Blood pressure‐lowering efficacy of monotherapy with thiazide diuretics for primary hypertension
Abstract
Background
Hypertension is a modifiable cardiovascular risk factor. Although it is established that low‐dose thiazides reduce mortality as well as cardiovascular morbidity, the dose‐related effect of thiazides in decreasing blood pressure has not been subject to a rigorous systematic review. It is not known whether individual drugs within the thiazide diuretic class differ in their blood pressure‐lowering effects and adverse effects.
Objectives
To determine the dose‐related decrease in systolic and/or diastolic blood pressure due to thiazide diuretics compared with placebo control in the treatment of patients with primary hypertension. Secondary outcomes included the dose‐related adverse events leading to patient withdrawal and adverse biochemical effects on serum potassium, uric acid, creatinine, glucose and lipids.
Search methods
We searched the Cochrane Central Register of Controlled Trials (CENTRAL 2014, Issue 1), Ovid MEDLINE (1946 to February 2014), Ovid EMBASE (1974 to February 2014) and ClinicalTrials.gov.
Selection criteria
We included double‐blind, randomized controlled trials (RCTs) comparing fixed‐dose thiazide diuretic monotherapy with placebo for a duration of 3 to 12 weeks in the treatment of adult patients with primary hypertension.
Data collection and analysis
Two authors independently screened articles, assessed trial eligibility, extracted data and determined risk of bias. We combined data for continuous variables using a mean difference (MD) and for dichotomous outcomes we calculated the relative risk ratio (RR) with 95% confidence interval (CI).
Main results
We included 60 randomized, double‐blind trials that evaluated the dose‐related trough blood pressure‐lowering efficacy of six different thiazide diuretics in 11,282 participants treated for a mean duration of eight weeks. The mean age of the participants was 55 years and baseline blood pressure was 158/99 mmHg. Adequate blood pressure‐lowering efficacy data were available for hydrochlorothiazide, chlorthalidone and indapamide. We judged 54 (90%) included trials to have unclear or high risk of bias, which impacted on our confidence in the results for some of our outcomes.
In 33 trials with a baseline blood pressure of 155/100 mmHg, hydrochlorothiazide lowered blood pressure based on dose, with doses of 6.25 mg, 12.5 mg, 25 mg and 50 mg/day lowering blood pressure compared to placebo by 4 mmHg (95% CI 2 to 6, moderate‐quality evidence)/2 mmHg (95% CI 1 to 4, moderate‐quality evidence), 6 mmHg (95% CI 5 to 7, high‐quality evidence)/3 mmHg (95% CI 3 to 4, high‐quality evidence), 8 mmHg (95% CI 7 to 9, high‐quality evidence)/3 mmHg (95% CI 3 to 4, high‐quality evidence) and 11 mmHg (95% CI 6 to 15, low‐quality evidence)/5 mmHg (95% CI 3 to 7, low‐quality evidence), respectively.
Direct comparison of doses did not show evidence of dose dependence for blood pressure‐lowering for any of the other thiazides for which RCT data were available: bendrofluazide, chlorthalidone, cyclopenthiazide, metolazone or indapamide.
In seven trials with a baseline blood pressure of 163/88 mmHg, chlorthalidone at doses of 12.5 mg to 75 mg/day reduced average blood pressure compared to placebo by 12.0 mmHg (95% CI 10 to 14, low‐quality evidence)/4 mmHg (95% CI 3 to 5, low‐quality evidence).
In 10 trials with a baseline blood pressure of 161/98 mmHg, indapamide at doses of 1.0 mg to 5.0 mg/day reduced blood pressure compared to placebo by 9 mmHg (95% CI 7 to 10, low‐quality evidence)/4 (95% CI 3 to 5, low‐quality evidence).
We judged the maximal blood pressure‐lowering effect of the different thiazides to be similar. Overall, thiazides reduced average blood pressure compared to placebo by 9 mmHg (95% CI 9 to 10, high‐quality evidence)/4 mmHg (95% CI 3 to 4, high‐quality evidence).
Thiazides as a class have a greater effect on systolic than on diastolic blood pressure, therefore thiazides lower pulse pressure by 4 mmHg to 6 mmHg, an amount that is greater than the 3 mmHg seen with angiotensin‐converting enzyme (ACE) inhibitors, angiotensin receptor blockers (ARBs) and renin inhibitors, and the 2 mmHg seen with non‐selective beta‐blockers. This is based on an informal indirect comparison of results observed in other Cochrane reviews on ACE inhibitors, ARBs and renin inhibitors compared with placebo, which used similar inclusion/exclusion criteria to the present review.
Thiazides reduced potassium, increased uric acid and increased total cholesterol and triglycerides. These effects were dose‐related and were least for hydrochlorothiazide. Chlorthalidone increased serum glucose but the evidence was unclear for other thiazides. There is a high risk of bias in the metabolic data. This review does not provide a good assessment of the adverse effects of these drugs because there was a high risk of bias in the reporting of withdrawals due to adverse effects.
Authors' conclusions
This systematic review shows that hydrochlorothiazide has a dose‐related blood pressure‐lowering effect. The mean blood pressure‐lowering effect over the dose range 6.25 mg, 12.5 mg, 25 mg and 50 mg/day is 4/2 mmHg, 6/3 mmHg, 8/3 mmHg and 11/5 mmHg, respectively. For other thiazide drugs, the lowest doses studied lowered blood pressure maximally and higher doses did not lower it more. Due to the greater effect on systolic than on diastolic blood pressure, thiazides lower pulse pressure by 4 mmHg to 6 mmHg. This exceeds the mean 3 mmHg pulse pressure reduction achieved by ACE inhibitors, ARBs and renin inhibitors, and the 2 mmHg pulse pressure reduction with non‐selective beta‐blockers as shown in other Cochrane reviews, which compared these antihypertensive drug classes with placebo and used similar inclusion/exclusion criteria.
Thiazides did not increase withdrawals due to adverse effects in these short‐term trials but there is a high risk of bias for that outcome. Thiazides reduced potassium, increased uric acid and increased total cholesterol and triglycerides.
PICOs
Plain language summary
Thiazide diuretics for the treatment of high blood pressure
Thiazide diuretics are a class of drugs commonly recommended as first‐line treatment for raised blood pressure because they significantly reduce death, stroke and heart attacks. This class includes bendrofluazide, chlorthalidone, cyclopenthiazide, hydrochlorothiazide, indapamide and metolazone. We asked by how much does this class of drugs lower blood pressure and whether there is a difference between individual drugs within the class. We searched the available scientific literature to find all the trials that had assessed this question. The data included in this review was up to date as of February 2014.
We found 60 trials that randomly assigned 11,282 adult participants, mean age 55 years, 53% male and 47% female, with blood pressure above 140/90 mmHg (mean blood pressure 158/99 mmHg) to take one of six thiazide diuretics or placebo for an average duration of eight weeks. Most of the trials (82%) were published before the year 2000 and most were found to have a high risk of bias in the adverse effect data. Co‐morbidities were not reported in most trials. The blood pressure‐lowering effect was modest. Thiazide diuretics reduced blood pressure by 9 points in the upper number (called systolic blood pressure) and 4 points in the lower number (called diastolic blood pressure). Different thiazide drugs have similar effects in lowering blood pressure and thiazides lower systolic blood pressure more than other classes of antihypertensive drugs.
This review could not provide a valid estimate of short‐term harms from all thiazide diuretics because there was incomplete reporting of metabolic effects (serum potassium, uric acid, creatinine, glucose, total cholesterol, low‐density cholesterol and triglycerides) and the number of participants who dropped out of the trials due to adverse drug effects.
Authors' conclusions
Summary of findings
| Hydrochlorothiazide compared with placebo for primary hypertension | |||||
| Patient or population: adults with primary hypertension Settings: outpatient Intervention: hydrochlorothiazide 3 to 100 mg/day Comparison: placebo | |||||
| Outcomes | Daily dose | MD (95% CI) mmHg | No of participants | Quality of the evidence | Comments |
| Systolic blood pressure | 3 to 6.25 mg | ‐3.6 (‐5.6 to ‐1.5) | 663 (8) | ⊕⊕⊕⊝ moderate1 | Similar to the effect as a second‐line drug (Chen 2009). This is an indirect comparison of the effect size with a wider confidence interval compared to the Chen review, which is based on 22 trials in 3283 patients using similar inclusion/exclusion criteria with systolic blood pressure‐lowering of ‐3.7 (‐4.6 to ‐2.8) mmHg |
| 12.5 mg | ‐6.3 (‐7.2 to ‐5.3) | 2645 (22) | ⊕⊕⊕⊕ | A narrow confidence interval based on a large sample size with a magnitude of lowering very similar to the effect as a second‐line drug (Chen 2009), which was ‐6.0 (‐6.5 to ‐5.4) mmHg | |
| 25 mg | ‐8.0 (‐9.0 to ‐7.0) | 3062 (25) | ⊕⊕⊕⊕ | A narrow confidence interval based on a large sample size with a magnitude of lowering very similar to the effect as a second‐line drug (Chen 2009), which was ‐8.0 (‐8.7 to ‐7.3) mmHg | |
| 50 to 100 mg | ‐10.2 (‐13.1 to ‐7.3) | 315 (2) | ⊕⊕⊝⊝ | The 2 included studies have a high risk of bias. The confidence interval is very wide with small a sample size providing insufficient data in both this review as well as in the Chen review comparing the effect as a second‐line drug (Chen 2009) | |
| Diastolic blood pressure | 3 to 6.25 mg | ‐2.4 (‐3.7 to ‐1.2) | 662 (8) | ⊕⊕⊕⊝ moderate1 | Similar to the effect as a second‐line drug (Chen 2009). This is an indirect comparison of the effect size with a wider confidence interval compared to the Chen review, which is based on 23 trials in 3364 patients using similar inclusion/exclusion criteria, with diastolic blood pressure‐lowering of ‐1.7 (‐2.2 to ‐1.2) mmHg. |
| 12.5 mg | ‐3.1 (‐3.7 to ‐2.5) | 2877 (25) | ⊕⊕⊕⊕ | Similar to the effect as a second‐line drug (Chen 2009), with a narrow confidence interval based on a large sample size with a magnitude of lowering similar to the effect as a second‐line drug (Chen 2009), which was ‐3.1 (‐3.4 to ‐2.8) mmHg | |
| 25 mg | ‐3.3(‐3.8 to ‐2.8) | 3359 (29) | ⊕⊕⊕⊕ | Similar to the effect as a second‐line drug (Chen 2009). This is an indirect comparison of the effect size with a wider confidence interval compared to the Chen review, which is based on 42 trials in 6153 patients using similar inclusion/exclusion criteria with diastolic blood pressure‐lowering of ‐4.0 (‐4.4 to ‐3.6) mmHg | |
| 50 to 100 mg | ‐4.7 (‐6.1 to ‐3.3) | 345 (4) | ⊕⊕⊝⊝ | The 4 included studies had a high risk of bias. The confidence interval is very wide with a small sample size providing insufficient data in both this review as well as in the Chen review comparing the effect as a second‐line drug (Chen 2009) | |
| GRADE Working Group grades of evidence CI: confidence interval; MD: mean difference | |||||
| 1Downgraded due to the small number of patients and wide confidence intervals. | |||||
| Thiazide compared with placebo for primary hypertension | ||||
| Patient or population: adults with primary hypertension Settings: outpatient Intervention: all thiazides^ Comparison: placebo | ||||
| Outcomes | MD (95% CI) mmHg | No of participants | Quality of the evidence | Comments |
| Systolic blood pressure | ‐9.1 (‐9.7 to ‐8.5) | 7733 (47) | ⊕⊕⊕⊕ | At doses achieving maximal effect and above |
| Diastolic blood pressure | ‐3.6 (‐4.0 to ‐3.3) | 8064 (51) | ⊕⊕⊕⊕ | At doses achieving maximal effect and above |
| Withdrawal due to adverse effects | RR 0.64 (95% CI 0.43 to 0.93) | 3698 (20) | ⊕⊝⊝⊝ | See comments 1 and 2 |
| GRADE Working Group grades of evidence | ||||
| ^Includes thiazide and thiazide‐like diuretics. 1Based on a high risk of selective reporting of outcome from 20 out of 40 trials meeting the inclusion criteria. 2Withdrawals due to inclusion of an increase in blood pressure as an adverse effect (AE) was the major reason for withdrawals in the placebo group. | ||||
Background
Description of the condition
Hypertension is a common and potentially serious problem. It is one of the risk factors for stroke, heart and blood vessel disease, kidney disease and early death that can most easily be reduced by treatment. Studies show a correlation between elevation of systolic or diastolic blood pressure and increased risk of stroke, heart failure, renal disease and coronary heart disease. There is considerable evidence that antihypertensive drugs reduce death, stroke and heart disease when given to people with moderate to severe hypertension (Musini 2009a; Wright 1999; Wright 2009). However, the magnitude of blood pressure reduction does not always parallel a reduction in mortality or cardiovascular morbidity. Other factors independent of blood pressure reduction may contribute to the beneficial and harmful effects of drug treatment. Nonetheless, the magnitude of blood pressure reduction is often considered an important surrogate or indicator of the likelihood that people will benefit from drug treatment.
Description of the intervention
Thiazide diuretics were developed during the 1950s, when chemists and physiologists tested derivatives of sulfonamide‐based carbonic anhydrase inhibitors, with the goal of discovering drugs that enhance the excretion of sodium with chloride rather than sodium bicarbonate. Thiazide diuretics have been widely used as pharmacological agents for the treatment of hypertension for over five decades. The members of this drug class are derived from benzothiadiazine. Drugs with a similar pharmacologic action on the kidney that do not have the thiazide chemical structure, such as indapamide, chlorthalidone and metolazone, are termed 'thiazide‐like diuretics'. Metolazone is a quinazoline. Chemically, metolazone is not a substitute for benzothiadiazine but it, and other drugs such as indapamide, act on the same co‐transporter in the kidney as thiazides. Therefore, they are appropriately grouped with thiazide diuretics despite not being thiazides themselves (Edwin 2006). In this review, we use the term 'thiazide' to encompass thiazides and 'thiazide‐like' diuretics including the following drugs: hydrochlorothiazide, chlorothiazide, buthiazide, bendroflumethiazide, hydroflumethiazide, trichlormethiazide, methyclothiazide, polythiazide, cyclothiazide, cyclopenthiazide, chlorthalidone, metolazone, quinethazone, fenquizone, clorexolone, clopamide, indapamide, diapamide, isodapamide, mefruside and xipamide. Thiazide diuretics were originally marketed and prescribed in starting doses much higher than the average starting and maximum doses that are currently used for the treatment of hypertension (Edwin 2006).
How the intervention might work
Physiological studies in people show that the early effect of diuretic therapy is to decrease the extracellular volume, plasma volume and cardiac output with relatively unchanged peripheral resistance (Edwin 2006). After several weeks of therapy cardiac output returns to normal and total peripheral resistance decreases. At the level of the kidney, thiazides inhibit reabsorption of sodium and chloride ions from the distal convoluted tubules by blocking the thiazide‐sensitive Na+Cl‐ co‐transporter (Hughes 2004). Basic research studies demonstrate mechanisms whereby the different thiazide and thiazide‐like drugs might differ in their actions and effects (Kurtz 2010). However, none of these differential effects have been demonstrated in people (Campbell 2004).
The thiazide diuretics have a unique adverse effect profile.They potentially affect blood lipids, glucose, potassium, calcium, magnesium, uric acid and chloride concentrations. Thiazides potentially increase Ca2+ concentrations by increasing calcium reabsorption at the distal tubule. This is thought to be due to lowering of the sodium concentration within the epithelial cells, and thus increase of the activity of the Na+/Ca2+‐ATPase on the basolateral membrane to pump more Ca2+ into the interstitium. Thiazides are also thought to increase the reabsorption of Ca2+ by a mechanism involving the reabsorption of sodium and calcium in the proximal tubule in response to sodium depletion. Some of this response is thought to be due to augmentation of the action of parathyroid hormone. Thiazides do not affect potassium transport directly; instead, they stimulate potassium urinary secretion indirectly. Hypokalemia results primarily from increased Na and fluid delivery to the distal tubule with an enhanced aldosterone effect. Thiazides also enhance potassium secretion by activating flow‐sensitive maxi‐K channels; these channels are molecularly distinct from the potassium secretory channels (Edwin 2006).
Why it is important to do this review
A previous review showed that treatment of primary hypertension with different antihypertensive drug classes, compared with placebo or no treatment, decreased stroke but had varied effects on coronary heart disease and all‐cause mortality (Wright 2009). Thiazide trials were classified as low‐dose (hydrochlorothiazide equivalent of less than 50 mg/day) or high‐dose (hydrochlorothiazide equivalent of > 50 mg/day) based on the starting dose in each trial. All trials used stepped care therapy with drugs from other drug classes, aiming to achieve blood pressure targets of < 140/90 mmHg. Interestingly enough, despite a similar magnitude of blood pressure reduction at one year (13/5 mmHg with first‐line, low‐dose thiazides versus 14/7 mmHg with first‐line high‐dose thiazides), there were dose‐related differences in the impact on mortality and coronary heart disease. Surprisingly, high‐dose thiazides (mean dose 90 mg/day hydrochlorothiazide equivalent) reduced stroke but not all‐cause mortality or coronary heart disease significantly, whereas low‐dose thiazides (mean dose of 24 mg/day hydrochlorothiazide equivalent) decreased all‐cause mortality, stroke and coronary heart disease. Several different drugs from the thiazide or thiazide‐like class were used in these trials, including bendrofluazide, chlorothiazide, chlorthalidone, hydrochlorothiazide, hydrochlorothiazide/amiloride, hydrochlorothiazide/triamterene, indapamide, methyclothiazide and trichlormethiazide.
Although it is now well established that first‐line, low‐dose thiazides reduce mortality as well as morbidity, the dose‐related effect of thiazides on systolic as well as diastolic blood pressure, when used as first‐line single drugs, has not been established. Using excessive doses of thiazides may increase toxicity, by causing hypokalemia, hyponatraemia or other metabolic disturbances without additional blood pressure‐lowering effect or commensurate clinical benefit. We also cannot assume that all diuretics will have the same efficacy in reducing blood pressure. The different classes of diuretics and individual drugs within each class might have differing efficacy and adverse effects. It is important to know whether the blood pressure‐lowering effect of thiazides and thiazide‐like diuretics as a class is different from other classes and to know the blood pressure‐lowering dose‐response relationship in relation to other effects of thiazides, such as their metabolic adverse effects.
The main aim of this systematic review is to quantify the dose‐response relationship of thiazides in lowering blood pressure in patients with hypertension. The information derived from this review should facilitate future reviews of head‐to‐head comparisons with other drug classes and assist clinicians in choosing a specific thiazide drug at an appropriate dose.
Objectives
Primary objective
To determine the dose‐related decrease in systolic and/or diastolic blood pressure due to thiazide diuretics compared with placebo control in the treatment of patients with primary hypertension.
Secondary objectives
To determine the dose‐related adverse events leading to patient withdrawal and adverse biochemical effects on serum potassium, uric acid, creatinine, glucose and lipids.
Methods
Criteria for considering studies for this review
Types of studies
Study design must meet the following criteria: placebo‐controlled; random allocation to thiazide diuretic group and parallel placebo group; duration of follow‐up of at least three weeks; office blood pressure measurement at baseline (following wash‐out) and at one or more time points between 3 and 12 weeks after starting treatment.
We included data from cross‐over trials if the authors reported data for the initial treatment period versus the parallel placebo control group followed by an adequate wash‐out period before crossing over to the other active treatment and if data were reported in a similar manner during the second treatment period.
Exclusion criteria
We excluded any of the following: non‐randomized trials; single‐blind as well as open‐label trials; trials using a thiazide diuretic in combination with other classes of drugs as first‐line treatment; abstracts without a complete trial report; trials reporting placebo blood pressure levels following a wash‐out period and comparing them with the treatment levels following randomization; and trials in patients with secondary causes of hypertension.
