Original Article
Antihypertensive treatment is not a risk factor for major cardiovascular events in the Gubbio residential cohort study Mariapaola Lanti a, Paolo E. Puddu b, Oscar Terradura Vagnarelli c, Martino Laurenzi c, Massimo Cirillo d, Mario Mancini e, Alberto Zanchetti f, and Alessandro Menotti a
Background and objectives: Demonstration of antihypertensive beneficial role in population settings is difficult. Relationships of antihypertensive treatment, blood pressure control, risk factors and cardiovascular outcomes were investigated in the Gubbio study. Material and methods: Among 2248 cardiovascular disease-free men and women aged 35–74 years, individuals were classified as nonhypertensive, controlled hypertensive, uncontrolled hypertensive and untreated hypertensive based on cut-off limits of 140/90 mmHg for SBP/DBP and/or the use of antihypertensive drugs. Endpoint was the first major coronary, cerebrovascular or peripheral hard event [cardiovascular disease (CVD)] during a 15-year average. Univariate and multivariate analyses were run. Results: Nonhypertensive individuals were about 10 years younger and had lower risk factor levels than the other categories. The relative risk (and 95% confidence interval) for CVD versus nonhypertension was 1.78 (1.02–3.10) for controlled hypertension, 3.76 (2.79–5.06) for uncontrolled hypertension and 3.30 (2.59–4.21) for untreated hypertension (UTH). After adjusting for covariates, such as sex, age, achieved blood pressure and other risk factors, the CVD risk of controlled hypertension was practically equal to that of nonhypertension, and remained unchanged even when blood pressure was excluded from the model (1.03, 0.58–1.82). The higher cardiovascular risk of uncontrolled hypertension and UTH was reduced after adjusting for covariates, but remained significantly higher than in nonhypertension, with no significant differences between uncontrolled hypertension and UTH. Conclusions: A higher level of baseline risk is not due to treatment per se, the risk being similar in uncontrolled hypertension and UTH. Adjustment for risk factors reduces the risk only in controlled hypertension, suggesting that there may be structural alterations scarcely reversible by antihypertensive treatment. Keywords: coronary heart disease fatalities, coronary heart disease, Cox model, Gubbio cohort study, hypertensive treatment, prediction, risk factors Abbreviations: CI, confidence interval; CKD, chronic kidney disease; CVD, cardiovascular disease; eGFR,
estimated glomerular filtration rate; HDL, high-density lipoprotein; LDL, low-density lipoprotein; UTH, untreated hypertension
INTRODUCTION
I
t is widely recognized that antihypertensive treatment in patients with high blood pressure provides significant benefit by preventing major cardiovascular events, such as stroke, myocardial infarction and heart failure. This is based on the evidence provided by a large number of randomized clinical trials in patients with high blood pressure in which different types of antihypertensive drugs have been compared with placebo treatment [1–5]. Nevertheless, a number of population studies report antihypertensive treatment as a risk factor for cardiovascular events, with variable amounts of relative risk [6–10]. In elderly men of three European countries, in whom mortality from cardiovascular and coronary heart disease (CHD) was investigated [11], the use of antihypertensive drugs was assessed in a multivariate model with other cardiovascular risk factors: there was a positive, sometimes significant, coefficient for the antihypertensive coding variable, suggesting an increased risk of cardiovascular mortality associated with antihypertensive treatment – a finding opposite to what expected from clinical trials with antihypertensive agents [1–5]. In another observational study, the risk of myocardial infarction and cardiovascular events was increased by 80–100% or more, when on top of standard risk factors, Journal of Hypertension 2015, 33:736–744 a Associazione per la Ricerca Cardiologica, Via Arco di Parma, bDepartment of Cardiovascular, Respiratory, Nephrological, Anesthesiological and Geriatric Sciences, Sapienza University of Rome, Viale del Policlinico, Rome, cCentre of Preventive Medicine, Gubbio, dDepartment of Medicine and Surgery, University of Salerno, Salerno, eDepartment of Clinical Medicine and Surgery, Federico II University, Naples and fIstituto Auxologico Italiano and University of Milan, Milan, Italy
Correspondence to Paolo E. Puddu, MD, PhD, FESC, FACC, Laboratory of Biotechnologies Applied to Cardiovascular Medicine, Department of Cardiovascular, Respiratory, Nephrological, Anesthesiological and Geriatric Sciences, Sapienza University of Rome, Viale del Policlinico, 155, Rome 00161, Italy. Tel: +39 06 49972659; fax: +39 06 4453891; e-mail: [email protected] Received 16 July 2014 Revised 14 November 2014 Accepted 14 November 2014 J Hypertens 33:736–744 Copyright ß 2015 Wolters Kluwer Health, Inc. All rights reserved. DOI:10.1097/HJH.0000000000000490
736
www.jhypertension.com
Volume 33 Number 4 April 2015
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
Antihypertensive treatment is not a risk factor
antihypertensive drugs were coded as present [6], again conflicting with the acknowledged benefits from clinical trials [1–5]. In most of the above-mentioned studies, no interpretation is given of this unexpected finding [12], although in some studies, a tentative explanation is advanced, suggesting a potential role of adverse effects of some antihypertensive drugs on blood levels of cholesterol and glucose [13,14]. Only one study clearly analyzed the problem, reaching the conclusion that the excess risk of people on antihypertensive treatment, as compared with untreated individuals, is mainly due to higher SBP levels, thus suggesting that the excess risk found in treated hypertensive patients may be drastically reduced if SBP was brought under control [15]. Hypertension experts have repeatedly expressed concern about the small proportion of treated patients among those with high blood pressure, and the even smaller proportion of ‘controlled hypertensive patients’ [16,17]. An accurate picture of the status of antihypertensive treatment and blood pressure control can be given only by studies conducted in general population samples since other approaches automatically cause a selection bias of peculiar categories of patients. Population studies also allow investigation of the potential treatment benefits in a more natural environment defined by the everyday life situation [18,19]. However, the assessment of the impact of antihypertensive drugs in the general population commonly makes use of a dichotomous variable, such as a yes-no, or 1-0 variable termed ‘antihypertensive treatment’, adopted to run both univariate and multivariate analyses to show the role of treatment. This is a biased approach since it assumes treated hypertensive patients should be at the same cardiovascular risk as untreated hypertensive and nonhypertensive patients, disregarding the fact that hypertensive patients are usually older, have often associated risk factors and organ damage, treatment is frequently reserved to more severe patients and elevated blood pressure is rarely brought to entirely normal values. Therefore, two different but highly related questions might be raised: why is the crude entry of a variable such as antihypertensive treatment in a multivariate model often associated with an apparent increase of cardiovascular risk, and how is it possible to show the benefits of antihypertensive treatment in a general population out of the context of a randomized interventional trial. These problems were investigated with a specific and dedicated analysis of the Gubbio population study data [18–20].
