Cochrane Database Syst Rev. 2018; 2018(4): CD001841. Monitoring Editor: James M Wright, University of British Columbia, Department of Anesthesiology, Pharmacology and Therapeutics, 2176 Health Sciences Mall, VancouverBCCanada, V6T 1Z3 Manipal University, Department of Pharmacology, ManipalIndia AbstractBackgroundThis is the first update of a review published in 2009. Sustained moderate to severe elevations in resting blood pressure leads to a critically important clinical question: What class of drug to use first‐line? This review attempted to answer that question. ObjectivesTo quantify the mortality and morbidity effects from different first‐line antihypertensive drug classes: thiazides (low‐dose and high‐dose), beta‐blockers, calcium channel blockers, ACE inhibitors, angiotensin II receptor blockers (ARB), and alpha‐blockers, compared to placebo or no treatment. Secondary objectives: when different antihypertensive drug classes are used as the first‐line drug, to quantify the blood pressure lowering effect and the rate of withdrawal due to adverse drug effects, compared to placebo or no treatment. Search methodsThe Cochrane Hypertension Information Specialist searched the following databases for randomized controlled trials up to November 2017: the Cochrane Hypertension Specialised Register, the Cochrane Central Register of Controlled Trials (CENTRAL), MEDLINE (from 1946), Embase (from 1974), the World Health Organization International Clinical Trials Registry Platform, and ClinicalTrials.gov. We contacted authors of relevant papers regarding further published and unpublished work. Selection criteriaRandomized trials (RCT) of at least one year duration, comparing one of six major drug classes with a placebo or no treatment, in adult patients with blood pressure over 140/90 mmHg at baseline. The majority (over 70%) of the patients in the treatment group were taking the drug class of interest after one year. We included trials with both hypertensive and normotensive patients in this review if the majority (over 70%) of patients had elevated blood pressure, or the trial separately reported outcome data on patients with elevated blood pressure. Data collection and analysisThe outcomes assessed were mortality, stroke, coronary heart disease (CHD), total cardiovascular events (CVS), decrease in systolic and diastolic blood pressure, and withdrawals due to adverse drug effects. We used a fixed‐effect model to to combine dichotomous outcomes across trials and calculate risk ratio (RR) with 95% confidence interval (CI). We presented blood pressure data as mean difference (MD) with 99% CI. Main resultsThe 2017 updated search failed to identify any new trials. The original review identified 24 trials with 28 active treatment arms, including 58,040 patients. We found no RCTs for ARBs or alpha‐blockers. These results are mostly applicable to adult patients with moderate to severe primary hypertension. The mean age of participants was 56 years, and mean duration of follow‐up was three to five years. High‐quality evidence showed that first‐line low‐dose thiazides reduced mortality (11.0% with control versus 9.8% with treatment; RR 0.89, 95% CI 0.82 to 0.97); total CVS (12.9% with control versus 9.0% with treatment; RR 0.70, 95% CI 0.64 to 0.76), stroke (6.2% with control versus 4.2% with treatment; RR 0.68, 95% CI 0.60 to 0.77), and coronary heart disease (3.9% with control versus 2.8% with treatment; RR 0.72, 95% CI 0.61 to 0.84). Low‐ to moderate‐quality evidence showed that first‐line high‐dose thiazides reduced stroke (1.9% with control versus 0.9% with treatment; RR 0.47, 95% CI 0.37 to 0.61) and total CVS (5.1% with control versus 3.7% with treatment; RR 0.72, 95% CI 0.63 to 0.82), but did not reduce mortality (3.1% with control versus 2.8% with treatment; RR 0.90, 95% CI 0.76 to 1.05), or coronary heart disease (2.7% with control versus 2.7% with treatment; RR 1.01, 95% CI 0.85 to 1.20). Low‐ to moderate‐quality evidence showed that first‐line beta‐blockers did not reduce mortality (6.2% with control versus 6.0% with treatment; RR 0.96, 95% CI 0.86 to 1.07) or coronary heart disease (4.4% with control versus 3.9% with treatment; RR 0.90, 95% CI 0.78 to 1.03), but reduced stroke (3.4% with control versus 2.8% with treatment; RR 0.83, 95% CI 0.72 to 0.97) and total CVS (7.6% with control versus 6.8% with treatment; RR 0.89, 95% CI 0.81 to 0.98). Low‐ to moderate‐quality evidence showed that first‐line ACE inhibitors reduced mortality (13.6% with control versus 11.3% with treatment; RR 0.83, 95% CI 0.72 to 0.95), stroke (6.0% with control versus 3.9% with treatment; RR 0.65, 95% CI 0.52 to 0.82), coronary heart disease (13.5% with control versus 11.0% with treatment; RR 0.81, 95% CI 0.70 to 0.94), and total CVS (20.1% with control versus 15.3% with treatment; RR 0.76, 95% CI 0.67 to 0.85). Low‐quality evidence showed that first‐line calcium channel blockers reduced stroke (3.4% with control versus 1.9% with treatment; RR 0.58, 95% CI 0.41 to 0.84) and total CVS (8.0% with control versus 5.7% with treatment; RR 0.71, 95% CI 0.57 to 0.87), but not coronary heart disease (3.1% with control versus 2.4% with treatment; RR 0.77, 95% CI 0.55 to 1.09), or mortality (6.0% with control versus 5.1% with treatment; RR 0.86, 95% CI 0.68 to 1.09). There was low‐quality evidence that withdrawals due to adverse effects were increased with first‐line low‐dose thiazides (5.0% with control versus 11.3% with treatment; RR 2.38, 95% CI 2.06 to 2.75), high‐dose thiazides (2.2% with control versus 9.8% with treatment; RR 4.48, 95% CI 3.83 to 5.24), and beta‐blockers (3.1% with control versus 14.4% with treatment; RR 4.59, 95% CI 4.11 to 5.13). No data for these outcomes were available for first‐line ACE inhibitors or calcium channel blockers. The blood pressure data were not used to assess the effect of the different classes of drugs as the data were heterogeneous, and the number of drugs used in the trials differed. Authors' conclusionsFirst‐line low‐dose thiazides reduced all morbidity and mortality outcomes in adult patients with moderate to severe primary hypertension. First‐line ACE inhibitors and calcium channel blockers may be similarly effective, but the evidence was of lower quality. First‐line high‐dose thiazides and first‐line beta‐blockers were inferior to first‐line low‐dose thiazides. Plain language summaryThiazides best first choice for hypertension Review Question(s) In this first update of a review published in 2009, we wanted to determine which drug class was the best first‐line choice in treating adult patients with raised blood pressure. We searched the available medical literature to find all the trials that compared the drugs to placebo or no treatment to assess this question. The data included in this review are up to date as of November 2017. Background High blood pressure or hypertension can increase the risk of heart attacks and stroke. One of the most important decisions in treating people with elevated blood pressure is what drug class to use first. This decision has important consequences in terms of health outcomes and cost. Study characteristics We found no new trials in this updated search. In the original review, we found 24 studies that randomly assigned 58,040 adult people (mean age 62 years) with high blood pressure, to four different drug classes or placebo. Duration of these studies ranged from three to five years. Drug classes studied included thiazide diuretics, beta‐blockers, ACE inhibitors, and calcium channel blockers. Key Results We concluded that most of the evidence demonstrated that first‐line low‐dose thiazides reduced mortality, stroke, and heart attack. No other drug class improved health outcomes better than low‐dose thiazides. Beta‐blockers and high‐dose thiazides were inferior. Conclusions High‐quality evidence supported that low‐dose thiazides should be used first for most patients with elevated blood pressure. Fortunately, thiazides are also very inexpensive. Quality of evidence The evidence for first‐line low dose thiazides was high quality. For the other classes, we judged the evidence to be moderate or low quality. Summary of findingsBackgroundElevated blood pressure (hypertension) is a chronic condition in which the blood pressure in the arteries is persistently elevated. It has been divided into three categories, based on resting blood pressures, measured in a standard way: mild hypertension (140 to 159/90‐99 mmHg), moderate hypertension (160 to 179/100 to 109 mmHg), and severe hypertension (180/110 mmHg or higher). Most people with high blood pressure have no signs or symptoms, even if blood pressure readings are very high. For most adults with primary or essential hypertension, there is no identifiable cause for the high blood pressure. Some people have high blood pressure, called secondary hypertension, caused by underlying conditions such as adrenal gland tumours, kidney problems, thyroid problems, excessive alcohol intake, or use of certain medications, such as birth control pills. Isolated systolic hypertension is a condition in which the diastolic pressure is normal (less than 90 mmHg), but systolic pressure is high (160 mmHg or greater). This is a common type of high blood pressure among older people. Blood pressure tends to increase with age. High blood pressure is more common in men in early middle age, more common in women after age 65, and more common in Blacks compared to Caucasians. The risk of high blood pressure is increased when there is a family history of high blood pressure, in the presence of obesity, or when physically inactive. High blood pressure is associated with smoking, too much salt in the diet, drinking excessive amounts of alcohol, high levels of stress, and chronic conditions such as diabetes, kidney disease, and sleep apnea. Uncontrolled persistent resting high blood pressure increases the risk of stroke, heart attack, heart failure, kidney damage, and vision loss. Description of the conditionWhen drug treatment is indicated in the management of patients with elevated blood pressure, an