Access the full text.
Sign up today, get DeepDyve free for 14 days.
BackgroundNational guidelines recommend consideration of step down or withdrawal of medication in patients with well-controlled hypertension, but knowledge of factors that predict or mediate success in achieving this goal is limited.ObjectiveTo identify patient characteristics associated with success in controlling blood pressure (BP) after withdrawal of antihypertensive medication.DesignThe Trial of Nonpharmacologic Interventions in the Elderly tested whether lifestyle interventions designed to promote weight loss or a reduced intake of sodium, alone or in combination, provided satisfactory BP control among elderly patients (aged 60-80 years) with hypertension after withdrawal from antihypertensive drug therapy. Participants were observed for 15 to 36 months after attempted drug withdrawal.Main Outcome MeasuresTrial end points were defined by (1) a sustained BP of 150/90 mm Hg or higher, (2) a clinical cardiovascular event, or (3) a decision by participants or their personal physicians to resume BP medication.ResultsProportional hazards regression analyses indicated that the hazard (±SE) of experiencing an end point among persons assigned to active interventions was 75%±9% (weight loss), 68%±7% (sodium reduction), and 55%±7% (combined weight loss/sodium reduction) that of the hazard for those assigned to usual care. Lower baseline systolic BP (P<.001), fewer years since diagnosis of hypertension (P<.001), fewer years of antihypertensive treatment (P<.001), and no history of cardiovascular disease (P=.01) were important predictors of maintaining successful nonpharmacological BP control throughout follow-up, based on logistic regression analysis. Age, ethnicity, baseline level of physical activity, baseline weight, medication class, smoking status, and alcohol intake were not statistically significant predictors. During follow-up, the extent of weight loss (P=.001) and urinary sodium excretion (P=.04) were associated with a reduction in the risk of trial end points in a graded fashion.ConclusionsWithdrawal from antihypertensive medication is most likely to be successful in patients with well-controlled hypertension who have been recently (within 5 years) diagnosed or treated, and who adhere to lifestyle interventions involving weight loss and sodium reduction. More than 80% of these patients may have success in medication withdrawal for longer than 1 year.NATIONAL guidelines recommend consideration of antihypertensive medication step down and withdrawal in patients with well-controlled hypertension.The rationale for this recommendation is that medications can be expensive,can cause bothersome adverse effects and undesirable metabolic changes,and may paradoxically increase the risk of clinical cardiovascular events.In addition, in many persons, hypertension occurs as a result of excess sodium intake or body weight, conditions that can potentially be reversed.Many authorshave examined the feasibility of withdrawing antihypertensive medications from patients with well-controlled hypertension. Reported long-term (≥1 year) success rates have ranged from 15% to 70%, with the higher rates having been achieved through lifestyle interventions that focus on antecedent risk factors for hypertension. Two randomized, controlled, clinical trialsconducted in middle-aged persons with hypertension identified weight loss (for overweight patients) and reduced sodium intake as interventions that can significantly increase the success rate for long-term withdrawal of antihypertensive medication. Similar results were observed in the Trial of Nonpharmacologic Interventions in the Elderly (TONE),which found weight loss and reduced sodium intake, alone and in combination, increased the likelihood of successful medication withdrawal. In TONE, those persons assigned to both weight loss and a reduced intake of sodium were about twice as likely to remain off antihypertensive medications and free of blood pressure (BP)–related events at 30 months compared with those assigned to usual care.The extent that these lifestyle change programs will be used in clinical practice may be enhanced by identifying patient profiles that are associated with high success rates. As Fletcher et alhave noted, the predictors of a successful outcome that have been identified for these profiles have varied among studies. Besides lifestyle interventions, studies have found lower BP before and during pharmacological treatment, more recent initiation of treatment, younger age, drug class, sex, and race to be associated with greater probability for successful antihypertensive medication withdrawal. However, the only factor that has been consistently identified across studies has been lower prewithdrawal BP.We used data from the 975 participants in TONE to assess the role of single and multiple patient characteristics in selecting suitable candidates for drug withdrawal. Because of its designed interventions, the clinical trial also allowed us to describe rigorously the role of 2 potential mediators of successful withdrawal—weight loss and reduced dietary sodium intake; our primary focus, however, was to identify baseline characteristics. Strengths of the TONE study included a large and diverse pool of participants, standard withdrawal and measurement protocols, and a substantial duration of systematic follow-up after attempted withdrawal from antihypertensive medication.MATERIALS AND METHODSSTUDY AIMS AND PARTICIPANT ELIGIBILITYDetailed descriptions of the trial's design, methods, recruitment, and randomized cohort have appeared elsewhere.In brief, the primary aims of TONE were to assess whether interventions