Flow chart of patients available at each stage of the protocol. BP indicates blood pressure.
Linden W, Lenz JW, Con AH. Individualized Stress Management for Primary HypertensionA Randomized Trial. Arch Intern Med. 2001;161(8):1071-1080. doi:10.1001/archinte.161.8.1071
Copyright 2001 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2001
To test the efficacy of individualized stress management for primary hypertension in a randomized clinical trial with the use of ambulatory blood pressure (BP) measures.
Men and women aged 28 to 75 years with mean ambulatory BP greater than140/90 mm Hg received 10 hours of individualized stress management by means of semistandardized treatment components. They were randomly assigned to immediate treatment (n = 27) or a wait list control group (n = 33). Participants on the wait list were subsequently offered treatment. Six-month follow-up data were available from 36 of the 45 participants who completed treatment. Measures were 24-hour ambulatory BP, lipid levels, weight, and psychological measures.
Blood pressure was significantly reduced in the immediate treatment group and did not change in control subjects (−6.1 vs +0.9 mm Hg for systolic and –4.3 vs +0.0 mm Hg for diastolic pressure). When the wait list control group was later treated, BP was similarly reduced by –7.8 and –5.2 mm Hg, and for the combined sample, total change at follow-up was –10.8 and –8.5 mm Hg. Level of BP at the beginning of treatment was correlated with BP change (r = 0.45 [P<.001] and 0.51 [P<.001], respectively), and amount of systolic BP change was positively correlated with reduction in psychological stress (r = 0.34) and change in anger coping styles (r = 0.35-0.41).
Individualized stress management is associated with ambulatory BP reduction. The effects were replicated and further improved by follow-up. Reductions in psychological stress and improved anger coping appear to mediate the reductions in BP change.
PSYCHOLOGICAL treatments for hypertension (consisting of relaxation, biofeedback, and/or stress management) have at best received mixed support from expert panels, which concluded that they are not very promising given inconclusive findings.1- 3 The current study was designed to integrate such cautious conclusions of consensus conferences with a more optimistic view that arises from meta-analytic reviews. We posit that the discrepancy between consensus group recommendations and conclusions from meta-analytic reviews is due to conceptual, measurement, and trial protocol differences that have been shown to affect blood pressure (BP) outcomes but were not appropriately dealt with in most study protocols.
Psychological stress is widely considered to contribute to the development of primary hypertension. The epidemiologic evidence of a link between stress and high BP is very convincing,4 yet the biopsychosocial pathway that would explain how stress can lead to disease is less clear.5,6 Consistent with these findings on stress-hypertension linkage, psychological treatments are designed to reduce stress by targeting deficient cognitive and behavioral stress-coping strategies and by reducing sympathetic arousal. Most reviewers report some clinical benefits associated with behavioral interventions to reduce arousal, and numerous reviews of this literature7- 10 suggest variable pretreatment to posttreatment effect sizes that range from d = 0.40 to d = 1.4.
Observed outcomes vary substantially as a function of study design. Jacob et al9 identified 75 controlled clinical trials of relaxation therapies for hypertension and noted that treatments starting with high initial BP also produced greater reductions (r = 0.75 for systolic BP [SBP] and r = 0.64 for diastolic BP [DBP]). Differential pretreatment levels had not been considered in the recommendations of the consensus groups1- 3 or previous reviews and may have led to an underestimation of the efficacy of psychosocial treatments.
Linden and Chambers10 conducted a meta-analysis of hypertension treatment outcomes including comparisons of nondrug treatments with pharmacologic agents. Ninety controlled trials of psychosocially based treatments were identified and were broken down into single-component and multicomponent relaxation therapy, and individualized, cognitive-behavioral therapies. Of the nondrug approaches, weight reduction–physical exercise and individualized, cognitive-behavioral psychological therapy were particularly effective and did not differ from drug treatments in observed raw effect sizes for SBP reductions. Drug therapies were initiated at higher initial levels of BP than nondrug therapies, with average pressures of 154.1 vs 145.4 mm Hg SBP and 101.5 vs 94.3 mm Hg DBP for drug and nondrug treatments, respectively. After adjustment for differences in initial pressure levels, the effects for nondrug therapies increased, and the effect size of individualized psychological therapy matched the effect sizes of drug treatments for SBP and DBP reduction. These findings suggest that the more comprehensive nondrug therapies can be effective especially when differences in pretreatment BP levels are accounted for.