Types of participants
Adults (18 years or older) with a baseline systolic blood pressure of at least 140 mmHg or a diastolic blood pressure of at least 90 mmHg, measured in a standard way. We excluded trials in which patients had significant renal insufficiency and a documented serum creatinine level more than 1.5 times the normal range from analysis. Participants were not further restricted by age, sex, cardiovascular baseline risk or any other co‐morbid conditions.
Types of interventions
We included monotherapy with any of the following thiazide and thiazide‐like diuretics: bendroflumethiazide, buthiazide, chlorthalidone, chlorothiazide, clopamide, clorexolone, cyclopenthiazide, cyclothiazide, diapamide, fenquizone, hydrochlorothiazide, hydroflumethiazide, indapamide, isodapamide, mefruside, methyclothiazide, metolazone, polythiazide, quinethazone, trichlormethiazide and xipamide. These drugs are referred to as 'thiazide diuretics' in this review for simplicity.
We did not include data from trials in which the thiazide was titrated to a higher dose in a subset of randomized patients to achieve target blood pressure levels because dose‐response relationships cannot be analyzed if patients within each randomized group are taking different doses of the same drug.
If all the patients in the trial were given a forced titrated dose irrespective of the blood pressure recorded, then we included the trial under the highest dose given.
Potassium supplementation was allowed in patients with low serum potassium levels.
Types of outcome measures
Primary outcomes
Change in systolic and diastolic blood pressure compared with placebo. If blood pressure measurements were available at more than one time during the 24‐hour period, we used the trough measurement. We defined peak level as within 12 hours of the dose and trough level as between 12 and 24 hours. If several blood pressure measurements were available within the acceptable window, we used the weighted mean of all blood pressure measurements reported by the study authors during the 3 to 12‐week range as the best estimate of treatment effect.
Secondary outcomes
-
Patient withdrawals due to adverse effects compared with placebo.
-
Change in the concentration of serum potassium, uric acid, creatinine, glucose and lipids compared with placebo. If several measurements were available within the acceptable window, then we used the weighted mean of all measurements reported by the study authors during the 3 to 12‐week range as the best estimate of treatment effect.
Search methods for identification of studies
Electronic searches
We searched the following electronic databases for primary studies: the Cochrane Central Register of Controlled Trials (CENTRAL 2014, Issue 1), Ovid MEDLINE (1946 to February 2014), Ovid EMBASE (1974 to February 2014), ClinicalTrials.gov and reference lists of included studies.
We searched the electronic databases using a strategy combining the Cochrane Highly Sensitive Search Strategy for identifying randomized trials in MEDLINE: sensitivity‐maximizing version (2008 revision) with selected MeSH terms and free‐text terms relating to the individual thiazide drugs and hypertension. We used no language restrictions. We used the standard search strategy of the Cochrane Hypertension Review Group with additional terms related to thiazide diuretics in the above listed databases to identify relevant articles. We translated the MEDLINE search strategy (Appendix 1) into EMBASE (Appendix 2) and CENTRAL (Appendix 3) using the appropriate controlled vocabulary as applicable.
Searching other resources
We also searched Web of Science and bibliographic citations. In case of incomplete reports, we used MEDLINE to search for related papers and contacted authors to retrieve missing information. We searched the bibliographies of pertinent articles, reviews and texts for additional citations. We used previously published meta‐analyses on the dose response of thiazide diuretics, as well as narrative reviews, to help identify references to trials. We assessed trials included in the Law et al systematic review and meta‐analysis for eligibility for this review (Law 2009). Several trials included in the Law meta‐analysis do not meet the inclusion criteria for this review and the reasons for exclusion are listed under Characteristics of excluded studies. We applied no language restrictions.
Data collection and analysis
Selection of studies
We screened all potentially relevant articles and rejected articles on the initial screen if the title or abstract was not a report of a randomized, placebo‐controlled trial or if it did not meet the minimum inclusion criteria. We retrieved the full texts of the remaining articles. Two independent review authors (VM and MN) assessed the eligibility of the trials using a trial selection form. The third review author (CJ) assessed the eligibility of non‐English trials. JMW or KB resolved discrepancies. We counted trials with more than one publication only once. See Figure 1.
Study flow diagram.
Data extraction and management
Once it was determined that the trials met the inclusion criteria, two independent review authors (VM and MN) abstracted data for all primary and secondary outcomes using a standard form, and then cross‐checked. If data were presented numerically (tables or text) or graphically (in figures), we preferred the numeric data because of possible measurement errors when estimating from the graphs. Both review authors (VM and MN) cross‐checked all numeric calculations and graphic interpolations.
The position of the patient during blood pressure measurement may affect the blood pressure‐lowering effect. However, in order not to lose valuable data, if only one position was reported, we collected data from that position. When blood pressure measurement data were available in more than one position, sitting blood pressure was the first preference, followed by standing and supine blood pressure measurements.
We extracted data for the following outcome measures:
-
The trough and/or peak systolic and diastolic blood pressure at baseline following the wash‐out period.
-
The trough and/or peak systolic and diastolic blood pressure measurement taken between 3 and 12 weeks of the treatment period. If more than one blood pressure measurement was available then we used the weighted mean blood pressure levels.
-
The number of patient withdrawals due to adverse effects for each drug dose during the specified period of time the patient is taking the drug.
-
The blood concentrations of serum potassium, uric acid, creatinine, glucose and lipids with standard deviation (SD) at baseline as well as between 3 and 12 weeks during the treatment period. If more than one measurement was available between 3 and 12 weeks then we used the weighted mean level.
Assessment of risk of bias in included studies
Two independent review authors (VM and MN) assessed the risk of bias of the included studies to create 'Risk of bias' tables as outlined in the Cochrane Handbook for Systematic Reviews of Interventions (Handbook 2011). We resolved any discrepancies between the review authors by discussion with a third review author (JW or KB).
We evaluated the following items: randomization; allocation concealment; blinding; incomplete outcome data reporting; selective reporting and other biases (e.g. industry sponsorship).
Measures of treatment effect
For continuous outcomes, we combined data for placebo‐corrected systolic and diastolic blood pressure reduction and for placebo‐corrected serum concentrations of potassium, uric acid, creatinine, glucose and lipids using a mean difference (MD) method, presented with 95% confidence intervals (CI). Blood pressure data are presented as systolic /diastolic blood pressure, with parentheses for 95% CI with accuracy up to one decimal point. Metabolic data are presented as mean difference with parentheses for 95% CI, with accuracy up to two decimal points.
For the dichotomous outcome withdrawals due to adverse effects, data are presented as risk ratio (RR) with 95% CI. We have not provided absolute risk difference and number needed to treat to benefit or harm due to very low‐quality evidence for the outcome 'withdrawal due to adverse effects' due to high risk of selective reporting bias.
Unit of analysis issues
For cross‐over trials that met the inclusion criteria, we used the first parallel‐group period when patients were randomized to thiazide diuretics or placebo in data analyses. For dose‐ranging trials that compared a single placebo group to several different doses, the forest plot includes the number of patients adjusted for placebo‐controlled treatment group in these trials for an accurate overall effect across all thiazide trials.
Dealing with missing data
In order to address missing data we attempted to contact the study's authors using the first author firstly then any of the co‐authors. We used the publication year 1990 as a cut‐off for verifying the authors' contact information. We did not check anything older than that. Most contacts were email addresses; the rest were telephone numbers, fax numbers or business addresses. We were not successful in obtaining additional data despite our efforts. In case missing information was not available, we included the best estimate based on the information in the same trial or from other trials using the same dose.
In case of missing SD for the change in blood pressure, we imputed the SD based on the information in the same trial or from other trials using the same dose. We used the following hierarchy (listed from high to low preference) to impute SD values:
-
SD calculated either from the t statistics corresponding to the exact P value reported or from the 95% CI of the mean difference between treatment groups.
-
SD of change in blood pressure from a different position than that of the blood pressure data used.
-
SD of blood pressure at the end of treatment.
-
SD of blood pressure at the end of treatment measured from a different position than that of the blood pressure data used.
-
SD of blood pressure at baseline (except if this measure is used for entry criteria) (Musini 2009b).
-
Mean SD of change in blood pressure from other trials using the same drug and dose.
-
Mean weighted SD of change available from all other trials using the same class of drug at any dose.
Assessment of heterogeneity
We calculated the heterogeneity of treatment effects between the trials using a standard Chi2 test in RevMan 5.2 (RevMan 2012). We applied the fixed‐effect model to obtain summary statistics of pooled trials, unless significant between‐study heterogeneity was present, in which case we used the random‐effects model. If a statistically significant difference compared to placebo control was still present using the random‐effects model, then we reported the fixed pooled estimate and CI because of the tendency for smaller trials, which are susceptible to publication bias, to be over‐weighted with a random‐effects analysis. We compared overall effect size using both a fixed‐effect and random‐effects model and determined that they were not significantly different from each other.
If the calculated I2 statistic value was greater than 50%, we explored the reasons for heterogeneity and the trials contributing to the heterogeneity (differences in baseline characteristics between trials and their possible impact on the magnitude of systolic and diastolic blood pressure reduction).
Data synthesis
We carried out data synthesis and analyses using the Cochrane RevMan 5.2 software (RevMan 2012). We directly compared the effect size between doses for each thiazide diuretic drug only using data from trials that randomized participants to different doses of the drug within the same trial. In case direct comparison was not possible, we did an 'adjusted indirect comparison' using the computer software developed by Wells 2009. We considered a P value less than 0.05 (P < 0.05) to be statistically significant for all comparisons.
Subgroup analysis and investigation of heterogeneity
We planned subgroup analyses at the protocol stage based on age (18 to 59 years versus 60 years or older), sex, race, comorbid conditions and severity of blood pressure at baseline. Due to a lack of available data, we could only perform subgroup analyses based on classifying trials reporting mean baseline severity of blood pressure level.
Sensitivity analysis
We intended to test the robustness of the results using sensitivity analyses (high versus low trial quality; fixed‐effect versus random‐effects model; position of blood pressure measurement; trough versus peak blood pressure measurement; reported versus imputed SD and industry versus non‐industry‐sponsored trials). Sufficient data were available to perform sensitivity analyses for overall thiazide diuretics compared to placebo including three drugs: chlorthalidone (14 RCTs were included), hydrochlorothiazide (44 RCTs were included) and indapamide (12 RCTs were included) for systolic and diastolic blood pressure‐lowering effects.
Results
Description of studies
Results of the search
The search strategy identified 18,293 citations; after de‐duplicating we screened 11,824 citations. Of these, 75 (0.7% of screened articles) were double‐blind, randomized, placebo‐controlled trials meeting the minimum inclusion criteria. Fifty‐nine RCTs were of a parallel‐group design yielding data that could be used in the evaluation of dose‐related blood pressure and metabolic effects of six thiazide diuretics (bendrofluazide, chlorthalidone, cyclopenthiazide, hydrochlorothiazide, indapamide and metolazone). See the PRISMA diagram (Figure 1).
Sixteen of the 75 RCTs were cross‐over trials, of which we excluded 15 because data were not reported for the initial parallel period. The single cross‐over trial which provided separate data for periods one and two compared indapamide 2.5 mg/day with placebo in 24 patients for a duration of eight weeks (Soltero 1989). See Characteristics of excluded studies.
Table 1: Included trials sorted according to the year of publication
| Year of publication | Trials included in meta‐analyses |
| 1946 to 1949 | 0 |
| 1950 to 1959 | 0 |
| 1960 to 1969 | 0 |
| 1970 to 1979 | 2 |
| 1980 to 1989 | 13 |
| 1990 to 1999 | 34 |
| 2000 to 2009 | 11 |
| 2010 to 2012 | 0 |
| Overall | 60 trials |
Forty‐nine of the 60 included trials (82%) were published before the year 2000.
Included studies
Please refer to Characteristics of included studies for details of each of the 60 included trials. Studies included adult patients with systolic blood pressure entry criteria of 140 mmHg or more and/or diastolic blood pressure entry criteria of 90 mmHg or more. Co‐morbidities were not reported in most trials. Across all 60 trials, the total number of randomized patients was 11,282; mean age was 55 years; mean blood pressure was 158/99 mmHg; and mean duration of treatment was eight weeks. Of the total population, 53% of patients were male and 47% were female.
Table 2: Overall summary of the 60 trials meeting the inclusion criteria
| Thiazide drug | Dose range | Number of trials | Total patients randomized | N (% males) and N (% females) | Mean duration (weeks) | Mean age (years) | Baseline systolic/diastolic blood pressure mmHg |
| Bendrofluazide | 1.25 to 10.0 mg/day | 1 | 257 | 103 (40%) 154 (60%) | 12 weeks | 57 years | 165/102 mmHg |
| Chlorthalidone | 12.5 to 100 mg/day | 8* | 1265 | 581 (50.4%) 684 (49.6%) | 7 weeks | 53 years | 163/88 mmHg |
| Cyclopenthiazide | 0.05 to 0.50 mg/day | 1 | 53 | 22 (41.5%) 31 (49.6%) | 8 weeks | 57 years | 164/97 mmHg |
| Hydrochlorothiazide | 3 to 100 mg/day | 40 | 7560 | 4152 (57%) 3408 (43%) | 8 weeks | 54 years | 155/100 mmHg |
| Indapamide | 1 to 5 mg/day | 10 | 2075 | 1018 (48.4%) 1057 (51.6%) | 10 weeks | 58 years | 161/98 mmHg |
| Metolazone | 0.5 to 2.0 mg/day | 1 | 105 | 46 (43.8%) 59 (56.2%) | 6 weeks | Not reported | 150/98 mmHg |
| Overall |
| 60 trials | 11,282* | 5922 (53.0%) 5360 (47.0%) | 8 weeks | 55 years | 158/99 mmHg |
*The Siegel 1992 RCT has both hydrochlorothiazide 50 mg/day and chlorthalidone 50 mg/day treatment arms therefore it is counted once in the overall included trial total. Also, the 33 placebo patients in the Siegel 1992 study are counted only once in the overall total.
For bendrofluazide , one randomized, double‐blind, parallel‐group, placebo‐controlled trial met the inclusion criteria (Carlsen 1990). Two hundred and fifty‐seven patients were randomized, with a mean age of 57 years and a mean baseline blood pressure of 165/102 mmHg; the percentage of male participants was 40% and of female participants 60%. The duration of treatment was 12 weeks.
For chlorthalidone , eight randomized, double‐blind, placebo‐controlled, parallel‐group trials met the inclusion criteria. There were 1265 randomized patients, with a mean age of 53 years and a mean baseline blood pressure of 163/88 mmHg; the percentage of male participants was 50% and of female participants 50%. The mean duration of treatment was seven weeks.
For cyclopenthiazide , one randomized, double‐blind, placebo‐controlled, parallel‐group trial met the inclusion criteria (McVeigh 1988). Fifty‐three patients were randomized, with a mean age 57 years and a mean baseline blood pressure of 164/97 mmHg; the percentage of male participants was 42% and of female participants 58%. The duration of treatment was eight weeks.
Forhydrochlorothiazide , 40 randomized, double‐blind, placebo‐controlled, parallel‐group trials met the inclusion criteria. There were 7560 randomized patients, with a mean age of 54 years and a mean baseline blood pressure of 155/100 mmHg; the percentage of male participants was 55% and of female participants 45%. The mean duration of treatment was eight weeks.
Forindapamide , 10 randomized, double‐blind, placebo‐controlled, parallel‐group trials met the inclusion criteria. There were 2075 randomized patients, with a mean age of 58 years and a mean baseline blood pressure of 161/98 mmHg; the percentage of male participants was 49% and of female participants 51%. The mean duration of treatment was 10 weeks.
For metolazone , one randomized, double‐blind, placebo‐controlled, parallel‐group trial met the inclusion criteria (Curry 1986). There were 105 randomized patients; mean age was not reported and the baseline blood pressure was 150/98 mmHg; the percentage of male participants was 44% and of female participants 56%. The duration of treatment was six weeks.
Excluded studies
We excluded 86 studies. See Characteristics of excluded studies for details. Reasons for exclusion include: not a randomized trial; had no parallel, placebo‐controlled treatment arm; used combination therapy; cross‐over trial design with no wash‐out between treatment periods either reporting data before the minimum three weeks duration period or not reporting data at the end of the first parallel placebo treatment period; improper blood pressure measurement; stepped up therapy only in non‐responders (not meeting the target goal blood pressure levels) and dose was not titrated in all randomized patients. Articles were also excluded if full‐text reports were not available or data were reported in a way that could not be used for analysis.
Risk of bias in included studies
We evaluated each trial that provided data for at least one of the outcome measures using the Cochrane 'Risk of bias' tool. See Figure 2 and Figure 3.
'Risk of bias' summary: 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.
For details of the following seven factors evaluated for risk of bias in each individual study, see the 'Risk of bias' tables in Characteristics of included studies.
Random sequence generation (selection bias)
We determined nine of the 60 included trials (15%) to have adequate random sequence generation (Chrysant 1996; Goldberg 1989; Jounela 1994; McVeigh 1988; McGill 2001; Morledge 1986; Papademetriou 2006; Schmieder 2009; Siegel 1992). We judged 49 (82%) of the trials as at unclear risk of bias as the technique of randomization was not reported and in one trial we judged reporting of randomization as at high risk of bias (Mroczek 1996).
Allocation
We judged four of the 60 trials (7%) as at low risk of bias for allocation concealment (Hulley 1985; Papademetriou 2006; Pool 1993; Soltero 1989). The remaining 56 trials (93%) did not report on how allocation concealment was performed and therefore we judged them as at unclear risk of bias, probably resulting in high risk of selection bias.
Blinding
We judged 11 of the 60 trials (18%) as at low risk of bias since they adequately described how patients and physicians and outcome assessors were blinded (Benz 1998; Bradley 1993; Brown 1990; Carlsen 1990; Chrysant 1996; Frishman 1994; Goldberg 1989; Hulley 1985; Jounela 1994; Materson 1978; McVeigh 1988). We judged three trials as at high risk of bias (Fernandez 1994; Fiddes 1997; Krantz 1988). We judged the remaining 46 trials (77%) as at unclear risk of bias due to lack of reporting.
Incomplete outcome data
We judged 19 of the 60 trials (32%) as at low risk of bias since they adequately described total withdrawals (which were less than 20% of the total randomized patients) and how these patients were accounted for in the analysis (Ambrosioni 1998; Brown 1990; Burris 1990; Canter 1994; Chrysant 1996; Drayer 1995; Fiddes 1997; Frei 1994; Frishman 1995; Jounela 1994; London 2006; McVeigh 1988; Moser 1991; Persson 1996; Schmieder 2009; Scholze 1993; Soltero 1989; Vardan 1987; Weir 1992). We judged 17 (28%) trials as at high risk of bias (Benz 1998; Capone 1983; Carlsen 1990; Chrysant 2004; Curry 1986; Ferrara 1984; Frishman 1994; Goldberg 1989; Hulley 1985; Lawton 1979; Morledge 1986; Papademetriou 2000; Philipp 1997; Prisant 2000; Roque 1996; Taylor 1988; Weidler 1995). We judged the remaining 24 trials (40%) as at unclear risk of bias.
Selective reporting
Systolic and diastolic blood pressure data were provided in all trials and therefore they were not subject to selective reporting bias. For other outcomes, we judged one out of 60 trials (2%) to be at low risk of bias since it adequately described all outcome measures specified in the publication (Myers 2000). We judged 21 of the 60 trials (35%) as at unclear risk and the remaining 38 trials (63%) as at high risk of bias since they did not report on outcomes such as total adverse events, withdrawal due to adverse effects or metabolic data even though these were collected, according to the methods sections of the publications.
Other potential sources of bias
A factor considered as another potential sources of bias was the source of funding for each included study. We judged 14 of the 60 trials (23%) as at low risk of other bias (Ambrosioni 1998; Benz 1998; Chrysant 1994; Chrysant 2004; Frishman 1994; Frishman 1995; Hulley 1985; Kochar 1999; London 2006; McGill 2001; Schmieder 2009; Schoenberger 1995; Scholze 1993; Vardan 1987). We judged nine trials (15%) as at high risk of other bias (Bradley 1993; Brown 1990; Curry 1986; Ferrara 1984; Fiddes 1997; Frei 1994; Lucas 1985; Papademetriou 2000; Prisant 2000), and 37 trials (62%) as at unclear risk of other bias. For details, please see the 'Risk of bias' tables in Characteristics of included studies.