important decision is which drug to choose first. The decision should be informed by the best available evidence on reduction of the outcomes that are important to the patient, i.e. the ability of the drug to reduce the adverse health outcomes associated with elevated blood pressure (disabling stroke, myocardial infarction, heart failure, and mortality). Description of the interventionHigh blood pressure should initially be managed with changing life style — eating a healthy diet with less salt, exercising regularly, quitting smoking, and maintaining a healthy weight. When these life‐style changes are not enough, treatment with antihypertensive drugs is recommended. Several different classes of medications are available to reduce blood pressure. The six main drug classes, included in this review, are thiazide diuretics, beta‐blockers, angiotensin converting enzyme (ACE) inhibitors, angiotensin receptor blockers, calcium channel blockers, and alpha blockers. How the intervention might workDifferent classes of antihypertensive drugs have different mechanisms of action. Thiazide and thiazide‐like diuretics: The mechanism of action by which thiazide diuretics lower blood pressure in the long term is not fully understood (Zhu 2005). After chronic use, thiazides lower peripheral resistance. The mechanism of these effects is uncertain, as it may involve effects on the whole body, renal autoregulation, or direct vasodilator actions (Hughes 2004). Thiazides act on the kidney to inhibit reabsorption of sodium (Na+) and chloride (Cl‐) ions from the distal convoluted tubules in the kidneys, by blocking the thiazide‐sensitive sodium‐chloride symporter (Duarte 2010). They also increase calcium reabsorption at the distal tubule, and increase the reabsorption of calcium ions (Ca2+), by a mechanism involving the reabsorption of sodium and calcium in the proximal tubule in response to sodium depletion. Beta‐blockers: Beta‐blockers are competitive antagonists that block the receptor sites for epinephrine (adrenaline) and norepinephrine on adrenergic beta‐receptors. Some block activation of all types of beta‐adrenergic receptors (β1, β2, and β3), and others are selective for one of the three types of beta receptors (Frishman 2005). Angiotensin converting enzyme (ACE) inhibitors: ACE inhibitors block the conversion of angiotensin I (AI) to angiotensin II (AII), and thus decrease the actions of angiotensin II. The end result is to lower arteriolar resistance and increase venous capacity; decrease cardiac output, cardiac index, stroke work, and volume; lower resistance in blood vessels in the kidneys; and increase excretion of sodium in the urine. Renin and AI increases in concentration in the blood as a result of negative feedback of the conversion of AI to AII. AII and aldosterone levels decrease. Bradykinin increases, because ACE is also responsible for inactivation of bradykinin. Angiotensin receptor blockers (ARBs): ARBs block the activation of angiotensin II AT1 receptors. Blockage of AT1 receptors directly causes vasodilation, reduces secretion of vasopressin, and reduces the production and secretion of aldosterone. Calcium channel blockers (CCBs): CCBs block calcium channel and inhibit calcium ion influx into vascular smooth muscle and myocardial cells. They reduce blood pressure through various mechanisms: by vasodilation, by reducing the force of contraction of the heart, by slowing the heart rate, and by directly reducing aldosterone production. Alpha blockers: α1 adrenergic receptor blockers inhibit the binding of norepinephrine (noradrenaline) to the α1 receptors on vascular smooth muscle cells. The primary effect of this inhibition is vasodilation, which decreases peripheral vascular resistance, leading to decreased blood pressure. Why it is important to do this reviewThere have been a number of reviews of the effectiveness of antihypertensive therapy, but most have emphasized effectiveness of all drug classes (Collins 1990; Gueyffier 1996), or effectiveness of all drug classes in special populations, such the elderly (Insua 1994; MacMahon 1993; Mulrow 1994; Mulrow 1998; Thijs 1992). When all drug therapies are included in one review, there is an underlying assumption that the benefits of lowering blood pressure are independent of the mechanism by which it is achieved. This assumption has not been proven, and it is likely that different classes of drugs will have different blood pressure lowering effects, and will have effects that are independent of the blood pressure lowering effect. A drug that lowers blood pressure could have pharmacological and physiological actions independent of blood pressure lowering, and these other actions (both known and unknown) could enhance or negate the effects on health outcomes associated with the decrease in blood pressure. This possibility is supported by a recent analysis that suggested that blood pressure lowering only explains about 50% of the treatment effect in antihypertensive trials (Boissel 2005). This review update aims to 1) document the best available evidence of effectiveness for different classes of drugs and doses used as first‐line therapy, compared to placebo or no treatment, and 2) present the outcome data in a way that best assists clinicians in the choice of a first‐line drug. ObjectivesPrimary objective
Secondary objectives
MethodsCriteria for considering studies for this reviewTypes of studiesRandomized controlled trials (RCT) of at least one year duration. The comparative group was a placebo, or an untreated control. We required the following data from the trial: baseline patient characteristics, clearly defined morbidity and mortality endpoints, and outcome data presented using the intention‐to‐treat principle. We excluded trials using other than randomized allocation methods, such as alternate allocation, week of presentation, or retrospective controls. We also excluded trials that compared two specific antihypertensive first‐line therapies without a placebo or untreated control. Types of participantsBlood pressure was measured using proper technique at least two times, with the patient resting for at least five minutes. All patients must have had a baseline resting blood pressure of at least 140 mmHg systolic or a diastolic blood pressure of at least 90 mmHg. Trials that included both hypertensive and normotensive patients were acceptable if the majority (> 70%) of patients had elevated blood pressure, or the trial separately reported outcome data on patients with elevated blood pressure. Trials were not limited by any other factor or baseline risk. It was assumed that age and comorbidities did not affect the risk ratio of outcomes associated with drug treatment. Types of interventionsTreatment was to be clearly defined as a specific class of first‐line antihypertensive therapy in
one of the following classes: thiazide diuretics, beta blockers, calcium channel blockers, angiotensin converting enzyme (ACE) inhibitors, angiotensin II receptor antagonists, or alpha adrenergic blockers. The majority (> 70%) of the patients in the treatment group were to be taking the drug class of interest after one year. We allowed initial combined therapies with drug classes not in the defined categories. We also allowed supplemental drugs from other drug classes of interest as stepped
therapy, but only as long as they were not taken by over 50% of the patients. We assumed that these supplemental drugs did not systematically interact to affect the occurrence of the outcomes studied. We classified groups according to the starting dose in the trial: High‐dose thiazide group: starting dose
Low‐dose thiazide group: starting dose
We calculated the average dose in the high‐dose and low‐dose group as a weighted average from the trials in which the average dose was reported, or could be estimated. Types of outcome measuresPrimary outcomes
When the primary trials did not report outcomes that fit the above definitions, we based our decisions on maximizing the inclusion of the data and maintaining concordance, with how the data were classified in previous reviews. We assumed that the effect of antihypertensive treatment on outcomes was independent of whether elevated blood pressure was defined in terms of systolic or diastolic pressure. Secondary outcomes
Search methods for identification of studiesElectronic searchesThe Cochrane Hypertension Information Specialist (DS) conducted systematic searches in the following databases for RCTs without language, publication year, or publication status restrictions:
The Cochrane Hypertension Information Specialist (DS) modelled subject strategies for databases on the search strategy designed for MEDLINE. Where appropriate, they were combined with subject strategy adaptations of the highly sensitive search strategy designed by Cochrane for identifying randomized controlled (as described in the Cochrane Handbook for Systematic Reviews of Interventions (Box 6.4.b; Higgins 2011a). The MEDLINE search strategy was translated into the other databases using the appropriate controlled vocabulary, as applicable (Appendix 1). Data collection and analysisSelection of studiesWe rejected articles on the initial screening if we could determine from the title or the abstract that the article was not a report of a randomized controlled trial, or that there was no possibility that the trial would fit the requirements of this review. Of the articles selected for further review, two reviewers (JMW and VM) independently assessed whether they would be included or excluded. Data extraction and managementThe data abstraction form included details of study design, randomization, blinding, duration of treatment, baseline characteristics, number of patients lost to follow‐up, outcomes, intervention, statistical analysis, and reporting. Two reviewers (JMW and VM) independently extracted the data, cross‐checked, and compared, whenever possible, to data from previously published meta‐analyses. We detailed trial characteristics in the 'Characteristics of included studies' table. We detailed trials that were excluded in the 'Characteristics of excluded studies' table. Assessment of risk of bias in included studiesTwo review authors (VM and RG) independently assessed risk of bias of each included trial; a third review author (JMW) sorted any disagreements. We assessed risk of bias according to Chapter 8 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011b). We assessed seven domains: randomization and allocation concealment to assess selection bias; blinding of the participants and physician to assess performance bias; blinding of the outcome assessor to assess detection bias; incomplete outcome reporting to assess attrition bias; selective reporting of outcomes to assess selective reporting bias; and we added an additional category ‐ other bias, to assess whether the study was funded by the manufacturer and conflict of interest was present, which we assessed as high risk of bias, since it has been shown to overestimate treatment effect. 