The choice of BP measurement protocols also influences the observed treatment effects. Significant decisions are where to sample (physician office vs ambulatory), who measures (physician, nurse, or patient), how many samples to take, and at what intervals. Studies with longer baselines resulted in smaller treatment effects, suggesting that high initial BP readings falsely boost observed treatment effects. Part of what appears to be a treatment effect is in fact habituation to measurement.11 Probably the most promising avenue for avoiding the reliability and validity problems of office measures is via the use of ambulatory BP devices to obtain 24-hour BP averages. Ambulatory BP monitoring (ABPM) is the approach recommended by the National High Blood Pressure Education Program12 because ambulatory BPs have (1) much improved test-retest stability given the increased number of measures and wider sampling and (2) a greater potential for differentiating patients with true hypertension from measurement-reactive patients, also known as "white-coat" responders.13,14 Because white-coat responders do not habituate to measurement, they cannot be detected in the office despite repeated office measures. Furthermore, ABPM is more clinically meaningful in that 24-hour averaged ambulatory BPs are better long-term predictors of the development of hypertension than resting measures in the laboratory,15,16 and they also relate more closely to target-organ damage than do laboratory measures.17,18 Disadvantages of ABPM are higher equipment cost and a more cumbersome protocol that requires high motivation of the patient.
As shown above, variations in study design and measurement protocol affect outcomes. Features that likely inflate the magnitude of observed BP reductions are office measures and short baselines.9,11 The magnitude of expectable change is low with low entrance BPs, and it is falsely reduced when the design does not permit exclusion of white-coat responders. In terms of technique-specific outcome, the adoption of a standardized rather than an individualized approach is associated with smaller reductions in BP.9,10
This study attempted to remedy past criticisms of hypertension trials by (1) including ABPM to test the generalizability of effects in the natural environment, (2) giving patients the apparently most potent intervention (ie, individualized psychological therapy based on a cognitive-behavioral stress management conceptualization), (3) including patients with sufficiently high initial BPs that improvement is biologically more likely, (4) testing for generalizability by including measures of multiple cardiovascular risk factors (ie, weight, exercise habits, lipid profiling, anger, and hostility),19- 22 (5) including follow-up measures, and (6) replicating a type of intervention that clinicians actually use in daily practice.
Patients were recruited via newspaper advertisements and screened on the telephone for inclusion criteria other than ABPM. They were asked to come to the clinic to give informed consent and undergo office BPs, ABPM, and the other measures. Patients who met the criteria for elevated ambulatory BP were sent to the commercial laboratory adjacent to the university for a blood sample (ie, lipid profiling).
Eligible patients were then randomly assigned to a delayed or an immediate treatment condition. All patients were treated for 10 weekly 1-hour sessions and then reassessed with ABPM and all other measures approximately 3 months later. The same test package was repeated at 6-month follow-up. Subjects assigned to delayed treatment were asked to come to the clinic for monthly checkups to ensure that their BP had not undergone significant increases that might require immediate treatment. If resting BPs had increased by more than 10%, we recommended a physician visit. Medication treatment status was monitored to determine whether patients continued to meet the inclusion criterion.
Initially eligible were 168 participants. After completing the consent phase, the first 24-hour BP monitoring, and the questionnaires, 11 patients decided that they did not want to continue with the study because of the discomfort and the inconvenience associated with ambulatory monitoring. Ninety-seven of the remaining 157 patients were excluded because their 24-hour BP mean was below 140 mm Hg SBP or 90 mm Hg DBP. This left 60 patients for random assignment into the 2 treatment conditions (n = 27 in the immediate treatment group and n = 33 in the wait list control group). Four patients in the treatment condition did not complete treatment and provided no posttreatment data. Of the wait-listed patients, 4 had changed their drug regimen and had to be excluded from the analyses, while 3 refused to participate in the posttest. This left a sample of 49 for the pretreatment-posttreatment vs control comparison (23 in the immediate treatment group and 26 in the wait list control group). One participant in the immediate treatment condition provided pertinent BP data but refused to complete the questionnaires, thus reducing the sample size for the questionnaire comparisons to 22. Five patients in the immediate treatment condition refused to return for follow-up, thus leaving 18 participants at follow-up.
