Meta-Analysis of the Hemodynamic Properties of Antihypertensive Medications

Background . While all antihypertensive medications lower blood pressure, hemodynamic properties of various classes of antihypertensive medications may diﬀer. Objective . To perform a meta-analysis to compare the hemodynamic properties of diﬀerent classes of antihypertensive medications. Methods . Studies involving the treatment of hypertension using the eﬀect of ACEIs, 𝛽𝛽 -blockers, CCBs and thiazide diuretics on plasma volume (PV), cardiac output (CO) or stroke volume (SV) were searched using online databases prior to May 2011. Studies had to be written in the English language, studying human subjects with a single pharmacological agent (monotherapy), and with a minimum duration of 4 weeks. Results . Seventy-�ve (75) studies that enrolled a total of 1522 subjects were included. All four antihypertensive classes lowered blood pressure. 𝛽𝛽 -blockers decreased heart rate; the other classes had no eﬀect upon heart rate. ACEIs increased PV; the other classes had no eﬀect upon PV. 𝛽𝛽 -blockers and thiazide diuretics decreased CO while ACEIs and CCBs had no eﬀect upon CO. 𝛽𝛽 -blockers and CCBs increased SV, thiazide diuretics decreased SV, and ACEIs did not change SV. Conclusion . In the treatment of uncomplicated hypertension, the various classes of antihypertensive medications diﬀer from each other in terms of their non-blood pressure lowering hemodynamic properties.


Introduction
While all antihypertensive medications lower blood pressure, the non-blood-pressure-lowering hemodynamic properties of the various classes of antihypertensive medications may differ. Numerous studies documenting the effects of different antihypertensive medications upon heart rate (HR), plasma volume (PV), cardiac output (CO), and stroke volume (SV) have been published, but the sample sizes of most of these studies have been small, typically less than 20 subjects each. Differences in hemodynamic properties that distinguish one class of antihypertensive medication from another might not be evident in studies with such small samples. Furthermore, if there are differences in hemodynamic properties among the different classes of antihypertensive drugs, such differences might help clinicians customize the selection of drugs for individual patients.
We performed this meta-analysis in order to clarify and compare the hemodynamic properties of four different classes of antihypertensive medications: angiotensin converting enzyme inhibitors (ACEIs), -adrenergic receptor antagonists ( -blockers), calcium channel blockers (CCBs), and thiazide diuretics.

Search Strategy and Selection
Criteria. Investigators (I. H. Iikhar, R. P. Blank�eld, N. Hassan) searched Pubmed for trials that evaluated the hemodynamic properties of the following classes of antihypertensive drugs: ACEIs, -blockers, CCBs and thiazide diuretics. We chose these medications because there is more hemodynamic data regarding these agents than other classes of antihypertensive drugs. Boolean operators (AND, OR) were used to combine the Medical Subject Heading (MeSH) text terms for the antihypertensive classes and the hemodynamic variables. In order to ensure that all relevant articles were identi�ed, additional searches for individual antihypertensive medications in each class were performed, and each individual medication was paired with one of the several hemodynamic variables in sequential fashion. Speci�cally, each drug was paired with the term "hypertension" along with either "plasma volume, " "cardiac output, " "stroke volume, " or "hemodynamics, " and a search was made for each of these pairings. In this manner, searches were made for all common drug names in each class (Table 1). Studies published prior to May 2011 were eligible for inclusion. Reference lists of all identi�ed articles were searched for additional articles that did not appear in the database searches. No attempt to identify unpublished studies was undertaken, and no attempt was made to contact authors.
Once potentially relevant articles were identi�ed by the initial search, two investigators (I. H. Iikhar, R. P. Blank�eld) independently evaluated studies for inclusion. Inclusion criteria were: (1) the published article had to be written in the English language, (2) the study had to include only human subjects with uncomplicated hypertension, (3) the study drug had to be administered only through the oral route, (4) a single pharmacological agent (monotherapy) was used to treat the hypertension, and (5) the duration of the study had to be a minimum period of 4 weeks. Studies were included if they contained data regarding PV, CO, and/or SV. Studies that provided data on cardiac index or stroke index were included if there was enough information to calculate CO or SV. Studies in which subjects used different antihypertensive medications at different times, but not simultaneously, were included. All case reports, case series, and review articles were excluded. Studies that included subjects with diabetes mellitus, heart failure, or coronary artery disease were excluded. If multiple publications of the same trial were identi�ed, only the most recent publication was chosen for inclusion. Studies that did not include standard deviation data were excluded. Disagreements were resolved through discussion among reviewers.