Of all 60 included studies, we only judged six (10%) as at low risk of bias in at least three of the six items evaluated using the 'Risk of bias' tool (Chrysant 1996; Jounela 1994; McVeigh 1988; Papademetriou 2006; Schmieder 2009; Vardan 1987). We judged the other 54 included trials (90%) as at unclear or high risk of bias. Thus, the overall evidence in this review is subject to a high risk of bias. This has to be taken into consideration in interpreting the findings. It is unclear what effects residual bias may have on our estimate of the efficacy of systolic and diastolic blood pressure reduction with thiazides, but it is likely that their harms are underestimated, including withdrawal due to adverse effects and adverse or potentially adverse metabolic changes.
Funding bias
Twenty‐eight (47%) of the 60 included studies were industry‐sponsored but in 26 of the 28 the bias would have been in favor of the other drug being tested and not in favor of the thiazide. Only five trials (8%) were sponsored by national agencies such as the National Institutes of Health, National Institute of Aging, National Heart and Lung Institute etc. and all of these were older trials and studied high doses of chlorthalidone and hydrochlorothiazide. For the remaining 26 trials (42.4%), the source of funding was not reported. See the 'Risk of bias' tables in Characteristics of included studies.
Publication bias
Publication bias is defined in this review as selective publication of studies with positive results and is another source of bias that may have skewed the results of this review. Examining the funnel plots for systolic and diastolic blood pressure in this review suggested asymmetry, indicating that there was a high risk of publication bias (Figure 4).
Funnel plot of comparison: 7 Thiazide versus placebo, outcome: 7.1 Systolic blood pressure.
Subgroup analyses based on risk of bias
In view of the fact that it was not possible to predict the direction of the bias in the industry‐funded trials and the trials where the source of funding was not reported, plus the higher doses used in the trials sponsored by national agencies, we did not attempt any subgroup comparisons.
Effects of interventions
See: Summary of findings for the main comparison Dose‐ranging blood pressure‐lowering efficacy of hydrochlorothiazide for primary hypertension; Summary of findings 2 Overall effects of thiazides for primary hypertension
Dose‐related systolic and diastolic blood pressure‐lowering
Three thiazide diuretics (bendrofluazide, cyclopenthiazide and metolazone) had only one trial each that met the minimum inclusion criteria. Therefore, there were insufficient data to present forest plots for these three drugs.
Bendrofluazide
Bendrofluazide dose ranged from 1.25 to 10 mg/day as monotherapy versus placebo control for a mean duration of treatment of 12 weeks in 257 patients in one trial (Carlsen 1990). See Analysis 1.1; Analysis 1.2.
Table 3: Dose‐related systolic and diastolic blood pressure‐lowering efficacy of bendrofluazide
| Bendrofluazide Carlsen 1990 | Systolic blood pressure mmHg (95% CI) | Diastolic blood pressure mmHg (95% CI) |
| 1.25 mg/day | ‐7.7 (‐15.0 to ‐0.4) | ‐5.8 (‐10.5 to ‐1.1) |
| 2.5 mg/day | ‐10.9 (‐18.1 to ‐3.7) | ‐6.9 (‐11.6 to ‐2.2) |
| 5 mg/day | ‐10.6 (‐17.8 to ‐3.4) | ‐6.2 (‐10.9 to ‐1.5) |
| 10 mg/day | ‐12.5 (‐19.8 to ‐5.2) | ‐7.0 (‐11.7 to ‐2.3) |
| Overall | ‐10.4(‐14.1 to ‐6.8) | ‐6.5 (‐8.8 to ‐4.1) |
The lowest dose of bendrofluazide that showed a statistically significant blood pressure‐lowering was 1.25 mg/day for both systolic and diastolic blood pressure. The overall placebo‐corrected systolic blood pressure‐lowering effect size across 1.25 to 10 mg/day doses was ‐10.4 mmHg (P value < 0.00001, with I2 = 0% and test for subgroup differences P value = 0.87, with I2 = 0%) and for diastolic blood pressure was ‐6.5 mmHg (P value < 0.00001, with I2 = 0% and test for subgroup differences P value = 0.98, with I2 = 0%). Direct comparison of doses in the Carlsen 1990 dose‐ranging trial showed no significant differences in systolic or diastolic blood pressure‐lowering between the different doses used.
The placebo‐corrected systolic/diastolic blood pressure‐lowering with bendrofluazide was 10.4/6.5 mmHg.
Cyclopenthiazide
Cyclopenthiazide doses ranged from 50 to 500 µg/day and were administered as monotherapy versus placebo control over eight weeks' duration in one trial (McVeigh 1988). See Analysis 2.1; Analysis 2.2.
Table 4: Dose‐related systolic and diastolic blood pressure‐lowering efficacy of cyclopenthiazide
| Cyclopenthiazide McVeigh 1988 | Systolic blood pressure mmHg (95% CI) | Diastolic blood pressure mmHg (95% CI) |
| 0.05 mg/day | ‐5.3 (‐18.7to 8.1) | ‐3.0 (‐11.7 to 5.7) |
| 0.125 mg/day | ‐12.0 (‐25.2 to 1.2) | ‐8.6 (‐17.1 to ‐0.1) |
| 0.5 mg/day | ‐14.9 (‐28.3 to ‐1.5) | ‐7.0 (‐15.7 to 1.7) |
| Overall | ‐10.8 (‐18.4 to ‐3.1) | ‐6.2 (‐11.2 to ‐1.3) |
The lowest dose of cyclopenthiazide that showed a statistically significant systolic blood pressure‐lowering was 500 µg/day. Cyclopenthiazide 0.125 mg/day showed a significant difference from placebo in lowering diastolic blood pressure. The overall placebo‐corrected systolic blood pressure‐lowering effect size across 0.05 to 0.5 mg/day doses was ‐10.8 mmHg (P value = 0.006, with I2= 0% and test for subgroup differences P value = 0.60, with I 2= 0%). For diastolic blood pressure it was ‐6.2 mmHg (P value = 0.01, with I 2 = 0% and test for subgroup differences P value = 0.65, with I 2= 0%). Direct comparison between doses showed no demonstrable dose response.
The placebo‐corrected systolic/diastolic blood pressure‐lowering with cyclopenthiazide was 10.8/6.2 mmHg.
Metolazone
Metolazone doses ranged from 0.5 to 2 mg/day and were administered as monotherapy versus placebo control over six weeks' duration in one trial (Curry 1986). See Analysis 3.1; Analysis 3.2.
Table 5: Dose‐related systolic and diastolic blood pressure‐lowering efficacy of metolazone
| Metolazone Curry 1986 | Systolic blood pressure mmHg (95% CI) | Diastolic blood pressure mmHg (95% CI) |
| 0.5 mg/day | ‐11.4 (‐20.5 to ‐2.3) | ‐5.9 (‐11.8 to ‐0.0) |
| 1.0 mg/day | ‐11.6 (‐20.8 to ‐2.5) | ‐6.4 (‐12.3 to ‐0.5) |
| 2 mg/day | ‐11.9 (‐21.0 to ‐2.9) | ‐5.2 (‐11.0 to 0.6) |
| Overall | ‐11.6 (‐16.9 to ‐6.4) | ‐5.8 (‐9.2 to ‐2.4) |
The lowest dose of metolazone that showed a statistically significant systolic and diastolic blood pressure‐lowering was 0.5 mg/day. The overall placebo‐corrected systolic blood pressure‐lowering effect across all doses was ‐11.6 mmHg (P value < 0.0001, with I2= 0% and the test for subgroup differences P value = 1.00, with I2 = 0%). For diastolic blood pressure it was ‐5.8 mmHg (P value = 0.0007, with I2 = 0% and the test for subgroup differences P value = 0.96, with I2 = 0%). Direct comparison of doses did not show any significant differences in systolic or diastolic blood pressure‐lowering between the different doses used.
The placebo‐corrected systolic/diastolic blood pressure‐lowering with metolazone was 11.6/5.8 mmHg.
Three thiazides: chlorthalidone, hydrochlorothiazide and indapamide had sufficient trials to pool data in meta‐analyses and conduct subgroup and sensitivity analyses.
Chlorthalidone
Seven trials met the inclusion criteria, enrolling a total of 1297 patients who were treated for a mean duration of seven weeks (Bradley 1993; Ferrara 1984; Hulley 1985; Lawton 1979; Materson 1978; Morledge 1986; Vardan 1987). These trials compared chlorthalidone doses ranging from 12.5 to 75 mg/day to a placebo control. See Analysis 4.1; Analysis 4.2.
Table 6: Dose‐related systolic and diastolic blood pressure‐lowering efficacy of chlorthalidone
| Chlorthalidone | Systolic blood pressure mmHg (95% CI) | Diastolic blood pressure mmHg (95% CI) |
| 12.5 to 15 mg/day | ‐10.1 (‐13.9 to ‐6.3) | ‐2.6 (‐5.1 to ‐0.0) |
| 25 mg/day | ‐13.6 (‐16.0 to ‐11.3) | ‐4.0 (‐5.7 to ‐2.3) |
| 50 mg/day | ‐9.9 (‐13.4 to ‐6.4) | ‐4.9 (‐7.3 to ‐2.5) |
| 75 mg/day | ‐12.9 (‐24.7 to ‐1.2) | ‐5.5 (‐13.1 to 2.1) |
| Overall | ‐12.0 (‐13.7 to ‐10.2) | ‐3.9 (‐5.1 to ‐2.7) |
The lowest dose of chlorthalidone that showed a statistically significant blood pressure‐lowering was 12.5 to 15 mg/day. Based on seven RCTs, the overall placebo‐corrected systolic blood pressure‐lowering effect for 12.5 to 75 mg/day doses was ‐12.0 mmHg (P value < 0.00001, with I 2 = 0%, test for subgroup differences P value = 0.24, with I2 = 27.8%). For diastolic blood pressure it was ‐3.9 mmHg (P value < 0.00001, with I2 = 0%, test for subgroup differences P value = 0.58, with I 2 = 0%).
Dose‐related systolic blood pressure‐lowering response of chlorthalidone by direct comparison
Two dose‐ranging trials allowed direct dose comparisons to be performed (N = 276) (Materson 1978; Morledge 1986). Chlorthalidone 25 mg/day did not lower systolic blood pressure more than 12.5 mg/day. Chlorthalidone 75 mg/day did not lower systolic blood pressure more than 50 mg/day, and 50 mg/day did not lower systolic blood pressure more than 25 mg/day or 12.5 to 15 mg/day. Therefore the maximum systolic blood pressure‐lowering dose of chlorthalidone is likely to be 12.5 mg/day, which lowered systolic blood pressure by ‐10.1 (95% CI ‐13.9 to ‐6.3) mmHg. Systolic blood pressure‐lowering efficacy at doses > 12.5 mg/day was ‐12.0 (95% CI ‐13.7 to ‐10.2) mmHg.
Dose‐related diastolic blood pressure‐lowering response of chlorthalidone by direct comparison
Two dose‐ranging trials allowed direct dose comparisons to be performed (N = 276) (Materson 1978; Morledge 1986). Both chlorthalidone 75 mg and the 50 mg/day dose did not significantly lower diastolic blood pressure compared to 12.5 mg/day. However, 25 mg/day significantly lowered diastolic blood pressure more than 12.5 mg/day by ‐2.2 (95% CI ‐4.3 to ‐0.2) mmHg. Chlorthalidone 75 mg/day dose was not significantly different from 50 mg/day dose nor was 50 mg dose significantly different from 25 mg/day. Therefore chlorthalidone 25 mg/day is likely to result in maximum diastolic blood pressure‐lowering efficacy. The maximum diastolic blood pressure‐lowering efficacy at doses > 25 mg/day was ‐4.3 (95% CI ‐5.7 to ‐3.0) mmHg.
Thus the dose to produce maximal systolic/diastolic blood pressure‐lowering with chlorthalidone is between 12.5 and 25 mg/day with an average blood pressure reduction of 12.0/3.9 mmHg.
Hydrochlorothiazide
Forty trials of hydrochlorothiazide met the inclusion criteria, with doses ranging from 3 to 100 mg/day for a mean treatment duration of eight weeks.
Since there were 15 dose‐ranging trials that compared a single placebo group to several different doses, the forest plot includes the number of patients adjusted for placebo‐controlled treatment group in these trials for an accurate overall effect across all thiazide trials (Benz 1998; Canter 1994; Chrysant 1994; Goldberg 1989; Jounela 1994; Kochar 1999; Lucas 1985; McGill 2001; Papademetriou 2000; Papademetriou 2006; Philipp 1997; Pool 1997; Pool 2007; Scholze 1993; Villamil 2007). See Analysis 5.1; Analysis 5.2.
Table 7: Dose‐related systolic and diastolic blood pressure‐lowering efficacy of hydrochlorothiazide
| Hydrochlorothiazide | Systolic blood pressure mmHg (95% CI) | Diastolic blood pressure mmHg (95% CI) |
| 3 to 6.25 mg/day | ‐3.6 (‐5.6 to ‐1.5) | ‐2.4 (‐3.7 to ‐1.2) |
| 12.5 mg/day | ‐6.3 (‐7.2 to ‐5.3) | ‐3.1 (‐3.7 to ‐2.5) |
| 25 mg/day | ‐8.0 (‐9.0 to ‐7.0) | ‐3.3 (‐3.8 to ‐2.8) |
| 37.5 mg/day | ‐7.3 (‐16.3 to 1.7) | ‐3.7 (‐9.3 to 1.9) |
| 50 mg/day | ‐10.5 (‐14.6 to ‐6.4) | ‐5.0 (‐6.7 to ‐3.3) |
| 100 mg/day | ‐9.9 (‐14.1 to ‐5.8) | ‐3.9 (‐6.6 to ‐1.2) |
| Overall | ‐6.9 (‐7.6 to ‐6.3) | ‐3.3 (‐3.6 to ‐2.9) |
The lowest dose of hydrochlorothiazide (3 to 6.25 mg/day) statistically significantly reduced both systolic and diastolic blood pressure in eight trials with 663 patients. Based on 33 trials, for a mean duration of treatment of eight weeks, the overall placebo‐corrected systolic blood pressure‐lowering effect size across 3 to 100 mg/day doses was ‐6.9 mmHg in 6725 patients (P value < 0.00001, with I2 = 21% and test for subgroup differences P value = 0.0005, with I2 = 77.4%). For diastolic blood pressure it was ‐3.3 mmHg in 7284 patients (P value < 0.00001, with I 2 = 8% and test for subgroup differences P value = 0.29, with I2 = 19.6%). The significant heterogeneity is explained by the significant differences in systolic blood pressure‐lowering between doses.
Dose‐related systolic blood pressure‐lowering response of hydrochlorothiazide by direct comparison
Hydrochlorothiazide 12.5 mg lowered systolic blood pressure more than hydrochlorothiazide 3 to 6.25 mg/day by ‐2.2 (95% CI ‐3.8 to ‐0.6) mmHg based on seven dose‐ranging trials in 920 patients (P value = 0.008; heterogeneity was not significant, with I 2 = 0%) (Canter 1994; Jounela 1994; Kochar 1999; McGill 2001; Papademetriou 2006; Pool 1997; Villamil 2007). Hydrochlorothiazide 25 mg/day lowered systolic blood pressure more than hydrochlorothiazide 12.5 mg/day by ‐2.7 (95% CI ‐3.9 to ‐1.5) mmHg based on 14 trials in 2019 patients (P value < 0.0001; heterogeneity was significant: P value = 0.007, with I 2 = 55%). Doses of 50 and 100 mg/day appeared to lower systolic blood pressure more (> 10 mmHg) but because the systolic blood pressure reduction at these doses was not statistically significantly greater than the 25 mg/day dose, hydrochlorothiazide 25 mg/day was chosen as the lowest dose with maximum systolic blood pressure‐lowering efficacy. The systolic blood pressure‐lowering efficacy at doses > 25 mg/day was ‐8.2 (95% CI ‐9.1 to ‐7.3) mmHg in 3417 patients (P value < 0.0001, and no significant heterogeneity was present, with I 2 = 25%; test of subgroup differences P value = 0.57, with I2 = 0%).
Dose‐related diastolic blood pressure‐lowering response of hydrochlorothiazide by direct comparison
Hydrochlorothiazide 12.5 mg/day significantly reduced diastolic blood pressure compared with 3 to 6.25 mg/day by ‐1.1 (95% CI ‐2.1 to ‐0.1) mmHg based on seven trials in 920 patients. Hydrochlorothiazide 25 mg significantly reduced diastolic blood pressure more than hydrochlorothiazide 12.5 mg/day by ‐1.00 (95% CI ‐1.6 to ‐0.4) mmHg based on 17 trials in 2315 patients and also compared to 3 to 6.25 mg/day by ‐1.6 (95% CI ‐2.6 to ‐0.6) mmHg based on seven trials in 917 patients.
Hydrochlorothiazide 37.5 mg/day versus 12.5 mg/day; hydrochlorothiazide 50 to 25 mg/day or hydrochlorothiazide 100 mg versus 50 mg/day were not significantly different from each other. Therefore hydrochlorothiazide 25 mg/day was chosen as the lowest dose with maximum diastolic blood pressure‐lowering efficacy. The maximum diastolic blood pressure‐lowering efficacy at doses > 25 mg/day was ‐3.4 (95% CI ‐3.9 to ‐3.0) mmHg in 3744 patients (P value < 0.0001; no significant heterogeneity was present, with I 2 = 26%; test of subgroup difference P value = 0.29, with I2 = 19.9%).
The placebo‐corrected systolic/diastolic blood pressure‐lowering with hydrochlorothiazide 3 to 100 mg/day was 6.9/3.3 mmHg.
We also plotted a weighted log dose‐response curve using individual data points from each study and the resulting curve showed a significant dose response for systolic blood pressure (slope ‐6.16 (‐8.75 to ‐3.56) and r = ‐ 0.58 but not for diastolic blood pressure slope ‐0.82 (‐3.44 to 1.79) and r = ‐0.43). See Figure 5 and Figure 6.
Dose‐related effect of hydrochlorothiazide on systolic blood pressure
Dose‐related effect of hydrochlorothiazide on diastolic blood pressure
Indapamide
Ten trials compared indapamide at doses ranging from 1 to 5 mg/day treated for a mean duration of 10 weeks (Ambrosioni 1998; Capone 1983; Fiddes 1997; Hall 1994; London 2006; Myers 2000; Prisant 2000; Soltero 1989; Taylor 1988; Weidler 1995). See Analysis 6.1; Analysis 6.2.
Table 8: Dose‐related systolic and diastolic blood pressure‐lowering efficacy of indapamide
| Indapamide | Systolic blood pressure mmHg (95% CI) | Diastolic blood pressure mmHg (95% CI) |
| 1.0 mg/day | ‐9.7 (‐19.9 to 0.5) | ‐3.0 (‐9.6 to 3.6) |
| 1.25 mg/day | ‐7.4 (‐9.2 to ‐5.5) | ‐3.6 (‐4.6 to ‐2.5) |
| 1.5 mg/day | ‐9.4 (‐11.5 to ‐7.3) | ‐4.1 (‐5.4 to ‐2.7) |
| 2.0 mg/day | ‐8.7 (‐17.4 to ‐0.0) | ‐3.6 (‐8.3 to 1.1) |
| 2.5 mg/day | ‐11.9 (‐15.9 to ‐8.0) | ‐5.3 (‐7.7 to ‐3.0) |
| 5.0 mg/day | ‐9.6 (‐19.4 to 0.2) | ‐4.0 (‐10.3 to 2.3) |
| Overall | ‐8.7 (‐10.0 to ‐7.4) | ‐3.9 (‐4.6 to ‐3.1) |
Based on the 10 trials (N = 2150 patients), the lowest dose of indapamide that statistically significantly lowered both systolic and diastolic blood pressure was 1.25 mg/day. The overall placebo‐corrected systolic blood pressure‐lowering effect across 1 to 5 mg/day doses was ‐8.7 mmHg (P value < 0.00001, with I2 = 31% and the test for subgroup differences P value = 0.41, with I2 = 1.5%). For diastolic blood pressure it was ‐3.9 mmHg (P value < 0.00001, with I2 = 11% and the test for subgroup differences P value = 0.85, with I2 = 0%).