Completion of the second 24-hour ABPM in the wait list control group showed that only 22 still had ABPM levels greater than 140/90 mm Hg. All 22 were offered treatment; they all accepted and completed treatment. Four of these 22 delayed treatment completers refused to return for a follow-up test. Altogether, 45 patients received treatment and 36 patients also completed the follow-up. The overall protocol and the number of patients available at each step is outlined in Figure 1.
A unique feature of this study was that patients did not receive a fully standardized intervention. To guarantee a high level of quality and maximal treatment benefit, (1) the interventions were delivered by 3 PhD-level psychotherapists (including J.W.L.) with specific training in cognitive-behavioral intervention for psychosomatic patients, (2) the therapists used a set of techniques and a theoretical orientation that was supported as efficacious in the psychotherapy literature at large,22,23 (3) each therapist first conducted a thorough assessment of psychological risk factors for cardiovascular disease present in a given patient, and (4) treatment relied as much as possible on manual-type descriptions of interventions. The most frequently offered standardized therapy components were Autogenic Training,24,25 thermal biofeedback,26 cognitive therapy,27 anxiety management,28 and type A hostile behavior reduction.29 This individualized approach has been used for about 10 years in our clinic, and we have a good track record with successful case studies. Therapists were instructed to record which problems were targeted with which interventions. Analysis of the patient records showed that, on average, each patient received 3.8 interventions. Most often used were treatment of anger or hostility (39/45), Autogenic Training (37/45), and discussion of relationship or existential issues (29/45); less often used were biofeedback (20/45) and cognitive therapy for anxiety (18/45) or depression (10/45).
Patients were hypertensive with 24-hour mean ambulatory BP of or exceeding 140 mm Hg SBP and 90 mm Hg DBP. Both drug-free and drug-treated patients were included, given that a medicated patient who met criteria effectively had uncontrolled hypertension (this is consistent with expert panel recommendations1). Patients taking medication were asked to maintain their dosage at a stable level throughout the study. This, however, did not preclude medication changes when the patient and his or her physician saw an urgent need for change, and appropriate qualifying statements were included with the consent form. Exclusion criteria were type 1 diabetes mellitus, hypertension of known organic origin, and congestive heart failure. There was no upper age limit for eligibility.
Measures included 24-hour ABPM, office resting BP, a lipids profile, psychological scales (daily stress, trait anger, preferred anger coping style, hostility, anxiety, and depression), weight, and exercise habits. Psychological scales included a 1-to-9 Likert scale asking patients to rate the average stress level on the day of ABPM, the Cook-Medley Hostility Inventory,30 the Beck Depression Inventory,31 the State-Trait Anxiety Inventory,32 the State-Trait Anger Scale,33 the Interpersonal Support Evaluation List social support scale,34 the Balanced Inventory of Desirable Responding,35 and the Behavioral Anger Response Questionnaire.36 These scales were chosen because of their satisfactory psychometric properties and the availability of norms. Subjects were also asked how many alcoholic beverages per week they consumed and how much time per week they spent exercising (defined as "exercising to the point of sweating"). For clarification of the notion of "1 alcoholic beverage," type and quantity of beverages were defined.
The Cook-Medley Hostility Inventory is derived from the Minnesota Multiphasic Personality Inventory, has 50 self-descriptive items, and refers to feelings of distrust toward others (high scores refer to elevated hostility). The State-Trait Anxiety Inventory contains 20 items scored on a 1 to 4 scale. The Beck Depression Inventory is a 21-item self-report tool with responses scored from 1 to 3. The State-Trait Anger Scale (20 items) taps the overall level of anger or predisposition to react angrily. The Interpersonal Support Evaluation List is a 36-item questionnaire assessing emotional, instrumental, and self-esteem support that people perceive as available from others. The Balanced Inventory of Desirable Responding has 2 subscales (20 items each) that tap impression management (a tendency to present oneself in a positive light) and self-deception (a more unconscious chronic habit of underestimating stress and personal flaws). The Behavioral Anger Response Questionnaire is a newly developed and extensively validated tool with 37 items and a 6-factor structure; each factor forms a subscale. The subscales describe different preferred styles of responding to anger provocation: aggressive responding, assertion, social support seeking, diffusion, avoidance, and rumination.
