Effects of HIV status and antiretroviral therapy on blood pressure

Authors


Dr Sarah Wilson, International Centre for Circulatory Health, Imperial College London, London W2 1PG, UK. Tel: +44 7979 612593; fax: +44 207 594 1145; e-mail: sarahpartridge.home@gmail.com

Abstract

Objective

High blood pressure is a major risk factor for cardiovascular disease and concerns have been raised over its possible association with antiretroviral drugs. The objective of this study was to explore the associations among blood pressure, HIV status and two predefined highly active antiretroviral therapy (HAART) regimens: treatment with and without nonnucleoside reverse transcriptase inhibitors (NNRTIs) (NNRTI- and non-NNRTI-based HAART).

Method

A cross-sectional survey was conducted among 612 adults attending the Sexual Health Outpatient Department at St Mary's NHS Hospital Trust, London.

Results

HIV-infected patients treated with NNRTIs had a blood pressure that was 4.6/4.2 mmHg higher than those who were HIV positive but treatment naïve. The diastolic difference remained statistically significant after adjusting for potential confounders of this association (2.4 mmHg; P=0.03). There was no difference in blood pressure between those treated with non-NNRTI-based regimens and those who were HIV positive but treatment naïve.

Conclusion

NNRTIs may be associated with an increase in blood pressure. Pending further more robust evidence from randomized clinical trials it would be prudent for clinicians to monitor blood pressure in all HIV-infected patients, particularly after initiating treatment with NNRTIs, and to commence antihypertensive therapy whenever appropriate.

Introduction

Deaths from AIDS have declined dramatically since the introduction of highly active antiretroviral therapy (HAART) in patients infected with HIV [1]. However, the incidence of myocardial infarction has increased fourfold and cardiovascular disease is now one of the major causes of death in HIV-infected patients in Western industrialized countries [2,3]. Anecdotal reports from the Peart-Rose (hypertension) Clinic at St Mary's NHS Hospital Trust, Paddington, London, raised awareness of an increasing number of young HIV-positive adults being referred with hypertension, a major risk factor for cardiovascular disease. This raised the possibility that HIV and/or its treatment may be associated with raised blood pressure. However, the published evidence concerning these associations is both limited and conflicting [4–15]. While the current studies do provide useful insights into a potentially emerging problem, methodological problems, including the choice of control group [4–10], inadequate blood pressure measurement [11–15] and crude categorical outcome measures [4,5,7,14,15], may have generated bias and confounded results. Another potential problem is the consideration of HAART as a single entity, despite it being an umbrella term for different combinations of antiretroviral drugs [4,5,11,13–15]. Standard HAART consists of a backbone of two nucleoside reverse transcriptase inhibitors (NRTIs) to which either a nonnucleoside reverse transcriptase inhibitor (NNRTI) or a boosted protease inhibitor (PI) is added [16]. If antiretroviral drug classes have different effects on blood pressure, this may explain the conflicting reports in the literature. In several studies, subgroup analyses have been conducted to explore the association between blood pressure and specific antiretroviral drug classes. Bergerson suggested that the use of NNRTIs may be associated with a higher prevalence of hypertension [5]; however, these findings were not reproduced in the DAD study [15]. These conflicting reports led to the design of a new cross-sectional survey to explore the associations among blood pressure, HIV status and two predefined antiretroviral treatment regimens, NNRTI- and non-NNRTI-based HAART.

Methods

Study population

Patients eligible for inclusion in this cross-sectional survey were those attending routine appointments at the Sexual Health Outpatient Department at St Mary's NHS Hospital Trust, Paddington, London, between September 2005 and September 2006 and who had had a previous HIV test. A random sample of patients was selected from the out-patient clinic lists. The pre-selected patients were then approached as they booked in for appointments and invited to participate in the study. Of those who were approached, more than 99% agreed to participate and of those only 13 patients (2%) were excluded as they had never had an HIV test. Written informed consent was obtained from each participant and the study was conducted in accordance with the Declaration of Helsinki [17] and Good Clinical Practice guidelines [18]. The protocol was approved by central UK and local hospital ethics review boards.