'Summary of findings' tableWe used GRADEpro GDT software to present the 'Summary of findings' table (GRADEpro GDT). We decided to include all clinically relevant primary and secondary outcomes: total mortality, total stroke, total coronary heart disease, total cardiovascular events, and withdrawal due to adverse events. We did not include the magnitude of systolic and diastolic blood pressure reduction. We considered five factors in grading the overall quality of evidence: limitations in study design and implementation, indirectness of evidence, unexplained heterogeneity or inconsistency of results, imprecision in results, and high probability of publication bias. This approach specifies four levels of quality: high‐, moderate‐, low‐, and very low‐quality evidence. The highest quality rating is for randomized trial evidence. Quality rating is downgraded by one level for each factor, up to a maximum of three levels for all factors. If there are severe problems for any one factor (when assessing limitations in study design and implementation, in concealment of allocation, loss of blinding, or attrition over 50% of participants during follow‐up), randomized trial evidence may fall by two levels due to that factor alone. Measures of treatment effectWe used Review Manager 5 for data synthesis and analyses (RevMan 2014). We based quantitative analyses of outcomes on intention‐to‐treat results. We used risk ratios (RR) with 95% confidence intervals (CI) to combine outcomes across trials. If there was a significant difference in any outcome measure, we presented an absolute risk reduction (ARR), and number needed to treat for an additional beneficial (NNTB) or harmful (NNTH) outcome in the 'Summary of findings' table. This estimate, with 95% confidence intervals (CI), is considered the best estimate of the average benefit and the range of that benefit in populations with different baseline risks. For continuous outcomes (systolic and diastolic blood pressure), we calculated the mean difference (MD) with 99% CI to combine outcomes across studies. Systolic and diastolic blood pressure readings were taken at one year, or the earliest time after one year, to include the data from the maximum number of randomized patients. Mean blood pressure values only reflected data for patients in whom blood pressure was measured. We used standard deviation of the change (SD) at one year if available. Unit of analysis issuesFor all outcomes measures reported, we used data from each trial at the end of the follow‐up period mentioned in each trial, which varied from 1.1 to 10 years. Dealing with missing dataWhen participants were lost to follow‐up, we used data as reported for participants who were followed until the end of the study, in the analyses. We outlined how data were accounted for and included in each study under assessment of attrition bias in the 'Risk of bias in included studies' table. For example, in the MRC‐TMH 1985 study, for events such as non‐fatal stroke or myocardial infarction, which terminated participation in the study, the investigators did not follow‐up these participants to the end of the study. In such instances, investigators included data available up to the time point during which participants were followed in the analyses. If the SD value for reduction in systolic and diastolic blood pressure was not reported at one year, we imputed the SD of the change at other time points during treatment. If the SD of the change was not available at all, we imputed the SD of the endpoint systolic or diastolic blood pressure. In cases where these values were also missing, we imputed the mean weighted SD of the change from other trials. This imputation is a limitation, and to reduce the impact of this limitation, we used a 99% level of significance for the blood pressure measurements, instead of the standard 95%. Assessment of heterogeneityWe tested heterogeneity of treatment effect between the trials using a standard Chi² statistic for heterogeneity. We used the fixed‐effect model to obtain summary statistics of pooled trials, unless there was significant between‐study heterogeneity, in which case we used the random‐effects model to test statistical significance. Assessment of reporting biasesFor each study, we evaluated whether selective reporting of outcomes was suspected. In the case of suspected reporting bias, we contacted study authors for clarification. We had planned to use a funnel plot to assess the possibility of publication bias for outcomes that were reported in 10 or more studies. A test for funnel plot asymmetry (small study effects) formally examines whether the association between estimated intervention effects and a measure of study size is greater than might be expected to occur by chance. Data synthesisWe used Review Manager 5 to perform data synthesis and analyses (RevMan 2014). We presented dichotomous outcomes as RR with 95% CI using a fixed‐effect model, and continuous outcomes (systolic and diastolic blood pressure) as MD with 99% CI. Subgroup analysis and investigation of heterogeneityWe planned the following subgroup analyses: different race: Black versus Caucasian versus Asian; and baseline severity of hypertension: mild, moderate, severe. Unfortunately, it was not possible to do any subgroup analyses, because there were a lack of data specific to race, and most trial participants had moderate to severe hypertension, but data were not available separately. When heterogeneity was significant (I² greater than 50%), we attempted to identify trials that would contribute to heterogeneity, and explore their population characteristics, baseline blood pressure, blinded or open‐label study design, use of antihypertensive drugs as fixed dose or stepped‐up therapy, or response to placebo that would possibly explain the reason for heterogeneity. As the decrease in systolic and diastolic blood pressure showed significant heterogeneity, we present results as MD with 99% CI using both a fixed‐effect as well as random effect model. Sensitivity analysisTo test for robustness of results, we conducted several sensitivity analyses. This was done by deselecting trials in the following categories: trials that were not placebo‐controlled and blinded, trials restricted to patients with isolated systolic hypertension, trials that enrolled more than 80% of patients with previous stroke, and myocardial infarction of peripheral vascular disease (secondary prevention), trials using combined starting drugs, and trials using supplemental drugs from other defined classes. ResultsDescription of studiesResults of the searchThe initial search up to 2009 resulted in 6232 citations, 5985 of which we excluded on reading titles and abstracts. We retrieved 247 citations for more detailed evaluation, 13 of which were review articles. We further evaluated 234 reports; we included 127 reports of 24 trials, and excluded 107 reports. The updated search in November 2017 resulted in 11,855 citations. We screened the titles and abstracts of these citations, and found 11,500 to be irrelevant. We requested the full text of 355 citations, but none of them met the minimum inclusion criteria. We found 6 additional reports of previously included studies. For this update of a total of 87 potential trials identified, we included 133 reports of 24 trials, (58,040 patients), and excluded 63 trials. Refer to Figure 1 for study flow diagram. Included studiesWe found 24 trials, with 28 active treatment arms, studying 58,040 patients, which met the inclusion criteria. Trials were only available to evaluate the effectiveness of four drug classes as first‐line drugs thiazides (19 trials in 39,713 patients), beta‐blockers (five trials in 19,313 patients), ACE inhibitors (three trials in 6002 patients), and calcium channel blockers (one trial in 4695 patients). Two trials evaluated thiazides as well as beta‐blockers versus placebo (MRC‐O 1992; MRC‐TMH 1985;); UKPDS 39 1998 evaluated beta‐blockers as well as ACE inhibitors versus placebo, and HYVET pilot 2003 evaluated thiazides as well as ACE inhibitors versus placebo, making 28 total comparisons from 24 trials. We did not identify any randomized controlled trials that compared first‐line alpha‐adrenergic blockers or angiotensin receptor blockers to placebo or untreated control group. The average age of participants across all included trials was 62 years. Six trials were limited to patients over 60 years of age (EWPHE 1985; HYVET 2008; HYVET pilot 2003; Kuramoto 1981; MRC‐O 1992; SHEP 1991; SHEP‐P 1989; SYST‐EUR 1997). The age range in other trials was between 21 and 80 years. The mean age of patients from the four classes was: thiazide ‒ 61 years, beta‐blockers ‒ 56 years, ACE inhibitors ‒ 67 years, and calcium channel blockers ‒ 70 years. Most participants were recruited from Western industrialized countries. Two trials did not report percentage of participants from different countries (HOPE HYP 2000; HYVET 2008). In 22 trials reporting recruitment of participants, 7750 (15.5%) were from USA, 32,907 (66%) from Europe; 3427 (6.9%) from Australia, and 91 (0.2%) from Japan. Females represented 45% of the population studied. Four