Weight was determined via a standard clinic scale (balance model), at the same time of day, with light clothing. Blood pressure and heart rate activity in the natural environment were monitored (Spacelabs Model 90207 monitors; Spacelabs, Redmond, Wash). The ABPM monitors (weighing 700 g) were fitted in the morning to the subject, pretested (ie, readings were compared with those of Dinamap [Critikon Corp, Tampa, Fla] laboratory monitors), and returned at the same time on the next day for analysis. Validation work suggests that the Spacelabs 90207 monitor is a reliable and accurate device.37
Office resting BP measurements were taken after a 5-minute rest period without human interaction as recommended by Linden et al,38 in a comfortably seated position, with the arm fully supported, by means of an automated monitor (Dinamap 845). Five measures were taken and averaged to maximize reliability.
Lipid profiles (providing levels of cholesterol, low- and high-density lipoproteins, ratio of low-density to high-density lipoproteins, and triglycerides) were obtained by sending patients to a commercial laboratory adjacent to campus where a nurse oversaw standardized sampling and storage of the blood samples. The assays followed standard laboratory procedures. These blood samples were initially taken after the ABPM but before therapy was begun.
Means and SDs for demographic and lifestyle factors for the immediate treatment and the wait list control groups at the time of randomization are given in Table 1. Group differences were tested via χ2 tests for categorical variables and via 2-tailed t tests for interval-scaled variables.
As Table 1 indicates, there were no meaningful differences between these groups at the point of random assignment. On the whole, this was a sample with fairly healthy lifestyles, ie, few smokers, modest reported alcohol intake, and a typically moderate physical exercise habit.
Office resting BP, ambulatory means (broken down into 24-hour mean and daytime [8 AM to 7 PM] vs nighttime [7 PM to 8 AM]), and lipid levels at baseline as well as treatment-induced changes from pretreatment to posttreatment and pretreatment to follow-up are given in Table 2. Heart rate data were also available but, consistent with other treatment study results, there was no significant variability over time in any condition, so that detailed reporting appeared redundant.
The treatment response was analyzed via residualized change score analysis, which is a type of covariance analysis that individually adjusts for any potential confound of differences in baselines that may affect subsequent degree of change. Residualized change scores are derived by calculating the predicted change score as a function of the correlation between baseline and subsequent change scores. Residualized change scores are superior to covariance analyses because they have no requirement of parallel regression slopes and it is not necessary to have high intercorrelations of baselines values with change scores.39,40 Multivariate analyses were rejected because the cardiovascular and lipid variables are different classes of biologic end points, and also because the daytime and nighttime ABPM means were a function of the 24-hour means; inclusion in the same multivariate analysis would have introduced an overly liberal bias toward finding statistical significance. A different analysis strategy was used for posttreatment to follow-up comparisons because the wait list group had now been treated and there was no longer a 2-group design; simple F tests for repeated measures were conducted instead.
Residualized change score analyses showed significantly greater 24-hour ABPM change in the treated group relative to controls (SBP, F1,47 = 6.02, P = .02; DBP, F1,47 = 7.5, P = .009). Further analyses of daytime BP changes also demonstrated treatment benefits for SBP (F1,47 = 4.3, P = .04) and DBP (F1,47 = 4.2, P = .04) reductions. The same was true for nighttime BPs, with F1,47 = 4.3, P = .04 for SBP and F1,47 = 5.6, P = .02 for DBP. The office BP readings in the treated group changed about as much as did the ABPM readings, but, because the control group showed BP declines parallel to that of the treated group, there was no significant effect of treatment vs control (F1,47 = 0.81, P = .37 for SBP; F1,47 = 1.17, P = .28 for DBP). There was no change in any of the lipid variables or in body weight (all F values were <1).
In the slightly smaller sample (n = 18) available for follow-up, there was a further reduction in 24-hour DBP (F1,16 = 9.51, P = .007) and a trend toward lower SBP (F1,16 = 2.9, P = .10). No further changes were observed in office BPs. Given that changes in the pretreatment to posttreatment phase are equally apparent in day and night readings, no further analysis on this feature was executed for the follow-up data. To determine the generalizability of the follow-up changes, the amount of BP change for those not completing follow-up was compared with the full sample data. As Table 2 shows, the full sample showed reductions of –6.1 and –4.3 mm Hg from posttreatment to follow-up, and the corresponding numbers for the noncompleters were –3.5 and –3.1 mm Hg, suggesting that they were not a particularly distinct subgroup. There was no change in body weight or any of the lipid variables during follow-up.