Measurements

An administered questionnaire was used to collect data on patient demographic characteristics, and social and medical history. Blood pressure was measured using a validated semi-automatic device, the OMRON HEM-705CP (OMRON Corporation, Kyoto, Japan), according to a standard protocol [19]. Biochemistry and virology were only available for those who were HIV positive and the most recent result was taken from the patient's medical record. All measurements were taken by a single trained researcher.

Power calculation

The power calculation was based on the hypothesis that people who were HIV positive and treated with HAART would have a diastolic blood pressure that was, on average, 4 mmHg higher than that of people who were HIV positive but untreated with HAART (HAART naïve). The standard deviation was estimated from previous studies to be 11 mmHg [4–15]. It was estimated that the study would have 80% power at the 0.05 significance level if 120 individuals were included in each of the two groups to be compared. We used the same power calculation to calculate the number of patients to be included in the HIV-negative group and the groups on different HAART regimens (NNRTI- and non-NNRTI-based HAART).

Statistical methods

The associations among blood pressure, HIV status and HIV treatment were evaluated using univariate and multivariate linear regression. Firstly, the whole data set (n=612) was used to identify factors that were associated with blood pressure by univariate analysis of the patients' demographic and lifestyle factors (e.g. age, sex, body mass index and alcohol intake) with systolic and diastolic blood pressure. Secondly, crude associations between the main outcomes of interest (HIV status/antiretroviral therapy) and systolic and diastolic blood pressure were determined using univariate linear regression. A process of forward stepwise modelling was then used to build multivariate linear regression models. Factors were included in the models if they confounded the association between exposure and outcome or were independently associated (P≤0.05) with systolic or diastolic blood pressure (e.g. age, sex, body mass index, alcohol intake, statin therapy, antihypertensive therapy, CD4 cell count and creatinine). When the final models were built, interactions were assessed by fitting an interaction term between each variable in the model and the main exposure of interest and assessing the impact using a likelihood ratio test. The appropriateness of using a linear model was explored by residual analysis and plotting of the standardized residuals against the fitted values for each of the models. The strength of evidence for the associations was assessed using 95% confidence intervals (CIs) and P-values.

Results

Characteristics of the study population

In this study, 612 individuals were recruited, of whom 154 were confirmed HIV negative and 458 were HIV positive. Of those who were HIV positive, 141 were HAART naïve and 317 were treated with HAART. Of those who were treated with HAART, 193 were on NNRTI-based regimens and 124 were on non-NNRTI-based regimens. The characteristics of the study population are shown in Table 1.

Table 1.   Characteristics of the study population
 HIV-negative
(n=154)
HIV positive,
HAART naïve
(n=141)
HIV positive,
HAART treated
(n=317)
  • Data presented are mean (standard deviation) or number (%).

  • *

    Family history is defined as a fatal or nonfatal stroke or heart attack in a father aged less than 55 years or mother aged less than 65 years at the time of the event.
    DVT, deep vein thrombosis; MI, myocardial infarction.

Age (years)33.9 (11.2)35.8 (8.5)43.2 (9.3)
Male133 (86%)112 (79%)269 (85%)
Gay/bisexual47 (31%)97 (69%)233 (74%)
Heterosexual107 (69%)44 (31%)84 (26%)
White108 (70%)96 (68%)233 (74%)
Black African9 (6%)30 (21%)58 (18%)
Other37 (24%)15 (11%)26 (8%)
Treatment for hypertension4 (3%)3 (2%)25 (8%)
Treatment for raised cholesterol (statin)3 (2%)3 (2%)62 (20%)
Diabetes1 (0.7%)2 (1.4%)9 (3%)
Previous MI/Stroke/DVT2 (1.3%)2 (1.4%)16 (5%)
Family history of cardiovascular disease*15 (10%)23 (16%)39 (12%)
Current smoker59 (38%)64 (45%)126 (40%)
Ex/never smoked95 (62%)77 (55%)191 (60%)
Alcohol units/week
 0 units32 (21%)46 (33%)101 (32%)
 1–21 units80 (52%)60 (42%)151 (48%)
 22 units or more42 (27%)35 (25%)65 (20%)
Recreational drug use in the last year52 (34%)68 (48%)98 (31%)
Body mass index24.9 (3.8)24.5 (4.4)24.4 (3.6)
Waist/hip ratio0.94 (0.06)0.93 (0.06)0.96 (0.07)