designed to reduce sodium intake and body weight, alone or in combination, could successfully substitute for pharmacological treatment of hypertension among older persons whose BP was well-controlled while taking a single medication. Participants were recruited from the communities near 4 academic medical institutions (Wake Forest University School of Medicine, Winston-Salem, NC; The Johns Hopkins Medical Institutions, Baltimore, Md; The University of Medicine and Dentistry of New Jersey–Robert Wood Johnson Medical School, New Brunswick; and the University of Tennessee at Memphis). The principal recruitment strategies (with percentage of prescreen contacts accounted for) included mass mailings (73%), mass media (14%), and community screening (5%). Major inclusion criteria were current pharmacological treatment that was based on a prior diagnosis of hypertension (from self-report or clinical history), age between 60 and 80 years, average BP less than 145 mm Hg (systolic) and less than 85 mm Hg (diastolic) (mean of 9 measurements, 3 at each of 3 separate visits) while taking a single antihypertensive medication, willingness to alter diet and physical activity in accordance with the requirements of the TONE interventions, and approval from the participant's personal physician. Major exclusion criteria were history of heart attack or stroke within the preceding 6 months; current angina pectoris; congestive heart failure (CHF); type 1 diabetes mellitus; serious psychiatric or physical illness; unexplained or involuntary weight loss of 4.5 kg or more during the previous year; a body mass index (calculated as the weight in kilograms divided by the height in meters squared) of less than 21 (men and women) or greater than 33 (men) or 37 (women), which was based on expected sex-specific 15th and 95th percentiles of the general elderly population; hypercreatinemia (creatine level >152.5 µmol/L); hyperkalemia (potassium level >5.5 mmol/L); hyperglycemia (nonfasting glucose level >14.4 mmol/L [>260 mg/dL]); or anemia (hemoglobin level <110 g/L). Individuals treated with 1 or 2 antihypertensive medications could participate in screening visits; however, those taking 2 medications had to meet the trial's BP criteria following step down from 2 medications to 1 medication before randomization. Combination drugs consisting of 2 diuretics or a diuretic and a nondiuretic agent were considered to be a single medication.INTERVENTION AND FOLLOW-UPRandomization stratified participants by body mass index: women with a body mass index greater than 27.3 and men with a body mass index greater than 27.8 were classified as obese and were randomized to 1 of 4 conditions (usual care, sodium reduction, weight loss, or combined sodium reduction and weight loss). Nonobese participants were randomized to 1 of 2 conditions (usual care or sodium reduction). Within 1 month of randomization, participants assigned to active interventions began their lifestyle modification counseling, which was conducted by means of individual and small group meetings. The goal of the weight loss intervention was achievement and maintenance of 4.5 kg or more of weight loss; the goal of the sodium reduction intervention was achievement and maintenance of a 24-hour urinary sodium excretion of 80 mmol or less. Participants assigned to usual care met periodically for educational sessions unrelated to nutrition and cardiovascular disease.Antihypertensive medication withdrawal began 90 days (±14 days) after the first intervention session and at a corresponding time for those randomized to usual care. The medication withdrawal process was standardized across the 4 clinical centers and used drug-specific tapering regimens. As medication was tapered, participants were examined weekly. After discontinuation of antihypertensive medication, participants had 3 additional biweekly visits to confirm that their BP remained lower than 150 mm Hg (systolic) and 90 mm Hg (diastolic).Every 3 months, participants attended data collection visits that were held at separate times and places from their intervention visits. During these visits, BP and other measurements were obtained by staff who were kept unaware of intervention assignments. Follow-up ended after a mean of 27.6 months (range, 15.6-35.9 months) following randomization. For the 364 participants who remained end point free throughout follow-up, close-out occurred a mean of 23.7 months (range, 11.9-31.6 months) after the completion of drug withdrawal.Data from all 975 randomized TONE participants are included in this report. Participants were classified as being end point free if (1) their antihypertensive medication was successfully withdrawn, (2) they remained without need of antihypertensive medication at close-out and their BP was less than 150/90 mm Hg throughout follow-up, and (3) they had no intervening cardiovascular events that may have been related to BP (eg, stroke, myocardial infarction, arrhythmia, CHF, or angina).MEASUREMENT OF PREDICTORSThis report examines baseline predictors and 2 intercurrent mediators of successful medication withdrawal and long-term maintenance of BP control (ie, without need for medication or occurrence of cardiovascular events). Data on these factors were collected using procedures detailed previously.Blood pressure was measured by trained and certified observers using a random-zero sphygmomanometer. At each visit, 3 BP measurements were obtained in the seated position after participants had rested for at least 5 minutes. Baseline BP was based on the average of 9 measurements at these 3 separate visits.Urinary excretion of sodium and potassium was assessed from standardized 24-hour urine collections obtained