Direct comparison of doses from one dose‐ranging trial did not show any significant differences in systolic or diastolic blood pressure between different doses used (McVeigh 1988).
The placebo‐corrected systolic/diastolic blood pressure‐lowering with indapamide was 8.4/3.8 mmHg.
Thiazides (all six drugs)
When the lowest dose of each of the six thiazide drugs that achieved maximal systolic and diastolic blood pressure reduction and all doses above it were pooled, the overall systolic blood pressure reduction for thiazide diuretics as a class was ‐9.1 (95% CI ‐9.7to ‐8.5) mmHg (heterogeneity: Chi2 = 68.49, df = 46 (P value = 0.02); I2 = 33%; test for overall effect: Z = 28.86 (P value < 0.00001); test for subgroup differences: Chi2 = 14.53, df = 5 (P value = 0.01), I2 = 65.6% was significant). For diastolic blood pressure it was ‐3.6 (95% CI ‐4.0 to ‐3.3) mmHg (heterogeneity: Chi2 = 81.78, df = 50 (P value = 0.003); I2 = 39%; test for overall effect: Z = 20.63 (P value < 0.00001); test for subgroup differences: Chi2 = 11.23, df = 5 (P value = 0.05), I2 = 55.5%). See Analysis 7.1; Analysis 7.2.
Subgroup analyses
Due to lack of data being reported in each trial for individual participants based on age (18 to 59 and 60 or older); sex; race (black, white and others); co‐morbid conditions; or baseline severity of hypertension (mild, moderate or severe) subgroup analyses could not be performed. However, based on trials reporting the mean baseline blood pressure levels of all included participants, we classified trials according to systolic and diastolic blood pressure and performed subgroup analyses.
Table 9: Subgroup analyses based on baseline mean systolic blood pressure
| Based on systolic blood pressure at baseline | # of trials | Systolic blood pressure decrease mmHg (95% CI) |
| < 140 mmHg | 2 | ‐6.7 (‐11.0 to ‐2.3) |
| 140 to 149 mmHg | 4 | ‐10.9 (‐13.4 to ‐8.4) |
| 150 to 159 mmHg | 20 | ‐9.1 (‐10.1 to ‐8.1) |
| 160 mmHg or > | 13 | ‐10.3 (‐11.5 to ‐9.2) |
| Overall | 39 | ‐9.6 (‐10.4 to ‐8.9) |
Table 9 summarizes the finding based on baseline systolic blood pressure. Heterogeneity and subgroup differences were not significant. There were no significant differences in systolic blood pressure‐lowering based on systolic blood pressure at baseline.
Table 10: Subgroup analyses based on baseline mean systolic blood pressure
| Based on diastolic blood pressure at baseline | # of trials | Diastolic blood pressure decrease mmHg (95% CI) |
| < 90 mmHg | 3 | ‐2.9 (‐4.7 to ‐1.1) |
| 90 to 99 mmHg | 14 | ‐2.8 (‐3.3 to ‐2.3) |
| 100 to 109 mmHg | 25 | ‐4.7 (‐5.2 to ‐4.1) |
| Overall | 42 | ‐3.6 (‐4.0 to ‐3.3) |
Table 10 summarizes the finding based on baseline diastolic blood pressure. Heterogeneity was significant (P value < 0.00001, with I 2 = 56%) and subgroup difference significant (P value < 0.0001, with I2 = 91.1%). There was a significantly greater magnitude of diastolic blood pressure‐lowering (by 1.8 mmHg) in trials with the highest mean diastolic blood pressure (between 100 and 109 mmHg at baseline).
Withdrawals due to adverse effects
Withdrawals due to adverse effects were reported for bendrofluazide, chlorthalidone, hydrochlorothiazide and indapamide trials. Please see Analysis 1.3; Analysis 4.3; Analysis 5.3 and Analysis 6.3
There is selective outcome reporting of withdrawals due to adverse effects. This outcome was reported only in 31 of the 60 trials (52%) meeting the inclusion criteria.
Bendrofluazide resulted in significantly lower withdrawals due to adverse effects compared with placebo based on one trial in 257 patients (risk ratio (RR) 0.19, 95% CI 0.07 to 0.57) (Carlsen 1990). Chlorthalidone resulted in significantly lower withdrawals due to adverse effects compared with placebo based on five out of eight trials in 1058 patients (RR 0.49, 95% CI 0.28 to 0.87) (Bradley 1993; Hulley 1985; Materson 1978; Morledge 1986; Vardan 1987). Hydrochlorothiazide resulted in significantly lower withdrawals due to adverse effects compared with placebo based on 20 out of 40 trials in 3698 patients (RR 0.64, 95% CI 0.43 to 0.93). Indapamide did not significantly change withdrawals due to adverse effects compared with placebo based on six out of nine trials in 1874 patients (RR 0.87, 95% CI 0.52 to 1.46)) (Ambrosioni 1998; Capone 1983; Fiddes 1997; Hall 1994; London 2006; Weidler 1995). The major reason for withdrawals in the placebo group was an increase in blood pressure, which was reported as an adverse effect and therefore included in the number of patients who withdrew due to adverse effects.
The selective reporting of adverse effects across trials (e.g. data on the number of patients with serious adverse events, the nature of these events or reporting of only drug‐related adverse events) and specific reasons for withdrawal due to adverse effects are provided in detail in the 'Risk of bias' tables (in the attrition and selective reporting sections). Due to the very low quality of evidence resulting from selective reporting bias, data are not reported as absolute risk difference or as number needed to treat to benefit or harm.
Metabolic data
The reporting on metabolic data is very limited and due to the high risk of selective reporting bias strong conclusions cannot be made.
Data on serum potassium, uric acid, creatinine, glucose, total cholesterol, low‐density cholesterol and triglycerides were limited to a minority of the trials:
-
bendrofluazide (Carlsen 1990);
-
cyclopenthiazide (McVeigh 1988);
-
chlorthalidone(Bradley 1993; Hulley 1985; Materson 1978; Morledge 1986; Siegel 1992; Vardan 1987);
-
hydrochlorothiazide(Chrysant 1994; Drayer 1995; Goldberg 1989; Jounela 1994; Mersey 1993; Pool 2007; Pool 1993; Saruta 2007; Scholze 1993; Schoenberger 1995; Siegel 1992);
-
indapamide (Capone 1983; Fiddes 1997; Hall 1994; Soltero 1989; Taylor 1988).
Dose‐related serum potassium levels (mmol/L)
Serum potassium levels were reported for bendrofluazide, chlorthalidone, cyclopenthiazide, hydrochlorothiazide and indapamide. No data were reported for metolazone. The overall decrease in serum potassium levels for each thiazide drug is shown below.
Table 11: Serum potassium levels (mmol/L)
| Drug and dose range | Number of trials reporting data | MD (95% CI) mmol/L |
| Bendrofluazide 1.25 to 10 mg/day | 1 | ‐0.37 (‐0.50 to ‐0.24) |
| Cyclopenthiazide 0.05 to 0.5 mg/day | 1 | ‐0.18 (‐0.42 to 0.07) |
| Chlorthalidone 12.5 to 100 mg/day | 4 | ‐0.40 (‐0.45 to ‐0.34) |
| Hydrochlorothiazide 3 to 100 mg/day | 11 | ‐0.23 (‐0.26 to ‐0.19) |
| Indapamide 1 to 5 mg/day | 6 | ‐0.32 (‐0.38 to ‐0.26) |
| Overall^ | 23 | ‐0.25 (‐0.28 to ‐0.22) |
^Heterogeneity: Chi2 = 22.07, df = 6 (P value = 0.001), I2 = 73%; test for overall effect: Z = 10.06 (P value < 0.00001); test for subgroup differences: Chi2 = 10.06, df = 3 (P value = 0.02), I2 = 70.2%.
Bendrofluazide ‐ See Table 11 and Analysis 1.4. Heterogeneity between doses was not significant. The 10 mg dose lowered serum potassium significantly more than the 2.5 mg/day dose by ‐0.25 (95% CI ‐0.40 to ‐0.10) mmol/L by direct dose comparison.
Cyclopenthiazide ‐ See Table 11 and Analysis 2.3. Heterogeneity between doses was not significant. Cyclopenthiazide 0.5 mg/day lowered serum potassium significantly more compared to 0.05 mg/day by ‐0.60 (95% CI ‐0.87 to ‐0.33) mmol/L and by ‐0.40 (95% CI ‐0.62 to ‐0.18) mmol/L compared to 0.125 mg/day by direct dose comparison.
Chlorthalidone ‐ See Table 11 and Analysis 4.4. Heterogeneity between doses was significant (P value < 0.0001, with I2 = 93.3%). Chlorthalidone 25 mg lowered serum potassium significantly more compared to a 12.5 to 15 mg/day dose ‐0.20 (95% CI ‐0.32 to ‐0.08) mmol/L based on 252 patients. However, doses higher than 25 mg did not differ significantly from 25 mg/day by direct dose comparison. We could not determine the cause of heterogeneity.
Hydrochlorothiazide ‐ See Table 11 and Analysis 5.4. A dose‐related significant decrease in serum potassium levels was observed: ‐0.16 (95% CI ‐0.21 to ‐0.11) mmol/L at a 12.5 mg/day dose; ‐0.30 (95% CI ‐0.36 to ‐0.24) mmol/L at 25 mg/day; and ‐0.48 (95% CI ‐0.68 to ‐0.29) mmol/L at 50 mg/day. Heterogeneity between doses was significant (P value < 0.00001, with I2 = 60%).
In direct dose comparisons, hydrochlorothiazide 25 mg/day lowered serum potassium significantly more compared to hydrochlorothiazide 12.5 mg/day by ‐0.15 (95% CI ‐0.22 to ‐0.09) mmol/L based on four trials in 642 patients. No significant differences were observed between other direct dose comparisons.
Indapamide ‐ See Table 11 and Analysis 6.4. Indapamide 1.25 mg/day lowered serum potassium significantly more than indapamide 1.0 mg/day by ‐0.43 (95% CI ‐0.66 to ‐0.20) mmol/L based on one trial in 47 patients. No significant differences were observed between other doses.
The overall reduction in serum potassium for all thiazide drugs compared to placebo, based on available data (based on 23 trials; N = 3868), was ‐0.25 (95% CI ‐0.28 to ‐0.22) mmol/L. See Analysis 7.3.
Indirect comparison shows significantly greater lowering of serum potassium with chlorthalidone compared to hydrochlorothiazide (‐0.10, 95% CI ‐0.17 to ‐0.30) mmol/L, but no other statistically significant difference was observed between different drugs.
Dose‐related serum uric acid levels (µmol/L)
Serum uric acid levels were reported for bendrofluazide, chlorthalidone, cyclopenthiazide, hydrochlorothiazide and indapamide. No data were reported for metolazone.
Table 12: Serum uric acid levels (µmol/L)
| Drug and dose range | Number of trials reporting data | MD (95% CI) µmol/L |
| Bendrofluazide 1.25 to 10.0 mg/day | 1 | 46.6 (33.2to 59.9) |
| Cyclopenthiazide 0.05 to 0.5 mg/day | 1 | 19.5 (‐36.2 to 75.3) |
| Chlorthalidone 12.5 to 100 mg/day | 2 | 64.2 (45.7 to 82.6) |
| Hydrochlorothiazide 3 to 100 mg/day | 5 | 32.9 (26.1 to 39.7) |
| Indapamide 1 to 5 mg/day | 4 | 39.8 (33.5 to 46.1) |
| Overall^ | 13 | 38.2 (34.2 to 42.2) |
^Heterogeneity: Chi2 = 23.13, df = 12 (P value = 0.03); I2 = 48%; test for overall effect: Z = 18.82 (P, 0.0001); test for subgroup differences: Chi2 = 13.76, df = 4 (P value = 0.008), I2 = 70.9%.
Bendrofluazide ‐ See Table 12 and Analysis 1.5. Bendrofluazide 5 mg/day dose showed a significant increase compared to 2.5 mg/day of 34 (95% CI 2.2to 65.8) µmol/L based on one trial in 104 patients. Bendrofluazide 10 mg/day also showed a significant increase compared to bendrofluazide 1.25 mg/day and bendrofluazide 2.5 mg/day by direct dose comparison. Bendrofluazide 10 mg was not significantly different from 5 mg/day.
Cyclopenthiazide ‐ See Table 12 and Analysis 2.4. Heterogeneity between doses was not significant (P value = 0.96, with I2 = 0%). No significant differences were observed between doses by direct dose comparison.
Chlorthalidone ‐ See Table 12 and Analysis 4.5. Direct comparison between doses showed no significant differences between doses.
Hydrochlorothiazide ‐ See Table 12 and Analysis 5.5. No significant differences between doses were observed by direct dose comparison.
Indapamide ‐ See Table 12 and Analysis 6.5. Direct comparison between doses showed no significant difference.
The overall increase in serum uric acid for all thiazide drugs compared to placebo, based on available data, was 38.2 (34.2 to 42.2) mmol/L. See Analysis 7.4.
Indirect comparison shows a significant increase in serum uric acid with chlorthalidone compared to hydrochlorothiazide (32.0, 95% CI 12.17 to 51.83 µmol/L) and compared to indapamide (26.4, 95% CI 6.91 to 45.89 µmol/L).
Dose‐related serum creatinine levels (µmol/L)
Serum creatinine levels were reported for bendrofluazide, hydrochlorothiazide and indapamide. Data were not reported for chlorthalidone, cyclopenthiazide and metolazone.
Table 13: Serum creatinine levels (µmol/L)
| Drug and dose range | Number of trials reporting data | MD (95% CI) µmol/L |
| Bendrofluazide 1.25 to 10.0 mg/day | 1 | 5.5 (1.9to 9.11) |
| Hydrochlorothiazide 3 to 100 mg/day | 3 | 0.32 (‐2.63 to 3.26) |
| Indapamide 1 to 5 mg/day | 1 | 0.00(‐2.43 to 2.43) |
| Overall^ | 5 | 1.34 (‐0.31 to 2.99) |
^Heterogeneity: Chi2 = 7.14, df = 4 (P value = 0.13); I2 = 44%; test for overall effect: Z = 1.59 (P value = 0.11); test for subgroup differences: Chi2 = 7.06, df = 2 (P value = 0.03), I2 = 71.7%.
See Analysis 7.5.
Bendrofluazide is the only drug that increased creatinine (see Table 13 and Analysis 1.6). Direct comparison between doses showed no significant difference.
Dose‐related serum glucose levels (mmol/L)
Serum glucose levels were reported for bendrofluazide, chlorthalidone, hydrochlorothiazide and indapamide. Data were not reported for cyclopenthiazide and metolazone.
Table 14: Serum glucose levels (mmol/L)
| Drug and dose range | Number of trials reporting data | MD (95% CI) mmol/L |
| Bendrofluazide 1.25 to 10.0 mg/day | 1 | 0.13 (‐0.06 to 0.33) |
| Chlorthalidone 12.5 to 100 mg/day | 3 | 0.34 (0.12 to 0.55) |
| Hydrochlorothiazide 3 to 100 mg/day | 6 | ‐0.11 (‐0.22 to 0.01) |
| Indapamide 1 to 5 mg/day | 3 | 0.13 (‐0.11 to 0.37) |
| Overall^ | 13 | 0.03 (‐0.05to 0.12) |
^Heterogeneity: Chi2 = 20.88, df = 12 (P value = 0.05), with I2 = 43%; test for overall effect: Z = 0.81 (P value = 0.42); test for subgroup differences: Chi2 = 16.0, df = 3 (P value = 0.001), with I2 = 81.3%. See Analysis 7.6.
No significant difference from placebo was noted with bendrofluazide, hydrochlorothiazide and indapamide.
Chlorthalidone ‐ a statistically significant increase in serum glucose was observed with chlorthalidone 25 mg/day (0.58, 95% CI 0.23 to 0.93 mmol/L). The overall increase from 12.5 to 75 mg/day was also significant (see Table 14). Heterogeneity between doses was not significant (P value = 0.39, with I2 = 4%) (see Analysis 4.6). There were no significant differences between doses.
Indirect comparison showed a significant increase in serum glucose with chlorthalidone compared to hydrochlorothiazide (0.45, 95% CI 0.21 to 0.69 mmol/L).
Dose‐related total cholesterol levels (mmol/L)
Total cholesterol levels were reported for bendrofluazide, chlorthalidone, cyclopenthiazide, hydrochlorothiazide and indapamide.
Table 15: Serum total cholesterol levels (mmol/L)
| Drug and dose range | Number of trials reporting data | MD (95% CI) mmol/L |
| Bendrofluazide 1.25 to 10.0 mg/day | 1 | 0.15 (‐0.05 to 0.35) |
| Chlorthalidone 12.5 to 100 mg/day | 2 | 0.41 (0.18 to 0.64) |
| Cyclopenthiazide 0.05 to 0.5 mg/day | 1 | 0.79 (0.36 to 1.23) |
| Hydrochlorothiazide 3 to 25 mg/day | 4 | 0.20 (0.17 to 0.22) |
| Indapamide 1 to 5 mg/day | 3 | 0.11 (0.01 to 0.21) |
| Overall^ | 11 | 0.21 (0.18 to 0.23) |
^Heterogeneity: Chi2 = 26.03, df = 10 (P value = 0.004); I2 = 62%; test for overall effect: Z = 15.59 (P value < 0.0001); test for subgroup differences: Chi2 = 14.01, df = 4 (P value = 0.007), I2 = 71.5%.
See Analysis 7.7.
Bendrofluazide ‐ See Table 15 and Analysis 1.8. Based on one trial in 257 patients there were no significant differences in serum total cholesterol compared to placebo or between different doses.
Chlorthalidone ‐ See Table 15 and Analysis 4.8. Based on one trial in 213 patients there was a significant increase in total serum cholesterol (0.41, 95% CI 0.18 to 0.64 mmol/L), but no difference between 25 mg versus 15 mg.
Cyclopenthiazide ‐ See Table 15 and Analysis 2.5. Based on one trial in 47 patients, there was a statistically significant increase in total serum cholesterol (0.79, 95% CI 0.36 to 1.23 mmol/L), but no significant difference between doses.
Hydrochlorothiazide ‐ See Table 15 and Analysis 5.9. Based on four trials in 450 patients there was a statistically significant increase in total serum cholesterol (0.20, 95% CI 0.17 to 0.22 mmol/L). A significant difference between doses was observed. Heterogeneity between doses was significant (P value = 0.0002, with I2 = 77%).
Indapamide ‐ See Table 15 and Analysis 6.9. Based on two trials in 398 patients, there was a statistically significant increase in total serum cholesterol at a 1.25 mg/day dose (0.11, 95% CI 0.01 to 0.21 mmol/L)
Indirect comparison showed a significant increase in total cholesterol with chlorthalidone compared to indapamide (0.3, 95% CI 0.05 to 0.55 mmol/L), but not compared to hydrochlorothiazide.
Dose‐related high‐density lipoprotein (HDL) cholesterol levels (mmol/L)
HDL cholesterol levels were reported for hydrochlorothiazide, chlorthalidone and indapamide. Data were not reported for bendrofluazide, cyclopenthiazide and metolazone.
Table 16: Serum HDL cholesterol levels (mmol/L)
| Drug and dose range | Number of trials reporting data | MD (95% CI) mmol/L |
| Chlorthalidone 45 mg/day | 1 | ‐0.11 (‐0.22 to 0.00) |
| Hydrochlorothiazide 3 to 25 mg/day | 1 | ‐0.17 (‐0.53 to 0.19) |
| Indapamide 1 to 5 mg/day | 1 | ‐0.04 (‐0.03 to ‐0.00) |
| Overall^ | 3 | ‐0.06 (‐0.10 to ‐0.02) |
^Heterogeneity: Chi2 = 1.06, df = 2 (P value = 0.59); I2 = 0%; test for overall effect: Z = 2.76 (P value = 0.006); test for subgroup differences: Chi2 = 1.06, df = 2 (P value = 0.59); I2 = 0%.