Data Extraction.
For studies that met the inclusion criteria, extracted data included the �rst author�s name, year of publication, number of subjects, pre-and posttreatment mean arterial pressure with standard deviations, pre-and posttreatment mean HR with standard deviations, pre-and posttreatment mean PV with standard deviations, pre-and posttreatment mean CO with standard deviations, and preand posttreatment mean SV with standard deviations. For studies that reported mean systolic and mean diastolic blood pressures but not mean arterial pressures, mean arterial pressures were calculated. Data was extracted from articles that reported mean cardiac index (CI) or mean stroke index  (SI) only if mean CO and/or mean SV could be calculated based upon mean body surface area data or mean height and weight data. is method of calculating CO or SV provides an approximation to the individual data [1]. We chose to use CO and SV rather than CI and SI because there was considerably more data available for CO and SV than there was for CI and SI. All included studies were peer reviewed and were assumed to represent valid information regarding drug effect.

Quantitative Data
Synthesis. e absolute effectiveness of each drug regimen was quanti�ed by estimating the mean difference of outcomes before and aer intervention. Effectiveness was pooled across studies using the DerSimonian and Laird random effects meta-analyses models.
Heterogeneity was assessed with 2 index and the tausquared test. To assess the risk of publication bias, funnel plots of standard error and difference in means were constructed. A funnel plot of the SV data for thiazide diuretics ( Figure 1) is representative of the funnel plot data. NCSS soware version 2007 was used to analyze the data.

Trial Flow.
A total of 227 potentially relevant articles were identi�ed, of which 152 were excluded, primarily due to insufficient outcome data. Figure 2 summarizes the results of the selection and exclusion process. Seventy-�ve (75) articles that enrolled a total number of 1522 subjects treated with ACEIs, -blockers, CCBs and thiazide diuretics were included .
For studies that measured PV, 125 Iodine-labeled albumin was used. For studies that measured CO, the methodologies varied. Twenty-six studies used a dye dilution technique, 8 used Doppler ultrasound, 7 used the Fick technique, 5 used impedance cardiography, 1 used the indirect Fick technique, 1 used thermodilution, and 1 used technetium labeled albumin. For studies that measured SV, the methodologies also varied. Ten used echocardiography, and 9 used multiple gated acquisition scans. Many of the studies that measured CO calculated SV. Likewise, many of the studies that measured SV calculated CO. Table 2 summarizes the total number of studies and the total number of subjects for each hemodynamic variable according to class of antihypertensive medication. A few studies included data on more than one medication. Other studies reported more than one trial of a single medication, or else reported subgroups of a single medication trial that were best analyzed as separate trials.

Study
Characteristics. e studies included in the metaanalysis enrolled men and women with uncomplicated hypertension. e majority of the subjects were men. Most of the studies did not mention the ethnicity or race of the subjects. When race or ethnicity was identi�ed, the majority of the subjects were Caucasian. Subjects ranged from 18-87 years old, but most were between 30-60 years of age. Most of the studies did not mention exclusion criteria, but some of the studies speci�ed exclusion criteria that included coronary artery disease, heart failure, chronic obstructive lung disease, renal disease, and diabetes mellitus.

Heterogeneity
Analysis. e tests for heterogeneity were signi�cant in the majority of analyses, indicating that a nonrandom model would be inadequate for the present study. erefore, random effect estimates were used throughout the meta-analysis to protect against possible undetected heterogeneities.
2 index was used to account for variability in effect size estimates across the studies. e Tau-squared test was used to account for variance between the studies. e 2 index for the data on SV for the different antihypertensive classes is 70% for the SV data on CCBs, 94% for the SV data on ACEIs, 33% for the data on -blockers, and 50% for the SV data on thiazides. e SV data suggests low-to-moderate heterogeneity for the CCBs and ACEIs, and medium heterogeneity for theblockers and thiazides [77]. Table 3 summarizes the changes in hemodynamic variables from baseline according to class of antihypertensive drug. A forest plot of the thiazide diuretic SV data ( Figure 3) is representative of the data analysis used to construct Table 3