trials included only men (OSLO 1986; VA‐I 1967; VA‐II 1970; VA‐NHLBI 1978) . Fourteen trials did not report ethnicity (Barraclough 1973; Carter 1970; Dutch TIA 1993; EWPHE 1985; HOPE HYP 2000; HYVET 2008; HYVET pilot 2003; Kuramoto 1981; MRC‐TMH 1985; MRC‐O 1992; OSLO 1986; PATS 1996; SYST‐EUR 1997; TEST 1995). Ten trials reported ethnicity. African‐Americans comprised the following percentages in these trials: ATTMH 1980 (0%), HSCSG 1974 (80%), SHEP 1991 (13.8%), SHEP‐P 1989 (18%), UKPDS 39 1998 (7.6%), USPHSHCSG 1977 (28%), VA‐I 1967 ( 53.8%), VA‐II 1970 (42%), VA‐NHLBI 1978 (25%), and Wolff 1966 (89.6%). The study population consisted of predominantly ambulatory patients recruited from the community, primary care centres, or hospital‐based clinics in 22 trials (57,982 patients, 99.7% of all patients included in this review). In the Kuramoto 1981 trial, 91 (0.2% of total) subjects were recruited from a home for the aged. Carter 1970 recruited 97 (0.3% of total) participants admitted to the hospital, who had survived an ischemic‐type major stroke. In most trials, it was possible to determine whether the participants in the trials represented primary or secondary prevention. All trials excluded patients with angina and congestive heart failure, as these conditions would require use of antihypertensive drugs for reasons independent of their antihypertensive action. Some trials allowed patients with prior myocardial infarction or stroke, as long as they were not recent (e.g. within the previous three months). Thus, by determining the baseline prevalence of stroke and myocardial infarction, it was possible to calculate the percentage that represented secondary prevention. Three trials did not report prevalence of stroke or myocardial infarction, but it was likely low in these trials (Barraclough 1973; Kuramoto 1981; VA‐II 1970; 587 participants, (1.0% of total randomized participants). Six trials (11,157 patients) were primary prevention with less than 1% secondary prevention patients (ATTMH 1980 (0.4%); MRC‐O 1992 (0%); OSLO 1986 (0%); UKPDS 39 1998 (0%); USPHSHCSG 1977 (0%); and VA‐NHLBI 1978 (0%). Six trials (12,042 patients) were secondary prevention (Dutch TIA 1993 (100%); HOPE HYP 2000 (88%); HSCSG 1974 (96%); OSLO 1986 (0%); PATS 1996 (100%); and TEST 1995 (100%). Nine trials (34,041 patients) were mostly primary prevention patients (EWPHE 1985 (it was reported that the baseline prevalence of cardiovascular complications was 36% and these included conditions other than proven myocardial infarction and stroke); HYVET 2008 (12%); HYVET pilot 2003 (7%); MRC‐TMH 1985 (2.2%); SHEP 1991 (6.4%); SHEP‐P 1989 (5.5%); SYST‐EUR 1997 (reported as 30% patients with cardiovascular complications); VA‐I 1967 (7%);and Wolff 1966 (29%). The percentage of secondary patients in the 10 mostly primary prevention trials was 3212 (5.6%) of total randomized patients. Thus, since 42,196 (72.7%) of total randomized people were primary prevention, the conclusions from this review are primarily relevant to the primary prevention setting. Baseline prevalence of diabetes was reported in 8 trials as follows: HOPE HYP 2000 (38%); HYVET 2008 (7%); MRC‐O 1992 (0%); SHEP 1991 (10.1%); UKPDS 39 1998 (100%); USPHSHCSG 1977 (0%); VA‐I 1967 (9.1%); and Wolff 1966 (16.0%). Baseline prevalence of smoking was as follows: ATTMH 1980 (25.0%); EWPHE 1985 (16.0%); HYVET 2008 (6.5%); OSLO 1986 (41.7%); MRC‐TMH 1985 (28.5%); MRC‐O 1992 (36.0%); SHEP 1991 (13.0%); SHEP‐P 1989 (11.0%); SYST‐EUR 1997 (7.0%); UKPDS 39 1998 (22.3%); and USPHSHCSG 1977 (46.7%). Recent trials defined stroke as the presence of neurological deficit lasting for more than 24 hours. It includes some patients with no disability. Older trials, like the HSCSG 1974, defined stroke as a neurological deficit lasting more than 24 hours, or a marked increase in transient ischemic attacks (twice the weekly pre‐randomization level of occurrence, more than four per week, or deterioration of more than eight points in neurological score). VA‐NHLBI 1978 defined stroke as typical weakness or paralysis. In some trials, stroke was not defined. In our opinion, lumping all strokes, including reversible ischemic neurological deficit (RIND), into one outcome is not optimal. More clinically relevant interpretations could be made if strokes were subdivided into three groups: strokes with no disability, strokes with mild disability, and strokes with severe disability. Myocardial infarction and sudden death were defined consistently across most trials. Myocardial infarction was defined as typical chest pain with ECG changes or increased cardiac enzymes; sudden death was defined as death within 24 hours of first evidence of acute cardiovascular disease, or unrelated to other known pre‐existing diseases. Five trials restricted recruitment to persons with systolic hypertension; defined as systolic pressure 160 to 219 mmHg, and diastolic pressure less than 90 mmHg (SHEP 1991; SHEP‐P 1989), diastolic pressure less than 95 mmHg (SYST‐EUR 1997), or systolic pressure higher than 140 mmHg (TEST 1995), or higher than 160 mmHg (HYVET 2008). Six trials based entry on diastolic hypertension (Barraclough 1973; USPHSHCSG 1977; VA‐I 1967; VA‐II 1970; VA‐NHLBI 1978; Wolff 1966); and 10 trials based entry on either systolic or diastolic hypertension (ATTMH 1980; Carter 1970; EWPHE 1985; HSCSG 1974; HYVET pilot 2003; Kuramoto 1981; MRC‐TMH 1985; MRC‐O 1992; OSLO 1986; UKPDS 39 1998). HOPE HYP 2000 represented the subgroup of the HOPE trial that had a baseline blood pressure higher than 140 mmHg systolic, or higher than 90 mmHg diastolic. Two trials were included because more than 70% of patients at entry had a systolic BP higher than 140 mmHg (Dutch TIA 1993; PATS 1996). Weighted mean baseline blood pressure for all the trials was 168/94 mmHg. When this was broken down into those that used systolic blood pressure as entry criteria, it was 173/84 mmHg; for those using diastolic pressure as entry criteria, it was 162/106 mmHg; and for those using both systolic and diastolic pressure as entry criteria, it was 167/97 mmHg. Two trials did not report baseline systolic pressure levels (Barraclough 1973;VA‐NHLBI 1978), and one trial did not report baseline diastolic pressure levels (VA‐NHLBI 1978). For complete description of the blood pressure inclusion criteria for each study, see 'Participants' in the 'Characteristics of included studies' table. A stepped approach to antihypertensive drug administration was used in 18 of the 24 trials. The exceptions were Dutch TIA 1993; HOPE HYP 2000; HSCSG 1974; PATS 1996; TEST 1995; and USPHSHCSG 1977, which used a standard dose of drug in the intervention arm. In 19 of the trials, a thiazide was the first‐line therapy in one of the arms of the trial. Because of a relatively large amount of data for thiazides, we were able to divide these 19 trials into those in which the thiazide starting dose was defined as low (8/19 trials, 874 patients) or high (11/19 trials, 19,839 patients), as explained in the methods. Three of the trials did not specify the thiazide dose, but were included in the high‐dose group because prescribing high doses of thiazides was common when those trials were conducted (Barraclough 1973; Carter 1970; OSLO 1986). The weighted mean dose of thiazide, in hydrochlorothiazide equivalents, was 90 mg for the high‐dose trials and 24 mg for the low‐dose trials. In five trials, a beta‐blocker was used as first‐line therapy in one of the arms of the trial (Dutch TIA 1993; MRC‐O 1992; MRC‐TMH 1985; TEST 1995; UKPDS 39 1998). Three trials used an angiotensin converting enzyme inhibitor, (HOPE HYP 2000; HYVET pilot 2003; UKPDS 39 1998). One trial used the calcium channel blocker nitrendipine (SYST‐EUR 1997). Second‐ and third‐line drugs included beta‐blockers, centrally‐acting drugs, peripherally‐acting anti‐adrenergic agents, vasodilators, thiazides, ACE inhibitors, alpha blockers, calcium channel blockers, and loop diuretics. See 'Interventions' in the 'Characteristics of included studies' table for a complete description of each study's drug treatment protocol. Mean duration of follow‐up ranged from 1.1 years for the HYVET pilot 2003 trial to 10 years for the USPHSHCSG 1977 trial. The weighted average follow‐up was 4.1 years for the thiazide trials, 5.3 years for the beta‐blocker trials, 4.9 years for the ACE Inhibitor trials, and 2.5 years for the one calcium‐channel blocker trial. Risk of bias in included studiesWe assessed risk of bias for each included study using the Cochrane 'Risk of bias' tool for RCTs, described in Chapter 8 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011b). Potential parameters of methodological quality listed in the 'Risk of bias' table include: method used to randomize participants, whether randomization was completed in an appropriate and blinded manner; whether participants, providers, outcome assessors, or a combination of these, were blinded to assigned therapy; whether the control group received a placebo or no treatment; percent of participants who did not complete follow‐up (dropouts); percent of participants not on assigned active or placebo therapy at study completion; selective reporting of outcomes; and other bias, in terms of funding of the trial by the manufacturer. Refer to Figure 2 for the 'Risk of bias' graph, which provides review authors' judgements about each risk of bias item, presented as percentages across all included studies. Refer to Figure 3 for the 'Risk of bias' summary, which provides review authors' judgements about each risk of bias item for each included study. Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies. Risk of bias summary: review authors' judgements about each risk of bias item for each included study. Effects of interventionsSee: Table 1; Table 2; Table 3; Table 4; Table 5 Summary of findings for the main comparisonFirst‐line low‐dose thiazide compared to placebo for hypertension
Summary of findings 2First‐line high‐dose thiazide compared to placebo for hypertension
Summary of findings 3First‐line beta‐blocker compared to placebo for hypertension