HIV-positive, HAART-naïve individuals compared with HIV-negative individuals

Comparison between the HIV-positive HAART-naïve group and the HIV-negative group suggested that they were similar in terms of age, gender, current cigarette smoking, alcohol intake, body mass index and waist–hip ratio (Table 1). Those who were HIV positive and HAART naïve were more likely to be homosexual or bisexual and were more likely to be of Black African origin, which reflects the known associates of a positive HIV status. The reported use of recreational drugs in the last year was higher in the HIV-positive HAART-naïve group (48%) than in the HIV-negative group (34%). However, both groups reported a much higher prevalence of recreational drug use than the UK national average for adults aged 15 to 64 years, estimated to be 11.8% in 2005 [20]. The numbers of individuals with previous cardiovascular disease, diabetes and treatment for hypertension or raised cholesterol were very low in both groups, as expected from other population surveys in this age group [21]. The median time since the HIV-negative group had their HIV test result was 4 weeks [interquartile range (IQR) 2 to 42 weeks]. The HIV-positive HAART-naïve group had been confirmed HIV positive for a median of 2 years (IQR 0.5 to 6 years). Their median viral load was 11 489 HIV-1 RNA copies per mL of plasma (IQR 1261 to 75 778 copies/mL plasma) and their median CD4 count was 480 cells/μL of whole blood (IQR 350–650 cells/μL whole blood).

HIV-positive, HAART-treated individuals compared with HIV-positive, HAART-naïve individuals

Comparison between the HIV-positive, HAART-treated group and the HAART-naïve group suggested that they were similar in terms of gender, sexual orientation, ethnic origin, current cigarette smoking, alcohol consumption, body mass index and creatinine. As expected, compared with the HAART-naïve patients, HAART-treated patients had been HIV positive for longer (9 years compared with 2 years); were older (43.2 years compared with 35.8 years); were more likely to be on drug treatment for hypertension and raised cholesterol; had higher waist–hip ratios and were less likely to have taken recreational drugs in the last year. The HAART-treated patients had been on antiretroviral therapy for a median of 6.8 years (IQR 3.7–9.2 years).

HIV-positive individuals treated with NNRTI- and non-NNRTI-based HAART regimens

More patients were treated with a HAART regimen that included an NNRTI than with a regimen that did not (61%vs. 39%, respectively). The characteristics of the two groups were similar with the exception of viral load and CD4 cell count. Viral load was undetectable (<50 copies/mL plasma) in 83% of those treated with an NNRTI-based HAART regimen and in 66% of those on a non-NNRTI-based HAART regimen. The median CD4 count was 535 cells/μL whole blood (IQR 365–725 cells/μL whole blood) in the NNRTI group and 470 cells/μL whole blood (IQR 340–660 cells/μL whole blood) in the non-NNRTI group. Changes to the antiretroviral treatment regimen were very common in both groups of patients, with 71% of those currently on NNRTI-based HAART and 90% of those on non-NNRTI-based HAART having taken the alternative class in the past. However, of those who were currently treated with an NNRTI, 94% of patients had been taking this class of drug for more than 6 months prior to blood pressure measurement. Of those not currently treated with an NNRTI, only one individual had taken an NNRTI in the 6 months prior to blood pressure measurement.

Blood pressure results

HIV-positive, HAART-naïve individuals compared with HIV-negative individuals

The mean blood pressure was 122.0/75.0 mmHg in the HIV-positive, HAART-naïve group and 122.9/75.5 mmHg in the HIV-negative group (the reference group) (Fig. 1). The mean blood pressure was therefore 0.9/0.5 mmHg lower in the HAART-naïve group (95% CI −3.6 to +1.7/−2.6 to 1.7; P=0.49/0.68) compared with the HIV-negative group. After adjusting for age, sex, body mass index and alcohol intake, the difference was −0.1/−0.4 mmHg (95% CI −2.4 to 2.3/−2.3 to 1.6; P=0.95/0.71).