MATERIAL AND METHODS Population, risk factors and mortality data Gubbio is a small town located in central Italy. All individuals living or working within the medieval walls and a subsample of those living outside the walls were invited at baseline examination. A total of 5831 individuals were invited and 5376 were seen at entry examination held between 1983 and 1985, with a participation rate of 92.2% [18–21]. Verbal consent was obtained from participants in compliance with the Helsinki Declaration. A questionnaire on lifestyle and health problems was administered, and a number of anthropometric, Journal of Hypertension
biochemical, biophysical and medical measurements were taken. For the purpose of this analysis, only individuals aged 35 through 74 years were considered. The following information and measurements taken at entry examination were used for the present analysis: sex, coded as 0 (female) or 1 (male); age in years; smoking habits were elicited from a standard questionnaire and the average number of cigarettes currently smoked per day was recorded; height in cm and weight in kg, following the WHO Cardiovascular Survey Methods Manual (WHO Manual) [22] and derived from these, BMIs (in kg/m2) were computed; SBP and DBP in mmHg, measured by trained personnel in a sitting position, after a 5 min rest, on the right arm positioned at heart level, by mercury sphygmomanometers, with the use of appropriate-sized cuffs; three measurements were taken following the procedure described by the WHO Manual [22] and the average of the second and third measurement was used for analysis; the fifth phase of the Korotkoff sound was recorded as DBP; the same procedure was used in both screenings; observers were trained and tested for standardized procedures including the use of audio-visual material provided by the London School of Hygiene and Tropical Medicine; mean blood pressure, expressed as DBP þ [(SBP DBP)/3)] in mmHg; heart rate in beats per min derived from the pulse count during 1 min; total serum cholesterol in mg/dl; high-density lipoprotein (HDL) serum cholesterol in mg/dl (for the purpose of this analysis we used HDL cholesterol and non-HDL cholesterol, the latter obtained by simple difference from total minus HDL cholesterol, as an approximation of low-density lipoprotein (LDL) cholesterol); j) fasting blood glucose in mg/dl; serum uric acid in mg/dl; serum creatinine in mg/dl and estimated glomerular filtration rate (eGFR) was derived in ml/min per 1.73 m2 of body surface area; urine creatinine in mEq/ (mmol/l); and finally a score of organ damage derived from data reported by the patients on the basis of a questionnaire and not directly measured; the presence of kidney disease, or retinal vascular changes or ECG abnormalities including left ventricular hypertrophy was classified as organ damage (expressed as 0 ¼ absent, 1 ¼ present). Biochemical measurements were estimated by enzymatic methods [23,24] and partly under external control of the WHO Lipid Reference Center of Prague. eGFR was estimated using the formula for chronic kidney disease (CKD)-EPI [25]. Information on the current use of antihypertensive drugs was also collected. Hypertension was defined as SBP at least 140 mmHg, or DBP at least 90 mmHg, or use of antihypertensive drugs. For the purpose of this analysis, individuals were classified into four categories: nonhypertensive individuals: SBP below 140 mmHg and DBP below 90 mmHg and no use of antihypertensive drugs; controlled hypertensive patients: SBP below 140 mmHg and DBP below 90 mmHg and use of antihypertensive drugs; uncontrolled hypertensive patients: SBP at least 140 mmHg or DBP at least 90 mmHg and use of antihypertensive drugs; untreated hypertensive patients: SBP at least 140 mmHg, or DBP at least 90 mmHg, and no use of antihypertensive drugs. Individuals carrying any cardiovascular disease of atherosclerotic or hypertensive origin (any manifestation of CHD, cerebrovascular diseases, peripheral artery disease) were defined as ‘prevalent’ individuals and were excluded www.jhypertension.com
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
737
Lanti et al.
from the analysis. The follow-up was conducted for 15 years after the entry examination using various techniques, such as field re-examination, checks of hospital records, interviews with patients, their family physicians and relatives, telephone interviews and postal questionnaires. The same procedures and checks of the local mortality registers were used for the evaluation of deaths and their causes. Each recorded event was accompanied by a date. For the purpose of this analysis, major fatal and nonfatal cardiovascular events were considered, including cases of CHD, cerebrovascular disease and peripheral artery disease [hard cardiovascular events: cardiovascular disease (CVD)]. The first event of any type occurring during an average of 15 years and a maximum of 18 years of follow-up was used for the analysis. Both deaths and major nonfatal cardiovascular events were coded following the 9th Revision of the WHO-International Classification of Diseases [26]. Details on the diagnostic procedures and criteria are reported elsewhere [25,27].
Statistical analysis Mean levels of the selected cardiovascular risk factors were computed for all individuals classified according to the four categories (nonhypertensive individuals; controlled hypertensive patients; uncontrolled hypertensive patients; untreated hypertensive patients), and cross comparisons were performed by one-way analysis of variance (ANOVA) and chi-square test for dichotomous variables. T tests on means or proportions between nonhypertensive patients versus each of the hypertensive categories were also performed. Kaplan–Meier graphs and statistics of the four categories were done. Proportional hazards models were solved using CVD incidence as end-point and the following covariates: model A – the four categories; model B – the four categories, sex, age; model C – the four categories, sex, age, mean blood pressure; model D – the four categories, sex, age, mean blood pressure and all the other covariates (cigarettes, BMI, blood glucose, serum cholesterol, serum HDL cholesterol, serum uric acid, eGFR, urine creatinine and the organ damage score); model E – same as model D, but excluding mean blood pressure. In all models, nonhypertensive patients were used as the reference. In all models, the four categories were treated as dummy variables using nonhypertensive patients as reference. Models were computed for all individuals and, independently, for those free from CVD events and still alive after the first 5 years of follow-up, starting from the assumption that people with uncontrolled hypertension and more organ damage may develop new events within shorter time. Mean blood pressure was considered with the purpose of exploiting the informativeness of both SBP and DBP.