The overall significant decrease in HDL cholesterol levels compared to placebo (‐0.06, 95% CI ‐0.10 to ‐0.02 mmol/L) was based on three trials in 348 patients. Heterogeneity and subgroup differences were not significant (see Table 16 and Analysis 7.8).
Dose‐related serum triglycerides levels (mmol/L)
Serum triglyceride levels were reported for bendrofluazide, chlorthalidone, cyclopenthiazide, hydrochlorothiazide and indapamide.
Table 17: Serum triglyceride levels (mmol/L)
| Drug and dose range | Number of trials reporting data | MD (95% CI) mmol/L |
| Bendrofluazide 1.25 to 10.0 mg/day | 1 | 0.26 (‐0.06 to 0.58) |
| Chlorthalidone 45 mg/day | 1 | 0.69 (0.05 to 1.33) |
| Cyclopenthiazide 0.05 to 0.5 mg/day | 1 | 0.20 (‐0.17 to 0.57) |
| Hydrochlorothiazide 3 to 25 mg/day | 2 | 0.09 (‐0.11 to 0.30) |
| Indapamide 1 to 5 mg/day | 1 | 0.21 (0.02 to 0.40) |
| Overall^ | 6 | 0.21 (0.08 to 0.33) |
^Heterogeneity: Chi2 = 5.63, df = 5 (P value = 0.34); I2 = 11%; test for overall effect: Z = 3.17 (P value = 0.002); test for subgroup differences: Chi2 = 3.88, df = 4 (P value = 0.42), I2 = 0%.
Bendrofluazide ‐ See Table 17 and Analysis 1.9. Based on one trial in 257 patients there was no significant difference in serum triglycerides compared to placebo. Heterogeneity between doses was not significant (P value = 0.79 and I2 = 0%).
Cyclopenthiazide ‐ See Table 17 and Analysis 2.6. Based on one trial in 48 patients there was a significant increase in serum triglycerides compared to placebo. Heterogeneity between doses was not significant (P value = 0.53 and I2 = 0%).
Chlorthalidone ‐ See Table 17 and Analysis 4.9. Based on one trial in 36 patients there was a significant increase in serum triglycerides compared to placebo with chlorthalidone 45 mg/day.
Hydrochlorothiazide ‐ See Table 17 and Analysis 5.10. Based on two trials in 255 patients there was no significant difference in serum triglycerides compared to placebo. Heterogeneity between doses was not significant (P value = 0.75 and I2 = 0%).
Indapamide ‐ See Table 17 and Analysis 6.10. Based on one trial in 203 patients there was a significant increase in serum triglycerides compared to placebo with indapamide 1.25 mg/day.
For the overall increase in serum triglycerides across drugs (see Table 17), heterogeneity and subgroup differences were not significant. See Analysis 7.9.
Sensitivity analyses
Sufficient data were available to carry out sensitivity analyses for systolic and diastolic blood pressure data for thiazide diuretics overall.
1. Quality of trials
Trials of high quality versus poor quality: since only two trials had adequate randomization, allocation concealment and blinding as opposed to unclear or high risk of bias in the other 58 trials (Papademetriou 2006; Pool 1993), this analysis was not meaningful.
2. Fixed‐effect versus random‐effects model
Sensitivity analyses using a fixed‐effect versus a random‐effects model showed no significant difference in the overall effect estimate (data not shown).
3. Position in which blood pressure was measured
Robustness of the overall effect size in relation to the position in which blood pressure was measured: trials with blood pressure data measured in a sitting position versus other measurements:
Blood pressure data in a sitting position were available in 30 trials. The overall magnitude of systolic blood pressure‐lowering was ‐9.1 (95% CI ‐9.8 to ‐8.3) mmHg and diastolic blood pressure reduction was ‐3.6 (95% CI ‐4.0 to ‐3.1) mmHg. Blood pressure data in a standing position were available in 11 trials: the magnitude of systolic blood pressure‐lowering was ‐10.2 (95% CI ‐11.6 to ‐8.4) mmHg and diastolic blood pressure reduction was ‐3.7 (95% CI ‐4.7 to ‐2.7) mmHg. Blood pressure data in a supine position were available in seven trials: the magnitude of systolic blood pressure‐lowering was ‐8.7 (‐95% CI 10.2 to ‐7.2) mmHg and diastolic blood pressure reduction was ‐3.8 (95% CI ‐4.7 to ‐3.0) mmHg. No significant differences were observed based on the position in which blood pressure was measured.
4. Peak versus trough ‐ magnitude of blood pressure‐lowering
Thirty trials reported trough systolic blood pressure measurement and the magnitude of systolic blood pressure‐lowering was ‐9.1 (95% CI ‐9.8 to ‐8.4) mmHg. In the remaining trials timing of systolic blood pressure measurement was not reported.
Thirty‐three trials reported trough diastolic blood pressure measurement and the magnitude of diastolic blood pressure‐lowering was ‐3.4 (95% CI ‐3.8 to ‐3.1) mmHg. In the remaining trials the timing of diastolic blood pressure measurement was not reported.
5. Available versus imputed standard deviation
Trials with published standard deviations of blood pressure change versus imputed standard deviations showed no significant difference in the overall estimate of effect for both systolic and diastolic blood pressure (data not shown).
6. Industry versus non‐industry‐sponsored trials
Twenty‐eight trials (47%) reported industry‐sponsored funding and overall the systolic blood pressure‐lowering of thiazide diuretics was ‐8.9 (95% CI ‐9.5 to ‐8.0) mmHg, which was similar in magnitude to five non‐industry‐sponsored trials (8.3%) (‐8.6, 95% CI ‐9.5 to ‐7.6 mmHg). The remaining trials did not report source of funding.
Twenty‐four (40%) trials reported industry‐sponsored funding and overall the diastolic blood pressure‐lowering of thiazide diuretics was ‐3.2 (95% CI ‐3.7 to ‐2.8) mmHg, which was similar in magnitude to five non‐industry sponsored trials (8.3%) (‐3.8, 95% CI ‐5.5 to ‐2.1 mmHg). The remaining trials did not report source of funding.
Discussion
This review provides the best available evidence of the blood pressure‐lowering effect of thiazide monotherapy for the treatment of elevated blood pressure. The drug for which we have the most data is hydrochlorothiazide: 35 randomized controlled trials (RCTs) in 6725 patients. It is the only drug for which we have sufficient RCTs over the commonly prescribed dose range to demonstrate a clear dose‐response effect. There is a very clear dose response for systolic blood pressure over the range 6.25, 12.5 and 25 mg/day (Analysis 5.1; Analysis 5.2). The test for subgroup differences for systolic blood pressure was significant(Chi2 = 22.16, df = 5 (P value = 0.0005), I2 = 77.4%), but not for diastolic blood pressure (Chi2 = 6.22, df = 5 (P value = 0.29), I2 = 19.6%). This means that systolic blood pressure‐lowering at higher doses of hydrochlorothiazide was significantly greater than at lower doses, but diastolic blood pressure‐lowering was not significantly different between the higher or lower dose.
See also Figure 5 and Figure 6. We plotted the log dose‐response curve using individual data points from each study and the resulting curve showed a significant dose response for systolic blood pressure (slope ‐6.16 (‐8.75 to ‐3.56) and r = ‐ 0.58) but not for diastolic blood pressure (slope ‐0.82 (‐3.44 to 1.79) and r = ‐0.43). The significant dose response for hydrochlorothiazide over the dose range 6.25 to 50 mg demonstrates that for each doubling of the dose there is a 2 mmHg greater reduction in systolic blood pressure. There are not enough data at 50 and 100 mg/day but the data that are available suggest that the maximal effect is achieved with a dose of 50 mg/day and that at least 80% of the blood pressure‐lowering effect occurs with 25 mg/day. The confidence intervals for the estimates for 12.5 mg/day and 25 mg/day are narrow, demonstrating that the findings are robust and unlikely to be changed by further RCTs. Further RCTs studying the blood pressure‐lowering effect of 50 mg/day are necessary to have a better estimate of the effect for that dose.
We compared the results obtained in this review with the published review of hydrochlorothiazide given as a second‐line drug (Chen 2009) (see Table 18). The Chen 2009 review has more data on hydrochlorothiazide, from 53 RCTs in 15,129 hypertensive patients with baseline blood pressure of 156/101 mmHg, compared to the 40 RCTs in 7284 patients with baseline blood pressure of 155/100 mmHg included in this review. Results based on both of these reviews show that the magnitude of the systolic blood pressure‐lowering effect is the same whether the drug is given alone or added as a second‐line drug to another antihypertensive drug. This provides strong evidence for the dose‐response relationship and the average magnitude of effect for each dose. Since 6.25 to 50 mg/day is also the common dose range prescribed, the review provides valuable support for dose titration over this range of doses for physicians treating hypertension.
Table 18: Comparing systolic blood pressure reduction with different doses of hydrochlorothiazide
| Hydrochlorothiazide dose mg/day | This review Monotherapy versus placebo | Chen 2009 review Second‐line drug versus placebo | Weighted combined systolic blood pressure‐lowering effect from both reviews | ||||
| RCTs | # of patients | MD (95% CI) mmHg | RCTs | # of patients | MD (95% CI) mmHg | mmHg | |
| 3.0 to 6.25 | 8 | 663 | ‐3.6 (‐5.6 to ‐1.5) | 22 | 3283 | ‐3.7 (‐4.6 to ‐2.8) | 3.7 |
| 12.5 | 22 | 2645 | ‐6.3 (‐7.2 to ‐5.3) | 53 | 8482 | ‐6.0 (‐6.5 to ‐5.4) | 6.1 |
| 25 | 25 | 3062 | ‐8.0 (‐9.0 to ‐7.0) | 39 | 5799 | ‐8.0 (‐8.7 to ‐7.3) | 8.0 |
| 50 | 2 | 169 | ‐10.5 (‐14.6 to ‐6.4) | 3 | 189 | ‐17.8 (‐21.6 to ‐14.0) | 14.4 |
| 100 | 2 | 146 | ‐9.9 (‐14.1 to ‐5.8) | No data | 9.9 | ||
Reduction in systolic blood pressure shows a dose‐response relationship based on 37 RCTs in 6685 patients. It is similar whether the drug is given as monotherapy (this review) or as a second‐line drug (Chen 2009). Hydrochlorothiazide lowers systolic blood pressure by 4, 6 and 8 mmHg at 6.25, 12.5 and 25 mg/day respectively. See Figure 5.
Table 19: diastolic blood pressure reduction with different doses of hydrochlorothiazide
| Hydrochlorothiazide dose mg/day | This review Monotherapy versus placebo | Chen 2009 review Second‐line drug versus placebo | Weighted combined diastolic blood pressure‐lowering effect from both reviews | ||||
| RCTs | # of patients | MD (95% CI) mmHg | RCTs | # of patients | MD (95% CI) mmHg | mmHg | |
| 5.0 to 6.25 | 8 | 663 | ‐2.4 (‐3.7 to ‐1.2) | 23 | 3364 | ‐1.7 (‐2.2 to ‐1.2) | 1.8 |
| 12.5 | 25 | 2877 | ‐3.1 (‐3.7 to ‐2.5) | 55 | 8659 | ‐3.1 (‐3.4 to ‐2.8) | 3.1 |
| 25 | 29 | 3359 | ‐3.3 (‐3.8 to ‐2.8) | 42 | 6153 | ‐4.0 (‐4.4 to ‐3.6) | 3.8 |
| 37.5 to 50 | 3 | 239 | ‐4.5 (‐6.7 to ‐3.3) | 3 | 189 | ‐8.3 (‐10.7 to ‐6.0) | 6.4 |
| 100 | 2 | 146 | ‐3.9 (‐6.6 to ‐1.2) | ‐ | ‐ | No data | 3.9 |
Reduction in diastolic blood pressure does not show a significant dose‐response relationship based on 40 RCTs in 7284 patients (Figure 6). Based on the results of both of these reviews, there is a trend toward a dose‐response relationship, and a similar magnitude of effect whether the drug is given as monotherapy or as a second‐line drug.
The greater reduction and dose‐related reduction in systolic compared to diastolic blood pressure means that hydrochlorothiazide has a dose‐related reduction of pulse pressure by 2 to 6 mmHg over the dose range of 6.25 to 50 mg/day. At maximal doses the average reduction in pulse pressure is 5.5 mmHg. This magnitude and pattern is different from other drug classes. Angiotensin receptor blockers (ARBs), angiotensin‐converting enzyme (ACE) inhibitors and renin inhibitors lower pulse pressure by 3 mmHg on average and there is no dose‐response relationship (Heran 2009a; Heran 2009b; Musini 2008). Non‐selective beta‐blockers have little or no effect on pulse pressure: at most 2 mmHg and no dose‐response effect (Wong 2014).
The thiazide with the second most data is indapamide, which lowered blood pressure by 7.4/3.6 mmHg at the lowest dose studied (1.25 mg/day) and 9/4 mmHg for all doses combined. The lowest dose result is very similar to the results with 25 mg hydrochlorothiazide. There was no dose‐response relationship for doses of indapamide higher than 1.25 mg/day and unfortunately there were no RCTs for doses of one‐half or one‐quarter of 1.25 mg/day. These data for indapamide show that the maximal blood pressure lowering effect is achieved with the lowest doses of 1 to 2 mg/day and there is no rationale for using higher doses.
The thiazide with the third most data is chlorthalidone. Like indapamide, chlorthalidone showed no dose response over the doses studied. It appeared that the lowest dose studied(12.5 mg/day) had the maximum blood pressure‐lowering effect. On first look it appears that chlorthalidone lowers blood pressure more than hydrochlorothiazide and indapamide. However, the baseline systolic blood pressure was significantly higher and the baseline diastolic blood pressure was significantly lower in the chlorthalidone trials. When the isolated systolic blood pressure trials were deselected, the blood pressure‐lowering effect of chlorthalidone was not significantly different from the other thiazides and, furthermore, the overall blood pressure‐lowering effect of chlorthalidone was not different from hydrochlorothiazide at the maximally effective doses of 50 mg and above.
The fact that chlorthalidone is more potent than hydrochlorothiazide (12.5 mg of chlorthalidone being equivalent to 50 mg of hydrochlorothiazide) is likely due to pharmacokinetic differences between it and hydrochlorothiazide. Hydrochlorothiazide has a half‐life of 8 to 15 hours with long‐term dosing. However, several studies show that the pharmacodynamic response is much longer than predicted by the half‐life (Allen 1982; Lutterodt 1980). Chlorthalidone in comparison has a half‐life ranging from 45 to 60 hours with long‐term dosing. Interindividual variability in half‐life is large. Chlorthalidone serum concentrations after 100 mg are only twice those of a 25 mg dose, indicating a flat dose‐serum concentration curve (Carter 2004; Riess 1977; Russell 1981). However, the longer half‐life does not mean that chlorthalidone is a superior antihypertensive compound. The data in this review suggest that doses of chlorthalidone that should be prescribed are 12.5 mg/day and lower. RCTs of the blood pressure‐lowering effect of chlorthalidone at doses lower than 12.5 mg/day would be useful.
The other three thiazides had only one RCT each so no conclusions can be made about dose‐response effects or relative blood pressure‐lowering potency. From the data available there is no reason to suggest that they are any different from hydrochlorothiazide in their blood pressure‐lowering effect. Any subgroup differences between the different thiazides in blood pressure‐lowering effect are more likely due to differences in patient population (baseline blood pressure, etc.) or biases in trial conduct than to any pharmacological differences in the magnitude or pattern of the blood pressure‐lowering effect.
Harms of thiazides
The limited data on withdrawal due to adverse effects suggested a decrease in the treatment group compared to the placebo group. This does not mean that placebo is more harmful than a thiazide. The likely reason for this is that withdrawal due to an increase in blood pressure was inappropriately counted as an adverse effect. As can be seen in the table below, the proportion of RCTs reporting withdrawals due to adverse effects (32/60 (53%)) is much less than the total number reporting blood pressure data (48/60 (80%)). This suggests that the trials in which withdrawals due to adverse effects were higher for the drug than for placebo were selectively under reported and this represents a high risk of selective reporting bias. In this review, therefore, no conclusions can be drawn as to whether thiazides increase withdrawals due to adverse effects with short‐term use.
Table 20: Withdrawals due to adverse effects
| Drug and dose range | Number of trials reporting data | WDAEs RR (95% CI) |
| Bendrofluazide 1.25 to 10.0 mg/day | 1/1 (N = 257) | 0.19 (0.07 to 0.57) |
| Chlorthalidone 12.5 to 100 mg/day | 5/7 (N = 1058) | 0.49 (0.28 to 0.87) |
| Cyclopenthiazide 0.05 to 0.5 mg/day | 0/1 | Not reported |
| Hydrochlorothiazide 3 to 100 mg/day | 20/40 (N = 3698) | 0.64 (0.43 to 0.93) |
| Indapamide 1 to 5 mg/day | 6/10 (N = 1874) | 0.83 (0.49 to 1.42) |
| Metolazone 0.5 to 2 mg/day | 0/1 | Not reported |
Metabolic data were also only reported in a minority of trials (see Table 21 below).
The available data show a clear reduction in serum potassium, an increase in serum uric acid and an increase in serum total cholesterol and triglycerides. All of these effects are well known to occur with thiazides. New findings from this review are that the metabolic effects were greater with higher doses and were less, in general, with hydrochlorothiazides than the other thiazides. In addition, serum glucose was not increased by thiazides overall. In fact the only thiazide associated with an increase in glucose in this review was chlorthalidone. However, in this review, the high risk of selective reporting bias for the metabolic data and indirect comparison between different thiazide drugs makes drawing strong conclusions impossible.
Table 21: Metabolic data
| Drug and dose range | Number of trials reporting data (N) | MD with 95% CI all thiazide trials |
| Serum potassium mmol/L | 22/59 (3868) | ‐0.25 (‐0.28 to ‐0.22) |
| Serum uric acid µmol/L | 13/59 (2332) | 38.2 (34.2 to 42.2) |
| Serum creatinine µmol/L | 5/59 (987) | 1.34 (‐0.31 to 2.99) |
| Serum glucose mmol/L | 13/59 (1989) | 0.03 (‐0.05 to 0.12) |
| Serum total cholesterol mmol/L | 11/59 (431) | 0.20 (0.18 to 0.23) |
| Serum high‐density lipoprotein cholesterol mmol/L | 3/59 (348) | ‐0.06 (‐0.10 to ‐0.02) |
| Serum triglycerides mmol/L | 6/59 (697) | 0.21 (0.08 to 0.33) |
Summary of main results
See summary of findings Table for the main comparison and summary of findings Table 2.
The magnitude of the dose‐related systolic and diastolic blood pressure‐lowering is from a low to high quality of evidence for hydrochlorothiazide. We determined the evidence for the overall blood pressure‐lowering effect of maximal doses of thiazides to be of high quality. We judged the data on withdrawals due to adverse effects to be of very low quality since there is selective reporting bias for this outcome and because withdrawal due to an increase in blood pressure was inappropriately included as an adverse event in the placebo group.
Overall completeness and applicability of evidence
In this review we systematically searched various databases from 1946 until February 2014 and it is unlikely that any RCTs have been missed. The robustness of the blood pressure‐lowering efficacy data is validated by comparing and demonstrating the similar magnitude of blood pressure reduction in this systematic review to that by Chen 2009 on the blood pressure‐lowering efficacy of hydrochlorothiazide as second‐line therapy for primary hypertension. The Chen 2009 review has more data on hydrochlorothiazide (from 53 RCTs in 15,129 hypertensive patients with baseline blood pressure of 156/101 mmHg) compared to the 40 RCTs in 7284 patients with baseline blood pressure of 155/100 mmHg included in this review.
Quality of the evidence
We assessed the risk of bias for each of the RCTs included in this review (Figure 2). The majority of RCTs (82% of included studies) were published before the year 2000, prior to standardization of reporting of RCTs. However, it is clear that thiazides lower blood pressure, that this is dose‐related and that the magnitude is probably approximately what is reported here. Although the magnitude of systolic and diastolic blood pressure‐lowering is similar to the effect when using thiazide as a second‐line drug (Chen 2009), we downgraded the quality of evidence from high to moderate for some doses as it is an indirect comparison of the effect size with wider confidence intervals compared to the Chen review, which used similar inclusion/exclusion criteria.