Figure 1.

 (a) Systolic and (b) diastolic blood pressure in patients who were HIV negative (hatched bars), HIV positive and treatment naïve (open bars), HIV positive and treated with nonnucleoside reverse transcriptase inhibitor (NNRTI)-based highly active antiretroviral therapy (HAART) (solid bars), and HIV positive and treated with non-NNRTI-based HAART (grey bars).

HIV-positive, HAART-treated individuals compared with HIV-positive, HAART-naïve individuals

The mean blood pressure was 125.2/77.7 mmHg in the HIV-positive, HAART-treated group and 122.0/75.0 mmHg in the HIV-positive, HAART-naïve group (the reference group) (Fig. 1). The mean blood pressure was therefore 3.2/2.7 mmHg higher in the HAART-treated group (95% CI 0.5–5.9/0.8–4.6; P=0.018/0.006) compared with the HAART-naïve group. After adjusting for age, sex, body mass index, alcohol intake, statin treatment, antihypertensive treatment, CD4 cell count and creatinine, the difference was 0.3/1.2 mmHg (95% CI −2.3 to 2.8/−0.8 to 3.2; P=0.84/0.24).

HIV-positive, NNRTI and non-NNRTI groups and the HIV-positive, treatment-naïve group

The mean blood pressure was 126.6/79.2 mmHg in the NNRTI-based HAART group, 123.1/75.4 mmHg in the non-NNRTI-based HAART group and 122.0/75.0 mmHg in the HAART-naïve group (the reference group) (Fig. 1). The mean blood pressure was therefore 4.6/4.2 mmHg higher in the NNRTI-treated group (95% CI 1.7–7.5/2.1–6.2; P=0.002/<0.001) and 1.1/0.4 mmHg higher in the non-NNRTI-treated group (95% CI −2.2 to 4.3/−1.9 to 2.7; P=0.52/0.74) compared with the HAART-naïve group (Tables 2 and 3). After adjusting for age, sex, body mass index, alcohol intake, statin treatment, antihypertensive treatment, CD4 cell count and creatinine, the mean blood pressure was 1.3/2.4 mmHg higher in the NNRTI-treated group (95% CI −1.5 to 4.0/0.3 to 4.6; P=0.37/0.026) and −1.3/−0.7 mmHg lower in the non-NNRTI-treated group (95% CI −4.4 to 1.7/−3.1 to 1.6; P=0.39/0.54) compared with the HAART-naïve group (Tables 2 and 3).

Table 2.   Comparison of systolic blood pressure between HIV-positive, treatment-naïve patients and those on highly active antiretroviral therapy (HAART) treated with and without nonnucleoside reverse transcriptase inhibitors (NNRTIs)
 Systolic
blood pressure
(mmHg)
Crude difference in systolic
blood pressure
(95% confidence interval and P-value)
Adjusted difference in systolic
blood pressure*
(95% confidence interval and P-value)
  • *

    Adjusted for age, sex, body mass index, alcohol intake, statin treatment, antihypertensive treatment, CD4 cell count and creatinine.

HIV positive, treatment naïve122.0ReferenceReference
HIV positive, treated with NNRTI-based HAART126.64.6 (1.7 to 7.5; P=0.002)1.3 (−1.5 to 4.0; P=0.37)
HIV positive, treated with non-NNRTI-based HAART123.11.1 (−2.2 to 4.3; P=0.52)−1.3 (−4.4 to 1.7; P=0.39)
Table 3.   Comparison of diastolic blood pressure between HIV-positive, treatment-naïve patients and those treated with HAART with and without nonnucleoside reverse transcriptase inhibitors (NNRTIs)
 Diastolic
blood pressure
(mmHg)
Crude difference in diastolic
blood pressure
(95% confidence interval and P-value)
Adjusted difference in diastolic
blood pressure*
(95% confidence interval and P-value)
  • *

    Adjusted for age, sex, body mass index, alcohol intake, statin treatment, antihypertensive treatment, CD4 cell count and creatinine.