Discussion
While all antihypertensive drugs lower BP, the various classes of antihypertensive medications differ in their effects upon other hemodynamic variables. -blockers reduce HR but none of the other classes of antihypertensive drugs do so.blockers and thiazide diuretics reduce cardiac output while ACEIs and CCBs have no effect upon CO. CCBs andblockers increase SV, thiazide diuretics decrease SV, and ACEIs have no effect upon SV. Our �ndings regarding ACEIs and PV, and our �ndings regarding thiazide diuretics and SV may be noteworthy.  ACEIs  18  277  16  231  6  78  14  212  14  198   -blockers  29  591  25  591  8  150  25  581  16  411   CCBs  16  312  13  256  7  127  11  198  11  194   iazides  11  124  8  103  7  102  7  87  7 95 * = total number of studies. * * = total number of subjects. ACEIs increase PV but none of the other classes of antihypertensive drugs change PV. e effect of ACEIs upon PV is surprising because ACEIs are not recognized as causing �uid retention [78]. To the extent that an increase in plasma volume results in an increase in intra-arterial volume, one would expect an increase in plasma volume to be accompanied by an increase in CO, assuming that there is a corresponding increase in intra-arterial volume [79][80][81]. Since our study found that ACEIs are not associated with an increase in CO despite an increase in PV, perhaps the vasodilatation that accompanies angiotensin converting enzyme inhibition involves the venous circulation rather than the arterial circulation [78].
If all antihypertensive drug classes reduce the risk of adverse cardiovascular events comparably, as a recent metaanalysis determined [82], then distinguishing hemodynamic differences between the different categories of antihypertensive medications has little clinical signi�cance. On the other hand, since a network meta-analysis found that lowdose diuretics are the most effective �rst-line treatment for preventing the occurrence of cardiovascular disease morbidity and mortality [83], then factors apart from BP reduction may be clinically relevant. Because all categories of antihypertensive medications lower BP equivalently [84], differences in other hemodynamic properties may be sig-ni�cant. In particular, our �nding that diuretics lower SV, whereas other antihypertensive medications do not, may be noteworthy because reductions in SV would be expected to reduce the likelihood of turbulent blood �ow, thereby slowing the progression of atherosclerotic cardiovascular disease [85].
ere are several limitations of this meta-analysis. Most of the studies included in the meta-analysis enrolled less than 20 subjects. �hen subjects are strati�ed by hemodynamic variable and drug class, the number of subjects in many of the categories is modest: many of the categories include less than 200 subjects, and several categories include less than 100 subjects. Another limitation is that most of the studies included in the meta-analysis compared an antihypertensive drug to placebo or no treatment, but few studies compared one antihypertensive medication with another. Accordingly, the data allow one to make qualitative comparisons of the hemodynamic properties of the different classes of antihypertensive medications. As a result, one can say that one class of medication decreases CO while another class does not. However, the data do not allow quantitative comparisons of the magnitude of the effect between the different classes of antihypertensive medications. Consequently, the results do not allow one to say that one class of medication lowers BP more than another, nor do the results allow one to say that one class of medication raises SV more than another. It is a limitation of this study that the variability in the CO and SV measures due to a wide variation in methodologies used could not be taken into account in the model. e lack of uniform drug dosage within all four categories of antihypertensive medications limits identifying precisely the hemodynamic effect of the different medications. Variability in dosage may be especially relevant to thiazide diuretics because the dose of these drugs in many of the trials was higher than the doses that are typically used in contemporary medical practice [86]. Accordingly, the magnitude of the hemodynamic effects of thiazide diuretics as they are currently prescribed are unlikely to be the same as the magnitude of the hemodynamic effects of thiazide diuretics at the doses used in the studies included in this metaanalysis. Even though the magnitude of any hemodynamic effect would be expected to be dose dependent, the qualitative direction of that effect would not be expected to be dose dependent.
An additional limitation is that we excluded studies that enrolled subjects with diabetes, heart failure, and coronary artery disease. erefore, our results may not be generalizable to hypertensive patients with these comorbidities. It is possible that we missed some studies by limiting the study to English language publications. However, there were only a dozen non-English language studies that were identi�ed by the initial search strategy, only two of which contained data that appeared as if it might be relevant to the metaanalysis, and we could not be certain that we could accurately extract the relevant data. It is also a limitation that our results re�ect a publication bias since we did not attempt to locate unpublished data. Finally, it is a limitation that we could not incorporate the many studies that reported cardiac index and stroke volume index data, but which did not report cardiac output or stroke volume data.
In summary, in the treatment of uncomplicated hypertension, the various classes of antihypertensive medications differ from each other in terms of their non-blood-pressurelowering hemodynamic properties. ��n��c� �f �n�eres�s e authors report no potential con�ict of interests with any companies/organizations whose products or services may be discussed in this paper.