Comparative effectiveness of monotherapies and combination therapies for patients with hypertension: protocol for a systematic review with network meta-analyses

Background Hypertension has been cited as the most common attributable risk factor for death worldwide, and in Canada more than one of every five adults had this diagnosis in 2007. In addition to different lifestyle modifications, such as diet and exercise, there exist many pharmaco-therapies from different drug classes which can be used to lower blood pressure, thereby reducing the risk of serious clinical outcomes. In moderate and severe cases, more than one agent may be used. The optimal mono- and combination therapies for mild hypertension and moderate/severe hypertension are unclear, and clinical guidelines provide different recommendations for first line therapy. The objective of this review is to explore the relative benefits and safety of different pharmacotherapies for management of non-diabetic patients with hypertension, whether of a mild or moderate to severe nature. Methods/Design Searches involving MEDLINE and the Cochrane Database of Systematic Reviews will be used to identify related systematic reviews and relevant randomized trials. The outcomes of interest include myocardial infarction, stroke, incident diabetes, heart failure, overall and cardiovascular related death, and important side effects (cancers, depression, syncopal episodes/falls and sexual dysfunction). Randomized controlled trials will be sought. Two reviewers will independently screen relevant reviews, titles and abstracts resulting from the literature search, and also potentially relevant full-text articles in duplicate. Data will be abstracted and quality will be appraised by two team members independently. Conflicts at all levels of screening and abstraction will be resolved through team discussion. Random effect pairwise meta-analyses and network meta-analyses will be conducted where deemed appropriate. Analyses will be geared toward studying treatment of mild hypertension and moderate/severe hypertension separately. Discussion Our systematic review results will assess the extent of currently available evidence for single agent and multi-agent pharmacotherapies in patients with mild, moderate and severe hypertension, and will provide a rigorous and updated synthesis of a range of important clinical outcomes for clinicians, decision makers and patients. Trial registration PROSPERO Registration Number: CRD42013004459


Background
Blood pressure (BP) is a measure of the pressure that blood places against the walls of blood vessels during circulation [1], measured in terms of systolic BP (SBP; normal values <130 mmHg) and diastolic BP (DBP; normal values <85 mmHg). When SBP rises above 140 mmHg or DBP rises above 90 mmHg, the patient is considered to have hypertension [1,2]. Hypertension is a chronic condition which places increased stress on the heart and blood vessels, and represents a critical risk factor for clinically significant events including myocardial infarction (MI), heart failure, stroke and death. Hypertension is the most common attributable risk factor for death worldwide, and an independent predictor of stroke mortality and ischemic heart disease mortality [3,4]. In 2007, more than one of every five adult Canadians suffered from high BP [5][6][7].
Reduction of elevated BP is associated with a reduction in the risk of clinically significant events. Evidence supports a direct correlation between the magnitude of BP reduction and the rate of major events, and thus treatment is geared toward lowering BP to below 140 SBP/90 DSP in most patients [8,9]. In addition to lifestyle changes, there exists a number of classes of antihypertensive pharmacotherapies which can be used to manage BP. Classes of agents include diuretics, angiotensin-converting enzyme (ACE) inhibitors, angiotensin receptor blockers, calcium channel blockers, beta blockers and alpha blockers, among others. Guidance from the Canadian Hypertension Education Program (CHEP) [2] suggests initial therapy should consist of monotherapy using any one among thiazide diuretics, beta-blockers, ACE inhibitors, longacting calcium channel blockers or angiotensin receptor blockers (ARB). It is further recommended that use of a second agent should be considered if target BP is not achieved with monotherapy, with a choice made among the agents considered for first line monotherapy. More than 66% of patients with hypertension cannot have their BP adequately lowered using monotherapy, and so the addition of a second agent (and third) may be required [10].
A first question in clinical practice is about choice of therapy, and which treatment should be given to a patient first to minimize risk of undesirable outcomes. Pharmacotherapies are associated with different mechanisms of action, harm profiles and costs. The optimal choice of a first line agent remains unclear, as some works suggest that thiazide diuretics are best [11,12], while others suggest calcium channel blockers should be considered as first line therapy in patients over 55 years of age or of Caribbean or African descent [13]. In Canada, all but alpha blockers are considered reasonable first line therapies, with patients' demographics and comorbidities playing a role in selection [2]. There also has been relatively little study to assess the relative effectiveness of combination therapies, which may be used in moderate to severe cases of hypertension.
To address these research gaps and discrepancies in guidance, the planned study will be a systematic review incorporating network meta-analyses to explore the relative effectiveness and safety of different monotherapies and combination therapies for hypertension. The planned review will include thiazide diuretics, ACE inhibitors, ARBs, calcium channel blockers, beta blockers, alpha blockers, placebo, no treatment and a combination antihypertensive therapy. Separate analyses to address benefits in different degrees of hypertension will be pursued.

Selection of studies
The research question for this systematic review was specified according to the Population-Intervention-Comparator-Outcome-Study Design framework. The following criteria related to the study population, interventions and comparators, clinical outcomes and study designs of interest will be sought.