HIV positive, treatment naïve75.0ReferenceReference
HIV positive, treated with NNRTI based HAART79.24.2 (2.1 to 6.2; P<0.001)2.4 (0.3 to 4.6; P=0.026)
HIV positive, treated with non-NNRTI-based HAART75.40.4 (−1.9 to 2.7; P=0.74)−0.7 (−1.3 to 1.6; P=0.54)

Discussion

Overall we found that HIV-positive patients treated with any HAART regimen had a blood pressure that was on average 3.2/2.7 mmHg higher than that of HIV-positive patients who were HAART naïve. However, after adjusting for potential confounders, particularly age, there was only weak evidence to support an increase in diastolic blood pressure (mean increase 1.2 mmHg; 95% CI −0.8 to 3.2; P=0.24) and no evidence for an increase in systolic blood pressure (mean increase 0.3 mmHg; 95% CI −2.3 to 2.8; P=0.84) associated with the use of HAART. However, the unique finding arising from this study was that HIV-positive patients treated with NNRTI-based HAART had a blood pressure that was on average 4.6/4.2 mmHg higher than that of HIV-positive patients who were HAART naïve. After adjusting for potential confounders, a significant increase in diastolic blood pressure (mean increase 2.4 mmHg; 95% CI 0.3 to 4.6; P=0.026) and a nonsignificant increase in systolic blood pressure (mean increase 1.3 mmHg; 95% CI −1.5 to 4.0; P=0.37) were apparent in association with the use of NNRTI-based HAART. We found no evidence to support a difference in blood pressure between HIV-positive patients treated with non-NNRTI-based HAART and HAART-naïve patients before or after adjusting for potential confounders. Furthermore, we found no evidence to support a difference in blood pressure between HIV-positive HAART-naïve patients and HIV-negative patients before or after adjusting for potential confounders of an association between blood pressure and HIV status.

The findings of this study could help to explain the conflicting reports in the literature regarding the association between HAART and blood pressure. Our study suggests that the association between raised blood pressure and HAART may depend on the combination of antiretroviral drugs used in the regimen. Thus the reported prevalence of raised blood pressure could be expected to vary depending on what antiretroviral drugs were being taken by the study population. In our study, 61% of the HIV-positive treated patients were on NNRTI-based HAART regimens and we found an overall positive association between blood pressure and HAART, although this was not statistically significant after adjustment for potential confounding factors.

The inconsistent evidence in the literature may also have arisen from the choice of control group. For example, HAART-treated HIV-positive patients were found to have higher blood pressures when compared with blood donor controls [4] and sick HIV-positive treatment-naïve patients [8–10], lower blood pressures when compared with patients from a medical out-patient clinic [6] and population controls [5], but no difference in blood pressure when compared with university and hospital staff [7]. Ideally the control group should be recruited from the same population from which the cases arose, as seen in the United States Multicenter AIDS Cohort Study (MACS) [13]. In our study all patients were recruited from the same Sexual Health Outpatient Department. Thus both HIV-positive and HIV-negative groups had made similar and potentially confounding lifestyle choices which had put them at risk of HIV infection and on analysis were found to have many similar demographic and lifestyle characteristics. The HIV-positive, HAART-naïve patients in this study were clinically well at the time of the survey and their immune function was comparable to that of the HAART-treated patients [median CD4 count 480 (IQR 350 to 650) cells/μL whole blood, vs. 500 (IQR 350–700) cells/μL whole blood]. This is important as previous studies have shown that patients with low CD4 counts (<200 cells/μL whole blood) have lower blood pressures than those with normal CD4 counts [8–10]. Other sources of error that may have contributed to the conflicting reports in the literature include selection bias and missing data. In this study patients were randomly selected from the out-patient clinic lists and, of those who were identified, 99% agreed to participate in the survey. Missing data were very few in this study, with the exception of viral load in HIV-positive, HAART-naïve patients. Viral load was measured infrequently in these individuals as treatment decisions were based on CD4 cell count and clinical symptoms. Comparison between HAART-naïve patients with and without missing data on viral load suggested that there were no systematic differences in demographic characteristics or blood pressure and the missing data were considered to be nondifferential. Inadequate blood pressure measurement may also have contributed to the inconsistent results reported in previous studies. However, in this study blood pressure was measured by a single trained researcher, using a validated semi-automatic device, following a standard protocol [19]. To improve the precision of the measurement, three readings were taken. In keeping with many previous blood pressure studies the first reading was discarded as it is often spurious and the mean of the second two readings was used in the analysis. Measurement error in studies where standardized equipment and protocols are not used is likely to be nondifferential but the resulting additional variability can lead to the masking of true blood pressure differences and the underestimation of effect estimates.