twice before randomization and once each at 9 months postrandomization, 18 months postrandomization, and close-out. Type of antihypertensive medication was ascertained from review of current medications that participants brought to the clinic and was grouped into 5 classes: angiotensin-converting enzyme inhibitors, β-blockers, calcium channel blockers, diuretics, and other (which included combination agents). Standardized measurements of weight and height were obtained at baseline; weight was remeasured every 3 months during follow-up. At baseline, participants were asked to report the average numbers of hours spent in vigorous, moderate, and mild physical activity per day. These domains were analyzed separately; similar and more concise results were obtained by creating a rough activity index defined by the sum of 3 times the vigorous hours, 2 times the moderate hours, and the mild hours. This graded the overall amount of physical activity; participants were grouped into tertiles according to this index. Other data (demographic and history) were collected from questionnaires. History of cardiovascular disease was based on the results of the participant's interview, echocardiogram, and physical examination.STATISTICAL METHODSKaplan-Meierplots were used to portray the overall distributions of time until failure of drug withdrawal. Participants who refused or failed the initial medication withdrawal attempt, or for whom it had become inappropriate to attempt drug withdrawal at 4 months after randomization, were classified as instantaneous failures; for others, time until failure was marked beginning from the date that use of medication ceased. Proportional hazards regression modelswere used to identify factors that influenced the rates of success and failure across the varying lengths of follow-up among participants. These yielded estimates for the relative chance (hazard) of failure, at any given time, among participants in subgroups defined by potential predictors. To portray these, we arbitrarily assigned a value of 1.00 to 1 subgroup and computed the hazards for the other subgroups relative to this reference. To identify factors that influenced the occurrence of specific types of failures, such as those that were BP-related (ie, based on elevated BP measurements or clinical events) or due to decisions by participants and their personal health care providers, separate proportional hazards regression models were used in which failures were limited to the subset of interest (and participants who otherwise failed were censored at these times).To identify the clinically important outcomes of long-term success, participants were grouped according to their end point status at the time of their exit from the study (15-36 months after attempted drug withdrawal), and logistic regression analysis was used to develop multivariable predictor models using forward and backward stepwise variable selection procedures. The final multivariable logistic regression modelwas used to project estimates of the success rates.RESULTSAcross all 4 centers, a total of 975 participants were randomly assigned to 1 of the following 6 study groups: nonoverweight usual care (n=194), nonoverweight sodium reduction (n=196), overweight usual care (n=147), overweight sodium reduction (n=144), overweight weight loss (n=147), and overweight combined sodium reduction and weight loss (n=147). Table 1provides descriptive data at baseline for the 975 participants. Nearly 80% of the cohort was aged 60 to 69 years; approximately 75% of the cohort had been diagnosed and treated for hypertension for 5 years or longer. Eighty-five percent of the participants were taking a single antihypertensive drug at their first screening contact, while the remainder had to be withdrawn from a second active medication before randomization. The mean (±SD) BP at baseline was 128.2±9.3 mm Hg (systolic) and 71.4 ± 7.3 mm Hg (diastolic). The cohort was nearly evenly split between women and men.Table 1. Baseline Characteristics of the 975 Randomized TONE Participants*CharacteristicsNo. (%) of ParticipantsAge, y60-69769 (78.9)70-80206 (21.1)Body weightNot overweight†390 (40.0)Overweight585 (60.0)Class of BP drugsACE inhibitor210 (21.5)β-blocker106 (10.9)Calcium channel blocker272 (27.9)Diuretic314 (32.2)Other73 (7.5)BP drugs at screening1824 (84.5)2151 (15.5)Years diagnosed with high BP‡0-4229 (23.5)5-9538 (55.2)≥10207 (21.3)Years taking BP medications‡0-4267 (27.4)5-9495 (50.8)≥10212 (21.8)History of CVDNo738 (75.7)Yes237 (24.3)SexFemale466 (47.8)Male509 (52.2)EthnicityAfrican American230 (23.6)Other745 (76.4)Smoking statusCurrent52 (5.3)Former392 (40.2)Never531 (54.5)Alcoholic intake, drinks/wkNone580 (59.5)1-6280 (28.7)≥7115 (11.8)*The baseline urine sodium concentration was 148.5 ± 54.2 mmol/24 h; potassium, 59.7 ± 21.3 mmol/24 h. The baseline physical activity levels were as follow: vigorous, 1.1 ± 2.0 h/d; moderate, 5.1 ± 3.2 h/d; and mild, 10.0 ± 3.7 h/d. (All values are given as mean ± SD.) TONE indicates Trial of Nonpharmacologic Interventions in the Elderly; BP, blood pressure; ACE, angiotensin-converting enzyme; and CVD, cardiovascular disease.†Overweight is defined as a body mass index of 27.3 kg/m2or greater for women and 27.8 kg/m2or greater for men. Body mass index is calculated as the weight in kilograms divided by the height in meters squared.