RESULTS The denominator used was 2248 individuals: men were 1005 (45%) and women 1243 (55%). During 18 years, 354 major CVD events occurred. For the analysis excluding the first 5 years of follow-up, the denominator was made of 2103 individuals – 919 men (44%) and 1184 women (57%) – with 281 major CVD events in the subsequent 13 years. Of the total denominator, 1246 individuals were classified as nonhypertensive individuals (55%), 98 controlled 738
www.jhypertension.com
hypertensive patients (4.3%), 265 uncontrolled hypertensive patients (11.8%) and 639 untreated hypertensive patients (28.4%). Their general characteristics are reported in Table 1, whereas the Kaplan–Meier graph is illustrated in Fig. 1. Nonhypertensive individuals had, by definition, lower blood pressure levels, but they were also about 10 years younger than the other three categories. Other risk factors tended also to show lower or less dangerous levels in nonhypertensive patients. Among controlled hypertensive patients, SBP and DBP values, although being by definition within the normotensive range, were significantly higher than in the nonhypertensive group. Among patients with an antihypertensive treatment (controlled þ uncontrolled), 41% took a diuretic drug, alone (15%) or in combination with other medications. The other 59% were treated with a variety of other drugs, including beta-blockers, calcium antagonists, reserpine and other unspecified substances. Crude incidence of major CVD events in 18 years was lowest for nonhypertensive individuals (8.8%), higher for controlled (14.3%) and untreated hypertensive patients (24.7%), and highest for uncontrolled hypertensive patients (27.2%). The situation was similar, or even more favourable, for the controlled hypertensive patients, when individuals who underwent a major CVD event or death during the first 5 years were excluded. Altogether, the category showing the worst risk factor profile was that of uncontrolled hypertensive patients. Systematic comparisons across the four categories are of little interest due to the confounding effect of most risk factor levels and mainly of age. Therefore, multivariate analysis including the four selected categories was performed and the results are shown in Table 2. It is evident that in comparison with nonhypertensive individuals, both during the entire follow-up and when considering only the survivors after the fifth year, uncontrolled and untreated hypertensive patients were at a significantly higher risk, whereas the risk in controlled hypertensive patients was not significantly different, except when no covariate adjustment was made. Furthermore, these conclusions were independent of whether or not mean blood pressure stayed in the models, a finding indicating that categorizing the population by characteristics such as sex, age, risk factors and organ damage may be even more important than the current blood pressure values in determining long-term CVD risk. Among the 10 risk factors tested in addition to sex, age and mean blood pressure, a significant role was played by uric acid (P ¼ 0.0217) and organ damage score (P ¼ 0.0371), and marginally (P ¼ 0.0870) by urine creatinine, whereas eGFR was constantly not contributory (P ¼ 0.5110). In a secondary analysis, two types of antihypertensive agents (diuretics and all others) were also fed into the models, but their coefficients were not significant. The same was found when the use of allopurinol (used quite rarely) was tested as a possible risk factor. Moreover, eGFR was treated in quintile classes (instead of as a continuous variable) and Q2 was the reference to account for potential J shape in the carried risk, but this did not allow to identify the role of eGFR as a significant risk factor. Finally, it is of note that average eGFR among the 18 individuals who Volume 33 Number 4 April 2015
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
Antihypertensive treatment is not a risk factor TABLE 1. General characteristics of individuals classified into four categories based on blood pressure levels and antihypertensive treatment
a
Sex (no. of women/no. of men and men proportion) Age (years) BMI (kg/m2) Cigarettes (no. per day) SBPb (mmHg) DBPb (mmHg) MBP (mmHg) HR (beats/min) FBG (mg/dl) Cholesterol (mg/dl) HDL (mg/dl) Uric acid (mg/dl) Creatinine (mg/dl) b eGFR (ml/min per 1.73 m2) Urine creatinine mEq/(mmol/l) Organ damage scorea
Nonypertensive individuals (n ¼ 1246)
Controlled hypertensive patients (n ¼ 98)
Mean
Mean
660/586 (0.47) 49.31 25.35 5.77 120.44 74.38 89.74 71.29 88.13 213.46 47.92 4.72 0.90 86.88 139.67 0.006
SD ¼
SD
62/36 (0.37)
10.07 3.84 9.55 10.46 7.34 7.37 10.18 18.69 40.74 12.67 1.47 1.17 15.69 56.59 0.0799
Uncontrolled hypertensive patients (n ¼ 265)
57.79 29.17 4.63 128.81 78.60 95.34 70.14 92.24 228.35 46.60 5.60 0.94 75.99 115.60 0.010
Mean
¼
167/98 (0.37)
8.33 4.66 9.89 7.48 7.39 6.03 10.83 19.62 34.51 12.25 1.70 0.18 14.70 49.98 0.0102
59.67 29.58 2.90 160.08 88.27 112.21 71.86 97.84 233.59 47.23 5.77 0.93 76.14 113.40 0.0189
Untreated hypertensive patients (n ¼ 639)
SD
Mean
SD
P < (ANOVA)
¼
365/274 (0.44)
¼
0.113
9.52 4.14 3.99 15.47 9.96 9.07 10.55 22.32 41.14 12.34 1.48 0.17 15.48 49.35 0.0044
0.000 0.016 0.000 0.000 0.016 0.040 0.040 0.000 0.000 0.975 0.000 0.973 0.000 0.000 0.801
9.25 4.67 6.59 18.21 10.40 10.63 12.66 23.76 43.91 11.89 1.65 0.17 16.39 46.61 0.0053
59.71 27.81 3.89 154.38 85.24 108.28 77.57 93.92 229.53 49.00 5.11 0.91 79.09 114.54 0.0125
T test on means or proportions provided P < 0.05 () or P < 0.001 () versus nonhypertensive patients. ANOVA, analysis of variance; eGFR, glomerular filtration rate measured by CKDEPI formula; FBG, fasting blood glucose; HDL, high-density lipoprotein; HR, heart rate; MBP, mean blood pressure; SE, standard error. a Proportion and SE, analyzed by chi-square test, instead of ANOVA. b Not used in the multivariate models.
reported a chronic renal disease was not significantly different from that of the remaining 2230 who reported to be apparently CKD-free (79.00 versus 82.95 ml/min per 1.73 m2 of body surface area). In another secondary analysis, we found that in this population, SBP was more predictive of CVD events than mean blood pressure (MBP). However, the reported models were run with MBP since the classification of patients into four categories was made using both SBP and DBP, then compacted into MBP. In a solution run with
SBP, instead of MBP, the role of SBP was greater and there was a slight decrease of the hazard ratio for untreated hypertensive patients, but the overall likelihood of the model was unchanged. Figure 2 provides the adjusted cumulative incidence of CVD events during the entire follow-up in the four selected categories, and Fig. 3 illustrates the CVD incidence in survivors after the fifth year. CVD risk of controlled hypertensive patients was superimposable to that of nonhypertensive patients, whereas untreated and especially
Survival probability (%)
100 95 90
1
85
Groups 1 2 3 4
2
80 75 4
70
3
65 0
5
Number at risk Group: 1, nonhypertensives 1197 1246 Group: 2, controlled hypertensives 88 98 Group: 3, uncontrolled hypertensives 233 265 Group: 4, untreated hypertensives 585 639
10
15
20
1121
1021
0
79
66
0
187
148
0
483
379
0
Time (years)
FIGURE 1 Kaplan–Meier graph of crude major CVD incidence during 18 years among 2248 individuals from the Gubbio residential cohort distributed into four categories (groups). Numbers at risk at 0, 5, 10 and 15 years are shown. Log-rank test had x2 ¼ 126.65 (P < 0.0001). CVD, cardiovascular disease.
