SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Study Limitations
  7. Conclusions
  8. Acknowledgments
  9. Disclosure
  10. References

The authors used pooled data from 6 valsartan-related studies including 3983 adherent and 10,663 nonadherent patients to evaluate blood pressure (BP) outcomes in both groups after 90 days of treatment, applying hierarchical linear and logistic regression to identify determinants of BP outcomes. The principal findings were that: (1) BP outcomes were consistently better in adherent patients; (2) approximately a quarter of the variance in 90-day BP values was attributable to a physician class effect; (3) common and unique patient- and physician-related variables were associated with BP outcomes in both groups; (4) physician vigilance was associated with better outcomes, especially in adherent patients; and (5) adherent patients were more likely to exhibit target organ damage and associated events while being prescribed more complex medication regimens. Adherence to antihypertensive medication may be a function of prior line treatment failure, severity of illness, and sequelae, and the ensuing patient resolution to change medication behavior.

The prevalence of nonadherence has been estimated at 25% in general,[1] 23% among patients with cardiovascular disease[1] and 50% among patients with hypertension.[2, 3] Nonadherence to antihypertensive medication is associated with a 30% loss of treatment effectiveness,[4] while adherence is associated with a 38% decreased risk of cardiovascular events.[5] While blood pressure (BP) control is a challenge in general,[6-10] achieving BP targets is particularly difficult in nonadherent patients.[11] Prior studies have focused mainly on identifying patient variables associated with nonadherence to antihypertensive medication,[12, 13] even though the determinants of nonadherence are believed to be multifactorial and to include, among others, health care providers and treatment-related factors.[14]

Adherence behavior was a consistent determinant of BP outcomes in multivariate analyses in a series of large observational studies on second-line treatment of hypertension with various valsartan regimens conducted in Belgium.[15] We pooled the data of 6 of these studies to examine whether there were differences in BP outcomes between adherent and nonadherent patients after 90 days of treatment, the proportion of variance in BP outcomes attributable to a physician class effect, and the patient- and physician-related determinants of BP outcomes common across and unique in each of these groups. The 14,646 evaluable patients were classified as nonadherent if they recalled not having taken their medication at some time in the 4 weeks prior to the 90-day follow-up visit; if not, they were classified as adherent. While perhaps a crude criterion potentially subject to bias, we have reported elsewhere that such a simple query is highly predictive of BP outcomes.[16]

Methods

  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Study Limitations
  7. Conclusions
  8. Acknowledgments
  9. Disclosure
  10. References

Table 1 summarizes the key characteristics of the studies from which the data were pooled. The common methodology of these studies has been described in detail elsewhere.[15] Essential elements are summarized below.

Table 1. Key Characteristics and Samples of Studies Included in Pooled Analysis
 PREVIEWIMPROVEINSISTeNOVABSCOREEXCELLENTTotal
Study characteristics
Year initiated200420042006200620082008 
Number of patients3194395010147333497354615,934
Number of physicians5046843082843546982832
 PREVIEWIMPROVEINSISTeNOVABSCOREEXCELLENTWeighted Average
  1. Abbreviations: HCTZ, hydrochlorothiazide; SD, standard deviation.

Patient characteristics
Age, y, mean (±SD)63.4±11.963.2±12.363.6±12.064.0±11.463.8±12.063.8±11.763.6±12.0
Male sex, % 47.748.748.549.052.353.950.5
Diabetes mellitus, %20.419.232.340.223.627.023.9
Valsartan formulations included
80 mg   
160 mg   
80/12.5 mg HCTZ   
160/12.5 mg HCTZ    
160/25 mg HCTZ    
80/5 mg amlodipine      
160/5 mg amlodipine      
160/10 mg amlodipine      

Design

We pooled the data from 6 similarly designed prospective, multicenter, pharmacoepidemiologic studies of 90-day second-line treatment with valsartan in hypertensive patients in whom first-line treatment failed or was not tolerated.[15] This yielded a dataset with a combined enrollment sample of 15,934 patients recruited by 2832 general practitioners. The evaluable sample included patients for whom data were available on the adherence stratification criterion and for whom BP values were available at both baseline and 90 days. The evaluable sample included 14,646 patients contributed by 2609 general practitioners.

Each study included a baseline assessment at the time of valsartan treatment initiation, and a follow-up assessment approximately 90 days later. The decision to treat with valsartan was made by the prescribing physician per best clinical judgment. Being observational studies, there were no required tests and all data collected were available from routine clinical practice.

Classification of Adherence Status

Adherence was assessed by asking patients whether they recalled not having taken their valsartan within the past 4 weeks (yes or no). This item was adapted from the first item of the Basel Assessment of Adherence Scale[17] and the Morisky Medication Adherence Scale[18] and has been shown to be highly predictive of BP outcomes.[16]

Variables and Measurements

Physicians

Physician data were collected via self-report using an investigator-designed survey. Variables included physician demographics and practice type, practice location/setting, patient mix, sources of information, and knowledge related to hypertension and of practice guidelines. Knowledge was defined by responses to 3 questions related to evidence-based hypertension management, scored correct (1) or incorrect (0), thus ranging from 0 to 3. In addition, hypertension management practices, prescription patterns, management of side effects, systolic BP (SBP) and diastolic BP (DBP) thresholds for treatment initiation and intensification, and perceptions of patient adherence were examined.

Patients

All patient data were collected as available in routine clinical practice. At baseline, patient demographics, hypertension and cardiovascular history, comorbidities, lifestyle, prior antihypertensive medications, clinical status, starting valsartan dose, class of concomitant antihypertensive medications, and self-reported adherence within the past 4 weeks were assessed. BP was instructed to be measured 3 times at 1- to 2-minute intervals, in a sitting position after 5 minutes of rest. Follow-up data (90 days after baseline) included SBP and DBP, as well as concomitant medication(s) taken or changed since previous visit, cholesterol levels, self-reported adherence with valsartan therapy within the past 4 weeks, and changes in valsartan dose since previous visit.