Pretreatment values and changes in psychological variables are displayed in Table 3. With the use of a traditional cutoff of P<.05, none of the psychological variables changed significantly more in the treated group than in the control group. Trends toward treatment-related changes were apparent in daily stress (F1,47 = 3.03, P = .09) and avoidance as an anger coping style (F1,47 = 2.91, P = .09), suggesting that treated patients reported slightly less stress and more use of avoidance. Analyses of the posttreatment to follow-up changes (with the use of simple F tests for repeated measures) for dependent measures showed no significant additional changes on any psychological variable.
Given the study protocol, there was no control condition for the evaluation of the wait list control group once it had become the delayed treatment condition. However, there was no change in the ambulatory BP of the wait list control group during the main treatment phase (Table 2). Therefore, simple repeated-measures F tests were used to determine treatment-related changes. Also, because some patients receiving treatment in phase 2 did not complete the follow-up (n = 5), the analyses of the posttreatment to follow-up changes (and pretreatment to follow-up changes) needed to be conducted separately with a correspondingly smaller sample size.
The biological and psychological pretreatment characteristics as well as treatment- and follow-up–related changes are displayed in Table 4.
Statistical analyses showed that SBP and DBP dropped significantly as a function of treatment (F1,20 = 9.71, P = .005, and F1,20 = 13.7, P<.001, respectively) and improved further from the end of treatment to follow-up (F1,16 = 17.8, P<.001, and F1,16 = 22.4, P<.001). There also were improvements on some of the psychological end points. Social support increased (F1,20 = 4.74, P = .04) and use of diffusion in angering situations increased (F1,20 = 6.68, P = .02). Some other effects approached traditional levels of significance: depression was reduced (F1,20 = 4.96, P = .05), as was trait anger (F1,20 = 4.11, P = .06). Small increases in use of diffusion and assertion as preferred anger coping styles were also noted (F1,20 = 4.02, P = .06, and F1,20 = 3.08, P = .10, respectively).
Given that ABPM levels of patients in the wait list group had not changed from the pretreatment to posttreatment measures in phase 1 (see Table 2 for detail), the effectiveness of treatment for phase 1 vs phase 2 could easily and meaningfully be compared by computing effect sizes for change. The effect size for SBP changes from pretreatment to posttreatment was d = 0.60, and d = 0.91 for the pretreatment to follow-up comparison; the corresponding scores in phase 2 were d = 0.68 and d = 1.00. For DBP, the scores were d = 0.59 for pretreatment to posttreatment and d = 1.24 for pretreatment to follow-up. The corresponding figures in phase 2 were d = 0.80 and d = 1.13. These effect sizes suggest that the overall treatment effect was very similar irrespective of whether patients were in the immediate or the delayed treatment conditions.
Of interest was how many patients completed treatment but not follow-up and how these patients may have differed from the others. For reasons of parsimony and maximal statistical power, these questions were analyzed on the combined samples (ie, those treated immediately together with those receiving delayed treatment). Of 36 patients completing follow-up, 12 reported changes in their antihypertensive drug regimen during this time; 2 decreased their dosages, 3 replaced one type of medication with another, and 7 either increased the dosage or added another drug to their regimen. The 7 patients with increased medication intake represent a potential threat to the interpretability of follow-up data because the medication changes represent a confounding treatment. For this reason, the average amount of BP change during follow-up was compared for the full sample of 36 completers with a reduced sample of 29 after removal of those in whom potential treatment confounds were present. The average BP change during follow-up was −3.9 mm Hg and –3.9 mm Hg (SBP and DBP, respectively). When the sample was reduced to 29 by removing those with medication increases, the average change was –3.0 and –3.4, respectively. These differences were not considered a serious threat to the interpretability of the DBP results from the full sample of 36, whereas they do suggest a slightly weakened effect for SBP.
Although treatment was associated with significant mean group changes in ABPM, reporting of group means can hide considerable variability in treatment responses. If not all patients benefit alike, then it is important to learn who can benefit so that valuable resources are not wasted on hypertensive patients who are not responsive to psychological intervention.41 With the use of a cutoff of a 5–mm Hg reduction in ABPM values (from pretreatment to posttreatment), only 55% of treated patients were above this cutoff for SBP change and only 50% were above the cutoff for DBP change, suggesting that only about half the treated patients showed clinically meaningful changes.