There is a high risk of bias in the adverse effect data (Figure 3), therefore the available evidence for adverse events is likely not an accurate reflection of reality.
The evidence is very low quality for the metabolic data due to the high risk of selective outcome reporting bias and a weak interpretation due to multiple indirect comparisons between different thiazide drugs.
Potential biases in the review process
One limitation of this review is that it is restricted to published trials and it is possible that smaller trials are missing (publication bias). See funnel plot (Figure 4). Our judgement of an unclear to high risk of bias in most of the included trials could also have led to an overestimation of blood pressure‐lowering effect and underestimation of adverse metabolic effects. The finding in this review that withdrawals due to adverse effects are lower with a thiazide compared to placebo is unlikely to be true. The short duration of trials in this review make it good for estimating the blood pressure‐lowering effect, but not accurate for estimating long‐term benefits or harms of thiazides.
Although we have followed the method of analysis for multiple armed studies described in the Cochrane Handbook for Systematic Reviews of Interventions (section 16.5.4) (Handbook 2011), since many of the included studies had multiple dose‐ranging treatment arms being compared to the same placebo group, the resulting comparisons remain co‐related and this method only partially overcomes unit of analysis error. The approach we used to overcome this unit of analysis error was to split the shared placebo group in to two or more groups with smaller sample sizes, to include two or more reasonably independent comparisons in the meta‐analysis as if they were from different studies.
We have followed the usual convention to round up numbers if > 0.5 and round down if < 0.5, therefore the magnitude of the blood pressure‐lowering effect is reported up to one decimal point and for metabolic data it is reported up to two decimal points in the Results section of this review. However, in the Abstract, for the sake of simplicity, using the usual convention the overall systolic/diastolic blood pressure‐lowering effect in mmHg is rounded to the nearest number (mmHg) without decimal points.
Agreements and disagreements with other studies or reviews
Messerli 2011 conducted a systematic review of the blood pressure‐lowering efficacy of hydrochlorothiazide for a minimum treatment period of four weeks in randomized trials comparing hydrochlorothiazide to other antihypertensive drug classes. Both office and ambulatory blood pressure measurements were available in eight studies (488 patients in total) using hydrochlorothiazide 12.5 to 25 mg/day for a mean duration of eight weeks. The mean baseline office systolic/diastolic blood pressure was 163/98 mmHg. The reduction in office systolic blood pressure from baseline was 12/7 mmHg, whereas the reduction in ambulatory 24‐hour blood pressure was 8/4 mmHg. The Messerli review is misleading and has been used to suggest that hydrochlorothiazide lowers office blood pressure more than ambulatory blood pressure. The ambulatory blood pressure reduction observed in the Messerli review is very similar to the reduction in placebo‐corrected office systolic/diastolic blood pressure in our review(6 to 8/3 to 4 mmHg). Our review is a more robust treatment effect estimate based on a greater number of RCTs and patients compared to the Messerli 2011 review.
Law 2009 included 354 randomized, double‐blind, placebo‐controlled trials of thiazides, beta‐blockers, ACE inhibitors, ARBs and calcium channel‐blockers administered either singly or in combination. They included both parallel and cross‐over trials together in their analysis. However, they presented placebo‐adjusted reductions in systolic and diastolic blood pressure according to dose expressed as a multiple of standard (recommended) doses of the drugs. What were considered as the standard doses for thiazide drugs were not described in the review. They combined trial data for specified equivalent daily doses of different drugs as the "usual maintenance dose" in reference pharmacopoeias. When a range was given they considered the lower dose as standard dose. The Law review did not present the magnitude of blood pressure‐lowering efficacy of each of the drugs within the thiazide diuretic class at all available doses from the included trials but presented data from all thiazide drugs together as a multiple of standard doses. Therefore, unlike the Law review, our review provides evidence that there is a dose‐related effect for hydrochlorothiazide and thus combining all doses is not rational.
The overall estimated reduction in systolic/diastolic blood pressure with thiazide diuretics in this review (‐9/‐4 mmHg) is similar to the treatment effect estimate in a systematic review of loop diuretics (‐8/‐4 mmHg) (Musini 2009c). This estimate was based on nine trials in 460 patients with a baseline blood pressure of 162/103 mmHg for a mean duration of nine weeks and was likely an overestimate due to the high risk of bias in the included studies. The loop diuretics review also did not provide a good estimate of the incidence of associated harms because of the short duration of the trials and the lack of reporting of adverse effects in many of the them. This review shows that thiazides lower systolic blood pressure and pulse pressure more and diastolic blood pressure less than shown in other Cochrane reviews on ARBs (Heran 2009b), ACE inhibitors (Heran 2009a), renin inhibitors (Musini 2008) and non‐selective beta‐blockers (Wong 2014). This observation is based on an indirect comparison between different antihypertensive drug class reviews compared to placebo control using similar inclusion/exclusion criteria. Therefore the common belief that different classes of antihypertensive drugs have the same blood pressure‐lowering effect is likely to be wrong.
'Risk of bias' summary: 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.
Funnel plot of comparison: 7 Thiazide versus placebo, outcome: 7.1 Systolic blood pressure.
Dose‐related effect of hydrochlorothiazide on systolic blood pressure
Dose‐related effect of hydrochlorothiazide on diastolic blood pressure
Comparison 1 Bendrofluazide versus placebo, Outcome 1 Systolic blood pressure.
Comparison 1 Bendrofluazide versus placebo, Outcome 2 Diastolic blood pressure.
Comparison 1 Bendrofluazide versus placebo, Outcome 3 Withdrawals due to adverse effects.
Comparison 1 Bendrofluazide versus placebo, Outcome 4 Serum potassium mmol/L.
Comparison 1 Bendrofluazide versus placebo, Outcome 5 Serum uric acid µmol/L.
Comparison 1 Bendrofluazide versus placebo, Outcome 6 Serum creatinine µmol/L.
Comparison 1 Bendrofluazide versus placebo, Outcome 7 Serum blood glucose mmol/L.
Comparison 1 Bendrofluazide versus placebo, Outcome 8 Total cholesterol mmol/L.
Comparison 1 Bendrofluazide versus placebo, Outcome 9 Triglycerides mmol/L.
Comparison 2 Cyclopenthiazide versus placebo, Outcome 1 Systolic blood pressure.
Comparison 2 Cyclopenthiazide versus placebo, Outcome 2 Diastolic blood pressure.
Comparison 2 Cyclopenthiazide versus placebo, Outcome 3 Serum potassium µmol/L.
Comparison 2 Cyclopenthiazide versus placebo, Outcome 4 Serum uric acid µmol/L.
Comparison 2 Cyclopenthiazide versus placebo, Outcome 5 Total cholesterol mmol/L.
Comparison 2 Cyclopenthiazide versus placebo, Outcome 6 Triglycerides mmol/L.
Comparison 3 Metolazone versus placebo, Outcome 1 Systolic blood pressure.
Comparison 3 Metolazone versus placebo, Outcome 2 Diastolic blood pressure.
Comparison 4 Chlorthalidone versus placebo, Outcome 1 Systolic blood pressure.
Comparison 4 Chlorthalidone versus placebo, Outcome 2 Diastolic blood pressure.
Comparison 4 Chlorthalidone versus placebo, Outcome 3 Withdrawals due to adverse events.
Comparison 4 Chlorthalidone versus placebo, Outcome 4 Serum potassium mmol/L.
Comparison 4 Chlorthalidone versus placebo, Outcome 5 Serum uric acid µmol/L.
Comparison 4 Chlorthalidone versus placebo, Outcome 6 Serum blood glucose mmol/L.
Comparison 4 Chlorthalidone versus placebo, Outcome 7 HDL cholesterol mmol/L.
Comparison 4 Chlorthalidone versus placebo, Outcome 8 Total cholesterol mmol/L.
Comparison 4 Chlorthalidone versus placebo, Outcome 9 Triglycerides mmol/L.
Comparison 5 Hydrochlorothiazide versus placebo, Outcome 1 Systolic blood pressure.
Comparison 5 Hydrochlorothiazide versus placebo, Outcome 2 Diastolic blood pressure.
Comparison 5 Hydrochlorothiazide versus placebo, Outcome 3 Withdrawals due to adverse events.
Comparison 5 Hydrochlorothiazide versus placebo, Outcome 4 Serum potassium mmol/L.
Comparison 5 Hydrochlorothiazide versus placebo, Outcome 5 Serum uric acid µmol/L.
Comparison 5 Hydrochlorothiazide versus placebo, Outcome 6 Serum creatinine µmol/L.
Comparison 5 Hydrochlorothiazide versus placebo, Outcome 7 Serum blood glucose mmol/L.
Comparison 5 Hydrochlorothiazide versus placebo, Outcome 8 HDL cholesterol mmol/L.
Comparison 5 Hydrochlorothiazide versus placebo, Outcome 9 Total cholesterol mmol/L.
Comparison 5 Hydrochlorothiazide versus placebo, Outcome 10 Triglycerides mmol/L.
Comparison 6 Indapamide versus placebo, Outcome 1 Systolic blood pressure.
Comparison 6 Indapamide versus placebo, Outcome 2 Diastolic blood pressure.
Comparison 6 Indapamide versus placebo, Outcome 3 Withdrawals due to adverse events.
Comparison 6 Indapamide versus placebo, Outcome 4 Serum potassium mmol/L.
Comparison 6 Indapamide versus placebo, Outcome 5 Serum uric acid µmol/L.
Comparison 6 Indapamide versus placebo, Outcome 6 Serum creatinine µmol/L.
Comparison 6 Indapamide versus placebo, Outcome 7 Serum blood glucose mmol/L.
Comparison 6 Indapamide versus placebo, Outcome 8 HDL cholesterol mmol/L.
Comparison 6 Indapamide versus placebo, Outcome 9 Total cholesterol mmol/L.
Comparison 6 Indapamide versus placebo, Outcome 10 Triglycerides mmol/L.
Comparison 7 Thiazide versus placebo, Outcome 1 Systolic blood pressure.
Comparison 7 Thiazide versus placebo, Outcome 2 Diastolic blood pressure.
Comparison 7 Thiazide versus placebo, Outcome 3 Serum potassium mmol/L.
Comparison 7 Thiazide versus placebo, Outcome 4 Serum uric acid µmol/L.
Comparison 7 Thiazide versus placebo, Outcome 5 Serum creatinine µmol/L.
Comparison 7 Thiazide versus placebo, Outcome 6 Serum blood glucose mmol/L.
Comparison 7 Thiazide versus placebo, Outcome 7 Serum total cholesterol mmol/L.
Comparison 7 Thiazide versus placebo, Outcome 8 Serum HDL mmol/L.
Comparison 7 Thiazide versus placebo, Outcome 9 Triglycerides.
| Hydrochlorothiazide compared with placebo for primary hypertension | |||||
| Patient or population: adults with primary hypertension Settings: outpatient Intervention: hydrochlorothiazide 3 to 100 mg/day Comparison: placebo | |||||
| Outcomes | Daily dose | MD (95% CI) mmHg | No of participants | Quality of the evidence | Comments |
| Systolic blood pressure | 3 to 6.25 mg | ‐3.6 (‐5.6 to ‐1.5) | 663 (8) | ⊕⊕⊕⊝ moderate1 | Similar to the effect as a second‐line drug (Chen 2009). This is an indirect comparison of the effect size with a wider confidence interval compared to the Chen review, which is based on 22 trials in 3283 patients using similar inclusion/exclusion criteria with systolic blood pressure‐lowering of ‐3.7 (‐4.6 to ‐2.8) mmHg |
| 12.5 mg | ‐6.3 (‐7.2 to ‐5.3) | 2645 (22) | ⊕⊕⊕⊕ | A narrow confidence interval based on a large sample size with a magnitude of lowering very similar to the effect as a second‐line drug (Chen 2009), which was ‐6.0 (‐6.5 to ‐5.4) mmHg | |
| 25 mg | ‐8.0 (‐9.0 to ‐7.0) | 3062 (25) | ⊕⊕⊕⊕ | A narrow confidence interval based on a large sample size with a magnitude of lowering very similar to the effect as a second‐line drug (Chen 2009), which was ‐8.0 (‐8.7 to ‐7.3) mmHg | |
| 50 to 100 mg | ‐10.2 (‐13.1 to ‐7.3) | 315 (2) | ⊕⊕⊝⊝ | The 2 included studies have a high risk of bias. The confidence interval is very wide with small a sample size providing insufficient data in both this review as well as in the Chen review comparing the effect as a second‐line drug (Chen 2009) | |
| Diastolic blood pressure | 3 to 6.25 mg | ‐2.4 (‐3.7 to ‐1.2) | 662 (8) | ⊕⊕⊕⊝ moderate1 | Similar to the effect as a second‐line drug (Chen 2009). This is an indirect comparison of the effect size with a wider confidence interval compared to the Chen review, which is based on 23 trials in 3364 patients using similar inclusion/exclusion criteria, with diastolic blood pressure‐lowering of ‐1.7 (‐2.2 to ‐1.2) mmHg. |
| 12.5 mg | ‐3.1 (‐3.7 to ‐2.5) | 2877 (25) | ⊕⊕⊕⊕ | Similar to the effect as a second‐line drug (Chen 2009), with a narrow confidence interval based on a large sample size with a magnitude of lowering similar to the effect as a second‐line drug (Chen 2009), which was ‐3.1 (‐3.4 to ‐2.8) mmHg | |
| 25 mg | ‐3.3(‐3.8 to ‐2.8) | 3359 (29) | ⊕⊕⊕⊕ | Similar to the effect as a second‐line drug (Chen 2009). This is an indirect comparison of the effect size with a wider confidence interval compared to the Chen review, which is based on 42 trials in 6153 patients using similar inclusion/exclusion criteria with diastolic blood pressure‐lowering of ‐4.0 (‐4.4 to ‐3.6) mmHg | |
| 50 to 100 mg | ‐4.7 (‐6.1 to ‐3.3) | 345 (4) | ⊕⊕⊝⊝ | The 4 included studies had a high risk of bias. The confidence interval is very wide with a small sample size providing insufficient data in both this review as well as in the Chen review comparing the effect as a second‐line drug (Chen 2009) | |
| GRADE Working Group grades of evidence CI: confidence interval; MD: mean difference | |||||