Statistical Modeling

We hypothesized that BP outcomes in each subgroup were related to patient- and physician-level variables. Each participating physician recruited several patients; therefore, patients could not be considered to be statistically independent but instead “nested” under their treating physician.[19] Thus, we assumed that the nj patients recruited by the jth physician shared some proportion of variance in BP values attributable to their common physician and that this might impact BP values prior to any patient-specific variables. We first performed conditional and unconditional two-level (physician- and patient-level) hierarchical linear modeling using residual maximum likelihood for each subgroup.[20] Conditional modeling identified physician- and patient-level determinants of BP values at 90 days. The unconditional models yielded the intraclass correlation coefficient (ICC), which quantified the variability in 90-day patient outcome attributable to between-physician variability (class effect) before any patient-level determinants were considered. Next, we performed logistic regression for each subgroup to identify predictors of uncontrolled SBP (≥140 mm Hg; ≥130 mm Hg for diabetics), uncontrolled DBP (≥90 mm Hg; ≥80 mm Hg for diabetics), and uncontrolled combined SBP/DBP, whereby we used generalized estimating equations to account for the nested structure of the data.

Independent variables were entered as potential determinants of BP into the statistical model, which was then purged to retain only significant determinants by using manual backward deletion (an overview of all selected variables can be found in the first column of Table VII). Only those retained as statistically significant are reported in this article. As the studies from which data were pooled differed in terms of valsartan formulations studied, we also assessed whether in any model an effect associated with a particular study existed. Such effect was interpreted as a proxy of valsartan formulation and classified as a patient-related determinant. Statistical significance was set at P<.05 and all tests were two-tailed.

Results

  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Study Limitations
  7. Conclusions
  8. Acknowledgments
  9. Disclosure
  10. References

Patients

Of the 14,646 evaluable patients, 10,663 (72.8%) were classified as adherent and 3983 (27.2%) as nonadherent (Table 2). Although differences in proportions were relatively small and influenced by the large sample size, adherent patients tended to be slightly older than nonadherent patients (P<.0001) and there were proportionately more adherent patients of female sex (P=.0220). The adherent group counted fewer patients with claudication intermittens (P=.0012), fewer smokers (P<.0001), and fewer diabetics (P=.0122). As to antihypertensive regimens, fewer adherent patients were prescribed angiotensin-converting enzyme (ACE) inhibitors (P=.0297). The groups did not differ significantly on other demographic variables, cardiovascular risk factors and history, or antihypertensive treatment patterns (Table 2).

Table 2. Patient Characteristics by Adherence Status
 Adherent (n=3983Nonadherent (n=10,663)P Value
  1. Abbreviations: ACE, angiotensin-converting enzyme; HCTZ, hydrochlorothiazide; HDL, high-density lipoprotein; LDL, low-density lipoprotein; SD, standard deviation. Missing data not reported. aCategories are not mutually exclusive, thus total percentages may exceed 100%. bBased on eNOVA, IMPROVE, INSIST, and PREVIEW studies. cBased on eNOVA, IMPROVE, INSIST, PREVIEW, and EXCELLENT studies. dExcluding the EXCELLENT study (calcium antagonist amlodipine was given as a study drug).

Demographics
Age, y, mean (±SD)64.0±11.962.9±11.8<.0001
Female sex48.8%47.1%.0220
Cardiovascular risk factors and historya
Smoker28.2%29.6%<.0001
Claudicatio intermittensb5.7%8.0%0.0012
Diabetes mellitus23.1%24.9%0.0122
Total cholesterol, mg/dL, mean (±SD)216.8±40.2218.7±39.30.2595
LDL, mg/dL, mean (±SD)117.8±43.0118.7±42.70.4932
HDL, mg/dL, mean (±SD)68.2±33.166.9±33.90.7269
Microalbuminuriab8.3%9.3%0.3759
Proteinuria4.1%4.4%0.6968
Renal impairment (creatinine >1.5 mg/dL)3.8%3.6%0.4382
Diabetic nephropathy2.9%3.1%0.6163
Amputationb0.3%0.1%0.1526
Anginab14.7%14.6%0.7964
Transient ischemic attacksb7.6%7.3%0.4962
Peripheral bypass or stent5.9%6.0%0.1947
Coronary revascularization8.9%8.8%0.9231
Cerebrovascular accident (ischemic)5.3%5.3%0.8770
Myocardial infarct8.3%7.8%0.0713
Left ventricular hypertrophyb13.4%15.2%0.0948
Congestive heart failurec4.1%4.4%0.9052
Cerebrovascular accident (hemorrhagic)b0.7%0.9%0.3334
Antihypertensive treatment patterns
Valsartan 80 mg6.7%6.5%0.6048
Valsartan 160 mg38.6%41.2% 
Valsartan 80/12.5 mg HCTZ6.5%6.4% 
Valsartan 160/12.5 mg HCTZ14.3%14.8% 
Valsartan 160/25 mg HCTZ9.7%8.6% 
Valsartan 80/5 mg amlodipine2.7%2.5% 
Valsartan 160/5 mg amlodipine16.2%14.8% 
Valsartan 160/10 mg amlodipine5.2%5.1% 
Concomitant diuretic30.1%32.6%0.7432
Concomitant α-blocker3.6%3.9%0.5509
Concomitant β-blocker41.4%42.0%0.4315
Concomitant calcium antagonistd51.0%52.4%0.3248
Concomitant ACE inhibitor4.3%5.6%0.0297

BP Values and Control

At baseline, nonadherent patients tended to have significantly higher mean SBP/DBP than adherent patients, in general and among nondiabetics but not among diabetics (Table 3). Further, proportionately fewer nonadherent patients had controlled SBP, DBP, and combined SBP/DBP at baseline in general, among nondiabetics and diabetics.