‡Data do not total 975 because these data were missing for one participant.The lifestyle interventions implemented in TONE resulted in substantial and statistically significant (P<.001) reductions in sodium intake and body weight. Mean levels of sodium excretion (averaged across the three to four 24-hour urine collections during follow-up) and weight loss (baseline minus mean of all postrandomization weights) were computed for participants grouped by intervention assignment. During follow-up, the mean (±SE) sodium excretion for individuals assigned to sodium reduction and the combined sodium reduction/weight loss interventions was 98.7±2.4 and 125.6±3.9 mmol/24 h, respectively, compared with 140.0±2.4 mmol/24 h for usual care and 160.9±4.2 mmol/24 h for weight loss alone. The mean (±SE) weight losses for individuals assigned to weight loss and combined sodium reduction/weight loss interventions were 4.86±0.36 and 3.92±0.40 kg, respectively, compared with 0.09±0.18 kg for usual care and 1.12±0.18 kg for sodium reduction alone.Table 2groups individuals according to sodium excretion and changes in weight loss relative to the goals of the active interventions (<80 mmol/24 h for sodium and ≥4.5 kg loss for weight). Participants assigned to sodium restriction interventions were much more likely to have sodium excretion within 20% of the intervention goal than those not assigned. Participants assigned to weight loss interventions were also much more likely to demonstrate weight losses within 20% of the intervention goals than those not assigned to these interventions. Omnibus χ2tests indicated significant differences existed among intervention arms (P<.001).Table 2. Long-term Intervention Effects on Average Weight Loss and Reduction in Sodium Excretion*Intervention AssignmentAchievement Relative to Goal for Active Interventions†Within 20% of GoalFrom 20%-50% of GoalMore Than 50% From GoalUrinary Sodium ExcretionCombined sodium reduction/ weight loss (n = 141)23.827.249.0Sodium reduction (n = 319)188.8.131.52Weight loss (n = 140)6.114.379.6Usual care (n = 320)15.317.067.7Weight LossCombined sodium reduction/ weight loss (n = 144)42.915.042.2Sodium reduction (n = 327)16.713.270.3Weight loss (n = 142)56.514.329.2Usual care (n = 323)7.67.984.5*P<.001 for the differences among intervention assignments for urinary sodium excretion and weight loss.†Urinary sodium excretion goals for combined sodium reduction/weight loss and sodium reduction interventions were less than 80 mmol/24 h; weight loss goals for combined sodium reduction/weight loss and weight loss interventions were 4.5 kg or more.Of the 975 randomized TONE participants, 364 (37.3%) remained free of an end point at close-out. Of the 611 end points, 452 (74.0%) occurred as a result of high BP measurements, 34 (5.6%) were attributable to cardiovascular or other clinical events (16 angina, 3 myocardial infarction, 2 CHF, 1 stroke, and 12 other), and the remaining 125 (20.5%) were due to decisions by the patients or their physicians not to begin drug tapering or to resume medication therapy. Due to their relative infrequency, cardiovascular events were grouped with end points resulting from high BP measurements in subsequent analyses and henceforth are referred to collectively as BP-related. Figure 1portrays the incidence rate of TONE end points, both overall and for BP–related end points, during the first 30 months of follow-up. The initial withdrawal attempt was not completed in 20.6% of the cohort; these individuals appear as instantaneous failures. Based on Kaplan-Meier calculations, the overall success rate at 30 months was estimated to be 30.4%±2.1%. End points related to high BP measurements were most frequent, and closely approximated the overall end point distribution.Rates of the Trial of Nonpharmacologic Interventions in the Elderly end points throughout 30 months of follow-up: overall and separately by type of end point.Table 3presents the results of proportional hazards models to assess the effects on the end point rates for each individual predictor or mediator. Included are Pvalues from omnibus tests to detect statistically significant differences among subgroups. Assignment to an active intervention (combined weight loss/sodium reduction, sodium reduction only, and weight loss only) was associated with a significantly lower incidence rate of end points across follow-up. Compared with usual care (both overweight and nonoverweight participants), assignment to the weight intervention (overweight participants only) reduced the hazard of end points by a factor of 0.75, to the sodium intervention (both overweight and nonoverweight participants) by a factor of 0.68, and to the combined intervention (overweight participants only) by a factor of 0.55. After controlling for the experimental condition of intervention assignment, factors that were associated with improved success rates were lower baseline systolic and diastolic BP, use of 1 medication vs 2 medications at first screening contact, fewer years diagnosed with hypertension and taking medications, no history of cardiovascular disease, and successful weight loss and sodium reduction. Participants with a medication-controlled systolic BP lower than 135 mm Hg at baseline had approximately half the failure rate of participants with a higher systolic BP. Those with a diastolic BP lower than 80 mm Hg had about 70% to 80% the failure rate experienced by others. Participants taking 1 antihypertensive medication at first screening contact had 74% of the hazard rate of those taking 2 drugs. Those who were more recently diagnosed as hypertensive, more recently prescribed medications for hypertension, or had no history of cardiovascular disease were also less likely to encounter an end point. Even when intervention assignment was controlled for, strong graded relationships were observed between the achieved levels of weight loss and sodium reduction and the corresponding incidence of end points.Table 3. Results From Individual Proportional Hazards Regression