Journal of Hypertension
www.jhypertension.com
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
739
Lanti et al. TABLE 2. Cox proportional-hazards model with major cardiovascular disease events as dependent variables and different grouping of risk factors as predictors
Nonhypertensive individuals
Variables in model
Controlled hypertensive patients
Uncontrolled hypertensive patients
Untreated hypertensive patients
HR (95% CI)
HR (95% CI)
HR (95% CI)
CVD risk years 0–15 (1.02 3.10) 3.76 (0.72 2.22) 2.40 (0.70 2.17) 2.17 (0.83 3.09) 1.86
BP categories only BP categories þ sex þ age BP categories þ sex þ age þ MBP BP categories þ sex þ age þ MBP þ 10 other variablesa BP categories þ sex þ age þ 10 other variables (MBP excluded)
Reference Reference Reference Reference
1.78 1.27 1.24 1.01
Reference
1.03 (0.58 1.82)
BP categories only BP categories þ sex þ age BP categories þ sex þ age þ MBP BP categories þ sex þ age þ MBP þ 10 other variablesa BP categories þ sex þ age þ 10 other variables (MBP excluded)
Reference Reference Reference Reference
1.60 1.13 1.10 0.88
Reference
0.89 (0.46 1.73)
(0.83 (0.58 (0.57 (0.45
(2.79 (1.76 (1.45 (1.23
5.06) 3.28) 3.24) 2.82)
3.30 1.85 1.70 1.62
2.04 (1.46 2.85)
CVD risk years 5–15 3.09) 3.73 2.17) 2.34 2.13) 2.13 1.71) 1.86
(2.66 (1.65 (1.35 (1.16
(2.59 (1.42 (1.22 (1.17
4.21) 2.40) 2.37) 2.26)
1.95 (1.34 2.29)
5.23) 3.33) 3.35) 2.97)
3.46 1.89 1.75 1.68
2.00 (1.37 2.91)
(2.64 (1.41 (1.21 (1.16
4.54) 2.53) 2.54) 2.44)
1.78 (1.32 2.40)
BP, blood pressure; CI, confidence interval; CVD, cardiovascular disease; HR, hazard ratio; MBP, mean blood pressure. a The variables included were: cigarette smoked per day, heart rate, body mass index, fasting blood glucose, total serum cholesterol, HDL serum cholesterol, serum uric acid, eGFR, urine creatinine, organ damage score.
uncontrolled hypertensive patients were at a higher CVD risk, attaining an incidence of 16% in 18 years.
DISCUSSION The present analysis of a population study with a long-term (average 15 and maximum 18 years) follow-up confirms that, when unadjusted data are considered, presence of antihypertensive drug treatment is associated with a
0.25
significantly higher incidence of major cardiovascular events than in nonhypertensive patients, although in the Gubbio study, the excess risk of treated hypertensive patients was not as large as that reported in other studies [6]. The conclusion drawn from some studies that the inclusion of antihypertensive treatment should be included amongst cardiovascular risk factors [6,8,10], is likely to be biased, however, if consideration is not given to available evidence that prevalence of hypertension increases with age, high blood pressure often clusters with a number of other cardiovascular risk factors [28] and development of
3 0.25 4 3
0.20 2 1
0.15
Non-HT Controlled Uncontrolled Untreated
0.10
1 2 3 4
Risk of major CVD
Risk of major CVD
0.20
4
0.15
0.10
Non-HT
1
1
Controlled
2
2
Uncontrolled 3 Untreated
4
0.05
0.05
0.00 0.00
0 0
5
10
15
20
Follow-up (years) FIGURE 2 Risk of major CVD event between year 0 and year 18 of follow-up among 2248 individuals, as a function of four categories based on mean blood pressure and antihypertensive treatment. Adjusted for sex, age, blood pressure and 10 other risk factors (cigarette smoked per day, heart rate, BMI, fasting blood glucose, total serum cholesterol, HDL serum cholesterol, serum uric acid, eGFR,
740
www.jhypertension.com
5
10
15
20
Follow-up (years) FIGURE 3 Risk of major CVD event between year 5 and year 18 of follow-up among 2103 survivors at the end of the 5-year term, as a function of four categories based on mean blood pressure and antihypertensive treatment. Adjusted for sex, age, blood pressure and 10 other risk factors (cigarette smoked per day, heart rate, BMI, fasting blood glucose, total serum cholesterol, HDL serum cholesterol, serum uric acid, eGFR, urine creatinine, organ damage score). CVD, cardiovascular disease; eGFR, estimated glomerular filtration rate; HDL, highdensity lipoprotein; HT, hypertension.
Volume 33 Number 4 April 2015
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
Antihypertensive treatment is not a risk factor
asymptomatic and symptomatic organ damage [29]. Moreover, antihypertensive treatment normalizes blood pressure only in a minority of patients [16,17], and blood pressure normalization by treatment, though accompanied by reduced incidence of cardiovascular events, does not necessarily lead to the same low level of risk as that of naturally normotensive individuals, especially when treatment is initiated once blood pressure has been elevated for a long time and irreversible or scarcely reversible organ damage has developed [30,31]. More than a decade ago, Benetos et al. [15] were among the first to assess whether increased CVD mortality in treated hypertensive patients could be explained by blood pressure levels remaining higher than in normotensive individuals, or by the presence of associated risk factors and/or associated diseases. They investigated 8893 treated hypertensive men and women from a working cohort of people living around Paris, and 25 880 sex-matched and age-matched untreated individuals from the same cohort. Vital status was obtained for an 8–12-year period. Treated hypertensive individuals had higher SBP (þ15 mmHg) and DBP (þ9 mmHg) than controls, and a higher prevalence of associated risk factors and diseases. Treated hypertensive patients compared with untreated individuals presented a two-fold increase in the risk ratio for CVD mortality [risk ratio 1.96, 95% confidence interval (CI) 1.74–2.22] and coronary mortality (risk ratio 1.99, 95% CI 1.63–2.44). Adjustment for unmodifiable risk factors decreased the excess of cardiovascular risk observed in treated individuals only slightly (risk ratio 1.77, 95% CI 1.56–2.00 for CVD mortality; and risk ratio 1.7, 95% CI 1.44–2.16 for coronary mortality). After additional adjustment for modifiable associated risk factors, the increased mortality in treated individuals, albeit reduced, still persisted (risk ratio 1.52, 95% CI 1.33–1.74 for CVD mortality; and risk ratio 1.49, 95% CI 1.19–1.86 for coronary mortality). Only after additional adjustment for SBP, CVD and coronary mortalities became similar in the two groups of individuals (risk ratio 1.06, 95% CI 0.92–1.23; and risk ratio 1.06, 95% CI 0.85–1.35, respectively). A similar risk due to incomplete normalization of blood pressure by antihypertensive treatment was described several years ago in an Italian population in which 10-year CVD mortality was 36% higher in uncontrolled than in controlled hypertensive patients [32]. Our analysis of the Gubbio population study has taken advantage of the accurate nature of this study, a really prospective investigation with some measurement of hypertension-related organ damage, and information on all major fatal and nonfatal cardiovascular outcomes rather than simply information on the vital status of the individuals. The results of our analyses lead to conclusions partly superimposable to those drawn by Benetos et al. [15], but with some important differences that shed some further light on values and limitations of antihypertensive drug treatment. Our analysis was based on a stratification of the Gubbio population into four categories: nonhypertensive individuals, treated and controlled hypertensive patients, treated but uncontrolled hypertensive patients and untreated hypertensive patients, thus splitting the group of treated hypertensive patients into two categories based on Journal of Hypertension
achieved or unachieved blood pressure control (SBP