Table 3. SBP and DBP at Baseline and 90 Days by Adherence Status
 Adherent GroupNonadherent GroupP Groupsb
Baseline90 DaysΔP Value Δ 90 DaysaBaseline90 DaysΔP Value Δ 90 Daysa
  1. Abbreviations: DBP, diastolic blood pressure; SBP, systolic blood pressure; SD, standard deviation. aP values for the comparison of blood pressure (BP) values at baseline and after 90 days (adherent and nonadherent). bP values for the comparison of BP values at baseline between the adherent and nonadherent group.

SBP, mm Hg, mean (±SD)
All patients155.6±15.4136.5±11.6−19.1<.0001155.9±14.8139.3±12.2−16.6<.0001.0010
Nondiabetics155.6±15.3136.2±11.3−19.4<.0001156.2±15.0139.1±12.1−17.1<.0001.0002
Diabetics155.0±15.7136.9±12.3−18.1<.0001154.4±13.9139.6±12.6−14.8<.0001.4592
DBP, mm Hg, mean (±SD)
All patients91.2±9.781.3±7.4−9.9<.000191.8±9.283.1±7.7−8.7<.0001.0002
Nondiabetics91.5±9.681.3±7.3−10.2<.000192.2±9.283.1±7.6−9.1<.0001.0004
Diabetics90.1±10.081.0±7.4−9.1<.000190.3±9.182.9±7.9−7.4<.0001.3717
SBP, control, %
All patients7.147.0+39.9<.00015.537.9+32.4<.0001<.0001
Nondiabetics8.455.9+47.5<.00016.845.6+38.8<.0001.0015
Diabetics2.616.9+14.3<.00011.713.8+12.1<.0001.0429
DBP, control, %
All patients24.970.0+45.1<.000121.561.8+40.3<.0001<.0001
Nondiabetics30.082.9+52.9<.000127.076.1+49.1<.0001.0048
Diabetics6.925.0+18.1<.00015.418.0+12.6<.0001.0804
SBP and DBP, control, %
All patients5.542.2+36.7<.00013.932.9+29.0<.0001<.0001
Nondiabetics6.752.1+45.4<.00015.041.6+36.6<.0001.0004
Diabetics1.18.5+7.4<.00010.75.5+4.8<.0001.3467

Both adherent and nonadherent patients showed statistically significant reductions in mean SBP and DBP values from baseline to 90 days, in general and stratified by diabetic status (all P<.0001). However, adherent patients showed consistently greater reductions in BP values than did nonadherent patients during this period, in general, and stratified by diabetic status. While both groups showed significant increases in the proportions of patients achieving BP control at 90 days, proportionately fewer nonadherent patients did so, in general and stratified by diabetic status.

Hierarchical Linear Modeling of BP Values

SBP at 90 days

In the adherent group, 25% of the variance in 90-day SBP values was attributable to a physician class effect (ICC, 0.25), and the remaining 75% was accounted for by patient-level factors. In the nonadherent group, these percentages were 27% (ICC, 0.27) and 73%, respectively (Table 4).

Table 4. Hierarchical Linear Modeling of SBP at 90 Days by Adherence Status
 AdherentNonadherent
EstimateSE t P ValueEstimateSE t P Value
  1. Abbreviations: ACE, angiotensin-converting enzyme; HTN, hypertension; MI, myocardial infarct; SBP, systolic blood pressure; SE, standard error. aReferent study: EXCELLENT.

Intercept106.131.589166.78<.0001107.552.854437.68<.0001
Patient determinants
SBP at diagnosis of HTN, per 1 mm Hg0.11810.007615.51<.00010.14300.014329.99<.0001
Diabetes mellitus1.45470.52562.77.0057
Renal impairment−4.30191.1550−3.72.0002
Cardiovascular disease (MI and coronary)−1.32510.3546−3.74.0002
Total cholesterol, per 1 mg/dL0.01640.00305.51<.00010.018710.005643.31.0009
Body mass index, per 1 kg/m20.048420.021552.25.0248
Valsartan dose (0/80/160 mg)1.39860.28005.00<.0001
Hydrochlorothiazide dose (0/12.5/25 mg)2.46180.28758.56<.00013.71060.49367.52<.0001
Concomitant drug: α-blocker2.59520.70803.67.0002
Concomitant drug: β-blocker0.96410.25863.73.0002
Concomitant drug: calcium channel blocker0.80600.33562.40.0164
Concomitant drug: ACE inhibitor2.10450.69353.03.0024
IMPROVE studya1.59840.57772.77.0057
INSIST studya−7.37521.5808−4.67<.0001
PREVIEW studya3.33400.56935.86<.0001
Physician determinants
Years in practice, per 1 y0.10490.01885.59<.0001
HTN patients seen in past year (per 1 patient)−0.002050.00091−2.24.0251
Male sex2.23280.84222.65.0081
Intraclass correlation coefficient0.25  <.00010.27  <.0001

At the patient level, SBP-elevating determinants retained in both the adherent and nonadherent models were a higher SBP at the time that hypertension was first diagnosed, higher total cholesterol levels, and higher hydrochlorothiazide dose (Table 4). Most unique SBP-elevating factors in the adherent group were associated with the antihypertensive regimen prescribed: a higher valsartan dose prescribed at the start of the study and concomitant prescriptions for α-blockers, β-blockers, calcium channel blockers, and ACE inhibitors. Higher SBP was further associated with study effects for the IMPROVE and PREVIEW studies. SBP-lowering determinants in the adherent group were a history of myocardial infarction and/or coronary disease. SBP-elevating determinants unique to nonadherent patients included comorbid diabetes and higher BMI. SBP-mitigating determinants in this group were renal impairment and a study effect for the INSIST study.