Summary of findings 4First‐line angiotensin converting enzyme inhibitor compared to placebo for hypertension
Summary of findings 5First‐line calcium channel blocker compared to placebo for hypertension
Refer to five Summary of findings tables: Table 1; Table 2; Table 3; Table 4; Table 5. Primary outcomes by drug classThiazidesBecause of the large number of trials, and the high heterogeneity in the effect between low dose and high dose thiazides on coronary heart disease events, we presented data for the two subgroups only, and not overall results for all thiazides. Low‐dose thiazidesFirst‐line low‐dose thiazide significantly reduced all the primary outcomes: mortality (risk ratio (RR) 0.89, 95% confidence interval (CI) 0.82 to 0.97; N = 19874; RCTs = 8; I² = 0%; Analysis 1.1), stroke (RR 0.68, 95% CI 0.60 to 0.77; N = 19874; RCTs = 8; I² = 0%; Analysis 1.2), coronary heart disease (RR 0.72, 95% CI 0.61 to 0.84; N = 19,022; RCTs = 7; I² = 0%; Analysis 1.3), and total cardiovascular events (RR 0.70, 95% CI 0.64 to 0.76; N = 19,022; RCTs = 7; I² = 0%; Analysis 1.4). The SHEP 1991 trial was the only one that reported total hospitalizations, an acceptable measure of total serious adverse events. It showed a numerical reduction with treatment (RR 0.95, 95% CI 0.89 to 1.01, N = 4736; RCT = 1; Analysis 1.5). Analysis Comparison 1 First‐line thiazide vs placebo, Outcome 1 Total mortality. Analysis Comparison 1 First‐line thiazide vs placebo, Outcome 2 Total stroke. Analysis Comparison 1 First‐line thiazide vs placebo, Outcome 3 Total coronary heart disease. Analysis Comparison 1 First‐line thiazide vs placebo, Outcome 4 Total cardiovascular events. Analysis Comparison 1 First‐line thiazide vs placebo, Outcome 5 Total hospitalizations. High‐dose thiazidesFirst‐line high‐dose thiazide, in contrast, significantly reduced stroke (RR 0.47, 95% CI 0.37 to 0.61; N = 19,839; RCTs = 11; I² = 46%; Analysis 1.2), and total cardiovascular events (RR 0.72, 95% CI 0.63 to 0.82; N = 19,839; RCTs = 11; I² = 35%; Analysis 1.4), but not mortality (RR 0.90, 95% CI 0.76 to 1.05; N = 19,839; RCTs = 11; I² = 21%; Analysis 1.1), or coronary heart disease (RR 1.01, 95% CI 0.85 to 1.20; N = 19,839; RCTs = 11; I² = 0%; Analysis 1.3). Beta‐blockersFirst‐line beta‐blockers reduced stroke (RR 0.83, 95% CI 0.72 to 0.97; N = 19313; RCTs = 5; I² = 7%; Analysis 2.2), and total cardiovascular events (RR 0.89, 95% CI 0.81 to 0.98; N = 19313; RCTs = 5; I² = 54%; Analysis 2.4), but not mortality (RR 0.96, 95% CI 0.86 to 1.07; N = 19313; RCTs = 5; I² = 25%; Analysis 2.1), or coronary heart disease (RR 0.90, 95% CI 0.78 to 1.03; N = 19313; RCTs = 5; I² = 4%; Analysis 2.3). Analysis Comparison 2 First‐line beta‐blocker vs placebo, Outcome 1 Total mortality. Analysis Comparison 2 First‐line beta‐blocker vs placebo, Outcome 2 Total stroke. Analysis Comparison 2 First‐line beta‐blocker vs placebo, Outcome 3 Total coronary heart disease. Analysis Comparison 2 First‐line beta‐blocker vs placebo, Outcome 4 Total cardiovascular events. Angiotensin converting enzyme (ACE) inhibitorsFirst‐line ACE inhibitors reduced mortality (RR 0.83, 95% CI 0.72 to 0.95; N = 6002; RCTs = 3; I² = 0%; Analysis 3.1), stroke (RR 0.65, 95% CI 0.52 to 0.82; N = 6002; RCTs = 3; I² = 0%; Analysis 3.2), coronary heart disease (RR 0.81, 95% CI 0.70 to 0.94; N = 5145; RCTs = 2; I² = 0%; Analysis 3.3), and total cardiovascular events (RR 0.76, 95% CI 0.67 to 0.85; N = 5145; RCTs = 2; I² = 0%; Analysis 3.4). Analysis Comparison 3 First‐line ACE inhibitor vs Placebo, Outcome 1 Total mortality. Analysis Comparison 3 First‐line ACE inhibitor vs Placebo, Outcome 2 Total stroke. Analysis Comparison 3 First‐line ACE inhibitor vs Placebo, Outcome 3 Total coronary heart disease. Analysis Comparison 3 First‐line ACE inhibitor vs Placebo, Outcome 4 Total cardiovascular events. Calcium channel blockersFirst‐line calcium channel blockers reduced stroke (RR 0.58, 95% CI 0.41 to 0.84; N = 4695; RCT = 1; Analysis 4.2), and total cardiovascular events (RR 0.71, 95% CI 0.57 to 0.87; N = 4695; RCT = 1; Analysis 4.4), but not mortality (RR 0.86, 95% CI 0.68 to 1.09; N = 4695; RCT = 1; Analysis 4.1), or coronary heart disease (RR 0.77, 95% CI 0.55 to 1.09; N = 4695; RCT = 1; Analysis 4.3). Analysis Comparison 4 First‐line calcium channel blocker vs Placebo, Outcome 1 Total mortality. Analysis Comparison 4 First‐line calcium channel blocker vs Placebo, Outcome 2 Total stroke. Analysis Comparison 4 First‐line calcium channel blocker vs Placebo, Outcome 3 Total coronary heart disease. Analysis Comparison 4 First‐line calcium channel blocker vs Placebo, Outcome 4 Total cardiovascular event. Secondary outcomesReduction in systolic and diastolic blood pressureAntihypertensive drug therapy significantly lowered both systolic and diastolic blood pressure, compared to the control group. Please refer to Table 6 for details. 1Blood pressure lowering efficacy with different drug classes
First‐line low‐dose thiazides, compared to placebo or no treatment, decreased systolic blood pressure (mean difference (MD) ‐12.56, 99% CI ‐13.22 to ‐11.91; N = 18,685; RCTs = 8; I² = 98%; Analysis 1.7), and diastolic blood pressure (MD ‐4.73, 99% CI ‐5.12 to ‐4.34; N = 18,685; RCTs = 8; I² = 98%; Analysis 1.8). Analysis Comparison 1 First‐line thiazide vs placebo, Outcome 7 Systolic blood pressure. Analysis Comparison 1 First‐line thiazide vs placebo, Outcome 8 Diastolic blood pressure. First‐line high‐dose thiazides, compared to placebo or no treatment, decreased systolic blood pressure (MD ‐13.66, 99% CI ‐14.40 to ‐12.91; N = 14,906; RCTs = 6; I² = 98%; Analysis 1.7), and diastolic blood pressure (MD ‐6.82, 99% CI ‐7.24 to ‐6.41; N = 19,347; RCTs = 10; I² = 97%; Analysis 1.8). First‐line beta‐blockers, compared to placebo or no treatment, reduced systolic blood pressure (MD ‐9.51, 99% CI ‐10.16 to ‐8.85; N = 18,833; RCTs = 5; I² = 92%; Analysis 2.6), and diastolic blood pressure (MD ‐5.64, 99% CI ‐6.06 to ‐5.22; N = 18,833; RCTs = 5; I² = 89%; Analysis 2.7). Analysis Comparison 2 First‐line beta‐blocker vs placebo, Outcome 6 Systolic blood pressure. Analysis Comparison 2 First‐line beta‐blocker vs placebo, Outcome 7 Diastolic blood pressure. First‐line ACE Inhibitors, compared to placebo or no treatment, decreased systolic blood pressure (MD ‐21.14, 99% CI ‐23.13 to ‐19.15; N = 1071; RCTs = 2; I² = 98%; Analysis 3.5), and diastolic blood pressure (MD ‐9.64, 99% CI ‐10.70 to ‐8.58; N = 1071; RCTs = 2; I² = 98%; Analysis 3.6). Analysis Comparison 3 First‐line ACE inhibitor vs Placebo, Outcome 5 Systolic blood pressure. Analysis Comparison 3 First‐line ACE inhibitor vs Placebo, Outcome 6 Diastolic blood pressure. First‐line calcium‐channel blockers, compared to placebo or no treatment, reduced systolic blood pressure (MD ‐8.90, 99% CI ‐10.14 to ‐7.66; N = 4695; RCT = 1; Analysis 4.6), and diastolic blood pressure (MD ‐4.50, 99% CI ‐5.10 to ‐3.90; N = 4695; RCT = 1; Analysis 4.7). Analysis Comparison 4 First‐line calcium channel blocker vs Placebo, Outcome 6 Systolic blood pressure. Analysis Comparison 4 First‐line calcium channel blocker vs Placebo, Outcome 7 Diastolic blood pressure. For each class of drugs, the blood pressure data were heterogeneous, however the effects remained highly significant, using the random‐effects model. Because of the high heterogeneity, and the fact that in many trials, other drugs were allowed, we did not think this was an accurate reflection of the blood pressure lowering effect of the first‐line drug. For the same reason, we did not present these data in the 'Summary of findings' tables, and no attempt was made to indirectly compare the different drugs for blood pressure. Withdrawal due to adverse effectsThis outcome was not reported in most of the trials. Where it was reported, drug therapy increased withdrawals due to adverse effects, compared to placebo or no treatment: low‐dose thiazides (RR 2.38, 95% CI 2.06 to 2.75; N = 8870; RCTs = 3; I² = 96%; Analysis 1.6), high‐dose thiazides (RR 4.48, 95% CI 3.83 to 5.24; N = 15,170; RCTs = 7; I² = 31%; Analysis 1.6), and beta‐blockers (RR 4.59, 95% CI 4.11 to 5.13; N = 18,565; RCTs = 4; I² = 96%; Analysis 2.5). Analysis Comparison 1 First‐line thiazide vs placebo, Outcome 6 Withdrawal due to adverse effects. Analysis Comparison 2 First‐line beta‐blocker vs placebo, Outcome 5 Withdrawal due to adverse effects. Because many of the trials did not report this outcome, and there was high heterogeneity between trials that did, we judged these data to have a high risk of bias. We could not calculate this information for the calcium channel blocker therapy, because withdrawals due to adverse drug effects were not reported in the SYST‐EUR 1997 trial, and authors declined to provide the information when requested. We could not use the data from the only ACE inhibitor trial that reported withdrawals due to adverse effects, as the untreated control group was not blinded (UKPDS 39 1998). DiscussionSummary of main resultsThis review, with a large amount of thiazide trial data (19 randomized controlled trials (RCTs), 39,713 participants), demonstrates the benefits of starting with a low‐dose thiazide as first‐line therapy for elevated blood pressure. The pooled data showed a reduction in total mortality when using a thiazide as the first‐line choice, and suggested that as first‐line therapy, low‐dose thiazide, reduced coronary heart disease events, whereas high‐dose thiazide did not. Five RCTs used a first‐line beta‐blocker (19,313 participants), and provided enough data to compare with first‐line thiazides. Analyses suggested that beta‐blockers reduced total stroke and total cardiovascular events less than all thiazides, and they reduced coronary heart disease less than a first‐line low‐dose thiazide. It is important to note that in four of the five beta‐blocker trials, atenolol was the beta‐blocker used. It is possible that the reduced effectiveness of first‐line beta blockers was limited to atenolol. First‐line angiotensin converting enzyme (ACE) inhibitors, in three RCTs and a smaller population (6002 