Antihypertensive treatment is not a risk factor for major cardiovascular events in the Gubbio residential cohort study Mariapaola Lanti a, Paolo E. Puddu b, Oscar Terradura Vagnarelli c, Martino Laurenzi c, Massimo Cirillo d, Mario Mancini e, Alberto Zanchetti f, and Alessandro Menotti a
Background and objectives: Demonstration of antihypertensive beneficial role in population settings is difficult. Relationships of antihypertensive treatment, blood pressure control, risk factors and cardiovascular outcomes were investigated in the Gubbio study. Material and methods: Among 2248 cardiovascular disease-free men and women aged 35–74 years, individuals were classified as nonhypertensive, controlled hypertensive, uncontrolled hypertensive and untreated hypertensive based on cut-off limits of 140/90 mmHg for SBP/DBP and/or the use of antihypertensive drugs. Endpoint was the first major coronary, cerebrovascular or peripheral hard event [cardiovascular disease (CVD)] during a 15-year average. Univariate and multivariate analyses were run. Results: Nonhypertensive individuals were about 10 years younger and had lower risk factor levels than the other categories. The relative risk (and 95% confidence interval) for CVD versus nonhypertension was 1.78 (1.02–3.10) for controlled hypertension, 3.76 (2.79–5.06) for uncontrolled hypertension and 3.30 (2.59–4.21) for untreated hypertension (UTH). After adjusting for covariates, such as sex, age, achieved blood pressure and other risk factors, the CVD risk of controlled hypertension was practically equal to that of nonhypertension, and remained unchanged even when blood pressure was excluded from the model (1.03, 0.58–1.82). The higher cardiovascular risk of uncontrolled hypertension and UTH was reduced after adjusting for covariates, but remained significantly higher than in nonhypertension, with no significant differences between uncontrolled hypertension and UTH. Conclusions: A higher level of baseline risk is not due to treatment per se, the risk being similar in uncontrolled hypertension and UTH. Adjustment for risk factors reduces the risk only in controlled hypertension, suggesting that there may be structural alterations scarcely reversible by antihypertensive treatment. Keywords: coronary heart disease fatalities, coronary heart disease, Cox model, Gubbio cohort study, hypertensive treatment, prediction, risk factors Abbreviations: CI, confidence interval; CKD, chronic kidney disease; CVD, cardiovascular disease; eGFR,
estimated glomerular filtration rate; HDL, high-density lipoprotein; LDL, low-density lipoprotein; UTH, untreated hypertension
INTRODUCTION
I
t is widely recognized that antihypertensive treatment in patients with high blood pressure provides significant benefit by preventing major cardiovascular events, such as stroke, myocardial infarction and heart failure. This is based on the evidence provided by a large number of randomized clinical trials in patients with high blood pressure in which different types of antihypertensive drugs have been compared with placebo treatment [1–5]. Nevertheless, a number of population studies report antihypertensive treatment as a risk factor for cardiovascular events, with variable amounts of relative risk [6–10]. In elderly men of three European countries, in whom mortality from cardiovascular and coronary heart disease (CHD) was investigated [11], the use of antihypertensive drugs was assessed in a multivariate model with other cardiovascular risk factors: there was a positive, sometimes significant, coefficient for the antihypertensive coding variable, suggesting an increased risk of cardiovascular mortality associated with antihypertensive treatment – a finding opposite to what expected from clinical trials with antihypertensive agents [1–5]. In another observational study, the risk of myocardial infarction and cardiovascular events was increased by 80–100% or more, when on top of standard risk factors, Journal of Hypertension 2015, 33:736–744 a Associazione per la Ricerca Cardiologica, Via Arco di Parma, bDepartment of Cardiovascular, Respiratory, Nephrological, Anesthesiological and Geriatric Sciences, Sapienza University of Rome, Viale del Policlinico, Rome, cCentre of Preventive Medicine, Gubbio, dDepartment of Medicine and Surgery, University of Salerno, Salerno, eDepartment of Clinical Medicine and Surgery, Federico II University, Naples and fIstituto Auxologico Italiano and University of Milan, Milan, Italy
Correspondence to Paolo E. Puddu, MD, PhD, FESC, FACC, Laboratory of Biotechnologies Applied to Cardiovascular Medicine, Department of Cardiovascular, Respiratory, Nephrological, Anesthesiological and Geriatric Sciences, Sapienza University of Rome, Viale del Policlinico, 155, Rome 00161, Italy. Tel: +39 06 49972659; fax: +39 06 4453891; e-mail: [email protected] Received 16 July 2014 Revised 14 November 2014 Accepted 14 November 2014 J Hypertens 33:736–744 Copyright ß 2015 Wolters Kluwer Health, Inc. All rights reserved. DOI:10.1097/HJH.0000000000000490
736
www.jhypertension.com
Volume 33 Number 4 April 2015
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
Antihypertensive treatment is not a risk factor
antihypertensive drugs were coded as present [6], again conflicting with the acknowledged benefits from clinical trials [1–5]. In most of the above-mentioned studies, no interpretation is given of this unexpected finding [12], although in some studies, a tentative explanation is advanced, suggesting a potential role of adverse effects of some antihypertensive drugs on blood levels of cholesterol and glucose [13,14]. Only one study clearly analyzed the problem, reaching the conclusion that the excess risk of people on antihypertensive treatment, as compared with untreated individuals, is mainly due to higher SBP levels, thus suggesting that the excess risk found in treated hypertensive patients may be drastically reduced if SBP was brought under control [15]. Hypertension experts have repeatedly expressed concern about the small proportion of treated patients among those with high blood pressure, and the even smaller proportion of ‘controlled hypertensive patients’ [16,17]. An accurate picture of the status of antihypertensive treatment and blood pressure control can be given only by studies conducted in general population samples since other approaches automatically cause a selection bias of peculiar categories of patients. Population studies also allow investigation of the potential treatment benefits in a more natural environment defined by the everyday life situation [18,19]. However, the assessment of the impact of antihypertensive drugs in the general population commonly makes use of a dichotomous variable, such as a yes-no, or 1-0 variable termed ‘antihypertensive treatment’, adopted to run both univariate and multivariate analyses to show the role of treatment. This is a biased approach since it assumes treated hypertensive patients should be at the same cardiovascular risk as untreated hypertensive and nonhypertensive patients, disregarding the fact that hypertensive patients are usually older, have often associated risk factors and organ damage, treatment is frequently reserved to more severe patients and elevated blood pressure is rarely brought to entirely normal values. Therefore, two different but highly related questions might be raised: why is the crude entry of a variable such as antihypertensive treatment in a multivariate model often associated with an apparent increase of cardiovascular risk, and how is it possible to show the benefits of antihypertensive treatment in a general population out of the context of a randomized interventional trial. These problems were investigated with a specific and dedicated analysis of the Gubbio population study data [18–20].