At the physician level, elevated SBP was associated with the treating physician's years in practice in the adherent group. In the nonadherent group, SBP was higher for patients treated by male physicians, while the volume of hypertensive patients seen in the past 12 months had an SBP-mitigating effect.

DBP at 90 Days

In the adherent patient group, 29% of the variance in 90-day DBP values was attributable to a physician class effect (ICC, 0.29), and the remaining 71% was attributable to patient-level factors. In the nonadherent group, these percentages were 23% and 77%, respectively (Table 5).

Table 5. Hierarchical Linear Modeling of DBP at 90 Days by Adherence Status
 AdherentNonadherent
EstimateSE t P ValueEstimateSE t P Value
  1. Abbreviations: ACE, angiotensin-converting enzyme; DBP, diastolic blood pressure; HTN, hypertension; MI, myocardial infarction; SE, standard error. aReferent study: EXCELLENT.

Intercept70.751.119463.20<.000165.12451.326849.08<.0001
Patient determinants
Age, per 1 y−0.04770.0067−7.17<.0001
DBP at diagnosis of HTN, per 1 mm Hg0.08980.007711.71<.00010.15550.0120812.88<.0001
Renal impairment−1.80960.6719−2.69.0071
Cardiovascular disease (MI and coronary)−0.55590.2329−2.39.0170
Total cholesterol, per 1 mg/dL0.01090.00205.57<.0001
Body mass index, per 1 kg/m20.02510.00992.54.01100.041560.013163.16.0016
Valsartan dose (0/80/160 mg)0.49170.18272.69.0071
Hydrochlorothiazide dose (0/12.5/25 mg)1.34490.18877.13<.00011.56790.27495.70<.0001
Concomitant drug: β-blocker0.39330.16892.33.0199
Concomitant drug: ACE inhibitor0.90440.45401.99.0464
IMPROVE studya−0.68890.3419−2.01.0440
INSIST studya−1.93090.5720−3.38.0007−1.86000.8494−2.19.0287
Physician determinants
Years in practice, per 1 y0.03850.01283.02.0026
Male sex1.20920.44032.75.0061
Intraclass correlation coefficient0.29  <.00010.23  <.0001

At the patient level, DBP-elevating determinants in both groups were higher DBP at diagnosis of hypertension, higher BMI, and higher hydrochlorothiazide dose (Table 5). A DBP-lowering effect was noted in association with the INSIST study. Unique DBP-elevating factors in the adherent group were elevated cholesterol levels, higher valsartan dose at study start, and concomitant treatment with β-blockers and with ACE inhibitors. Variables associated with lower DBP were increasing patient age, a history of cardiovascular disease, and a study effect of the IMPROVE study. The only DBP-lowering factor in the nonadherent group was renal impairment.

Among the physician determinants, higher DBP was associated with more years in practice, for the adherent group only. For the nonadherent group, higher DBP was associated with male physician sex.

Logistic Regression Modeling of BP Control

SBP Control at 90 Days

At the patient level, factors decreasing the likelihood of controlled SBP in both the adherent and nonadherent group were higher SBP at the diagnosis of hypertension, comorbid diabetes, higher cholesterol levels, and higher hydrochlorothiazide doses prescribed at the start of the study (Table 6). In the adherent group, the likelihood of controlled SBP increased when patients had a history of cardiovascular disease and decreased with higher valsartan doses. Study effects were associated with the PREVIEW and IMPROVE studies. At the physician level, the variables decreasing the likelihood of SBP control were having been in practice longer and male sex. No unique patient- or physician-related determinants were retained for the nonadherent group.

Table 6. Logistic Regression Modeling of Controlled 90-Day BP by Adherence Status
 AdherentNonadherent
OR (95% CI)P ValueOR (95% CI)P Value
  1. Abbreviations: BP, blood pressure; CI, confidence interval; DBP, diastolic blood pressure; HCTZ, hydrochlorothiazide; HTN, hypertension; MI, myocardial infarction; OR, odds ratio; SBP, systolic blood pressure. aEstimate of the intercept (not intercept of odds ratio).