Analyses*PredictorRelative Hazard Ratio ± SEEnd Points Related to Elevated BP or Events (n = 486)End Points for Other Reasons (n = 125)All End Points (N = 611)Intervention assignmentWeight loss0.75 ± 0.100.73 ± 0.200.75 ± 0.09Sodium reduction0.69 ± 0.070.64 ± 0.140.68 ± 0.07Combined sodium reduction/weight loss0.54 ± 0.080.59 ± 0.170.55 ± 0.07Usual care1.001.001.00P<.001.11<.001Baseline weightNot overweight0.90 ± 0.100.80 ± 0.170.87 ± 0.08Overweight1.001.001.00P.30.27.16Age, y60-690.94 ± 0.111.32 ± 0.320.93 ± 0.0970-791.001.001.00P.58.24.46Baseline systolic BP, mm Hg<1200.37 ± 0.050.62 ± 0.170.41 ± 0.05120-1340.52 ± 0.050.76 ± 0.150.56 ± 0.05135-1441.001.001.00P<.001.18<.001Baseline diastolic BP, mm Hg<700.64 ± 0.091.13 ± 0.360.72 ± 0.0970-790.76 ± 0.101.10 ± 0.340.82 ± 0.1080-841.001.001.00P.009.89.04SexFemale0.99 ± 0.091.94 ± 0.341.14 ± 0.09Male1.001.001.00P.98<.001.11EthnicityAfrican American1.06 ± 0.110.73 ± 0.170.99 ± 0.09Other1.001.001.00P.57.09.91Class of antihypertensive medicationACE inhibitor1.20 ± 0.241.16 ± 0.501.15 ± 0.20β-blocker0.83 ± 0.201.91 ± 0.841.10 ± 0.22Calcium channel blocker1.28 ± 0.251.38 ± 0.571.27 ± 0.22Diuretic1.23 ± 0.241.42 ± 0.581.20 ± 0.20Other1.001.001.00P.06.34.48No. of BP drugs during screening10.67 ± 0.081.10 ± 0.280.74 ± 0.0821.001.001.00P<.001.71.004Years taking hypertensive drugs0-40.54 ± 0.060.78 ± 0.180.58 ± 0.065-90.82 ± 0.100.88 ± 0.200.81 ± 0.08≥101.001.001.00P<.001.51<.001Years diagnosed with hypertension0-40.52 ± 0.060.78 ± 0.170.56 ± 0.065-90.78 ± 0.090.88 ± 0.200.80 ± 0.08≥101.001.001.00P<.001.51<.001History of CVDNo0.88 ± 0.090.56 ± 0.100.79 ± 0.07Yes1.001.001.00P.22.002.01Baseline current smoking statusNo1.02 ± 0.211.06 ± 0.411.03 ± 0.19Yes1.001.001.00P.98.93.94Baseline alcohol intake, drinks/dNone1.25 ± 0.191.04 ± 0.291.20 ± 0.16<11.20 ± 0.200.87 ± 0.271.13 ± 0.16≥11.001.001.00P.32.69.34Baseline activity level, tertileLowest0.92 ± 0.100.90 ± 0.190.91 ± 0.09Mid1.08 ± 0.120.86 ± 0.191.03 ± 0.03Highest1.001.001.00P.35.78.44Baseline urine sodium excretion, mmol/24 h†<120.40.99 ± 0.111.12 ± 0.251.01 ± 0.10120.4-165.01.14 ± 0.131.07 ± 0.171.12 ± 0.11>165.01.001.001.00P.36.88.43Baseline urine potassium excretion, mmol/24 h†<49.51.00 ± 0.111.05 ± 0.171.01 ± 0.1049.5-65.80.87 ± 0.100.75 ± 0.170.84 ± 0.08>65.81.001.001.00P.36.27.13Weight loss overweight participants only, kg‡≥3.60.68 ± 0.180.46 ± 0.240.63 ± 0.132.2-3.20.78 ± 0.150.64 ± 0.260.75 ± 0.13<2.21.001.001.00P.12.18.0424 h Sodium excretion during follow-up, mmol§≤960.76 ± 0.090.49 ± 0.120.69 ± 0.0897-1200.95 ± 0.120.37 ± 0.120.81 ± 0.09>1201.001.001.00P.08<.001.002*Analyses were performed to assess relationships between potential predictors and (1) all end points, (2) end points related to high blood pressure or clinical events, and (3) all other end points. Intervention assignment and clinical center are included as a covariate in all assessments of other predictors. All values of 1.00 indicate reference values. BP indicates blood pressure; ACE, angiotensin-converting enzyme; and CVD, cardiovascular disease.†The lowest value given is between zero and the 33rd percentile; mid value, between the 34th and 66th percentile; and highest value, at the 67th percentile or higher.‡The goal for weight loss was 4.5 kg. The highest weight given is greater than 80% of the goal; mid weight, between 50% and 79% of the goal; and lowest weight, less than 50% of the goal.§The goal for 24-h sodium excretion was 80 mmol. The lowest value given is less than 120% of the goal; mid value, between 120% and 150% of the goal; and highest value, greater than 150% of the goal.Age group, baseline weight, sex, race, medication type, smoking status, alcohol intake, baseline activity level, and baseline sodium and potassium excretion did not appear to be important univariable predictors. We conjectured that the dietary interventions may have obscured any relationships between baseline levels of sodium and potassium excretion and end points, and so repeated these analyses among the subset of participants randomized to usual care. Among these individuals, those with baseline urinary sodium excretion in the lowest tertile had a hazard ratio (±SE) of 0.92±0.15 (compared with the highest tertile), but this difference was not statistically significant (P=.62). Similarly, the relationship between end points and potassium excretion did not reach statistical significance in this subset of participants (P=.56; hazard ratio for lowest vs highest tertile, 0.91±0.14).The predictors of the subset of end points classified as being BP-related were generally similar to those for the overall end points. The subset of end points that were non–BP-related (participant or personal physician decisions) occurred more often among women (P<.001), among those with a history of cardiovascular disease (P=.002), and among those who did not alter their sodium intake (P<.001). Women were at approximately twice the risk of these end points compared with men, and those with a history of cardiovascular disease were also at about twice the risk of those with no history. Individuals who were not compliant with sodium reduction were at over twice the risk for these end points compared with those who reduced their sodium intake.Forward and backward selection algorithms yielded identical multivariable models for identifying factors predicting and/or mediating successful maintenance off drugs throughout close-out. Assignment to active intervention, weight loss, sodium reduction, lower baseline systolic BP, and baseline medication status (fewer medications and fewer years taking medications) were each identified as important (nominal P<.05) and independent predictors of success. Table 4provides