participants), were associated with similar benefits, but wider confidence intervals than first‐line low‐dose thiazides for all outcomes. The amount of data for first‐line calcium channel blockers in one trial (4695 participants) was insufficient to make any meaningful comparisons; this can be appreciated by noting the wide confidence intervals associated with the treatment effects for this drug class (SYST‐EUR 1997). Relative risk (risk ratio) is the best way to indirectly compare the effectiveness between different drug classes. When we compared risk ratios (RR), beta blockers (RR 0.89, 95% confidence interval (CI) 0.81 to 0.98) appeared to be less effective than low‐dose thiazides (RR 0.70, 95% CI 0.64 to 0.76) in reducing total cardiovascular events. First‐line ACE inhibitors (RR 0.76, 95% CI 0.67 to 0.85), and calcium channel blockers (RR 0.71, 95% CI 0.57 to 0.87), had less data, but could not be distinguished from low‐dose thiazides for the effect on total cardiovascular events, or other outcomes. However, for the patient, it is more meaningful to have a measure of the absolute risk reduction (ARR) over a specified period of time. We calculated this summary measure for total cardiovascular events for the four drug classes in three clinical settings, where it was possible: secondary prevention, primary prevention (moderate to severe hypertension), and primary prevention (mild to moderate hypertension). Secondary prevention: For the three secondary prevention trials using thiazides, and the one secondary trial using an ACE inhibitor, the average baseline blood pressure was approximately 155/94 mmHg. The average total cardiovascular event rate was 23.1% over five years in the control group, and the RR with treatment was 0.76. Therefore, the ARR over five years was 5.5% (23.1 X 0.24). For the two secondary prevention beta‐blocker trials, there was no clear reduction in total cardiovascular events (RR 1.01, 95% CI 0.84 to 1.21). There were no secondary prevention trials that used calcium channel blockers. Primary prevention (moderate to severe hypertension): For the seven low‐dose thiazide trials in this category, the baseline systolic blood pressure was 175 mmHg, and the average total cardiovascular event rate over five years in the control groups was 16%. The RR with treatment was 0.68, for an ARR over five years of 5.1% (16 X 0.32). This is similar to the 5.5% calculated for secondary prevention. For the two beta‐blocker trials in this category, there was no significant reduction in total cardiovascular events (RR 0.88, 95% CI 0.71 to 1.02). In the one calcium channel blocker trial, the ARR over five years was 4.6% (16 X 0.29), and in the one ACE inhibitor trial, the estimated ARR over five years was 3.7% (16 X 0.23; UKPDS 39 1998). Primary prevention (mild to moderate hypertension): There were five first‐line high‐dose thiazide trials in this category, with an average baseline systolic blood pressure of 160 mmHg. In these five trials, the average cardiovascular event rate in the control group over five years was 4.1%. The RR with first‐line high‐dose thiazide treatment was 0.80 (95% CI 0.69 to 0.94), and the ARR over five years was 0.82% (4.1 X 0.2). For the one beta‐blocker trial in this category, the RR with treatment was 0.82, and the ARR over five years was 0.74% (4.1 X 0.18). This demonstrates that the absolute benefits are at least as good for first‐line low dose thiazides as the other classes of antihypertensives. The number needed to treat for an additional beneficial outcome (NNTB) for a low‐dose thiazide in moderate to severe hypertension (average systolic 175 mmHg) is about 20 over a five‐year duration. The NNTB was only a little lower in a secondary prevention setting, though the baseline blood pressures (average systolic 155 mmHg) were also lower. Notably, the NNTB was much higher, about 120 over five years, for primary prevention patients with mild to moderate hypertension. The low absolute benefit in individuals with lower blood pressure at baseline reflected two differences: the lesser relative benefit of a RR of 0.8 versus a RR of 0.7, and the lower five‐year event rate in the control groups of 4%. One of the limitations of this analysis was that the first‐line drug used in this population was a high‐dose thiazide, which the evidence suggests is not as effective. It is possible that using a low‐dose thiazide in these trials would have improved the benefit to a RR of 0.7. However, even if that was true, the absolute benefit would still be small (ARR = 1.2%; NNTB 83 over five years). The low absolute benefit of antihypertensive therapy for mild to moderate elevations in blood pressure in primary prevention needs to be reflected by authors of hypertension guidelines. Many of the patients in these trials also had co‐morbidities, such as diabetes mellitus. In this review, it was not possible to assess these patients separately, but SHEP 1991 analyzed their diabetes subgroup, and found the relative benefit was the same, and the absolute benefit was greater in diabetic patients than in non‐diabetic patients. Overall completeness and applicability of evidenceSince 72.7% of participants in this review were primary prevention, the data are primarily relevant to a primary prevention population. Three of the included first‐line thiazide trials included participants with a prior stroke or transient ischemic attack (TIA; (Carter 1970; HSCSG 1974; PATS 1996)). When we deselected these trials from the thiazide analyses, the treatment effect estimates did not change. When we deselected all trials in which the baseline prevalence of myocardial infarction was either not reported (Barraclough 1973; Kuramoto 1981; MRC‐O 1992; VA‐II 1970), or was greater than 10% (Carter 1970; EWPHE 1985; HSCSG 1974; Wolff 1966), the treatment effect for total cardiovascular events was not different. For first‐line ACE inhibitors, HOPE HYP 2000 was predominantly secondary prevention, and the relative benefit with treatment was similar to the other two primary prevention trials (HYVET pilot 2003; UKPDS 39 1998). For sensitivity analyses, we deselected trials that were not placebo controlled and blinded, trials that were restricted to patients with isolated systolic hypertension, small trials, and trials using supplemental drugs from other defined classes. In all of these instances, there was no clinically important change in the treatment effect estimate. The data on withdrawals due to adverse effects were incomplete and heterogeneous. Therefore, it was difficult to compare the different drug classes for this outcome. However, there was nothing to suggest that other classes of drugs were better tolerated than first‐line low‐dose thiazides. The blood pressure data in this systematic review were heterogeneous. This was because the number of drugs used in the trials differed, and only 16 of the 24 trials were double‐blinded. Blood pressure measurements are subject to bias if the observer and the patient know what is being administered. Because of these factors, it was not possible to use the blood pressure data to assess the blood pressure lowering efficacy of the different classes of drugs, or to compare low‐dose and high‐dose first‐line thiazides. There is growing evidence that the surrogate marker ‐ the lowering of blood pressure ‐ is inadequate to predict health outcomes with antihypertensive therapy. Despite similar reductions in blood pressure, we found that first‐line low‐dose and high‐dose thiazide therapy had different impacts on the incidence of coronary heart disease. In the 15 of the 24 trials where it could be assessed, despite dose titration and the addition of supplemental drugs, about 40% of the patients did not achieve the target blood pressure. Four trials reported that 34% of the patients did not achieve the target systolic pressure of less than 160 mm Hg (Kuramoto 1981; MRC‐O 1992; SHEP 1991; SHEP‐P 1989). In two trials, over 50% did not achieve the target systolic pressure of less than 150 mmHg (HYVET 2008; HYVET pilot 2003). The other nine trials reported that 40% of the patients did not achieve the target diastolic pressure of less than 90 mmHg (ATTMH 1980; EWPHE 1985; HSCSG 1974; MRC‐TMH 1985; OSLO 1986; USPHSHCSG 1977; VA‐I 1967; VA‐II 1970; Wolff 1966). This observation is extremely important clinically, as it means that physicians can only expect to achieve blood pressure targets in about 60% of the patients they treat. However, this does not mean that the patients who do not achieve the target will benefit less than those who do, as blood pressure reduction is only partly responsible for the risk reduction from antihypertensive treatment (Boissel 2005). The evidence showed that starting with a low dose of thiazide, and only titrating up if necessary, significantly reduced the risk of coronary heart disease. In contrast, starting with a high dose (50 mg of hydrochlorothiazide or the equivalent) did not. Since the high‐dose therapy has no proven advantages, there is no justification for using high doses, or for comparing low‐dose and high‐dose thiazide therapies in a future trial. Therefore, the current recommendation for the use of thiazides in the management of hypertension is justifiable, based on the evidence presented here: start with a low dose (12.5 mg hydrochlorothiazide, or the equivalent), increase the dose if necessary, and do not exceed a dose of 50 mg of hydrochlorothiazide, or the equivalent. The fact that thiazides are similarly, or more effective than other drug classes in reducing mortality and morbidity, and that the evidence is most robust for thiazides, is reason by itself to prescribe them first for most patients with hypertension. The value of this approach is further supported by the fact that of the drug classes studied, the thiazides are the least expensive drug class in most countries. If more than one thiazide or thiazide‐like drug is available, then the least expensive one is the most rational choice. Additional information supporting the use of thiazides was recently published by Puttnam 2017. By using data from