Intercept4.993 (4.234–5.753)a<.00013.286 (2.315–4.257)a<.0001
SBP control at 90 days
Patient determinants
SBP at diagnosis of HTN, per 1 mm Hg0.981 (0.978–0.985)<.00010.984 (0.979–0.990)<.0001
Diabetes mellitus0.136 (0.117–0.158)<.00010.157 (0.125–0.198)<.0001
Total cholesterol, per 1 mg/dL0.998 (0.996–0.999)<.00070.997 (0.995–0.999).0009
Cardiovascular disease (MI and coronary)1.201 (1.029–1.401).0199
Valsartan dose (0/80/160 mg)0.737 (0.650–0.837)<.0001
HCTZ dose (0/12.5/25 mg)0.677 (0.596–0.768)<.00010.677 (0.574–0.798)<.0001
PREVIEW study0.639 (0.531–0.768)<.0001
INSIST study1.517 (1.062–2.167).02211.748 (1.041–2.934).0348
IMPROVE study0.824 (0.680–0.999).0488
Physician determinants
Years in practice, per 1 y0.989 (0.982–0.996).0022
Male sex0.834 (0.707–0.983).0303  
 AdherentNonadherent
OR (95% CI)P ValueOR (95% CI)P Value
Intercept4.208 (3.402–5.014)a<.00015.457 (4.562–6.351)a<.0001
DBP control at 90 days
Patient determinants
Age, per 1 y1.012 (1.007–1.017)<.0001
DBP at diagnosis of HTN, per 1 mm Hg0.975 (0.970–0.981)<.00010.957 (0.948–0.966)<.0001
Diabetes mellitus0.059 (0.051–0.069)<.00010.056 (0.045–0.070)<.0001
Cardiovascular disease (MI and coronary)1.256 (1.055–1.496).0104
Total cholesterol, per 1 mg/dL0.998 (0.997–0.999).0046
Valsartan dose (0/80/160 mg)0.774 (0.678–0.885).0002
HCTZ dose (0/12.5/25 mg)0.773 (0.678–0.881).00010.605 (0.508–0.719)<.0001
INSIST study2.264 (1.331–3.853).0026
PREVIEW study0.780 (0.637–0.955).0163
IMPROVE study1.420 (1.073–1.880).0026
 AdherentNonadherent
OR (95% CI)P ValueOR (95% CI)P Value
Intercept4.571 (3.764–5.377)a<.00013.490 (2.428–4.552)a<.0001
SBP/DBP control at 90 days
Patient determinants
SBP at diagnosis of HTN, per 1 mm Hg0.979 (0.975–0.983)<.00010.984 (0.979–0.990)<.0001
DBP at diagnosis of HTN, per 1 mm Hg1.009 (1.003–1.015).0019
Diabetes mellitus0.136 (0.117–0.158)<.00010.057 (0.039–0.083)<.0001
Cardiovascular disease (MI and coronary)1.206 (1.034–1.408).0172
Total cholesterol, per 1 mg/dL0.998 (0.996–0.999).00050.996 (0.994–0.998).0003
Valsartan dose (0/80/160 mg)0.735 (0.648–0.835)<.0001
HCTZ dose (0/12.5/25 mg)0.668 (0.589–0.759)<.00010.707 (0.593–0.844).0001
IMPROVE study0.822 (0.679–0.997).0464
INSIST study1.538 (1.077–2.196).0179
PREVIEW study0.642 (0.534–0.742)<.0001
Physician determinants
Years in practice, per 1 y0.988 (0.981–0.995).0011
Male sex0.828 (0.702–0.976).02430.767 (0.589–0.998).0480
DBP Control at 90 Days

At the patient level, in both adherent and nonadherent patients, the likelihood of DBP control was negatively affected by a higher DBP at the diagnosis of hypertension, comorbid diabetes, and higher hydrochlorothiazide dose. Factors decreasing the odds of controlled DBP in the adherent group included higher cholesterol levels and higher valsartan doses, as well as a study effect for the PREVIEW study. A history of cardiovascular disease and higher age increased the odds of DBP control. Factors enhancing DBP control in the nonadherent group were study effects for the INSIST and IMPROVE studies. No physician-level factors were associated with achieving DBP control.

Combined SBP/DBP Control at 90 Days

Four variables negatively impacted the likelihood of SBP/DBP control in both adherent and nonadherent patients: higher SBP at the diagnosis of hypertension, comorbid diabetes, elevated total cholesterol levels, and higher hydrochlorothiazide doses. In adherent patients, a higher DBP at the diagnosis of hypertension, a history of a cardiovascular disease, and the study effect of INSIST were related to higher probability of achieving SBP/DBP control. Higher valsartan doses and study effects of the IMPROVE and PREVIEW studies were related to a lower level of SBP/DBP control. No unique factors impacting SBP/DBP control were retained for the nonadherent group.

At the physician level, the likelihood of SBP/DBP control was affected negatively by physician years in practice in the adherent group. In both the adherent and nonadherent group, male physician sex negatively affected the probability of achieving SBP/DBP control.

Discussion

  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Study Limitations
  7. Conclusions
  8. Acknowledgments
  9. Disclosure
  10. References

The principal findings of this subanalysis comparing 3983 adherent and 10,663 nonadherent patients to valsartan-centric antihypertensive medication regimens are 5-fold. First, BP outcomes following 90 days of second-line treatment in patients in whom first-line treatment failed or was not tolerated were consistently better in adherent patients. While BP decreased in both the adherent and nonadherent group, reductions and control rates were higher among adherent patients. Significant differences were generally not noted between adherent and non-adherent patients with diabetes. Second, and understandably in line with related findings from each of the constituent studies,[15] approximately a quarter of the variance in BP values at 90 days was attributable to a physician class effect. This affirms that patients seen by the same clinician are affected by that physician's knowledge, experience, treatment perspectives, and practice patterns, among other factors. Third, common and unique patient- and physician-related variables were associated with BP values and BP control in both groups (see Table 7 for a summary). These determinants centered on BP values at initial diagnosis of hypertension, risk factors and comorbid diseases, and treatment-related variables. Fourth, physician vigilance was associated with better outcomes. Although not measured directly, it is inferred from such seemingly paradoxical findings[15] that cardiovascular disease and renal impairment improved but that complex antihypertensive medication regimens impaired BP outcomes, especially in adherent patients. Lastly, adherent patients were more likely to exhibit target organ damage and associated events, while being prescribed more complex medication regimens. This may suggest that adherence, whether in general or in relation to antihypertensive regimens, may be less of a behavioral trait but instead shaped by poor prior treatment outcomes if not prior adverse events. These are important findings considering the evidence that good antihypertensive therapy reduces morbidity and mortality and further corroborate that increasing the effectiveness of adherence interventions might have a far greater impact on the health of the population than any improvement in specific medical treatments.[21-24]