odds ratios (with confidence intervals) and parameter estimates for logistic equations (1 and 2) for projecting success rates for subgroups of participants identified by combinations of these factors. Since intervention assignment and changes in lifestyle were statistically confounded, inclusion of both these terms was used to express the overall impact of adherence to the TONE interventions.Table 4. Results From a Multivariable Logistic Regression Model for Predicting Rates of Successful Drug Withdrawal at Study Exit*PredictorParameter ± SEOR (95% CI) for SuccessPIntervention assignmentCombined sodium reduction/weight loss0.88 ± 0.232.41 (1.55-3.77)Sodium reduction0.50 ± 0.181.65 (1.15-2.37)<.001Weight loss0.53 ± 0.241.71 (1.08-2.71)Usual care0.001.00 (. . .)Lifestyle modificationWeight loss, kg≥3.60.68 ± 0.291.17 (1.12-1.25)2.2-3.20.34 ± 0.261.41 (0.85-2.32).04<2.20.001.00 (. . .)Sodium excretion, mmol/24 h≤960.62 ± 0.181.86 (1.30-2.65)97-1190.27 ± 0.191.30 (0.89-1.91).003≥1200.001.00 (. . .)Baseline systolic BP, mm Hg<1201.47 ± 0.224.34 (2.82-6.67)120-1341.01 ± 0.182.75 (1.95-3.89)<.001≥1350.001.00 (. . .)Medication statusNo. of BP drugs10.44 ± 0.211.55 (1.03-2.33).0420.001.00 (. . .)Years taking BP drugs0-40.75 ± 0.172.11 (1.52-2.93)5-90.26 ± 0.181.29 (0.90-1.85)<.001≥100.001.00 (. . .)*The fitted intercept for the multivariable regression model was −2.70 ± 0.28. OR indicates odds ratio; CI, confidence interval; BP, blood pressure; and ellipses, OR fixed at 1.00.Table 5lists examples of estimated success rates from the parameters of Table 4. Subgroups of patients for 2 conditions were compared: patients under usual care who made no changes in their lifestyle vs patients who were within 20% of compliance with an active intervention focusing on both weight loss and sodium reduction (ie, lost ≥3.6 kg and had urinary sodium excretions of ≤96 mmol/24 h). Depending on length of prior treatment, treated systolic BP, and number of medications, estimated success rates for patients under usual care ranged from 6.3% to 48.7%. For patients who were successful in responding to a combined intervention, estimated success rates were projected to range from 37.2% to 89.3%.Table 5. Predicted Success Rates From Multivariable Logistic Regression ModelYears of MedicationBaseline Systolic Blood Pressure, mm Hg<120120-134≥135Usual Care Participants* (1 Medication at Screening)0-448.737.617.95-936.726.911.8≥1031.022.29.4Usual Care Participants* (2 Medications at Screening)0-438.028.012.45-927.319.28.0≥1022.515.56.3Participants Assigned to Combined Sodium Reduction/Weight Loss Intervention† (1 Medication at Screening)0-489.384.265.95-983.776.554.2≥1079.971.647.8Participants Assigned to Combined Sodium Reduction/ Weight Loss Intervention† (2 Medications at Screening)0-484.477.555.65-976.867.843.3≥1071.961.937.2*No lifestyle modifications.†Successful lifestyle modifications.COMMENTThe results from TONE suggest that drug withdrawal may be an attractive treatment option, particularly for patients who have been recently and successfully treated with a single antihypertensive agent and who are able to adhere to lifestyle interventions. Under these circumstances, we estimate that maintaining satisfactory BP control following withdrawal of antihypertensive medication may be possible in more than 80% of patients. The mean BP of randomized participants who were end point free at close-out was 131 to 134 mm Hg (systolic) and 74 to 75 mm Hg (diastolic).If lifestyle interventions are not provided for such patients, long-term withdrawal may be feasible in less than half the patients.PREDICTORS OF SUCCESSThe predictors of successful drug withdrawal identified in TONE fall into 2 categories: those related to characteristics of the patient's clinical status (baseline level of BP, years since diagnosis and first pharmacological treatment, number of drugs required for treatment, and history of cardiovascular disease) and those related to prescription and success of weight loss and sodium reduction interventions. The markers of clinical status are generally consistent with findings in other studies and populations: the best potential candidates for antihypertensive drug withdrawal are those whose BP is well-controlled and who have less comorbidity. The results of TONE also suggest that withdrawal might be most appropriate during the first 5 years since diagnosis and treatment of hypertension.In TONE, assignment to an intervention program and success in meeting the goals of the program both independently enhanced the probability that medications could be withdrawn successfully. A successful lifestyle intervention program appears to be essential to increase success rates, regardless of the participant's clinical characteristics.These findings suggest that many patient characteristics may not contribute to making decisions regarding drug withdrawal for the cohort of patients represented by the TONE participants. These characteristics include sex, race, age, medication class, and several descriptors of current lifestyle (smoking status, alcohol intake, physical activity, and sodium and potassium excretion). Previous researchhas found that age, sex, and smoking status do not consistently impair drug withdrawal.One of the limitations of TONE was that the original diagnosis of hypertension, which led to pharmacological therapy, was made by participant's personal physicians sometime before the study, and was thus unstandardized. Any factor that was systematically related to misdiagnosis had the potential, in our analysis, of being identified as predicting successful withdrawal from (the unneeded) medications. Blood pressure measurement in TONE was performed by centrally