Assuming that NNRTIs do increase blood pressure, the question arises as to whether the relatively small observed blood pressure differences (mean 4.6/4.2 mmHg before adjustment for potential confounders and 1.4/2.3 mmHg after adjustment) are clinically important in the context of a relatively young, essentially normotensive population. Given that the blood pressure level of young people has a big impact on cardiovascular outcomes in later life [22] and that blood pressure has a multiplicative effect on the risk of heart disease and stroke in combination with other risk factors, including cigarette smoking, dyslipidaemia, insulin resistance and male gender [21], all of which are common in the HIV-positive population [14], such blood pressure differences may well be important. Furthermore, because the success of HAART in preventing AIDS deaths continues to increase the life expectancy of HIV-positive patients and age is one of the biggest independent risk factors for cardiovascular disease [21], small differences in blood pressure, as observed in this study, have the potential to have a big impact on future cardiovascular disease outcomes in the HIV-positive population.

Major limitations of this study include the fact that we cannot rule out the theoretical possibility of reverse causality – that is, patients may have been started preferentially on NNRTI-based HAART regimens because they had higher blood pressure, and although no rationale to support such a possibility is apparent, particularly as blood pressure is not routinely measured in HIV clinics, this cannot be ruled out. A second limitation of these data is that this study was powered to detect a mean difference of 4 mmHg in diastolic blood pressure between HIV-positive patients treated with and without HAART. We failed to find a statistically significant difference between these two groups after adjusting for potential confounders and a larger study is required to confirm our apparently negative findings. In particular, our significant finding that NNRTI-based HAART regimens were associated with raised blood pressure warrants further investigation, ideally in the context of a randomized controlled trial. Indeed, systematic blood pressure measurements are being made as part of the SPARTAC trial [23] which will allow more robust investigation of these findings.

The findings of our analyses are further complicated by the fact that our HIV-positive HAART-naïve group was significantly younger than those who were treated. This reflects the natural history of HIV disease whereby an individual usually remains well for several years after being infected, and in the United Kingdom treatment is only routinely initiated when the CD4 count falls below 200 cells/μL. Age is thus a major confounder of any association between blood pressure and HAART and, although this was taken into account in our analyses, residual confounding must be considered as a potential contributor to the observed blood pressure differences. Finally, we cannot be certain how well the findings from our single-centre hospital-based study would translate to other units in the United Kingdom and across the world, particularly in countries with different antiretroviral drug treatment policies from those in the United Kingdom.

Conclusion

This study suggests that HAART regimens that include an NNRTI may be associated with an increase in blood pressure. In light of the evidence from this and some other studies, we suggest that it would be prudent for HIV clinicians to routinely measure blood pressure, particularly after initiating NNRTI therapy, and where appropriate commence antihypertensive treatment as part of the ongoing management of HIV-positive patients.

Acknowledgements

We thank all of the doctors, nurses and receptionists working in the Winsland and Wharfside clinics at St Mary's Hospital, Paddington, London but most of all the patients for their important contribution. We thank Dr Lucy Pickard and Dr Natalie Atere-Roberts for conducting the pilot study. We thank Dr Tim Clayton for his advice on the statistical analysis. We thank The Hypertension Trust for funding this work.

Role of the Authors: Sarah Wilson conducted the study, carried out the statistical analysis and wrote the paper. Neil Poulter and George Scullard advised on the study design and supervised the project. Sarah Fidler and Jonathan Weber contributed to the ethics applications. All authors contributed to writing the paper.

Ancillary