MATERIAL AND METHODS Population, risk factors and mortality data Gubbio is a small town located in central Italy. All individuals living or working within the medieval walls and a subsample of those living outside the walls were invited at baseline examination. A total of 5831 individuals were invited and 5376 were seen at entry examination held between 1983 and 1985, with a participation rate of 92.2% [18–21]. Verbal consent was obtained from participants in compliance with the Helsinki Declaration. A questionnaire on lifestyle and health problems was administered, and a number of anthropometric, Journal of Hypertension
biochemical, biophysical and medical measurements were taken. For the purpose of this analysis, only individuals aged 35 through 74 years were considered. The following information and measurements taken at entry examination were used for the present analysis: sex, coded as 0 (female) or 1 (male); age in years; smoking habits were elicited from a standard questionnaire and the average number of cigarettes currently smoked per day was recorded; height in cm and weight in kg, following the WHO Cardiovascular Survey Methods Manual (WHO Manual) [22] and derived from these, BMIs (in kg/m2) were computed; SBP and DBP in mmHg, measured by trained personnel in a sitting position, after a 5 min rest, on the right arm positioned at heart level, by mercury sphygmomanometers, with the use of appropriate-sized cuffs; three measurements were taken following the procedure described by the WHO Manual [22] and the average of the second and third measurement was used for analysis; the fifth phase of the Korotkoff sound was recorded as DBP; the same procedure was used in both screenings; observers were trained and tested for standardized procedures including the use of audio-visual material provided by the London School of Hygiene and Tropical Medicine; mean blood pressure, expressed as DBP þ [(SBP DBP)/3)] in mmHg; heart rate in beats per min derived from the pulse count during 1 min; total serum cholesterol in mg/dl; high-density lipoprotein (HDL) serum cholesterol in mg/dl (for the purpose of this analysis we used HDL cholesterol and non-HDL cholesterol, the latter obtained by simple difference from total minus HDL cholesterol, as an approximation of low-density lipoprotein (LDL) cholesterol); j) fasting blood glucose in mg/dl; serum uric acid in mg/dl; serum creatinine in mg/dl and estimated glomerular filtration rate (eGFR) was derived in ml/min per 1.73 m2 of body surface area; urine creatinine in mEq/ (mmol/l); and finally a score of organ damage derived from data reported by the patients on the basis of a questionnaire and not directly measured; the presence of kidney disease, or retinal vascular changes or ECG abnormalities including left ventricular hypertrophy was classified as organ damage (expressed as 0 ¼ absent, 1 ¼ present). Biochemical measurements were estimated by enzymatic methods [23,24] and partly under external control of the WHO Lipid Reference Center of Prague. eGFR was estimated using the formula for chronic kidney disease (CKD)-EPI [25]. Information on the current use of antihypertensive drugs was also collected. Hypertension was defined as SBP at least 140 mmHg, or DBP at least 90 mmHg, or use of antihypertensive drugs. For the purpose of this analysis, individuals were classified into four categories: nonhypertensive individuals: SBP below 140 mmHg and DBP below 90 mmHg and no use of antihypertensive drugs; controlled hypertensive patients: SBP below 140 mmHg and DBP below 90 mmHg and use of antihypertensive drugs; uncontrolled hypertensive patients: SBP at least 140 mmHg or DBP at least 90 mmHg and use of antihypertensive drugs; untreated hypertensive patients: SBP at least 140 mmHg, or DBP at least 90 mmHg, and no use of antihypertensive drugs. Individuals carrying any cardiovascular disease of atherosclerotic or hypertensive origin (any manifestation of CHD, cerebrovascular diseases, peripheral artery disease) were defined as ‘prevalent’ individuals and were excluded www.jhypertension.com
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
737
Lanti et al.
from the analysis. The follow-up was conducted for 15 years after the entry examination using various techniques, such as field re-examination, checks of hospital records, interviews with patients, their family physicians and relatives, telephone interviews and postal questionnaires. The same procedures and checks of the local mortality registers were used for the evaluation of deaths and their causes. Each recorded event was accompanied by a date. For the purpose of this analysis, major fatal and nonfatal cardiovascular events were considered, including cases of CHD, cerebrovascular disease and peripheral artery disease [hard cardiovascular events: cardiovascular disease (CVD)]. The first event of any type occurring during an average of 15 years and a maximum of 18 years of follow-up was used for the analysis. Both deaths and major nonfatal cardiovascular events were coded following the 9th Revision of the WHO-International Classification of Diseases [26]. Details on the diagnostic procedures and criteria are reported elsewhere [25,27].
Statistical analysis Mean levels of the selected cardiovascular risk factors were computed for all individuals classified according to the four categories (nonhypertensive individuals; controlled hypertensive patients; uncontrolled hypertensive patients; untreated hypertensive patients), and cross comparisons were performed by one-way analysis of variance (ANOVA) and chi-square test for dichotomous variables. T tests on means or proportions between nonhypertensive patients versus each of the hypertensive categories were also performed. Kaplan–Meier graphs and statistics of the four categories were done. Proportional hazards models were solved using CVD incidence as end-point and the following covariates: model A – the four categories; model B – the four categories, sex, age; model C – the four categories, sex, age, mean blood pressure; model D – the four categories, sex, age, mean blood pressure and all the other covariates (cigarettes, BMI, blood glucose, serum cholesterol, serum HDL cholesterol, serum uric acid, eGFR, urine creatinine and the organ damage score); model E – same as model D, but excluding mean blood pressure. In all models, nonhypertensive patients were used as the reference. In all models, the four categories were treated as dummy variables using nonhypertensive patients as reference. Models were computed for all individuals and, independently, for those free from CVD events and still alive after the first 5 years of follow-up, starting from the assumption that people with uncontrolled hypertension and more organ damage may develop new events within shorter time. Mean blood pressure was considered with the purpose of exploiting the informativeness of both SBP and DBP.