Furthermore, to identify the characteristics of those who benefited from treatment and those who did not, we computed correlations for pretreatment variables with BP change from pretreatment to follow-up, and for BP change as a function of change in psychological end points. Only the treated sample for whom follow-up data were available was used for these analyses (n = 36). Given the exploratory nature of these tests, only some of the key findings are reported here.
Patients reporting infrequent use of aggression and assertion as preferred anger coping styles at baseline showed significantly greater SBP change due to treatment (correlations with DBP change showed a similar pattern but failed to meet a P<.05 criterion). Also, low levels of self-deception at pretreatment predicted greater SBP change (r = 0.33). Number of antihypertensive drugs was not a predictor of either SBP or DBP change. Interestingly, high levels of triglycerides at pretreatment could be statistically linked to greater 24-hour SBP. Overall, however, few baseline indices (demographic, biological, or psychological) predicted differential treatment responses.
Initial BP level was strongly predictive of subsequent change; SBP level at pretreatment (24-hour average) correlated with SBP change at posttreatment and follow-up (r = 0.45 [P<.001] and r = 0.51 [P<.001], respectively). The DBP level at pretreatment correlated with DBP change at posttreatment and follow-up (r = 0.18 [P = .23] and r = 0.58 [P<.001], respectively). When SBP improved, so did DBP (r = 0.91 for pretreatment to posttreatment and r = 0.90 for pretreatment to follow-up changes, respectively).
Finally, the interrelationships of change in psychological variables with change in 24-hour ABPM were studied to determine whether effective change of targeted psychological end points also accounted for change in ambulatory BP. There were some but overall few significant correlations. Patients who reported more frequent use of support seeking and avoidance after treatment and less use of aggression also showed greater systolic BP reductions (r = 0.41, P = .01; r = 0.35, P = .04; and r = −0.38, P = .02, respectively). Judging by the direction of correlations, DBP was similarly affected by the same psychological change variables, but none of these correlations was significant. Greater use of assertion as an anger coping style showed a trend toward an association with lower DBP (r = 0.30, P = .08). Given the small sample sizes and the exploratory nature of these tests, results should be interpreted with caution.
The goal of this study was to test whether a psychological intervention (individualized stress management) can be an efficacious treatment for primary hypertension. The findings confirm that 10 hours of such treatment, costing approximately US $800 per patient to deliver, can lead to significant and clinically meaningful reductions in both systolic and diastolic 24-hour mean BP in at least a subgroup of patients. In contrast, the untreated wait list control group showed no change in ambulatory BP means during the 3 months of observation. The positive treatment results were equally apparent in daytime and nighttime ambulatory readings. Furthermore, when the group initially placed on the wait list was treated later, the same positive treatment response was replicated. This was most apparent when BP changes were translated into effect sizes: they were almost identical for pretreatment to posttreatment changes and for the posttreatment to follow-up changes. In addition, treatment effects improved further during the 6 months of follow-up. These latter findings should be interpreted with caution, as 20% of participants were not available for the follow-up. Nevertheless, this does not appear to create a critical systematic bias because the noncompleters had been approximately as successful with treatment initially than those completing all measures. Whether treatment gains apparent at 6-month follow-up translate into even longer, stable benefits remains to be determined.
We posit that a powerful effect must have been present, because the study not only found a positive result with a relatively small sample but also succeeded in replicating the finding with another small sample. This observed replication of an effect supports the strength of the findings.
Interestingly, there was no corresponding treatment effect in office BPs because both controls and treated patients showed similar-sized, albeit small, reductions in BP from pretreatment to posttreatment. Hence, the use of ambulatory BP in this study produced results that are clearly different from office BP results. Given the growing literature on the superior criterion validity of ABPM, it was reassuring to see that our intervention showed the treatment benefits most clearly on ABPM. It appears that habituation to measurement is a serious threat to office readings and that it is likely to account for the differential results between office readings and ABPM.