Table 7. Summary of Determinants Retained in Hierarchical Linear and Logistic Regression Modeling (Any Occurrence of Determinant)
 AdherentNonadherent
BP ValuesBP ControlBP ValuesBP Control
SBPDBPSBPDBPSBP/DBPSBPDBPSBPDBPSBP/DBP
  1. Abbreviations: ACE, angiotensin-converting enzyme; BP, blood pressure; DBP, diastolic blood pressure; ESC-ESH, European Society of Cardiology/European Society of Hypertension; HCTZ, hydrochlorothiazide; HTN, hypertension; SBP, systolic blood pressure. Negative impact is denoted by a minus (−) sign: increases BP levels and decreases odds of BP control. Positive impact is denoted by a plus (+) sign: decreases BP levels and increases odds of BP control. aMyocardial infarct and coronary. bNumber of correct responses to 3 hypertension management questions.

Patient determinants
Demographics
Age, per 1 y + +      
Male sex          
Blood pressure
SBP at diagnosis of HTN, per 1 mm Hg    
DBP at diagnosis of HTN, per 1 mm Hg  +   
Risk factors
Diabetes mellitus   
Renal impairment     ++   
Cardiovascular diseasea+++++     
Body mass index, per 1 kg/m2       
Total cholesterol, per 1 mg/dL  
Treated for high cholesterol          
Smoker          
Valsartan regimen prescribed
Valsartan dose (0/80/160 mg)     
HCTZ dose (0/12.5/25 mg)
Concomitant antihypertensive treatment
α-Blocker         
β-Blocker        
Calcium channel blocker         
ACE inhibitor        
Diuretic          
Studies
PREVIEW      
IMPROVE+    + 
INSIST ++ +++++ 
eNOVA          
BSCORE          
EXCELLENT          
Physician determinants
Practices
Years in practice, per 1 y      
Medication visit duration for newly diagnosed HTN patient          
Experience with HTN
HTN patients in past year (per 1 patient)     +    
Male sex     
Knowledge
Knowledge test scoreb          
Notion of the ESC-ESH guidelines          
Hypertension education in past year          
Heard or read of ESC-ESH best practices          

A physician class effect accounted for between 23% and 29% of the difference in BP after 90 days of treatment. The determinants identified through hierarchical linear modeling clarify potential sources of this variance. In adherent patients, the length of time a physician had been practicing was associated with higher SBP and DBP values and a lower likelihood of SBP and SBP/DBP control. This might indicate that younger physicians are more likely to intensify therapy when observing poor BP outcomes, whereas their older colleagues may exhibit more therapeutic inertia.[25, 26] In contrast, this variable was not a determinant of BP in nonadherent patients. In this group, physicians' volume in hypertensive patients in the preceding year was retained as an SBP-mitigating factor, pointing at the role of routine clinical experience with hypertension. In line with a recent study, patients seen by female general practitioners may have better outcomes than those seen by male general practitioners.[27] Consistently, male physicians were associated with worse BP outcomes in both adherent and nonadherent patients. While modeling yielded some physician-related determinants, these are unlikely to fully explain the 23% to 29% of variance in BP values after 90 days of treatment accounted for by a physician class effect.

Concerning patient-related factors, modeling revealed that in both adherent and nonadherent patients, two factors were consistently associated with worse 90-day BP outcomes: higher BP at the time of diagnosis of hypertension and hydrochlorothiazide (HCTZ) dose. Both variables relate to the severity of hypertension and the need for combination therapy with at least HCTZ in second line if not in first line as well. Note that in adherent patients, valsartan dose followed the trend observed for HCTZ dose, underscoring the possible relationship with severity of disease and need for more aggressive treatment. Further, diabetes was confirmed as the single most powerful obstacle in achieving BP control, as evidenced by the ORs ranging from 0.053 to 0.156.[28, 29] Elevated total cholesterol was consistently related to worse BP outcomes, underscoring the influence of dyslipidemia in hypertension regardless of patients' medication behavior.[30] Body mass index was confirmed to be a predictor of higher DBP values in both groups.[31] Study-related DBP-lowering effects were noted in both the adherent and nonadherent groups in the INSIST study. As this study included a stronger valsartan formulation (Table 1), the effect was likely a proxy for treatment intensity.

While these various determinants were common to both the adherent and nonadherent groups, there was differentiation between them in terms of the determinants' relative impact—certainly in the linear models but also in the logistic models. For instance, the slope estimates of the determinants on BP for the nonadherent cohort obtained from the linear analyses were consistently greater than those for the adherent cohort, denoting a stronger negative impact on BP values after 90 days of treatment. This suggests that nonadherence amplifies the effect of variables known to be associated with poor BP outcomes. In contrast, adherence might mitigate the effect of these variables. Hence, it is not because a given determinant was retained in both the adherent and nonadherent models that its effect is constant. On the contrary, our findings reveal that the effect is expressed as a function of medication behavior: “bad” in nonadherent patients and “less bad” in adherent patients.

In addition to these common determinants, several determinants unique to each group were identified. In adherent patients, most determinants of BP after 90 days were related to medication (valsartan and concomitant antihypertensive agents) or to study-related effects differentiating studies with weaker valsartan formulations (such as PREVIEW and IMPROVE) from those including also stronger formulations, singularly or in single-pill combinations with HCTZ. This validates that complex treatment regimens with several antihypertensive agents are needed in patients in whom prior-line treatment failed and that treatment resistance should be considered. Yet, this also confirms that, while an essential behavior, patient adherence to less-intensive and less-effective treatment regimens is insufficient to achieve targets.[32]

Adherence behavior indeed may not necessarily be a trait but also a positive behavioral response shaped by patients' realization that hypertensive disease has progressed and that major clinical events either have occurred or are more likely. Note in this regard that the adherent group comprised proportionately more patients with microalbuminuria, left ventricular hypertrophy, cerebrovascular accident, claudicatio intermittens, and elevated low-density lipoprotein cholesterol, and that cardiovascular disease was retained as a determinant in nearly all models for adherent patients. The adherent group included indeed more patients with evidence of (advanced) target organ damage.