RESULTS The denominator used was 2248 individuals: men were 1005 (45%) and women 1243 (55%). During 18 years, 354 major CVD events occurred. For the analysis excluding the first 5 years of follow-up, the denominator was made of 2103 individuals – 919 men (44%) and 1184 women (57%) – with 281 major CVD events in the subsequent 13 years. Of the total denominator, 1246 individuals were classified as nonhypertensive individuals (55%), 98 controlled 738
www.jhypertension.com
hypertensive patients (4.3%), 265 uncontrolled hypertensive patients (11.8%) and 639 untreated hypertensive patients (28.4%). Their general characteristics are reported in Table 1, whereas the Kaplan–Meier graph is illustrated in Fig. 1. Nonhypertensive individuals had, by definition, lower blood pressure levels, but they were also about 10 years younger than the other three categories. Other risk factors tended also to show lower or less dangerous levels in nonhypertensive patients. Among controlled hypertensive patients, SBP and DBP values, although being by definition within the normotensive range, were significantly higher than in the nonhypertensive group. Among patients with an antihypertensive treatment (controlled þ uncontrolled), 41% took a diuretic drug, alone (15%) or in combination with other medications. The other 59% were treated with a variety of other drugs, including beta-blockers, calcium antagonists, reserpine and other unspecified substances. Crude incidence of major CVD events in 18 years was lowest for nonhypertensive individuals (8.8%), higher for controlled (14.3%) and untreated hypertensive patients (24.7%), and highest for uncontrolled hypertensive patients (27.2%). The situation was similar, or even more favourable, for the controlled hypertensive patients, when individuals who underwent a major CVD event or death during the first 5 years were excluded. Altogether, the category showing the worst risk factor profile was that of uncontrolled hypertensive patients. Systematic comparisons across the four categories are of little interest due to the confounding effect of most risk factor levels and mainly of age. Therefore, multivariate analysis including the four selected categories was performed and the results are shown in Table 2. It is evident that in comparison with nonhypertensive individuals, both during the entire follow-up and when considering only the survivors after the fifth year, uncontrolled and untreated hypertensive patients were at a significantly higher risk, whereas the risk in controlled hypertensive patients was not significantly different, except when no covariate adjustment was made. Furthermore, these conclusions were independent of whether or not mean blood pressure stayed in the models, a finding indicating that categorizing the population by characteristics such as sex, age, risk factors and organ damage may be even more important than the current blood pressure values in determining long-term CVD risk. Among the 10 risk factors tested in addition to sex, age and mean blood pressure, a significant role was played by uric acid (P ¼ 0.0217) and organ damage score (P ¼ 0.0371), and marginally (P ¼ 0.0870) by urine creatinine, whereas eGFR was constantly not contributory (P ¼ 0.5110). In a secondary analysis, two types of antihypertensive agents (diuretics and all others) were also fed into the models, but their coefficients were not significant. The same was found when the use of allopurinol (used quite rarely) was tested as a possible risk factor. Moreover, eGFR was treated in quintile classes (instead of as a continuous variable) and Q2 was the reference to account for potential J shape in the carried risk, but this did not allow to identify the role of eGFR as a significant risk factor. Finally, it is of note that average eGFR among the 18 individuals who Volume 33 Number 4 April 2015
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
Antihypertensive treatment is not a risk factor TABLE 1. General characteristics of individuals classified into four categories based on blood pressure levels and antihypertensive treatment
a
Sex (no. of women/no. of men and men proportion) Age (years) BMI (kg/m2) Cigarettes (no. per day) SBPb (mmHg) DBPb (mmHg) MBP (mmHg) HR (beats/min) FBG (mg/dl) Cholesterol (mg/dl) HDL (mg/dl) Uric acid (mg/dl) Creatinine (mg/dl) b eGFR (ml/min per 1.73 m2) Urine creatinine mEq/(mmol/l) Organ damage scorea
Nonypertensive individuals (n ¼ 1246)
Controlled hypertensive patients (n ¼ 98)
Mean
Mean
660/586 (0.47) 49.31 25.35 5.77 120.44 74.38 89.74 71.29 88.13 213.46 47.92 4.72 0.90 86.88 139.67 0.006
SD ¼
SD
62/36 (0.37)
10.07 3.84 9.55 10.46 7.34 7.37 10.18 18.69 40.74 12.67 1.47 1.17 15.69 56.59 0.0799
Uncontrolled hypertensive patients (n ¼ 265)
57.79 29.17 4.63 128.81 78.60 95.34 70.14 92.24 228.35 46.60 5.60 0.94 75.99 115.60 0.010
Mean
¼
167/98 (0.37)
8.33 4.66 9.89 7.48 7.39 6.03 10.83 19.62 34.51 12.25 1.70 0.18 14.70 49.98 0.0102
59.67 29.58 2.90 160.08 88.27 112.21 71.86 97.84 233.59 47.23 5.77 0.93 76.14 113.40 0.0189
Untreated hypertensive patients (n ¼ 639)
SD
Mean
SD
P < (ANOVA)
¼
365/274 (0.44)
¼
0.113
9.52 4.14 3.99 15.47 9.96 9.07 10.55 22.32 41.14 12.34 1.48 0.17 15.48 49.35 0.0044
0.000 0.016 0.000 0.000 0.016 0.040 0.040 0.000 0.000 0.975 0.000 0.973 0.000 0.000 0.801
9.25 4.67 6.59 18.21 10.40 10.63 12.66 23.76 43.91 11.89 1.65 0.17 16.39 46.61 0.0053
59.71 27.81 3.89 154.38 85.24 108.28 77.57 93.92 229.53 49.00 5.11 0.91 79.09 114.54 0.0125
T test on means or proportions provided P < 0.05 () or P < 0.001 () versus nonhypertensive patients. ANOVA, analysis of variance; eGFR, glomerular filtration rate measured by CKDEPI formula; FBG, fasting blood glucose; HDL, high-density lipoprotein; HR, heart rate; MBP, mean blood pressure; SE, standard error. a Proportion and SE, analyzed by chi-square test, instead of ANOVA. b Not used in the multivariate models.
reported a chronic renal disease was not significantly different from that of the remaining 2230 who reported to be apparently CKD-free (79.00 versus 82.95 ml/min per 1.73 m2 of body surface area). In another secondary analysis, we found that in this population, SBP was more predictive of CVD events than mean blood pressure (MBP). However, the reported models were run with MBP since the classification of patients into four categories was made using both SBP and DBP, then compacted into MBP. In a solution run with
SBP, instead of MBP, the role of SBP was greater and there was a slight decrease of the hazard ratio for untreated hypertensive patients, but the overall likelihood of the model was unchanged. Figure 2 provides the adjusted cumulative incidence of CVD events during the entire follow-up in the four selected categories, and Fig. 3 illustrates the CVD incidence in survivors after the fifth year. CVD risk of controlled hypertensive patients was superimposable to that of nonhypertensive patients, whereas untreated and especially
Survival probability (%)
100 95 90
1
85
Groups 1 2 3 4
2
80 75 4
70
3
65 0
5
Number at risk Group: 1, nonhypertensives 1197 1246 Group: 2, controlled hypertensives 88 98 Group: 3, uncontrolled hypertensives 233 265 Group: 4, untreated hypertensives 585 639
10
15
20
1121
1021
0
79
66
0
187
148
0
483
379
0
Time (years)
FIGURE 1 Kaplan–Meier graph of crude major CVD incidence during 18 years among 2248 individuals from the Gubbio residential cohort distributed into four categories (groups). Numbers at risk at 0, 5, 10 and 15 years are shown. Log-rank test had x2 ¼ 126.65 (P < 0.0001). CVD, cardiovascular disease.