Population
Non-diabetic patients with hypertension are the target population of interest. No restrictions will be used regarding gender. Studies in adult patients will be retained, and subgroup analyses based on age (for example, age >60 years, age >75 years) will be explored. Studies of both first line therapy (that is, mainly monotherapies) and second line therapy (that is, mainly combination therapies) will be included. Studies related to hypertension during pregnancy, hypertensive emergencies and hypertension treatment in acute stroke will be excluded. Studies conducted in the following sub-populations will be eligible if all patients in the study were required to have hypertension upon entry, and if the benefits of BP lowering on outcomes were examined in the study: patients with micro-and macroalbuminuria; patients with metabolic disease; patients with myocardial ischemia; patients with coronary artery disease/atherosclerosis.

Interventions/Comparators
We will seek studies evaluating the following classes in monotherapy and combination therapy regimens: calcium channel blockers, beta blockers, angiotensin receptor blockers, alpha blockers, thiazide diuretics, potassium sparing diuretics and ACE inhibitors. Fixed doses for each agent will not be sought as an inclusion criterion, as it is anticipated that the majority of trials will have variable doses titrated for each patient.

Outcomes
Studies that report one or more of the following clinical outcomes will be included: MI, stroke, incident diabetes, overall and cardiovascular related death; adherence to treatment; and important adverse events which may impact quality of life (for example, sexual dysfunction, depression, syncopal episodes/falls, cancers). Where outcome definitions vary across studies, appropriate sensitivity analyses will be undertaken.

Study design
Randomized controlled trials of a minimum, one-year duration will be included in this review. Only English language studies will be retained, as past research suggests that limiting included studies to the English language does not greatly affect summary estimates from meta-analyses [14,15].

Literature search strategy
Based on awareness of a large number of existing reviews and network meta-analyses that can be leveraged, an unlimited primary search for RCTs will not be performed. In its place, we will use a staged approach to study identification, beginning with the identification of relevant trials included in Cochrane systematic reviews searched for in the Cochrane Database of Systematic Reviews (publication years 2005 and onward) and existing network meta-analyses of which we are aware [16][17][18][19][20][21]. Medline (OVID interface) will next be searched to identify other relevant reviews, and we will systematically compare studies from relevant articles in reverse chronologic order until yield from this approach becomes low. The Medline search is presented in a supplemental Appendix, and was designed by a senior information specialist in consultation with the review team and peer reviewed using PRESS (Peer Review of Electronic Search Strategies) [22] by another experienced information specialist. To identify additional relevant RCTs published outside the time frames of these reviews, we will compile a list of the unique Medline identification numbers of all relevant articles, and perform a related articles search in Medline. Search results will be limited to the top general medicine journals and related specialty journals as determined by impact factor. This technique has been shown to be effective in identifying relevant studies [23], has been used recently by reviews in several fields including hypertension, and will increase efficiency in study identification in the presence of an already large and established evidence base.

Study selection and data extraction
Review of citations based on title, keywords and abstract (Level 1 screening) and full text articles (Level 2 screening) will be carried out independently by two reviewers. Level 1 citations deemed potentially relevant or unclear will be carried forward to Level 2. Study selection will be conducted using Distiller Systematic Review Software (DSR; Ottawa, ON, Canada). Where consensus is not achieved following discussion, a third member of the research team will be consulted to settle disagreements. The process of literature selection will be reported using a flow diagram as recommended by the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) statement [24].

Data extraction and risk of bias assessment
Primary data collection will be performed independently by two reviewers using a standardized data collection form implemented using SR Distiller Software (Evidence Partners Incorporated, Ottawa, Ontario, Canada); this software will also be used to compare collected data for accuracy and agreement, with disagreements being settled by discussion. The following information will be collected from all eligible studies: authorship list, year and journal of publication, countries of study, funding source, group sample sizes, study inclusion criteria, age distribution, gender distribution, ethnicity distribution, patient comorbidity history, past/present medication use and all relevant outcome information. To assess study risk of bias, all RCTs will be reviewed using the Cochrane Risk of Bias (RoB) tool [25]. Disagreements will be resolved through discussion or by third party adjudication.

Data analysis
Studies of first line therapy (that is, mainly monotherapies unless otherwise rationalized to consider combination therapies) and second line therapy (that is, mainly combination therapies) will be analyzed separately to maintain clinical homogeneity in terms of patient characteristics. Among combination therapy regimens, we will also explore analyses to assess whether there is a differential clinical benefit between combinations involving a thiazide diuretic versus those which do not.