It is striking that the models for the nonadherent subsample comprised fewer unique determinants. In fact, except for study effects of the INSIST and IMPROVE studies, there were none in the logistic models; only differences in the magnitude of the ORs. Renal impairment was retained as a determinant in some of the models for the nonadherent patients, but with a BP-lowering effect, so was cardiovascular disease (specifically, a myocardial infarction or coronary disease) in adherent patients. These results may seem paradoxical, as their impact would be expected to be negative. We believe that these variables are proxies of physician vigilance and that the general practitioners in the 6 pooled studies ended up paying closer clinical attention to patients with evident advanced target organ damage.

Study Limitations

  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Study Limitations
  7. Conclusions
  8. Acknowledgments
  9. Disclosure
  10. References

Our analyses had limitations. Our data were not from pooled randomized controlled trials impeding drawing causal inferences. Yet, the methodology of the studies included in this report was virtually identical, which strengthens the assumption that the observed results were indeed a function of the variables studied. Although we found that about one fourth of the variance in BP is explained by a physician class effect, modeling yielded few specific physician factors explaining BP variation. Future studies will need to expand the data model to identify other potential determinants.

Conclusions

  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Study Limitations
  7. Conclusions
  8. Acknowledgments
  9. Disclosure
  10. References

Physicians should closely follow up patients that present with higher BP values at the diagnosis of hypertension and with well-known risk factors, such as diabetes as they will be more likely to experience uncontrolled BP despite being treated for it. As the factors influencing BP vary in adherent and nonadherent patients, physicians need to differentiate between them in the management of hypertension. Simple self-report measures are practically feasible and indicative of patients' actual intake behavior. In adherent patients with persistent uncontrolled BP additional antihypertensive medications or hygienic measures are to be considered. In nonadherent patients the first aim should be to improve intake behavior, as nonadherence severely attenuates optimum clinical benefit and results in poorer outcomes. Further, it is important for physicians to realize that they too influence BP. When experiencing difficulties in controlling BP, after tackling possible patient-related factors and after intensifying treatment, referral should be considered.

This pooled analysis of 6 observational effectiveness studies of valsartan regimens in second-line antihypertensive treatment comparing outcomes and determinants in adherent vs nonadherent patients confirms some prior findings with regard to differential outcomes, but, in particular, sheds light on the behavioral dynamics of adherence and nonadherence to medication therapy. Adherent patients were more likely to exhibit target organ damage and associated events while being prescribed more complex medication regimens, whereas nonadherent patients presented with singular risk conditions but without manifest sequelae. Adherence behavior indeed may not necessarily (only) be a trait but (also) a positive behavioral response shaped by patients' realization that hypertensive disease has progressed and that major clinical events either have occurred or are more likely.

Acknowledgments

  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Study Limitations
  7. Conclusions
  8. Acknowledgments
  9. Disclosure
  10. References

The studies of which the data were pooled were sponsored by Novartis Pharma through research grants and contracts. The analyses reported in this manuscript were not done under these grants and contracts, but independently by the external authors and without funding. Statistical analysis and manuscript development were done independently from the sponsor. Any issues related to results or manuscripts were addressed by the external authors. The sponsor had right of review and comment. The authors employed by the sponsor refrained from undue influence. I. Abraham was supported as principal investigator of the Academic Fellowship Program in Clinical Outcomes and Comparative Effectiveness Research sponsored by the US Bureau of Health Professions through the Arizona Area Health Education Program. This program also supported D. Sun as a predoctoral fellow. Y. Van Camp was supported by an international predoctoral fellowship from the University of Antwerp through a grant from the Flemish Ministry of Innovation. L. Villa was supported by a doctoral Fulbright fellowship.

Disclosure

  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Study Limitations
  7. Conclusions
  8. Acknowledgments
  9. Disclosure
  10. References

I. Abraham, K. Denhaerynck, and K. MacDonald are employees of Matrix45. Y. Van Camp, L. Villa, and D. Sun were interns at Matrix45. Matrix45 was contracted by Novartis to conduct the individual studies from which the data were pooled for the analyses reported in this paper. These analyses were performed pro bono by Matrix45 and without contract or compensation. By company policy, employees of Matrix45 cannot hold equity in client organizations or perform services for these clients independently. Matrix45 provides similar services to other biopharmaceutical companies. Vancayzeele, H. Brié, A. Aerts, and C. Hermans are employees of Novartis.