Journal of Hypertension
www.jhypertension.com
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
739
Lanti et al. TABLE 2. Cox proportional-hazards model with major cardiovascular disease events as dependent variables and different grouping of risk factors as predictors
Nonhypertensive individuals
Variables in model
Controlled hypertensive patients
Uncontrolled hypertensive patients
Untreated hypertensive patients
HR (95% CI)
HR (95% CI)
HR (95% CI)
CVD risk years 0–15 (1.02 3.10) 3.76 (0.72 2.22) 2.40 (0.70 2.17) 2.17 (0.83 3.09) 1.86
BP categories only BP categories þ sex þ age BP categories þ sex þ age þ MBP BP categories þ sex þ age þ MBP þ 10 other variablesa BP categories þ sex þ age þ 10 other variables (MBP excluded)
Reference Reference Reference Reference
1.78 1.27 1.24 1.01
Reference
1.03 (0.58 1.82)
BP categories only BP categories þ sex þ age BP categories þ sex þ age þ MBP BP categories þ sex þ age þ MBP þ 10 other variablesa BP categories þ sex þ age þ 10 other variables (MBP excluded)
Reference Reference Reference Reference
1.60 1.13 1.10 0.88
Reference
0.89 (0.46 1.73)
(0.83 (0.58 (0.57 (0.45
(2.79 (1.76 (1.45 (1.23
5.06) 3.28) 3.24) 2.82)
3.30 1.85 1.70 1.62
2.04 (1.46 2.85)
CVD risk years 5–15 3.09) 3.73 2.17) 2.34 2.13) 2.13 1.71) 1.86
(2.66 (1.65 (1.35 (1.16
(2.59 (1.42 (1.22 (1.17
4.21) 2.40) 2.37) 2.26)
1.95 (1.34 2.29)
5.23) 3.33) 3.35) 2.97)
3.46 1.89 1.75 1.68
2.00 (1.37 2.91)
(2.64 (1.41 (1.21 (1.16
4.54) 2.53) 2.54) 2.44)
1.78 (1.32 2.40)
BP, blood pressure; CI, confidence interval; CVD, cardiovascular disease; HR, hazard ratio; MBP, mean blood pressure. a The variables included were: cigarette smoked per day, heart rate, body mass index, fasting blood glucose, total serum cholesterol, HDL serum cholesterol, serum uric acid, eGFR, urine creatinine, organ damage score.
uncontrolled hypertensive patients were at a higher CVD risk, attaining an incidence of 16% in 18 years.
DISCUSSION The present analysis of a population study with a long-term (average 15 and maximum 18 years) follow-up confirms that, when unadjusted data are considered, presence of antihypertensive drug treatment is associated with a
0.25
significantly higher incidence of major cardiovascular events than in nonhypertensive patients, although in the Gubbio study, the excess risk of treated hypertensive patients was not as large as that reported in other studies [6]. The conclusion drawn from some studies that the inclusion of antihypertensive treatment should be included amongst cardiovascular risk factors [6,8,10], is likely to be biased, however, if consideration is not given to available evidence that prevalence of hypertension increases with age, high blood pressure often clusters with a number of other cardiovascular risk factors [28] and development of
3 0.25 4 3
0.20 2 1
0.15
Non-HT Controlled Uncontrolled Untreated
0.10
1 2 3 4
Risk of major CVD
Risk of major CVD
0.20
4
0.15
0.10
Non-HT
1
1
Controlled
2
2
Uncontrolled 3 Untreated
4
0.05
0.05
0.00 0.00
0 0
5
10
15
20
Follow-up (years) FIGURE 2 Risk of major CVD event between year 0 and year 18 of follow-up among 2248 individuals, as a function of four categories based on mean blood pressure and antihypertensive treatment. Adjusted for sex, age, blood pressure and 10 other risk factors (cigarette smoked per day, heart rate, BMI, fasting blood glucose, total serum cholesterol, HDL serum cholesterol, serum uric acid, eGFR,
740
www.jhypertension.com
5
10
15
20
Follow-up (years) FIGURE 3 Risk of major CVD event between year 5 and year 18 of follow-up among 2103 survivors at the end of the 5-year term, as a function of four categories based on mean blood pressure and antihypertensive treatment. Adjusted for sex, age, blood pressure and 10 other risk factors (cigarette smoked per day, heart rate, BMI, fasting blood glucose, total serum cholesterol, HDL serum cholesterol, serum uric acid, eGFR, urine creatinine, organ damage score). CVD, cardiovascular disease; eGFR, estimated glomerular filtration rate; HDL, highdensity lipoprotein; HT, hypertension.
Volume 33 Number 4 April 2015
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
Antihypertensive treatment is not a risk factor
asymptomatic and symptomatic organ damage [29]. Moreover, antihypertensive treatment normalizes blood pressure only in a minority of patients [16,17], and blood pressure normalization by treatment, though accompanied by reduced incidence of cardiovascular events, does not necessarily lead to the same low level of risk as that of naturally normotensive individuals, especially when treatment is initiated once blood pressure has been elevated for a long time and irreversible or scarcely reversible organ damage has developed [30,31]. More than a decade ago, Benetos et al. [15] were among the first to assess whether increased CVD mortality in treated hypertensive patients could be explained by blood pressure levels remaining higher than in normotensive individuals, or by the presence of associated risk factors and/or associated diseases. They investigated 8893 treated hypertensive men and women from a working cohort of people living around Paris, and 25 880 sex-matched and age-matched untreated individuals from the same cohort. Vital status was obtained for an 8–12-year period. Treated hypertensive individuals had higher SBP (þ15 mmHg) and DBP (þ9 mmHg) than controls, and a higher prevalence of associated risk factors and diseases. Treated hypertensive patients compared with untreated individuals presented a two-fold increase in the risk ratio for CVD mortality [risk ratio 1.96, 95% confidence interval (CI) 1.74–2.22] and coronary mortality (risk ratio 1.99, 95% CI 1.63–2.44). Adjustment for unmodifiable risk factors decreased the excess of cardiovascular risk observed in treated individuals only slightly (risk ratio 1.77, 95% CI 1.56–2.00 for CVD mortality; and risk ratio 1.7, 95% CI 1.44–2.16 for coronary mortality). After additional adjustment for modifiable associated risk factors, the increased mortality in treated individuals, albeit reduced, still persisted (risk ratio 1.52, 95% CI 1.33–1.74 for CVD mortality; and risk ratio 1.49, 95% CI 1.19–1.86 for coronary mortality). Only after additional adjustment for SBP, CVD and coronary mortalities became similar in the two groups of individuals (risk ratio 1.06, 95% CI 0.92–1.23; and risk ratio 1.06, 95% CI 0.85–1.35, respectively). A similar risk due to incomplete normalization of blood pressure by antihypertensive treatment was described several years ago in an Italian population in which 10-year CVD mortality was 36% higher in uncontrolled than in controlled hypertensive patients [32]. Our analysis of the Gubbio population study has taken advantage of the accurate nature of this study, a really prospective investigation with some measurement of hypertension-related organ damage, and information on all major fatal and nonfatal cardiovascular outcomes rather than simply information on the vital status of the individuals. The results of our analyses lead to conclusions partly superimposable to those drawn by Benetos et al. [15], but with some important differences that shed some further light on values and limitations of antihypertensive drug treatment. Our analysis was based on a stratification of the Gubbio population into four categories: nonhypertensive individuals, treated and controlled hypertensive patients, treated but uncontrolled hypertensive patients and untreated hypertensive patients, thus splitting the group of treated hypertensive patients into two categories based on Journal of Hypertension
achieved or unachieved blood pressure control (SBP