a national database, a post‐trial cohort surveillance study of the Antihypertensive and Lipid‐Lowering Treatment to Prevent Heart Attack trial (ALLHAT) examined whether the use of thiazide diuretics for the treatment of hypertension was associated with reduced risk of fracture, compared with non‐use. ALLHAT was a large randomized clinical trial that compared the effect of first‐step therapy with different classes of antihypertensive drug therapy (chlorthalidone, amlodipine, and lisinopril) in preventing fatal coronary heart disease (CHD), nonfatal myocardial infarction (primary outcome), or other cardiovascular disease (CVD) events. Hospitalized hip and pelvic fractures were chosen as endpoints, because they are almost always associated with hospitalization. A total of 22,180 participants, with a mean age of 70.4 years, were followed for up to eight years (mean 4.9 years) during masked therapy. Participants randomized to receive chlorthalidone versus amlodipine or lisinopril had a lower risk of fracture on adjusted analyses (hazards ratio (HR) 0.79, 95% CI 0.63 to 0.98; P = 0.04). The authors concluded that "findings from a large randomized clinical trial provide evidence of a beneficial effect of thiazide‐type diuretic therapy in reducing hip and pelvic fracture compared with treatment with other antihypertensive medications". Quality of the evidenceWe graded the overall quality of the evidence and developed 'Summary of findings' tables, using GRADEpro GDTsoftware (GRADEpro GDT). We created five 'Summary of findings' tables to display the quality of evidence and summary of the effects on clinically important outcomes, for first‐line low‐dose thiazides (Table 1), high‐dose thiazides (Table 2), beta‐blockers (Table 3), ACE inhibitors (Table 4), and calcium channel blockers (Table 5). We found high‐quality evidence that first‐line low‐dose thiazides reduced mortality, stroke, CHD, and total cardiovascular events more than placebo. We found moderate‐quality evidence that high‐dose thiazides reduced stroke, and CVS events more than placebo, and low‐quality evidence that first‐line high‐dose thiazides did not reduce CHD more than placebo. We found low‐quality evidence that beta‐blockers reduced stroke, and total cardiovascular events more than placebo. We found moderate‐quality evidence that ACE inhibitors reduced mortality and total CVS events, and low‐quality evidence that they reduced stroke and CHD more than placebo. We found low‐quality evidence that calcium channel blockers reduced mortality, stroke, CHD, and CVS events more than placebo. We found low‐quality evidence that thiazides and beta‐blockers caused more participants to withdraw from the trials due to adverse effects than those who took placebo. We found low‐ to very low‐quality evidence that all four drug classes reduced systolic and diastolic blood pressure. Potential biases in the review processA potential limitation of attributing the benefit predominantly to the first‐line thiazide is the fact that in some of the thiazide trials, the first‐line therapy included the thiazide combined with another drug. In two trials, the first‐line therapy was a combination of a thiazide plus a potassium sparing diuretic (EWPHE 1985; MRC‐O 1992). Deselecting these two trials had no impact on the treatment effect. In five trials, a reserpine derivative was combined with the thiazide as first‐line therapy (HSCSG 1974; USPHSHCSG 1977; VA‐I 1967; VA‐II 1970; Wolff 1966. When these five trials were deselected, the treatment effect did not change either. Therefore, the data for first‐line thiazides appear robust, and the benefits achieved are most likely attributable to the first‐line thiazide therapy. Agreements and disagreements with other studies or reviewsIt is important in this type of analysis not to include trials in which there is significant contamination by drugs from other drug classes of interest. In our opinion, Psaty 1997 incorrectly included the trial by Coope 1986 and the STOP 1991 trial in their beta‐blocker group. In the trial by Coope 1986, 67% of patients in the active treatment group received bendrofluazide, and 70% received atenolol; in the STOP 1991 trial, more than 70% in the active treatment group received thiazides and more than 70% received beta‐blockers. Therefore, in those two trials, the treatment effect could not be attributed to first‐line thiazides or first‐line beta‐blockers. Other minor differences in the trials in this review and that by Psaty 1997 are as follows. We classified Kuramoto 1981 as a low‐dose trial, and they classified it as high‐dose trial. It fit into our low‐dose category, based on the starting dose of 1 mg of trichlormethiazide. Moving the trial from the low‐dose to the high‐dose group did not change our findings. Unlike Psaty 1997, we excluded the Hypertension Detection and Follow‐up Program trial because it did not have an untreated control group, and the intervention group was treated with life‐style therapies in addition to drug therapy (HDFP 1984). We included one small trial by Wolff 1966, not referenced by Psaty 1997. This trial was also included in the meta‐analysis in Collins 1990 and in Gueyffier 1996, The meta‐analysis by the blood pressure trialists group was incomplete and misleading because of their decision to exclude all trials prior to 1995 (BBLTTC 2005). The network meta‐analysis included most of the trials comparing treatment versus placebo or no treatment that were included in this review (Psaty 2003). However, they again incorrectly included two trials in their beta‐blocker group (Coope 1986; STOP 1991), and in our opinion, inappropriately included PROGRESS 2001, IDM 2001, and Lewis 2001 as evidence for first‐line treatment in hypertension, when most patients did not have hypertension. Most importantly, this review includes four trials not included in the Psaty 2003 review (HOPE HYP 2000; HYVET pilot 2003; UKPDS 39 1998). Our reasons for classifying UKPDS 39 1998 in this review deserve mention. First, in UKPDS 39 1998, the treatment target for the control group was over 200/105 mmHg for the first five years, after which it was lowered to over 180/105 mmHg. This was similar to the escape therapy in other trials that classified the control group as no treatment. For example, in the OSLO 1986 trial, drug treatment in the control group was started if blood pressure exceeded 179/109 mmHg and was sustained. Second, the UKPDS 39 1998 target in the intervention group (less than 150/85 mmHg) was very similar to the targets in the intervention groups of the included studies in this review (less than 140 to 160/less than 80 to 90 mmHg). Third, the difference in blood pressure achieved between the treatment and control group over the nine years of the trial (10 to 11/4 to 6 mmHg) is similar to the difference in BP for other included trials in this review (9 to 12/3 to 6 mmHg; (ATTMH 1980; Kuramoto 1981; MRC‐TMH 1985; SYST‐EUR 1997)). Wiysonge 2017 also included UKPDS 39 1998 in their beta‐blocker review, published in 2007. In contrast to Wiysonge 2017, we did not include IPPPSH 1985 as representing first‐line beta blockers, as more than 65% of the patients also received a thiazide. We did include Dutch TIA 1993 and TEST 1995, which Wiysonge 2017 did not. Other reviews did not include HOPE HYP 2000, as the data separated for the hypertensive group had not been published. We are thankful to the authors of HOPE for providing these data. The HYVET pilot 2003 trial and the HYVET 2008 trial are more recent trials included in our review. Despite these differences, the conclusions and interpretation of the findings in this review are concordant with Psaty 2003, who concluded that thiazides were the first‐line drug of choice, and Wiysonge 2017, who concluded that beta‐blockers were not an appropriate first‐line drug choice. Refer to Table 7 for further details. 2First‐line beta‐blocker compared to placebo. Comparison of this review with Wysonge 2017
The recent Taverny 2016 review assessed the effects of antihypertensive pharmacotherapy on three separate cardiac outcomes ‐ sudden cardiac death, fatal myocardial infarction, and nonfatal myocardial infarction in hypertensive individuals. In our review, these three outcomes were combined as total CHD events. In Taverny 2016, based on 15 randomized placebo‐controlled trials, in 39,908 patients, for a mean duration of 4.2 years, moderate‐quality evidence showed that antihypertensive therapy did not reduce sudden cardiac death (RR 0.96, 95% CI 0.81 to 1.15), but did reduce both nonfatal myocardial infarction (RR 0.85, 95% CI 0.74 to 0.98), and fatal myocardial infarction (RR 0.75, 95% CI 0.62 to 0.90). However, they did not report results separately for specific first‐line antihypertensive drug classes. Withdrawals due to adverse effects were increased in the drug treatment group to 12.8%, compared with 6.2% in the no treatment group. A Cochrane review, Pharmacotherapy for mild hypertension, has been published and is being updated (Diao 2012). A systematic review, Pharmacotherapy for hypertension in adults (18 to 59 years), has recently been published (Musini 2017). The Cochrane review, Pharmacotherapy for hypertension in the elderly (60 years or older), is being updated (Musini 2009). Since more trials comparing drug therapy with a placebo or no treatment are unlikely to be forthcoming, future evidence of effectiveness of first‐line therapy will need to come from head‐to‐head comparisons. Two Cochrane reviews have compared first‐line calcium channel blockers (Chen 2010) and first‐line drugs inhibiting the renin angiotensin system (Xue 2015) with first‐line thiazides. These reviews confirm the likely morbidity benefits of first‐line thiazides. We are also currently working on a meta‐analysis of head‐to‐head trials in a separate systematic review. That review will focus on head‐to‐head trials comparing first‐line thiazides with other classes of drugs (Reinhart 2011). Authors' conclusionsImplications for practiceChoice of first‐line treatment versus placebo or no treatment: the evidence for morbidity and mortality.
