preventing events. Although thesample size of TONE was large, the relatively lowincidence of events over the 3 years of follow-uplimits the statistical power of our analyses. However,despite these limitations, this analysis is consonantwith previous studies of antihypertensive drug with-drawal20–27 and suggests that antihypertensive medi-cation can be safely withdrawn in older persons with-out known cardiovascular disease provided that theydo not have BP.150/90 mm Hg at withdrawal, andgood BP control is maintained with nonpharmacologictherapy. The study supports the step-down recommen-dation of the Joint National Committee as well as thesafety of conducting similar clinical trials.

1. National High Blood Pressure Education Program. The Fifth Report of theJoint National Committee on Detection, Evaluation, and Treatment of HighBlood Pressure.Arch Intern Med1993;153:154–183.2. Cohn JN. The management of chronic heart failure.N Engl J Med1996;335:490–498.3. Taylor SH. Refocus on diuretics in the treatment of heart failure.Eur Heart J1995;16(suppl F):7–15.4. Pitt B. Pharmacologic treatment of heart failure with standard drugs.Curr OpinCardiol 1994;9:289–294.5. Thadani U, Chohan A. Chronic stable angina pectoris. Strategies for effectivedrug therapy.Postgrad Med1995;98:175–176, 179–183, 187–188.6. Kostis JB. Angiotensin converting enzyme inhibitors. II. Clinical use.AmHeart J 1988;116:1591–1605.7. Opie LH. Calcium channel antagonists in the management of anginal syn-dromes: changing concepts in relation to the role of coronary vasospasm.ProgCardiovasc Dis1996;38:291–314.8. Walma EP, Hoes AW, van Dooren C, Prins A, van der Does E. Withdrawal oflong term diuretic medication in elderly patients: a double blind randomised trial.Br Med J1997; 315:464–468.9. Appel LJ, Espeland M, Whelton PK, Dolecek T, Kumanyika S, Applegate WB,

Ettinger WH, Kostis JB, Wilson AC, Lacy C, Miller ST. Trial of Nonpharma-cologic Intervention in the Elderly (TONE): design and rationale of a bloodpressure control trial.Ann Epidemiol1995;5:119–129.10. Whelton PK, Bahnson J, Appel LJ, Charleston J, Cosgrove N, Espeland MA,Folmar S, Hoagland D, Krieger S, Lacy C, Lichtermann L, Oates-Williams F,Tayback M, Wilson AC. Recruitment in the Trial of Nonpharmacologic Inter-ventions in the Elderly (TONE).J Clin Geriatr Soc1997;45:185–193.11. Whelton PK, Appel L, Kostis J, Applegate W, Espeland MA, Kumanyika S,Ettinger W, Lacy C, Cutler JA, Folmar S. Efficacy of weight loss and sodiumreduction in the treatment of hypertension: main results of the Trial of Nonphar-macologic Interventions in the Elderly (TONE).JAMA 1998;279:839–846.12. SAS Institute Inc., SAS Technical Report P-243, SAS/STAT Software: TheGENMOD Procedure, Release 6.09, Cary, NC: SAS Institute Inc., 1993.13. Eddington ES. Randomization Tests. 2nd ed. New York: Marcel Dekker, Inc.,1987.14. Freeman MF, Tukey JW. Transformations related to the angular and thesquare root.Ann Math Statist1950;21:607–611.15. Kaplan E, Meier P. Nonparametric estimation from incomplete observations.J Am Statist Assoc1958;8:699–711.16. Cox DR. Regression models and life tables.J Roy Stat Soc1972;34B:187–197.17. Cox DR, Oakes D. Analysis of Survival Data. London: Chapman and Hall,1984.18. Kannel WB, Wilson PW. An update on coronary risk factors.Med Clin NorthAm 1995; 79:951–971.19. Thompson DW, Furlan AJ. Clinical epidemiology of stroke.Neurol Clin1996;14:309–315.20. Veterans Administration Cooperative Study Group on AntihypertensiveAgents. Return of elevated blood pressure after withdrawal of antihypertensivedrugs.Circulation 1975;51:1107–1113.21. Marland LJ, Lutz L, Castle CH. Effects of withdrawing diuretic therapy onblood pressure in mild hypertension.Hypertension1983;5:539–544.22. Stamler R, Stamler J, Grimm R, Gosch F, Dyer A, Berman R, Civinelli J,Elmer P, Fishman J, Van Heel N, McDonald A, McKeever P. Trial on control ofhypertension by nutritional means: three-year results.J Hypertension1984;2:167–170.23. Langford HG, Blaufox D, Oberman A, Hawkins M, Curb JD, Cutter GR,Wassertheil-Smoller S, Pressel S, Babcock C, Abernethy JD, Hotchkiss J, TylerM. Dietary therapy slows the return of hypertension after stopping prolongedmedication.JAMA 1985;253:657–66.24. Medical Research Council Working Party on Mild Hypertension. Course ofblood pressure in mild hypertensives after withdrawal of long term antihyperten-sive therapy.Br Med J1986;293:988–992.25. Stamler R, Stamler J, Grimm R, Gosch FC, Elmer P, Dyer A, Nerman R,Fishman J, Van Heel N, Civinelli J, MacDonald A. Nutritional therapy for highblood pressure: final report of a four-year randomized controlled trial–the Hy-pertension Control Program.JAMA 1987;257:1484–1488.26. Fletcher AE, Franks PJ, Bulpitt CJ. The effect of withdrawing antihyperten-sive therapy: a review.J Hypertension1988;6:431–436.27. Ekbom T, Lindholm LH, Oden A, Dahlof B, Hansson L, Westers P-O,Schersten B. A 5-year prospective, observational study of the withdrawal ofantihypertensive treatment in elderly people.J Intern Med1994;235:581–588.


does withdrawal of antihypertensive medication increase the risk of cardiovascular events?

Documents