| 1Downgraded due to the small number of patients and wide confidence intervals. | |||||
| Thiazide compared with placebo for primary hypertension | ||||
| Patient or population: adults with primary hypertension Settings: outpatient Intervention: all thiazides^ Comparison: placebo | ||||
| Outcomes | MD (95% CI) mmHg | No of participants | Quality of the evidence | Comments |
| Systolic blood pressure | ‐9.1 (‐9.7 to ‐8.5) | 7733 (47) | ⊕⊕⊕⊕ | At doses achieving maximal effect and above |
| Diastolic blood pressure | ‐3.6 (‐4.0 to ‐3.3) | 8064 (51) | ⊕⊕⊕⊕ | At doses achieving maximal effect and above |
| Withdrawal due to adverse effects | RR 0.64 (95% CI 0.43 to 0.93) | 3698 (20) | ⊕⊝⊝⊝ | See comments 1 and 2 |
| GRADE Working Group grades of evidence | ||||
| ^Includes thiazide and thiazide‐like diuretics. 1Based on a high risk of selective reporting of outcome from 20 out of 40 trials meeting the inclusion criteria. 2Withdrawals due to inclusion of an increase in blood pressure as an adverse effect (AE) was the major reason for withdrawals in the placebo group. | ||||
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Systolic blood pressure Show forest plot | 1 | 257 | Mean Difference (IV, Fixed, 95% CI) | ‐10.43 [‐14.06, ‐6.80] |
| 1.1 1.25 mg/day | 1 | 63 | Mean Difference (IV, Fixed, 95% CI) | ‐7.70 [‐14.98, ‐0.42] |
| 1.2 2.5 mg/day | 1 | 65 | Mean Difference (IV, Fixed, 95% CI) | ‐10.9 [‐18.14, ‐3.66] |
| 1.3 5.0 mg/day | 1 | 65 | Mean Difference (IV, Fixed, 95% CI) | ‐10.6 [‐17.84, ‐3.36] |
| 1.4 10 mg/day | 1 | 64 | Mean Difference (IV, Fixed, 95% CI) | ‐12.5 [‐19.76, ‐5.24] |
| 2 Diastolic blood pressure Show forest plot | 1 | 257 | Mean Difference (IV, Fixed, 95% CI) | ‐6.48 [‐8.82, ‐4.14] |
| 2.1 1.25 mg/day | 1 | 63 | Mean Difference (IV, Fixed, 95% CI) | ‐5.80 [‐10.49, ‐1.11] |
| 2.2 2.5 mg/day | 1 | 65 | Mean Difference (IV, Fixed, 95% CI) | ‐6.9 [‐11.57, ‐2.23] |
| 2.3 5.0 mg/day | 1 | 65 | Mean Difference (IV, Fixed, 95% CI) | ‐6.20 [‐10.87, ‐1.53] |
| 2.4 10 mg/day | 1 | 64 | Mean Difference (IV, Fixed, 95% CI) | ‐7.0 [‐11.68, ‐2.32] |
| 3 Withdrawals due to adverse effects Show forest plot | 1 | 257 | Odds Ratio (M‐H, Fixed, 95% CI) | 0.19 [0.07, 0.57] |
| 3.1 1.25 mg/day | 1 | 63 | Odds Ratio (M‐H, Fixed, 95% CI) | 0.23 [0.03, 1.81] |
| 3.2 2.5 mg/day | 1 | 65 | Odds Ratio (M‐H, Fixed, 95% CI) | 0.22 [0.03, 1.74] |
| 3.3 5.0 mg/day | 1 | 65 | Odds Ratio (M‐H, Fixed, 95% CI) | 0.11 [0.01, 1.30] |
| 3.4 10 mg/day | 1 | 64 | Odds Ratio (M‐H, Fixed, 95% CI) | 0.22 [0.03, 1.77] |
| 4 Serum potassium mmol/L Show forest plot | 1 | 257 | Mean Difference (IV, Fixed, 95% CI) | ‐0.37 [‐0.50, ‐0.24] |
| 4.1 1.25 mg/day | 1 | 63 | Mean Difference (IV, Fixed, 95% CI) | ‐0.25 [‐0.51, 0.01] |
| 4.2 2.5 mg/day | 1 | 65 | Mean Difference (IV, Fixed, 95% CI) | ‐0.29 [‐0.54, ‐0.04] |
| 4.3 5.0 mg/day | 1 | 65 | Mean Difference (IV, Fixed, 95% CI) | ‐0.42 [‐0.68, ‐0.16] |
| 4.4 10 mg/day | 1 | 64 | Mean Difference (IV, Fixed, 95% CI) | ‐0.54 [‐0.80, ‐0.28] |
| 5 Serum uric acid µmol/L Show forest plot | 1 | 257 | Mean Difference (IV, Fixed, 95% CI) | 46.57 [33.24, 59.91] |
| 5.1 1.25 mg/day | 1 | 63 | Mean Difference (IV, Fixed, 95% CI) | 24.0 [0.04, 47.96] |
| 5.2 2.5 mg/day | 1 | 65 | Mean Difference (IV, Fixed, 95% CI) | 34.0 [8.33, 59.67] |
| 5.3 5.0 mg/day | 1 | 65 | Mean Difference (IV, Fixed, 95% CI) | 68.0 [33.12, 102.88] |
| 5.4 10 mg/day | 1 | 64 | Mean Difference (IV, Fixed, 95% CI) | 73.0 [47.55, 98.45] |
| 6 Serum creatinine µmol/L Show forest plot | 1 | 257 | Mean Difference (IV, Fixed, 95% CI) | 5.50 [1.90, 9.11] |
| 6.1 1.25 mg/day | 1 | 63 | Mean Difference (IV, Fixed, 95% CI) | 5.1 [‐1.89, 12.09] |
| 6.2 2.5 mg/day | 1 | 65 | Mean Difference (IV, Fixed, 95% CI) | 5.2 [‐1.63, 12.03] |
| 6.3 5.0 mg/day | 1 | 65 | Mean Difference (IV, Fixed, 95% CI) | 5.00 [‐3.58, 13.58] |
| 6.4 10 mg/day | 1 | 64 | Mean Difference (IV, Fixed, 95% CI) | 6.5 [‐0.33, 13.33] |
| 7 Serum blood glucose mmol/L Show forest plot | 1 | 257 | Mean Difference (IV, Fixed, 95% CI) | 0.13 [‐0.06, 0.33] |
| 7.1 1.25 mg/day | 1 | 63 | Mean Difference (IV, Fixed, 95% CI) | ‐0.11 [‐0.49, 0.27] |
| 7.2 2.5 mg/day | 1 | 65 | Mean Difference (IV, Fixed, 95% CI) | 0.22 [‐0.15, 0.59] |
| 7.3 5.0 mg/day | 1 | 65 | Mean Difference (IV, Fixed, 95% CI) | 0.12 [‐0.25, 0.49] |
| 7.4 10 mg/day | 1 | 64 | Mean Difference (IV, Fixed, 95% CI) | 0.35 [‐0.09, 0.79] |
| 8 Total cholesterol mmol/L Show forest plot | 1 | 257 | Mean Difference (IV, Fixed, 95% CI) | 0.15 [‐0.05, 0.35] |
| 8.1 1.25 mg/day | 1 | 63 | Mean Difference (IV, Fixed, 95% CI) | 0.03 [‐0.38, 0.44] |
| 8.2 2.5 mg/day | 1 | 65 | Mean Difference (IV, Fixed, 95% CI) | 0.06 [‐0.34, 0.46] |
| 8.3 5.0 mg/day | 1 | 65 | Mean Difference (IV, Fixed, 95% CI) | 0.18 [‐0.22, 0.58] |
| 8.4 10 mg/day | 1 | 64 | Mean Difference (IV, Fixed, 95% CI) | 0.31 [‐0.08, 0.70] |
| 9 Triglycerides mmol/L Show forest plot | 1 | 257 | Mean Difference (IV, Fixed, 95% CI) | 0.26 [‐0.06, 0.58] |
| 9.1 1.25 mg/day | 1 | 63 | Mean Difference (IV, Fixed, 95% CI) | 0.13 [‐0.49, 0.75] |
| 9.2 2.5 mg/day | 1 | 65 | Mean Difference (IV, Fixed, 95% CI) | 0.52 [‐0.11, 1.15] |
| 9.3 5.0 mg/day | 1 | 65 | Mean Difference (IV, Fixed, 95% CI) | 0.08 [‐0.65, 0.81] |
| 9.4 10 mg/day | 1 | 64 | Mean Difference (IV, Fixed, 95% CI) | 0.27 [‐0.34, 0.88] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Systolic blood pressure Show forest plot | 1 | 53 | Mean Difference (IV, Fixed, 95% CI) | ‐10.75 [‐18.44, ‐3.05] |
| 1.1 0.05 mg/day | 1 | 17 | Mean Difference (IV, Fixed, 95% CI) | ‐5.3 [‐18.71, 8.11] |
| 1.2 0.125 mg/day | 1 | 19 | Mean Difference (IV, Fixed, 95% CI) | ‐12.0 [‐25.16, 1.16] |
| 1.3 0.5 mg/day | 1 | 17 | Mean Difference (IV, Fixed, 95% CI) | ‐14.90 [‐28.31, ‐1.49] |
| 2 Diastolic blood pressure Show forest plot | 1 | 53 | Mean Difference (IV, Fixed, 95% CI) | ‐6.23 [‐11.19, ‐1.27] |
| 2.1 0.05 mg/day | 1 | 17 | Mean Difference (IV, Fixed, 95% CI) | ‐3.0 [‐11.65, 5.65] |
| 2.2 0.125mg/day | 1 | 19 | Mean Difference (IV, Fixed, 95% CI) | ‐8.60 [‐17.09, ‐0.11] |
| 2.3 0.5 mg/day | 1 | 17 | Mean Difference (IV, Fixed, 95% CI) | ‐7.0 [‐15.65, 1.65] |
| 3 Serum potassium µmol/L Show forest plot | 1 | 53 | Mean Difference (IV, Fixed, 95% CI) | ‐0.18 [‐0.42, 0.07] |
| 3.1 0.05 mg/day | 1 | 17 | Mean Difference (IV, Fixed, 95% CI) | 0.10 [‐0.35, 0.55] |
| 3.2 0.125 mg/day | 1 | 19 | Mean Difference (IV, Fixed, 95% CI) | ‐0.10 [‐0.52, 0.32] |
| 3.3 0.5 mg/day | 1 | 17 | Mean Difference (IV, Fixed, 95% CI) | ‐0.50 [‐0.92, ‐0.08] |
| 4 Serum uric acid µmol/L Show forest plot | 1 | 53 | Mean Difference (IV, Fixed, 95% CI) | 19.54 [‐36.18, 75.26] |
| 4.1 0.05 mg/day | 1 | 17 | Mean Difference (IV, Fixed, 95% CI) | 30.0 [‐70.86, 130.86] |
| 4.2 0.125 mg/day | 1 | 19 | Mean Difference (IV, Fixed, 95% CI) | 20.0 [‐75.05, 115.05] |
| 4.3 0.5 mg/day | 1 | 17 | Mean Difference (IV, Fixed, 95% CI) | 10.0 [‐84.04, 104.04] |
| 5 Total cholesterol mmol/L Show forest plot | 1 | 47 | Mean Difference (IV, Fixed, 95% CI) | 0.79 [0.36, 1.23] |
| 5.1 0.05 mg/day | 1 | 16 | Mean Difference (IV, Fixed, 95% CI) | 1.2 [0.46, 1.94] |
| 5.2 0.125 mg/day | 1 | 17 | Mean Difference (IV, Fixed, 95% CI) | 0.90 [0.17, 1.63] |
| 5.3 0.5 mg/day | 1 | 14 | Mean Difference (IV, Fixed, 95% CI) | 0.2 [‐0.60, 1.00] |
| 6 Triglycerides mmol/L Show forest plot | 1 | 48 | Mean Difference (IV, Fixed, 95% CI) | 0.20 [‐0.17, 0.57] |
| 6.1 0.05 mg/day | 1 | 16 | Mean Difference (IV, Fixed, 95% CI) | 0.3 [‐0.30, 0.90] |
| 6.2 0.125 mg/day | 1 | 17 | Mean Difference (IV, Fixed, 95% CI) | 0.4 [‐0.29, 1.09] |
| 6.3 0.5 mg/day | 1 | 15 | Mean Difference (IV, Fixed, 95% CI) | ‐0.1 [‐0.74, 0.54] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Systolic blood pressure Show forest plot | 1 | 105 | Mean Difference (IV, Fixed, 95% CI) | ‐11.63 [‐16.89, ‐6.38] |
| 1.1 0.5 mg/day | 1 | 35 | Mean Difference (IV, Fixed, 95% CI) | ‐11.40 [‐20.50, ‐2.30] |
| 1.2 1.0 mg/day | 1 | 34 | Mean Difference (IV, Fixed, 95% CI) | ‐11.6 [‐20.75, ‐2.45] |
| 1.3 2 mg/day | 1 | 36 | Mean Difference (IV, Fixed, 95% CI) | ‐11.90 [‐20.95, ‐2.85] |
| 2 Diastolic blood pressure Show forest plot | 1 | 105 | Mean Difference (IV, Fixed, 95% CI) | ‐5.83 [‐9.22, ‐2.44] |
| 2.1 0.5 mg/day | 1 | 35 | Mean Difference (IV, Fixed, 95% CI) | ‐5.90 [‐11.77, ‐0.03] |
| 2.2 1.0 mg/day | 1 | 34 | Mean Difference (IV, Fixed, 95% CI) | ‐6.40 [‐12.30, ‐0.50] |
| 2.3 2.0 mg/day | 1 | 36 | Mean Difference (IV, Fixed, 95% CI) | ‐5.2 [‐11.04, 0.64] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Systolic blood pressure Show forest plot | 7 | 1153 | Mean Difference (IV, Fixed, 95% CI) | ‐11.98 [‐13.71, ‐10.24] |
| 1.1 12.5 to 15 mg/day | 3 | 185 | Mean Difference (IV, Fixed, 95% CI) | ‐10.09 [‐13.88, ‐6.30] |
| 1.2 25 to 30 mg/day | 5 | 752 | Mean Difference (IV, Fixed, 95% CI) | ‐13.64 [‐16.03, ‐11.25] |
| 1.3 45 to 50 mg/day | 4 | 192 | Mean Difference (IV, Fixed, 95% CI) | ‐9.92 [‐13.44, ‐6.39] |
| 1.4 75 mg/day | 1 | 24 | Mean Difference (IV, Fixed, 95% CI) | ‐12.9 [‐24.65, ‐1.15] |
| 2 Diastolic blood pressure Show forest plot | 7 | 1153 | Mean Difference (IV, Fixed, 95% CI) | ‐3.93 [‐5.13, ‐2.74] |
| 2.1 12.5 to 15 mg/day | 3 | 185 | Mean Difference (IV, Fixed, 95% CI) | ‐2.55 [‐5.07, ‐0.02] |
| 2.2 25 to 30 mg/day | 5 | 752 | Mean Difference (IV, Fixed, 95% CI) | ‐3.98 [‐5.69, ‐2.28] |
| 2.3 45 to 50 mg/day | 4 | 192 | Mean Difference (IV, Fixed, 95% CI) | ‐4.91 [‐7.29, ‐2.53] |
| 2.4 75 mg/day | 1 | 24 | Mean Difference (IV, Fixed, 95% CI) | ‐5.5 [‐13.08, 2.08] |
| 3 Withdrawals due to adverse events Show forest plot | 5 | 1058 | Odds Ratio (M‐H, Fixed, 95% CI) | 0.49 [0.28, 0.87] |
| 3.1 12.5 mg/day | 2 | 85 | Odds Ratio (M‐H, Fixed, 95% CI) | 0.11 [0.02, 0.55] |
| 3.2 15 mg/day | 1 | 101 | Odds Ratio (M‐H, Fixed, 95% CI) | 0.79 [0.13, 4.94] |
| 3.3 25 mg/day | 4 | 732 | Odds Ratio (M‐H, Fixed, 95% CI) | 0.68 [0.29, 1.57] |
| 3.4 45 mg/day | 1 | 34 | Odds Ratio (M‐H, Fixed, 95% CI) | 0.0 [0.0, 0.0] |
| 3.5 50 mg/day | 2 | 81 | Odds Ratio (M‐H, Fixed, 95% CI) | 0.42 [0.13, 1.43] |
| 3.6 75 mg/day | 1 | 25 | Odds Ratio (M‐H, Fixed, 95% CI) | 0.85 [0.03, 23.82] |
| 4 Serum potassium mmol/L Show forest plot | 5 | 1203 | Mean Difference (IV, Fixed, 95% CI) | ‐0.40 [‐0.45, ‐0.34] |
| 4.1 12.5 to 15 mg/day | 3 | 235 | Mean Difference (IV, Fixed, 95% CI) | ‐0.40 [‐0.51, ‐0.29] |
| 4.2 25 to 30 mg/day | 4 | 784 | Mean Difference (IV, Fixed, 95% CI) | ‐0.26 [‐0.34, ‐0.19] |
| 4.3 50 mg/day | 3 | 161 | Mean Difference (IV, Fixed, 95% CI) | ‐0.77 [‐0.90, ‐0.64] |
| 4.4 75 mg/day | 1 | 23 | Mean Difference (IV, Fixed, 95% CI) | ‐0.6 [‐1.05, ‐0.15] |
| 5 Serum uric acid µmol/L Show forest plot | 2 | 285 | Mean Difference (IV, Fixed, 95% CI) | 64.16 [45.69, 82.63] |
| 5.1 12.5 to 15 mg/day | 2 | 120 | Mean Difference (IV, Fixed, 95% CI) | 52.71 [26.11, 79.30] |
| 5.2 25 to 30 mg/day | 2 | 120 | Mean Difference (IV, Fixed, 95% CI) | 74.44 [47.85, 101.02] |
| 5.3 50 mg/day | 1 | 22 | Mean Difference (IV, Fixed, 95% CI) | 90.0 [‐47.99, 227.99] |
| 5.4 75 mg/day | 1 | 23 | Mean Difference (IV, Fixed, 95% CI) | 70.0 [‐70.76, 210.76] |
| 6 Serum blood glucose mmol/L Show forest plot | 3 | 394 | Mean Difference (IV, Fixed, 95% CI) | 0.34 [0.12, 0.55] |
| 6.1 12.5 mg/day | 1 | 24 | Mean Difference (IV, Fixed, 95% CI) | 0.17 [‐0.68, 1.02] |
| 6.2 15 mg/day | 1 | 122 | Mean Difference (IV, Fixed, 95% CI) | 0.05 [‐0.33, 0.43] |
| 6.3 25 to 30 mg/day | 2 | 147 | Mean Difference (IV, Fixed, 95% CI) | 0.58 [0.23, 0.93] |
| 6.4 50 mg/day | 2 | 77 | Mean Difference (IV, Fixed, 95% CI) | 0.43 [‐0.04, 0.91] |
| 6.5 75 mg/day | 1 | 24 | Mean Difference (IV, Fixed, 95% CI) | 0.17 [‐0.61, 0.95] |
| 7 HDL cholesterol mmol/L Show forest plot | 1 | Mean Difference (IV, Fixed, 95% CI) | Subtotals only | |
| 7.1 45 mg/day | 1 | 34 | Mean Difference (IV, Fixed, 95% CI) | ‐0.11 [‐0.22, 0.00] |
| 8 Total cholesterol mmol/L Show forest plot | 2 | 213 | Mean Difference (IV, Fixed, 95% CI) | 0.41 [0.18, 0.64] |
| 8.1 15 mg/day | 1 | 88 | Mean Difference (IV, Fixed, 95% CI) | 0.23 [‐0.24, 0.70] |
| 8.2 25 mg/day | 1 | 91 | Mean Difference (IV, Fixed, 95% CI) | 0.20 [‐0.24, 0.64] |
| 8.3 45 mg/day | 1 | 34 | Mean Difference (IV, Fixed, 95% CI) | 0.6 [0.28, 0.92] |
| 9 Triglycerides mmol/L Show forest plot | 1 | Mean Difference (IV, Fixed, 95% CI) | Subtotals only | |
| 9.1 45 mg/day | 1 | 34 | Mean Difference (IV, Fixed, 95% CI) | 0.69 [0.05, 1.33] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Systolic blood pressure Show forest plot | 35 | 6725 | Mean Difference (IV, Fixed, 95% CI) | ‐6.94 [‐7.56, ‐6.31] |
| 1.1 3.0 to 6.25 mg/day | 8 | 663 | Mean Difference (IV, Fixed, 95% CI) | ‐3.56 [‐5.57, ‐1.54] |
| 1.2 12.5 mg/day | 22 | 2645 | Mean Difference (IV, Fixed, 95% CI) | ‐6.27 [‐7.24, ‐5.31] |
| 1.3 25 mg/day | 25 | 3062 | Mean Difference (IV, Fixed, 95% CI) | ‐6.00 [‐8.96, ‐7.04] |
| 1.4 37.5 mg/day | 1 | 40 | Mean Difference (IV, Fixed, 95% CI) | ‐7.30 [‐16.30, 1.70] |
| 1.5 50 mg/day | 2 | 169 | Mean Difference (IV, Fixed, 95% CI) | ‐10.47 [‐14.60, ‐6.35] |
| 1.6 100 mg/day | 2 | 146 | Mean Difference (IV, Fixed, 95% CI) | ‐9.91 [‐14.05, ‐5.77] |
| 2 Diastolic blood pressure Show forest plot | 39 | 7284 | Mean Difference (IV, Fixed, 95% CI) | ‐3.25 [‐3.59, ‐2.90] |
| 2.1 3.0 to 6.25 mg/day | 8 | 663 | Mean Difference (IV, Fixed, 95% CI) | ‐2.43 [‐3.67, ‐1.19] |
| 2.2 12.5 mg/day | 25 | 2877 | Mean Difference (IV, Fixed, 95% CI) | ‐3.12 [‐3.71, ‐2.53] |
| 2.3 25 mg/day | 29 | 3359 | Mean Difference (IV, Fixed, 95% CI) | ‐3.28 [‐3.77, ‐2.79] |
| 2.4 37.5 mg/day | 1 | 40 | Mean Difference (IV, Fixed, 95% CI) | ‐3.7 [‐9.26, 1.86] |
| 2.5 50 mg/day | 3 | 199 | Mean Difference (IV, Fixed, 95% CI) | ‐4.97 [‐6.65, ‐3.30] |