Blood pressure measurement.
Population.
Implications for researchAt the present time, RCT data for antihypertensives as first‐line treatment versus placebo or no treatment are lacking for all classes other than thiazides. Since we have clear evidence of effectiveness of first‐line low‐dose thiazides in primary prevention, patients with blood pressure of 160/100 mmHg or higher, and in secondary prevention, in patients with lower blood pressures, it would be unethical to do further trials for this population compared to a placebo or untreated control group. Future RCTs in these populations should be done with low‐dose first‐line thiazides as the comparison group. The benefits and harms of primary prevention treatment of patients with lower blood pressures remains uncertain at the present time. Large trials recruiting primary prevention patients in lower blood pressure categories and using first‐line low‐dose thiazides compared to placebo are needed. FeedbackWas Kuramoto study randomized?, 17 April 2014SummaryIn the table of included studies, the Cochrane review authors state that the included study by Kuramoto, a Japanese study is a randomised, double‐blind placebo‐controlled trial conducted in Japan. Internal validity, risk of bias for allocation is low risk. It appears to me that the Kuramoto study may not have been randomized, but is a matched case control study. I agree with the conflict of interest statement below: I certify that I have no affiliations with or involvement in any organization or entity with a financial interest in the subject matter of my feedback. ReplyThank you for your comment. Kuramoto 1981 study is a placebo‐controlled single site study conducted in ambulatory patients in home for the aged in Tokyo, Japan. This study was included as a randomised study in the "Pharmacotherapy for hypertension in the elderly" original review by Mulrow 1998 based on unpublished information as per personal conversation with study author. Mulrow 1998 review stated that "Allocation of individuals within matched pairs to treatment and control groups was made by a blinded statistical coordinator, thought to be randomised but not entirely clear. Both patients and providers were blinded." Based on this information from Mulrow 1998 review we have decided to include this study and reassessed the risk of bias for the two domains ‐ allocation concealment and blinding of patient and physician as "Low risk" of bias. ContributorsKirsty Loudon, PhD Student Affiliation: University of Dundee What's new
HistoryProtocol first published: Issue 4, 1999
AcknowledgementsWe would like to thank Stephen Adams for finding references and Ciprian Jauca for assistance with formatting and copy‐editing. AppendicesAppendix 1. Search strategiesDatabase: Cochrane Hypertension
Specialised Register via Cochrane Register of Studies (CRS‐Web) *************************** Database: Cochrane Central Register of Controlled Trials via Cochrane Register of Studies (CRS‐Web) ***************************************************** Database: Ovid MEDLINE(R) 1946 to Present with Daily Update *************************** Database: Embase <1974 to 2017 November 22> *************************** Database: ClinicalTrials.gov *************************** Database: WHO International Clinical Trials Registry Platform (ICTRP) NotesNew search for studies and content updated (no change to conclusions) Data and analysesComparison 1First‐line thiazide vs placebo
Comparison 2First‐line beta‐blocker vs placebo
Comparison 3First‐line ACE inhibitor vs Placebo
Comparison 4First‐line calcium channel blocker vs Placebo
Analysis Comparison 4 First‐line calcium channel blocker vs Placebo, Outcome 5 Heart Failure. Characteristics of studiesCharacteristics of included studies [ordered by study ID]
Characteristics of excluded studies [ordered by study ID]
Differences between protocol and reviewAs the original version of this review did not include 'Risk of bias' assessment of all included studies, according to chapter 8 of the Cochrane Handbook for Systematic Reviews of Interventions, the protocol was updated to include it, using standard Cochrane methodology. Also, this update meets MECIR standards, including the addition of five 'Summary of findings' tables. A third author, Rupam Gill has been added to the updated review. RG independently assessed risk of bias of the 24 included studies as a second reviewer. Although the protocol did not specify primary and secondary outcome measures separately, we have categorized total mortality, total cardiovascular events, total stroke, and total CHD as the primary outcome measures. Reduction in systolic and diastolic blood pressure and withdrawal due to adverse effects are secondary outcome measures. Contributions of authorsJim Wright (JMW) was responsible for the design of the protocol, assessed all trials for inclusion or exclusion, checked the data entry, and contributed to the data analysis and to the interpretation and final draft of the review. Vijaya Musini (VM) assessed all trials for inclusion or exclusion, extracted data, checked data entry, assessed risk of bias of all included studies, contributed to data analysis, interpretation and final draft of the review. Summary of Findings table was also prepared by VM. Rupam Gill (RG) was the second reviewer who assessed risk of bias of all included studies. Sources of supportInternal sources
External sources
Declarations of interestJMW: none. VM: none. RG: none. ReferencesReferences to studies included in this reviewATTMH 1980 {published data only}
Barraclough 1973 {published data only}
Carter 1970 {published data only}
Dutch TIA 1993 {published data only}
EWPHE 1985 {published data only}
HOPE HYP 2000 {published data only}
HSCSG 1974 {published data only}
HYVET 2008 {published data only}
HYVET pilot 2003 {published data only}
Kuramoto 1981 {published data only}
MRC‐O 1992 {published data only}
MRC‐TMH 1985 {published data only}
OSLO 1986 {published data only}
PATS 1996 {published data only}
SHEP 1991 {published data only}
SHEP‐P 1989 {published data only}
SYST‐EUR 1997 {published data only}
TEST 1995 {published data only}
UKPDS 39 1998 {published data only}
USPHSHCSG 1977 {published data only}
VA‐I 1967 {published data only}
VA‐II 1970 {published data only}
VA‐NHLBI 1978 {published data only}
Wolff 1966 {published data only}
References to studies excluded from this reviewACTIVE I 2011 {published data only}
ADVANCE 2007 {published data only}
ALLHAT 2000 {published data only}
BENEDICT 2004 {published data only}
BENEDICT A 2006 {published data only}
Berglund 1981 {published data only}
CASTEL 1994 {published data only}
Coope 1986 {published data only}
DIABHYCAR 2004 {published data only}
DREAM 2006 {published data only}
EUROPA 2003 {published data only}
Fuchs 2011 {published data only}
GENERIC 2010 {published data only}
GENRES 2007 {published data only}
GLANT 1995 {published data only}
HAPPHY 1987 {published data only}
HDFP 1984 {published data only}
Hood 2007 {published data only}
HOPE 3 2016 {published data only}
HOT 1995 {published data only}
IDM 2001 {published data only}
IDNT 2003 {published data only}
IMAGINE 2008 {published data only}
Imai 2011 {published data only}
INSIGHT 1996 {published data only}
IPPPSH 1985 {published data only}
Kondo 2003 {published data only}
Kuramoto‐2 1994 {published data only}
Lewis 1993 {published data only}
Lewis 2001 {published data only}
MacMahon 2000 {published data only}
MAPHY 1988 {published data only}
Materson 1997 {published data only}
MIDAS 1996 {published data only}
Morgan 1980 {published data only}
NAVIGATOR 2010 {published data only}
NICOLE 2003 {published data only}
NORDIL 2000 {published data only}
PEACE 2004 {published data only}
Pool 2007 {published data only}
PRoFESS 2008 {published data only}
PROGRESS 2001 {published data only}
QUIET 2001 {published data only}
REIN 1997 {published data only}
RENAAL 2001 {published data only}
ROAD 2007 {published data only}
ROADMAP 2011 {published data only}
SCAST 2015 {published data only}
SCAT 2000 {published data only}
Schmieder 2012 {published data only}
SCOPE 2003 {published data only}
SHELL 1995 {published data only}
Sprackling 1981 {published data only}
STONE 1996 {published data only}
STOP 1991 {published data only}
STOP‐2 1993 {published data only}
Strandberg 1991 {published data only}
SYST‐CHINA 1993 {published data only}
TOMHS 1993 {published data only}
TRANSCEND 2008 {published data only}
VACS 1982 {published data only}
VHAS 1997 {published data only}
White 1995 {published data only}
Additional referencesBBLTTC 2005
Boissel 2005
BPLTTC 2000
Chen 2010
Collins 1990
Diao 2012
Duarte 2010
Frishman 2005
Goeres 2014
GRADEpro GDT [Computer program]
Gueyffier 1996
Gueyffier 1999
Higgins 2011a
Higgins 2011b
Hughes 2004
Insua 1994
Kang 2004
Kızılırmak 2017
Law 2009
MacMahon 1993
Mulrow 1994
Mulrow 1998
Musini 2009
Musini 2017
Nikolaus 2000
Parsons 2016
Pearce 1995
Psaty 1997
Psaty 2003
Puttnam 2017
Quan 1999
Reinhart 2011
RevMan 2014 [Computer program]
Sundstro¨m 2015
Tan 2016
Taverny 2016
Thijs 1992
Thomopoulos 2014
Thomopoulos 2016
Turnbull 2003
Wiysonge 2017
Xue 2015
Zanchetti 2015
Zhu 2005
References to other published versions of this reviewWright 1999
Articles from The Cochrane Database of Systematic Reviews are provided here courtesy of Wiley |