References

  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Study Limitations
  7. Conclusions
  8. Acknowledgments
  9. Disclosure
  10. References
  • 1
    DiMatteo MR. Variations in patients' adherence to medical recommendations – a quantitative review of 50 years of research. Med Care. 2004;42:200209.
  • 2
    Vrijens B, Vincze G, Kristanto P, et al. Adherence to prescribed antihypertensive drug treatments: longitudinal study of electronically compiled dosing histories. Br Med J. 2008;336:11141117.
  • 3
    Hill MN, Miller NH, DeGeest S, American Socociety of Hypertension Writing Group. ASH position paper: adherence and persistence with taking medication to control high blood pressure. J Clin Hypertens (Greenwich). 2010;12:757764.
  • 4
    DiMatteo MR, Giordani PJ, Lepper HS, Croghan TW. Patient adherence and medical treatment outcomes – a meta-analysis. Med Care. 2002;40:794811.
  • 5
    Mazzaglia G, Ambrosioni E, Alacqua M, et al. Adherence to antihypertensive medications and cardiovascular morbidity among newly diagnosed hypertensive patients. Circulation. 2009;120:15981605.
  • 6
    Berlowitz DR, Ash AS, Hickey EC, et al. Inadequate management of blood pressure in a hypertensive population. N Engl J Med. 1998;339:19571963.
  • 7
    Fagard RH, Van den Enden M, Leeman M, Warling X. Survey on treatment of hypertension and implementation of World Health Organization/International Society of Hypertension risk stratification in primary care in Belgium. J Hypertens. 2002;20:12971302.
  • 8
    Hajjar I, Kotchen TA. Trends in prevalence, awareness, treatment, and control of hypertension in the United States, 1988–2000. JAMA. 2003;290:199206.
  • 9
    Lloyd-Jones DM, Evans JC, Levy D. Hypertension in adults across the age spectrum – current outcomes and control in the community. JAMA. 2005;294:466472.
  • 10
    Van der Niepen P, Giot C, van de Borne P. Prevalence of isolated uncontrolled systolic blood pressure among treated hypertensive patients in primary care in Belgium: results of the I-inSYST survey. J Hypertens. 2008;26:10572063.
  • 11
    Rose AJ, Glickman ME, D'Amore MM, et al. Effects of daily adherence to antihypertensive medication on blood pressure control. J Clin Hypertens (Greenwich). 2011;13:416421.
  • 12
    Shelley D, Tseng TY, Andrews H, et al. Predictors of blood pressure control among hypertensives in community health centers. Am J Hypertens. 2011;24:13181323.
  • 13
    Morris AB, Li JJ, Kroenke K, et al. Factors associated with drug adherence and blood pressure control in patients with hypertension. Pharmacotherapy. 2006;26:483492.
  • 14
    World Health Organization. Adherence to Long-Term Therapies: Evidence for Action. World Health Organization: Geneva, Switzerland; 2003.
  • 15
    Abraham I, MacDonald K, Hermans C, et al. Real-world effectiveness of valsartan on hypertension and total cardiovascular risk: review and implications of a translational research program. Vasc Health Risk Manag. 2011;7:209235.
  • 16
    Villa L, Abraham I, Macdonald K, Denhaerynck K. Correlation of physician-rated adherence with therapeutical outcomes in antihypertensive treatment: pooled analysis findings from six valsartan studies including 15,583 available patients. Value Health. 2011;14:A383A384.
  • 17
    Dobbels F, Berben L, De Simone S, et al. The psychometric properties and practicability of self-report instruments to identify medication non-adherence in adult transplant patients: a systematic review. Transplantation. 2010;90:205219.
  • 18
    Morisky DE, Green LW, Levine DM. Concurrent and predictive validity of a self-reported measure of medication adherence. Med Care. 1986;24:6774.
  • 19
    Twisk JWR. Applied Multilevel Analysis. Cambridge: University Press; 2006.
  • 20
    Kreft I, De Leeuw J. Introducing Multilevel Modeling. Thousand Oaks, CA: Sage; 2002.
  • 21
    Haynes RB, Ackloo E, Sahota N, et al. Interventions for enhancing medication adherence. Cochrane Database Syst Rev. 2008;Art. No.: CD000011.
  • 22
    Planas LG, Crosby KM, Mitchell KD, Farmer KC. Evaluation of a hypertension medication therapy management program in patients with diabetes. J Am Pharm Assoc. 2009;49:164170.
  • 23
    Gutierrez-Angulo ML, Lopetegi-Uranga P, Sanchez-Martin I, Garaigordobil-Landazabal M. [Therapeutic compliance in patients with arterial hypertension and type 2 diabetes mellitus]. Rev Calid Asist. 2012;27:7277.
  • 24
    Stults B, Jones R. Management of hypertension in diabetes. Diabetes Spectr. 2006;19:2531.
  • 25
    Verpooten GA, Aerts A, Coen N, et al. Antihypertensive effectiveness of aliskiren for the ‘real-world’ management of hypertension: multilevel modelling of 180-day blood pressure outcomes (the Belgian DRIVER Study). Int J Clin Pract. 2011;65:5463.
  • 26
    Choudhry NK, Fletcher RH, Soumerai SB. Systematic review: the relationship between clinical experience and quality of health care. Ann Intern Med. 2005;142:260273.
  • 27
    Gouni-Berthold I, Berthold HK. Role of physician gender in drug therapy. Handb Exp Pharmacol. 2012;214:183208.
  • 28
    Scheen AJ, Philips JC, Krzesinski JM. [Hypertension and diabetes: about a common but complex association]. Rev Med Liege. 2012;67:133138.
  • 29
    Ruckert IM, Schunk M, Holle R, et al. Blood pressure and lipid management fall far short in persons with type 2 diabetes: results from the DIAB-CORE Consortium including six German population-based studies. Cardiovasc Diabetol. 2012;11:50.
  • 30
    Sakurai M, Stamler J, Miura K, et al. Relationship of dietary cholesterol to blood pressure: the INTERMAP study. J Hypertens. 2011;29:222228.
  • 31
    de Simone G, Devereux RB, Chinali M, et al. Risk factors for arterial hypertension in adults with initial optimal blood pressure – the strong heart study. Hypertension. 2006;47:162167.
  • 32
    Frampton JE, Scott LJ. Amlodipine/valsartan single-pill combination: a review of its use in the management of hypertension. Am J Cardiovasc Drugs. 2009;9:309330.