Approach to analysis
We will begin with a narrative overview of studies included in the review which will provide insights regarding the degree of clinical and methodologic homogeneity among included RCTs, thereby helping to explore the assumptions of homogeneity and similarity for network meta-analysis [26]. Study of statistical heterogeneity as measured by values of I 2 within pairwise meta-analyses will also contribute to this assessment. Where assumptions are judged reasonable, network meta-analysis (NMA) will be carried out for each clinical outcome separately; NMA is an approach to evidence synthesis which allows for the combination of direct and indirect evidence to compare three or more treatments in a unified analysis. Indirect comparisons between treatments A and B based on a common comparator C where no trials of A versus B exist (that is, no direct evidence) but trials of A versus C and B versus C exist (that is, indirect evidence) were originally proposed by Bucher et al. [27], and Lumley [28] and Lu and Ades [29] subsequently developed extensions of this methodology. In addition to estimating all pairwise comparisons between treatments in a network, this technique can also be used to estimate probabilities of treatment of superiority to rank the treatments. As part of this will involve analyses that include combination therapies, recent methods proposed by Mills and Thorlund may be explored [30].

Addressing clinical and methodological heterogeneity
Findings from risk of bias assessments of included studies will inform sensitivity analyses, including meta-regression or exclusion of higher risk studies to address the impact of perceived study deficiencies. Meta-regression and/or removal of studies from the treatment network to address clinically important variations between studies with regard to gender distribution, age distribution, history of chosen clinically relevant baseline comorbidities and other relevant factors will be considered. If several trials contain percentages of diabetic patients, sensitivity analyses excluding these studies will be performed.
Clinical practice for patients with hypertension has evolved over time, including the emergence of common use of co-interventions such as statins and aspirin. To explore the potential impact of such changes, we will explore meta-regression analyses using the network metaanalysis approach to adjust for year of initiation of patient enrollment in each study. This may be a helpful proxy in the absence of other information about background treatments.

Assessment of coherence
This review will consist of a mixture of both direct evidence (that is, head to head comparisons between different antihypertensive agents or classes) and indirect evidence (that is, inactively controlled studies of the different agents or classes). As described elsewhere [31,32], there is a need to verify that findings generated from synthesis of direct and indirect data do not differ more than one would expect by chance. We will employ methods described by Dias et al. [33] to assess the validity of this assumption in this work. As statistics alone cannot always be replied upon to identify important clinical or methodologic differences between studies, we will also employ evidence tables of study characteristics to further assess homogeneity.

Reporting of findings from analyses
Full graphical and numeric presentations of findings [34] will be provided to convey the results of our work. This will include network diagrams showing the structure of available evidence for all possible treatment comparisons, summary point estimates and 95% credible intervals for all pairwise comparisons, and estimated probabilities that each therapy is deemed 'best' for each outcome along with associated average rankings. Forest plots of summary estimates as well as rankograms [34] will be used to clearly delineate any important variations from the primary analysis for subgroup effects.

Discussion
The proposed review will add to the literature in several ways. To our knowledge, the planned review includes analyses for certain clinical outcomes which have not been the subject of past network meta-analyses, and others which are important for consideration of harms and patients' quality of life. The review will also pursue updates of comparisons of therapies using a network meta-analysis approach for several clinical outcomes which have not been studied using this technique since 2003, and thus for which considerable new data are likely available. The review will generate new information by addressing combination therapies for hypertension, both (i) in comparison to monotherapies in specific patient populations where either might be used, and (ii) in patients who have moderate to severe hypertension. To our knowledge, neither has been addressed using network meta-analysis. Finally, we will also improve upon limitations of past reviews by including improved assessment and accounting for the effects of clinical heterogeneity between studies.
There are potential challenges to the planned review. Clinical expertise and preliminary review of a sample of relevant trials for this research shows that in some studies, while patients are randomized to one active antihypertensive agent to address the study's research question, additional anti-hypertensive medications may be prescribed to patients at the physicians' discretion. While these additional treatments may have implications for the additional effects seen, outcome data for those remaining strictly on the prescribed treatments may not be available. This may be a limiting factor concerning the 'purity' of some included trial data. We anticipate at least a moderate degree of clinical heterogeneity with regard to the clinical populations enrolled from study to study. If the extent of included literature turns out to be small, the ability to explore heterogeneity could be limited; however, past works suggest considerable evidence will be found. There are many possible outcomes of interest to be studied in relation to hypertensive patients. We have attempted to identify outcomes that are clinically important and can be studied using network meta-analysis to produce important information in a timely fashion. ((Adrenergic or alpha-adrenergic or betaadrenergic) adj3 (block* or alpha-block* or betablock* or antagonist* or alpha-antagonist* or betaantagonist*)). tw. (17756) 139. ((alpha1 or "alpha-1" or alpha2 or "alpha-2" or beta or beta1 or "beta-1" or beta2 or "beta-2" or beta3 or "beta-3") adj2 (block* or antagonist*)). tw.  155. exp Doxazosin/ (1097) analy* or integrative research or integrative review* or integrative overview* or research integration or research overview* or collaborative review*). tw.
(49866) 274. (systematic review* or systematic overview* or evidence-based review* or evidence-based overview* or (evidence adj3 (review* or overview*)) or meta-review* or meta-overview* or "review of reviews" or technology assessment* or HTA or HTAs