| 2.6 100 mg/day | 2 | 146 | Mean Difference (IV, Fixed, 95% CI) | ‐3.90 [‐6.57, ‐1.23] |
| 3 Withdrawals due to adverse events Show forest plot | 20 | 3698 | Odds Ratio (M‐H, Fixed, 95% CI) | 0.64 [0.43, 0.93] |
| 3.1 3.0 mg/day | 1 | 27 | Odds Ratio (M‐H, Fixed, 95% CI) | 0.0 [0.0, 0.0] |
| 3.2 6.0 mg/day | 1 | 27 | Odds Ratio (M‐H, Fixed, 95% CI) | 0.0 [0.0, 0.0] |
| 3.3 6.25 mg/day | 1 | 259 | Odds Ratio (M‐H, Fixed, 95% CI) | 0.09 [0.02, 0.43] |
| 3.4 12.5 mg/day | 11 | 1676 | Odds Ratio (M‐H, Fixed, 95% CI) | 0.94 [0.53, 1.67] |
| 3.5 25 mg/day | 12 | 1645 | Odds Ratio (M‐H, Fixed, 95% CI) | 0.60 [0.33, 1.07] |
| 3.6 50 mg/day | 1 | 64 | Odds Ratio (M‐H, Fixed, 95% CI) | 0.45 [0.04, 5.26] |
| 4 Serum potassium mmol/L Show forest plot | 11 | 2036 | Mean Difference (IV, Fixed, 95% CI) | ‐0.22 [‐0.25, ‐0.18] |
| 4.1 3.0 mg/day | 1 | 27 | Mean Difference (IV, Fixed, 95% CI) | 0.03 [‐0.26, 0.32] |
| 4.2 6.0 mg/day | 1 | 27 | Mean Difference (IV, Fixed, 95% CI) | 0.10 [‐0.22, 0.42] |
| 4.3 12.5 mg/day | 7 | 1026 | Mean Difference (IV, Fixed, 95% CI) | ‐0.16 [‐0.21, ‐0.11] |
| 4.4 25 mg/day | 7 | 805 | Mean Difference (IV, Fixed, 95% CI) | ‐0.30 [‐0.36, ‐0.24] |
| 4.5 50 mg/day | 2 | 151 | Mean Difference (IV, Fixed, 95% CI) | ‐0.48 [‐0.68, ‐0.29] |
| 5 Serum uric acid µmol/L Show forest plot | 5 | 1043 | Mean Difference (IV, Fixed, 95% CI) | 32.88 [26.12, 39.65] |
| 5.1 3.0 mg/day | 1 | 27 | Mean Difference (IV, Fixed, 95% CI) | 10.40 [‐26.09, 46.89] |
| 5.2 6.0 mg/day | 1 | 27 | Mean Difference (IV, Fixed, 95% CI) | 11.5 [‐29.31, 52.31] |
| 5.3 12.5 mg/day | 5 | 835 | Mean Difference (IV, Fixed, 95% CI) | 33.00 [25.67, 40.34] |
| 5.4 25 mg/day | 2 | 154 | Mean Difference (IV, Fixed, 95% CI) | 47.15 [24.38, 69.92] |
| 6 Serum creatinine µmol/L Show forest plot | 3 | 527 | Mean Difference (IV, Fixed, 95% CI) | 0.32 [‐2.63, 3.26] |
| 6.1 12.5 mg/day | 2 | 258 | Mean Difference (IV, Fixed, 95% CI) | 0.57 [‐3.40, 4.55] |
| 6.2 25 mg/day | 2 | 269 | Mean Difference (IV, Fixed, 95% CI) | 0.0 [‐4.38, 4.38] |
| 6.3 50 mg/day | 0 | 0 | Mean Difference (IV, Fixed, 95% CI) | 0.0 [0.0, 0.0] |
| 7 Serum blood glucose mmol/L Show forest plot | 6 | 1041 | Mean Difference (IV, Fixed, 95% CI) | ‐0.12 [‐0.24, 0.01] |
| 7.1 3.0 mg/day | 1 | 27 | Mean Difference (IV, Fixed, 95% CI) | 0.07 [‐0.45, 0.59] |
| 7.2 6.0 mg/day | 1 | 27 | Mean Difference (IV, Fixed, 95% CI) | ‐0.18 [‐0.68, 0.32] |
| 7.3 12.5 mg/day | 4 | 605 | Mean Difference (IV, Fixed, 95% CI) | ‐0.10 [‐0.33, 0.13] |
| 7.4 25 mg/day | 3 | 298 | Mean Difference (IV, Fixed, 95% CI) | 0.01 [‐0.19, 0.20] |
| 7.5 50 mg/day | 1 | 84 | Mean Difference (IV, Fixed, 95% CI) | ‐0.5 [‐0.81, ‐0.19] |
| 8 HDL cholesterol mmol/L Show forest plot | 1 | 159 | Mean Difference (IV, Fixed, 95% CI) | ‐0.17 [‐0.53, 0.19] |
| 8.1 3.0 mg/day | 1 | 44 | Mean Difference (IV, Fixed, 95% CI) | ‐0.52 [‐1.15, 0.11] |
| 8.2 6.0 mg/day | 1 | 44 | Mean Difference (IV, Fixed, 95% CI) | 0.01 [‐0.68, 0.70] |
| 8.3 12.5 mg/day | 1 | 32 | Mean Difference (IV, Fixed, 95% CI) | ‐0.12 [‐1.03, 0.79] |
| 8.4 25 mg/day | 1 | 39 | Mean Difference (IV, Fixed, 95% CI) | 0.04 [‐0.67, 0.75] |
| 9 Total cholesterol mmol/L Show forest plot | 4 | 450 | Mean Difference (IV, Fixed, 95% CI) | 0.20 [0.17, 0.22] |
| 9.1 3.0 mg/day | 1 | 27 | Mean Difference (IV, Fixed, 95% CI) | ‐0.4 [‐0.67, ‐0.13] |
| 9.2 6.0 mg/day | 1 | 27 | Mean Difference (IV, Fixed, 95% CI) | 0.06 [‐1.15, 1.27] |
| 9.3 12.5 mg/day | 2 | 159 | Mean Difference (IV, Fixed, 95% CI) | ‐0.04 [‐0.34, 0.27] |
| 9.4 25 mg/day | 3 | 237 | Mean Difference (IV, Fixed, 95% CI) | 0.21 [0.18, 0.23] |
| 10 Triglycerides mmol/L Show forest plot | 2 | 255 | Mean Difference (IV, Fixed, 95% CI) | 0.09 [‐0.11, 0.30] |
| 10.1 3.0 mg/day | 1 | 27 | Mean Difference (IV, Fixed, 95% CI) | 0.03 [‐0.67, 0.73] |
| 10.2 6.0 mg/day | 1 | 27 | Mean Difference (IV, Fixed, 95% CI) | ‐0.17 [‐0.88, 0.54] |
| 10.3 12.5 mg/day | 1 | 28 | Mean Difference (IV, Fixed, 95% CI) | ‐0.11 [‐0.77, 0.55] |
| 10.4 25 mg/day | 2 | 173 | Mean Difference (IV, Fixed, 95% CI) | 0.16 [‐0.08, 0.40] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Systolic blood pressure Show forest plot | 10 | 2104 | Mean Difference (IV, Fixed, 95% CI) | ‐8.69 [‐9.96, ‐7.42] |
| 1.1 1.0 mg/day | 1 | 29 | Mean Difference (IV, Fixed, 95% CI) | ‐9.7 [‐19.89, 0.49] |
| 1.2 1.25 mg/day | 4 | 736 | Mean Difference (IV, Fixed, 95% CI) | ‐7.37 [‐9.21, ‐5.54] |
| 1.3 1.5 mg/day | 2 | 955 | Mean Difference (IV, Fixed, 95% CI) | ‐9.40 [‐11.51, ‐7.29] |
| 1.4 2.0 mg/day | 1 | 74 | Mean Difference (IV, Fixed, 95% CI) | ‐8.7 [‐17.38, ‐0.02] |
| 1.5 2.5 mg/day | 5 | 281 | Mean Difference (IV, Fixed, 95% CI) | ‐11.94 [‐15.88, ‐7.99] |
| 1.6 5.0 mg/day | 1 | 29 | Mean Difference (IV, Fixed, 95% CI) | ‐9.6 [‐19.40, 0.20] |
| 2 Diastolic blood pressure Show forest plot | 10 | 2104 | Mean Difference (IV, Fixed, 95% CI) | ‐3.88 [‐4.63, ‐3.14] |
| 2.1 1.0 mg/day | 1 | 29 | Mean Difference (IV, Fixed, 95% CI) | ‐1.00 [‐9.57, 3.57] |
| 2.2 1.25 mg/day | 4 | 736 | Mean Difference (IV, Fixed, 95% CI) | ‐3.55 [‐4.57, ‐2.52] |
| 2.3 1.5 mg/day | 2 | 955 | Mean Difference (IV, Fixed, 95% CI) | ‐4.05 [‐5.38, ‐2.72] |
| 2.4 2.0 mg/day | 1 | 74 | Mean Difference (IV, Fixed, 95% CI) | ‐3.60 [‐8.27, 1.07] |
| 2.5 2.5 mg/day | 5 | 281 | Mean Difference (IV, Fixed, 95% CI) | ‐5.32 [‐7.65, ‐2.98] |
| 2.6 5.0 mg/day | 1 | 29 | Mean Difference (IV, Fixed, 95% CI) | ‐4.0 [‐10.32, 2.32] |
| 3 Withdrawals due to adverse events Show forest plot | 6 | 1874 | Odds Ratio (M‐H, Fixed, 95% CI) | 0.83 [0.49, 1.42] |
| 3.1 1.0 mg/day | 1 | 29 | Odds Ratio (M‐H, Fixed, 95% CI) | 0.10 [0.00, 2.66] |
| 3.2 1.25 mg/day | 3 | 621 | Odds Ratio (M‐H, Fixed, 95% CI) | 1.07 [0.49, 2.32] |
| 3.3 1.5 mg/day | 2 | 958 | Odds Ratio (M‐H, Fixed, 95% CI) | 0.73 [0.30, 1.79] |
| 3.4 2.0 mg/day | 1 | 74 | Odds Ratio (M‐H, Fixed, 95% CI) | 0.33 [0.02, 5.61] |
| 3.5 2.5 mg/day | 2 | 163 | Odds Ratio (M‐H, Fixed, 95% CI) | 1.06 [0.17, 6.53] |
| 3.6 5.0 mg/day | 1 | 29 | Odds Ratio (M‐H, Fixed, 95% CI) | 0.0 [0.0, 0.0] |
| 4 Serum potassium mmol/L Show forest plot | 5 | 541 | Mean Difference (IV, Fixed, 95% CI) | ‐0.32 [‐0.38, ‐0.26] |
| 4.1 1.0 mg/day | 1 | 31 | Mean Difference (IV, Fixed, 95% CI) | ‐0.11 [‐0.38, 0.16] |
| 4.2 1.25 mg/day | 2 | 398 | Mean Difference (IV, Fixed, 95% CI) | ‐0.30 [‐0.37, ‐0.23] |
| 4.3 2.5 mg/day | 3 | 81 | Mean Difference (IV, Fixed, 95% CI) | ‐0.41 [‐0.57, ‐0.26] |
| 4.4 5.0 mg/day | 1 | 31 | Mean Difference (IV, Fixed, 95% CI) | ‐0.65 [‐0.91, ‐0.39] |
| 5 Serum uric acid µmol/L Show forest plot | 4 | 558 | Mean Difference (IV, Fixed, 95% CI) | 39.81 [33.54, 46.08] |
| 5.1 1.0 mg/day | 1 | 46 | Mean Difference (IV, Fixed, 95% CI) | 35.69 [14.15, 57.23] |
| 5.2 1.25 mg/day | 2 | 398 | Mean Difference (IV, Fixed, 95% CI) | 36.14 [28.70, 43.58] |
| 5.3 2.5 mg/day | 2 | 69 | Mean Difference (IV, Fixed, 95% CI) | 52.33 [34.50, 70.16] |
| 5.4 5.0 mg/day | 1 | 45 | Mean Difference (IV, Fixed, 95% CI) | 57.17 [35.17, 79.17] |
| 6 Serum creatinine µmol/L Show forest plot | 1 | Mean Difference (IV, Fixed, 95% CI) | Subtotals only | |
| 6.1 1.25 mg/day | 1 | 203 | Mean Difference (IV, Fixed, 95% CI) | 0.0 [‐2.43, 2.43] |
| 7 Serum blood glucose mmol/L Show forest plot | 3 | 490 | Mean Difference (IV, Fixed, 95% CI) | 0.13 [‐0.11, 0.37] |
| 7.1 1.0 mg/day | 1 | 31 | Mean Difference (IV, Fixed, 95% CI) | ‐0.08 [‐1.12, 0.96] |
| 7.2 1.25 mg/day | 2 | 398 | Mean Difference (IV, Fixed, 95% CI) | 0.12 [‐0.14, 0.38] |
| 7.3 2.5 mg/day | 1 | 30 | Mean Difference (IV, Fixed, 95% CI) | 0.34 [‐0.71, 1.39] |
| 7.4 5.0 mg/day | 1 | 31 | Mean Difference (IV, Fixed, 95% CI) | 0.25 [‐0.76, 1.26] |
| 8 HDL cholesterol mmol/L Show forest plot | 1 | Mean Difference (IV, Fixed, 95% CI) | Subtotals only | |
| 8.1 1.25 mg/day | 1 | 203 | Mean Difference (IV, Fixed, 95% CI) | ‐0.05 [‐0.10, ‐0.00] |
| 9 Total cholesterol mmol/L Show forest plot | 2 | Mean Difference (IV, Fixed, 95% CI) | Subtotals only | |
| 9.1 1.25 mg/day | 2 | 398 | Mean Difference (IV, Fixed, 95% CI) | 0.12 [0.02, 0.23] |
| 10 Triglycerides mmol/L Show forest plot | 1 | Mean Difference (IV, Fixed, 95% CI) | Subtotals only | |
| 10.1 1.25 mg/day | 1 | 203 | Mean Difference (IV, Fixed, 95% CI) | 0.23 [0.03, 0.43] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Systolic blood pressure Show forest plot | 47 | 7733 | Mean Difference (IV, Fixed, 95% CI) | ‐9.14 [‐9.76, ‐8.51] |
| 1.1 Bendrofluazide 1.25 to 10 mg/day | 1 | 257 | Mean Difference (IV, Fixed, 95% CI) | ‐10.4 [‐14.03, ‐6.77] |
| 1.2 Chlorthalidone 12.5 to 100 mg/day | 7 | 1167 | Mean Difference (IV, Fixed, 95% CI) | ‐11.79 [‐13.50, ‐10.08] |
| 1.3 Cyclopenthiazide 0.5 mg/day | 1 | 25 | Mean Difference (IV, Fixed, 95% CI) | ‐14.90 [‐24.58, ‐5.22] |
| 1.4 Hydrochlorothiazide 25 to 100 mg/day | 28 | 4099 | Mean Difference (IV, Fixed, 95% CI) | ‐8.68 [‐9.49, ‐7.87] |
| 1.5 Indapamide 1.0 to 5.0 mg/day | 9 | 2080 | Mean Difference (IV, Fixed, 95% CI) | ‐8.37 [‐9.65, ‐7.10] |
| 1.6 Metolazone 0.5 to 2.0 mg/day | 1 | 105 | Mean Difference (IV, Fixed, 95% CI) | ‐11.7 [‐16.95, ‐6.45] |
| 2 Diastolic blood pressure Show forest plot | 51 | 8064 | Mean Difference (IV, Fixed, 95% CI) | ‐3.63 [‐3.97, ‐3.28] |
| 2.1 Bendrofluazide 1.25 to 10 mg/day | 1 | 257 | Mean Difference (IV, Fixed, 95% CI) | ‐6.5 [‐8.84, ‐4.16] |
| 2.2 Chlorthalidone 12.5 to 100 mg/day | 7 | 1153 | Mean Difference (IV, Fixed, 95% CI) | ‐3.07 [‐4.27, ‐1.87] |
| 2.3 Cyclopenthiazide 0.125 to 0.5 mg/day | 1 | 40 | Mean Difference (IV, Fixed, 95% CI) | ‐7.90 [‐13.16, ‐2.64] |
| 2.4 Hydrochlorothiazide 25 to 100 mg/day | 32 | 4429 | Mean Difference (IV, Fixed, 95% CI) | ‐3.50 [‐3.92, ‐3.08] |
| 2.5 Indapamide 1.0 to 5.0 mg/day | 9 | 2080 | Mean Difference (IV, Fixed, 95% CI) | ‐3.77 [‐4.52, ‐3.02] |
| 2.6 Metolazone 0.5 to 2.0 mg/day | 1 | 105 | Mean Difference (IV, Fixed, 95% CI) | ‐5.8 [‐9.19, ‐2.41] |
| 3 Serum potassium mmol/L Show forest plot | 22 | 3868 | Mean Difference (IV, Fixed, 95% CI) | ‐0.25 [‐0.28, ‐0.22] |
| 3.1 BDFZ 1.25 to 10 mg/day | 1 | 218 | Mean Difference (IV, Fixed, 95% CI) | ‐0.38 [‐0.62, ‐0.14] |
| 3.2 CTD 12.5 to 100 mg/day | 4 | 1068 | Mean Difference (IV, Fixed, 95% CI) | ‐0.32 [‐0.38, ‐0.25] |
| 3.3 CYPTZ 0.05 to 0.5 mg/day | 1 | 53 | Mean Difference (IV, Fixed, 95% CI) | ‐0.20 [‐0.45, 0.05] |
| 3.4 HCTZ 3 to 100 mg/day | 11 | 1988 | Mean Difference (IV, Fixed, 95% CI) | ‐0.21 [‐0.24, ‐0.17] |
| 3.5 IND 1.0 to 5 mg/day | 5 | 541 | Mean Difference (IV, Fixed, 95% CI) | ‐0.32 [‐0.38, ‐0.26] |
| 4 Serum uric acid µmol/L Show forest plot | 13 | 2332 | Mean Difference (IV, Fixed, 95% CI) | 38.22 [34.24, 42.20] |
| 4.1 BDFZ 1.25 to 10 mg/day | 1 | 255 | Mean Difference (IV, Fixed, 95% CI) | 50.30 [36.62, 63.98] |
| 4.2 CTD 12.5 to 100 mg/day | 2 | 285 | Mean Difference (IV, Fixed, 95% CI) | 64.74 [46.11, 83.36] |
| 4.3 CYPTZ 0.05 to 0.5 mg/day | 1 | 53 | Mean Difference (IV, Fixed, 95% CI) | 20.0 [‐35.61, 75.61] |
| 4.4 HCTZ 3 to 100 mg/day | 5 | 1122 | Mean Difference (IV, Fixed, 95% CI) | 32.74 [26.00, 39.48] |
| 4.5 IND 1.0 to 5 mg/day | 4 | 617 | Mean Difference (IV, Fixed, 95% CI) | 37.77 [32.23, 43.31] |
| 5 Serum creatinine µmol/L Show forest plot | 5 | 987 | Mean Difference (IV, Fixed, 95% CI) | 1.34 [‐0.31, 2.99] |
| 5.1 BDFZ 1.25 to 10 mg/day | 1 | 257 | Mean Difference (IV, Fixed, 95% CI) | 5.45 [1.99, 8.91] |
| 5.2 HCTZ 3 to 100 mg/day | 3 | 527 | Mean Difference (IV, Fixed, 95% CI) | 0.32 [‐2.63, 3.26] |
| 5.3 IND 1.0 to 5 mg/day | 1 | 203 | Mean Difference (IV, Fixed, 95% CI) | 0.0 [‐2.43, 2.43] |
| 6 Serum blood glucose mmol/L Show forest plot | 12 | 1989 | Mean Difference (IV, Fixed, 95% CI) | 0.03 [‐0.05, 0.12] |
| 6.1 BDFZ 1.25 to 10 mg/day | 1 | 257 | Mean Difference (IV, Fixed, 95% CI) | 0.15 [‐0.04, 0.34] |
| 6.2 CTD 12.5 to 100 mg/day | 3 | 326 | Mean Difference (IV, Fixed, 95% CI) | 0.37 [0.14, 0.60] |
| 6.3 HCTZ 3 to 100 mg/day | 6 | 916 | Mean Difference (IV, Fixed, 95% CI) | ‐0.11 [‐0.22, 0.01] |
| 6.4 IND 1.0 to 5 mg/day | 3 | 490 | Mean Difference (IV, Fixed, 95% CI) | 0.13 [‐0.11, 0.37] |
| 7 Serum total cholesterol mmol/L Show forest plot | 11 | 1394 | Mean Difference (IV, Fixed, 95% CI) | 0.21 [0.18, 0.23] |
| 7.1 BDFZ 1.25 to 10 mg/day | 1 | 257 | Mean Difference (IV, Fixed, 95% CI) | 0.44 [0.24, 0.64] |
| 7.2 CTD 12.5 to 100 mg/day | 2 | 213 | Mean Difference (IV, Fixed, 95% CI) | 0.41 [0.18, 0.63] |
| 7.3 CYPTZ 0.05 to 0.5 mg/day | 1 | 53 | Mean Difference (IV, Fixed, 95% CI) | 0.79 [0.37, 1.21] |
| 7.4 HCTZ 3 to 100 mg/day | 4 | 450 | Mean Difference (IV, Fixed, 95% CI) | 0.20 [0.18, 0.23] |
| 7.5 IND 1.0 to 5 mg/day | 3 | 421 | Mean Difference (IV, Fixed, 95% CI) | 0.11 [0.01, 0.21] |
| 8 Serum HDL mmol/L Show forest plot | 3 | 348 | Mean Difference (IV, Fixed, 95% CI) | ‐0.06 [‐0.10, ‐0.02] |
| 8.1 CTD 45 mg/day | 1 | 34 | Mean Difference (IV, Fixed, 95% CI) | ‐0.11 [‐0.22, 0.00] |
| 8.2 HCTZ 25mg/day | 1 | 111 | Mean Difference (IV, Fixed, 95% CI) | ‐0.15 [‐0.72, 0.42] |
| 8.3 IND 1.25 mg/day | 1 | 203 | Mean Difference (IV, Fixed, 95% CI) | ‐0.05 [‐0.10, ‐0.00] |
| 9 Triglycerides Show forest plot | 6 | 697 | Mean Difference (IV, Fixed, 95% CI) | 0.21 [0.08, 0.33] |
| 9.1 BDFZ 1.25 to 10 mg/day | 1 | 157 | Mean Difference (IV, Fixed, 95% CI) | 0.37 [‐0.09, 0.83] |
| 9.2 CTD 45 mg/day | 1 | 34 | Mean Difference (IV, Fixed, 95% CI) | 0.69 [0.05, 1.33] |
| 9.3 CYPTZ 0.5 mg/day | 1 | 48 | Mean Difference (IV, Fixed, 95% CI) | 0.21 [‐0.16, 0.58] |
| 9.4 HCTZ 25 mg/day | 2 | 255 | Mean Difference (IV, Fixed, 95% CI) | 0.09 [‐0.11, 0.30] |
| 9.5 IND 1.25 mg/day | 1 | 203 | Mean Difference (IV, Fixed, 95% CI) | 0.23 [0.03, 0.43] |