trained personnel; multiple readings were taken at repeated examinations to reduce error. Some end points, however, were decisions made by the participant's personal physician based on BP measurements obtained outside of the trial. If any factors existed that were systematically related to errors in BP measurement, either within or extraneous to the TONE clinics, these might also be identified in our analyses as being predictive of success or failure.EXPECTED SUCCESS RATESSuccess rates developed from multivariable regression models in TONE must be viewed cautiously. First, the factors selected for inclusion in the model were chosen because they had strong associations with successful withdrawal. It is likely (and statistically expected) that predictions from an equation developed in this manner would overestimate how well the rule applies in other settings.It is difficult to quantify the expected bias of such estimation; statistical methods such as bootstrapping, jackknifing, and simulations are of some use.In general, factors that increase the amount of expected bias include a small sample size and a large number of potential predictors. We performed bootstrap procedures in an attempt to quantify the expected bias associated with our predictive equations. Our results indicate that this bias is likely to be small, ie, less than 5%, for populations represented by the TONE cohort. Bootstrapped SDs for estimated success rates range from 1% to 8%. This suggests that even though actual rates in clinical practice may differ from rates estimated in our analyses, the magnitude of these differences may be relatively small and the differences are not likely to modify decisions on whether to attempt drug withdrawal.A second reason that the predicted success rates appearing in Table 5may not be as precise as desired is that the model did not include any higher-order interaction terms and was applied to the examples reflecting the extremes of the distribution. We chose to include only main effects in the logistic regression model to "average" across many individual subgroups; the examples in Table 5were chosen purposefully to portray the potential range of differences in success rates.The TONE participants were motivated volunteers who may have been more healthy than patients in more typical clinic populations. The TONE interventions included frequent contact with interventionists providing more support for lifestyle changes than may be generally feasible. Because of this, the success rates observed in TONE may be higher than those obtainable in clinical practice. However, 20.5% of the overall end points were due to patient or physician decisions; perhaps this source of failures could be reduced given the encouraging results from TONE.CLINICAL IMPLICATIONSIn TONE, BP was monitored at regularly scheduled, 3-month intervals, to enhance safety and reassure participants. Feedback on weight loss and sodium intake was provided during interventions by nutritionists, and support came from the nutritionist and fellow participants.In general, participants tolerated drug withdrawal well. The major concerns of withdrawing hypertensive medications are unmasking coexistent diseases such as angina, arrhythmia, or CHF, for which the antihypertensive treatment was simultaneously providing treatment.This was not a problem in TONE, perhaps due to the selection of participants and the tapering regimen that was adopted for drug withdrawal.Patients were seen in a face-to-face visit on a weekly basis during drug withdrawal and biweekly for a 6-week period after medication was discontinued. Blood pressure was monitored and participants were queried about symptoms of disease and other possible adverse effects of drug withdrawal. Persons with angina, CHF, and recent myocardial infarction were excluded from participation in TONE. Patients with these conditions may not be appropriate for discontinuation of their medications.Hypertension is a common condition in older people. Treatment of hypertension reduces the morbidity and mortality from coronary heart disease and stroke. Our data suggest that many older people will remain normotensive after withdrawal of antihypertensive medication. Thus, the need for antihypertensive medication should be assessed on a periodic basis and, for those individuals who are willing to engage in lifestyle modifications, a trial of medication withdrawal can be safely and successfully undertaken.POTENTIAL PUBLIC HEALTH IMPACT AND CONCLUSIONSThe potential public health utility of substituting lifestyle changes for drug treatment will depend on several factors, including relative costs of lifestyle intervention programs and the medications to be discontinued. If the trend toward use of more expensive medications continues,then lifestyle modification becomes increasingly attractive. The public health benefits of reduced salt intake depend on the extent of future moderation of the sodium content of processed and commercially prepared foods. Such changes would not only facilitate efforts at replacing a portion of pharmacological management in the elderly but also further the primary prevention of hypertension,obviating the need for medication use in the first place.National High Blood Pressure Education ProgramThe fifth report of the Joint National Committee on detection, evaluation, and treatment of high blood pressure.Arch Intern Med.1993;153:154-183.MAbramowiczDrugs for hypertension.Med Lett Drugs Ther.1995;37:45-50.TAManolioJACutlerCDFurbergBMPsatyPKWheltonWBApplegateTrends in pharmacologic management of hypertension in the United States.Arch Intern Med.1995;155:829-837.CMonaneRJGlynnJHGurwitzTrends in medication choices for hypertension