The success of this study is in some contrast to previous reports1,3 and therefore warrants careful interpretation. Initially, it had been posited that this study might turn out promising results because of 3 design characteristics. First, it was argued that nondrug intervention trials typically start off with relatively low initial BP levels and that large subsequent reductions in BPs would be unlikely because of floor effects. This study confirmed that high initial levels of BP were also strongly predictive of degree of change (r = 0.51 and 0.58). Given that 24-hour ambulatory means are typically 5 to 10 points lower than office means1 and that our sample had ambulatory means of 153/97 mm Hg at pretreatment, this study had the potential to lead to substantial reductions in BP. The second promising feature was that of ABPMs being used as entry criteria, thus eliminating subjects with white-coat hypertension from inclusion. About half of all potential participants in this study had been told by their physicians in the recent past that their office BPs were greater than 140/90 mm Hg, but they failed to meet the ABPM entry criterion for this study. Because the incidence of white-coat hypertension is typically reported to be around 20% to 30%, there is strong reason to believe that subjects with white-coat hypertension were in fact excluded. The third argument was that individually tailored interventions would be more potent than standardized ones. The current protocol did not have a standardized treatment control group, and no direct test of standardized vs individualized treatment was possible. Nevertheless, the observed effect sizes conform closely to those reported previously,10 where it had been demonstrated that individualized treatment was superior to standardized treatments.
Neither lipid levels nor body weight changed in either group, suggesting that no generalization of BP change to these other biological end points had occurred. The same observation also implies that the observed treatment benefit for BP was not likely to be caused by any dramatic changes in eating or exercise behavior that could have confounded the psychological treatment benefits. We did not systematically assess potential confounding of BP changes caused by changes in salt intake; this could have an effect on outcomes, but it does not appear very likely because salt reduction (even when it is systematically implemented) tends to produce only small BP reductions.17
The psychological measures served primarily to explore the pathway of psychological characteristics leading to higher stress levels, which in turn were presumed to contribute to BP elevations. On the whole, the psychological changes were small, although all the signs point in the right direction. In the delayed treatment group, psychological changes were stronger and this may be partly because of the absence of a control group. The pattern of changes, however, was the same and most consistent for social support, perceived stress, and anger coping styles. Interestingly, stress reduction– and anger coping–related changes were also the ones that correlated with amount of BP change. Given the exploratory nature of these analyses and the potential for experiment-wise error, these results should be interpreted with caution. Nevertheless, it appears that changes in anger coping habits and stress reduction are the most promising targets for attempts to reduce BP. This observation is not entirely surprising given the long and fruitful history of studies linking stress and anger to hypertension.42- 44
The overall results were promising, but the variability in treatment outcome needs to be highlighted, because only half of all treated patients showed major improvements if a criterion of –5 mm Hg is used as a definition of success. Analyses of baseline characteristics as predictors of outcome showed that age, sex, and medication status were not significant predictors of BP reduction, but high SBP and DBP levels at pretreatment were. It was reassuring to see that age was not an impairment of treatment success. Of the psychological characteristics, only self-deception appears to predict lesser treatment benefits. It is therefore premature to make distinct recommendations about which patients should be offered the psychological intervention. At this time, we recommend considering patients for psychological treatment who report a great deal of subjective stress and/or find psychological interventions inherently appealing. The current sample consisted of relatively educated patients who were psychologically minded and willing to be active players in their health care. Other patients may prefer to control hypertension exclusively with medication or other lifestyle changes.
In terms of research, the logical next step would be to directly compare the individualized treatment offered herein with the best available standardized treatment, using ABPM and a relatively high BP entry criterion. In addition, we need to continue researching the link of psychological factors and BP change over time so as to strengthen the currently tenuous models of a pathway from emotion and behavior to hypertension and cardiovascular disease in general. In this regard, stress and anger coping styles appear to be a particularly promising area of research.
Accepted for publication December 11, 2000.
This study was supported by the Medical Research Council of Canada, Ottawa, Ontario.
I gratefully acknowledge the professional contributions of David Eveleigh, PhD, and Sheryl A. Tanco, PhD, who provided treatments (as did J.W.L.). Appreciation is also expressed to Brenda Hogan, MA, David Hammond, and David Paul, as well as Thomas Rutledge, PhD, and M. Jane Irvine, PhD, who provided valuable feedback on the manuscript.
Corresponding author: Wolfgang Linden, PhD, Psychology/UBC 2136 West Mall, Vancouver, British Columbia, Canada V6T 1Z4 (e-mail: firstname.lastname@example.org).