in the elderly: the decline of thiazides.Hypertension.1995;25:1045-1051.1988 Joint National CommitteeThe 1988 report of the Joint National Committee on detection, evaluation, and treatment of high blood pressure.Arch Intern Med.1988;48:1023-1038.RWGiffordRABorazanianTraditional first-line therapy: overview of medical benefits and side effects.Hypertension.1989;13:I-119–I-124.KJamersonPrevalence of complications and response to different treatments of hypertension in African Americans and white Americans in the US.Clin Exp Hypertens.1993;15:979-995.PHHeldSYusufCDFurbergCalcium channel blockers in acute myocardial infarction and unstable angina: an overview.BMJ.1989;299:1187-1192.BMPsatySRHeckbertTDKoespellThe risk of myocardial infarction associated with antihypertensive drug therapies.JAMA.1995;274:620-625.CDFurbergBMPsatyJVMeyerNifedipine: dose related increase in mortality in patients with coronary heart disease.Circulation.1995;92:1326-1331.Veterans Administration Cooperative Study Group on Antihypertensive AgentsReturn of elevated blood pressure after withdrawal of antihypertensive drugs.Circulation.1975;51:1107-1113.LJMalandLLutzCHCastleEffects of withdrawing diuretic therapy on blood pressure in mild hypertension.Hypertension.1983;5:539-544.FAFinnertyStep-down treatment of mild systemic hypertension.Am J Cardiol.1984;53:1304-1307.RStamlerJStamlerRGrimmTrial on control of hypertension by nutritional means: three-year results.J Hypertens.1984;2:167-170.HGLangfordDBlaufoxAObermanDietary therapy slows the return of hypertension after stopping prolonged medication.JAMA.1985;253:657-666.Medical Research Council Working Party on Mild HypertensionCourse of blood pressure in mild hypertensives after withdrawal of long term antihypertensive therapy.BMJ.1986;293:988-992.RStamlerJStamlerRGrimmNutritional therapy for high blood pressure: final report of a four-year randomized controlled trial—the Hypertension Control Program.JAMA.1987;257:1484-1488.AEFletcherPJFranksCJBulpittThe effect of withdrawing antihypertensive therapy: a review.J Hypertens.1988;6:431-436.TEkbomLHLindholmAOdenA 5-year prospective, observational study of the withdrawal of antihypertensive treatment in elderly people.J Intern Med.1994;235:581-588.BGagerbergJWikstrandGBerglundMHartfordSLyungmanIWendelhagWithdrawal of antihypertensive drug treatment: time course for re-development of hypertension and effects on left ventricular map.J Hypertens.1992;10:587-593.GLJenningsKSudhirMEslerCReidPIKornerDeterminants of the return of hypertension after cessation of long term therapy.J Hypertens.1992;10(suppl 4):S123.GLJenningsPIKornerELauferMEslerDBurtonABruceHow hypertension redevelops after the cessation of long term therapy.J Hypertens.1984;2(suppl 3):217-219.RHThurnWSmithOn re-setting the "barostats" in hypertensive patients.JAMA.1967;201:301-304.PKWheltonLAppelMAEspelandEfficacy of weight loss and sodium reduction in the treatment of hypertension: main results of the Trial of Nonpharmacologic Interventions in the Elderly (TONE).JAMA.1998;279:839-846.LJAppelMEspelandPKWheltonTrial of Nonpharmacologic Intervention in the Elderly (TONE): design and rationale of blood pressure control trial.Ann Epidemiol.1995;5:119-129.PKWheltonJBahnsonLJAppelRecruitment in the Trial of Nonpharmacologic Intervention in the Elderly (TONE).J Am Geriatr Soc.1997;45:185-193.JLBahnsonPKWheltonLJAppelBaseline characteristics of randomized participants in the Trial of Nonpharmacologic Intervention in the Elderly (TONE).Dis Manage Clin Outcomes.1997;1:2-9.ARFrisanchoNew standards of weight and body composition by frame size and height for assessment of nutritional status of adults and the elderly.Am J Clin Nutr.1984;40:808-819.EKaplanPMeierNonparametric estimation from incomplete observations.J Am Stat Assoc.1958;8:699-711.DRCoxRegression models and life tables.J R Stat Soc.1972;34B:187-197.DRCoxDOakesAnalysis of Survival Data.London, England: Chapman & Hall; 1984.DCollettModelling Binary Data.London, England: Chapman & Hall; 1991.MDFotherbyWithdrawal of antihypertensive therapy in the elderly: the issues.Drugs Aging.1995;6:436-444.PLund-JohansenStopping antihypertensive drug therapy in elderly people–a dangerous experiment?J Intern Med.1994;235:577-579.BEfronEstimating the error rate of a prediction rule: improvements on cross-validation.J Am Stat Assoc.1983;78:316-331.GGongCross-validation, the jackknife, and the bootstrap: excess error estimation in forward logistic regression.J Am Stat Assoc.1986;81:108-113.JKKirkSHJohnsonSafe discontinuation of antihypertensive therapy.Arch Fam Med.1995;4:266-270.JBKostisMAEspelandLAppelKJohnsonJPierceJWoffordDoes withdrawal of antihypertensive therapy medication increase the risk of cardiovascular events: results from the Trial of Nonpharmacologic Interventions in the Elderly (TONE).Am J Cardiol.1998;82:1501-1508.MAEspelandSKumanyikaJBKostisAntihypertensive medication use among recruits for the Trial of Nonpharmacologic Interventions in the Elderly (TONE).J Am Geriatr Soc.1996;44:1183-1189.Accepted for publication May 14, 1998.This study was supported by grants R01 AG-09799, R01 HL-48642, R01 HL-48642, R01 AG-09771, R01 AG-09773, R03 HL-60197, and P60AG-10484 from the National Institute of Aging and the National Heart, Lung, and Blood Institute, National Institutes of Health, Bethesda, Md.Reprints: Mark A. Espeland, PhD, Section on Biostatistics, Department of Public Health Sciences, Wake Forest University School of Medicine, Medical Center Boulevard, Winston-Salem, NC 27157 (e-mail: email@example.com).
Archives of Family Medicine – American Medical Association
Published: May 1, 1999
Access the full text.
Sign up today, get DeepDyve free for 14 days.