Hydrochlorothiazide vs chlorthalidone, indapamide, and potassium‐sparing/hydrochlorothiazide diuretics for reducing left ventricular hypertrophy: A systematic review and meta‐analysis

Abstract

Left ventricular hypertrophy develops in 36%‐41% of hypertensive patients and independently predicts cardiovascular events and total mortality. Moreover, drug‐induced reduction in left ventricular mass (LVM) correlates with improved prognosis. The optimal thiazide‐type diuretic for reducing LVM is unknown. Evidence regarding potency, cardiovascular events, sodium, and potassium suggested the hypothesis that “CHIP” diuretics (CHlorthalidone, Indapamide, and Potassium‐sparing diuretic/hydrochlorothiazide [PSD/HCTZ]) would reduce LVM more than HCTZ. Systematic searches of five databases were conducted. Among the 38 randomized trials, a 1% reduction in systolic blood pressure (SBP) predicted a 1% reduction in LVM, P = 0.00001. CHIP‐HCTZ differences in reducing LVM differed across trials (ie, heterogeneity), making interpretation uncertain. However, among the 28 double‐blind trials, heterogeneity was undetectable, and HCTZ reduced LVM (percent reduction [95% CI]) by −7.3 (−10.4, −4.2), P < 0.0001. CHIP diuretics surpassed HCTZ in reducing LVM: chlorthalidone −8.2 (−14.7, −1.6), P = 0.015; indapamide −7.5 (−12.7, −2.3), P = 0.005; and all CHIP diuretics combined −7.7 (−12.2, −3.1), P < 0.001. The comparison of PSD/HCTZ with HCTZ had low statistical power but favored PSD/HCTZ: −6.0 (−14.1, +2.1), P = 0.149. Thus, compared to HCTZ, CHIP diuretics had twice the effect on LVM. CHIP diuretics did not surpass HCTZ in reducing systolic or diastolic blood pressure: −0.3 (−5.0, +4.3) and −1.6 (−5.6, +2.4), respectively. The strength of evidence that CHIP diuretics surpass HCTZ for reducing LVM was high (GRADE criteria). In conclusion, these novel results have demonstrated that CHIP diuretics reduce LVM 2‐fold more than HCTZ among hypertensive patients. Although generally related to LVM, blood pressure fails to explain the superiority of CHIP diuretics for reducing LVM.

1. INTRODUCTION

Left ventricular hypertrophy (LVH) is found in 36%‐41% of patients with hypertension and predicts cardiovascular events (CVEs) and total mortality independently of traditional risk.1, 2, 3, 4, 5, 6, 7 LVH carries an increased risk of 1.6‐ to 3.5‐fold for CVEs,2, 3 cardiovascular deaths,2 and total mortality,2 and an 8‐fold risk for congestive heart failure (CHF).5 Among hypertensive patients, LVH contributes to about 30% of all deaths, 25% of CVEs, and 75% of CHF.1, 2, 5, 8 Conversely, numerous data demonstrate that reduction in LVH by antihypertensive medications correlates with reduced CVEs.9 Thus, it is critical to reduce LVH.10

As with other antihypertensive medications, thiazide‐type diuretics reduce LVH, that is, elevated left ventricular mass (LVM).11 However, differences among thiazide‐type diuretics for reducing LVM have never been examined. A recent article reviewed features of HCTZ vs its alternatives, which, for brevity, can be designated as CHIP diuretics (CHlorthalidone, Indapamide, and Potassium‐sparing/ hydrochlorothiazide [PSD/HCTZ] combinations).12

It was predicted a priori that these CHIP diuretics would surpass HCTZ for reducing LVM based on the following: (a) Chlorthalidone is associated with reduced LVH when measured by electrocardiogram.13 (b) Chlorthalidone, indapamide, and PSD/HCTZ are each associated with lower blood pressure relative to HCTZ.14, 15, 16, 17 (c) Chlorthalidone and triamterene/HCTZ are superior to HCTZ for reducing CVEs and CHF.12, 18, 19 (d) Reducing sodium reduces LVM,20, 21, 22, 23, 24, 25, 26, 27, 28 and chlorthalidone and PSD/HCTZ combinations reduce sodium more than HCTZ (Supplement 1 in Appendix S1),29, 30, 31 (e) Preserving potassium reduces LVM,20, 21, 22, 32, 33, 34 and PSD/HCTZ combinations are of course superior to HCTZ for preserving potassium.35, 36

To test the hypothesis that this set of diuretics is superior to HCTZ for reducing LVM, a systematic review and network meta‐analyses of randomized trials were conducted. The effects on systolic and diastolic blood pressure (SBP and DBP) from CHIP diuretics relative to HCTZ were also examined.

2. METHODS

This systematic review followed PRISMA guidelines. Inclusion criteria were as follows: (a) patients with hypertension; (b) use in the diuretic arm of HCTZ, chlorthalidone, indapamide, triamterene/HCTZ, amiloride/HCTZ, spironolactone/HCTZ, spironolactone, eplerenone, or canrenone; (c) use in its comparator arm of another diuretic or one of the nondiuretic classes commonly used to treat hypertension, namely a beta‐blocker (BB) or a first‐line drug class, including a dihydropyridine calcium channel blocker (CCB) and a renin‐angiotensin system inhibitor (RASI), which includes an angiotensin‐converting enzyme inhibitor or an angiotensin receptor blocker; (d) change in LVM or change in LVM indexed to height or to body surface area as outcomes; (e) trial duration 8 weeks to 1 year. Excluded were trials differing from those above (eg, a diuretic compared with two other drugs); LVM measured by ECG only; and studies of renal disease.

Five databases were searched: PUBMED, COCHRANE, SCOPUS, clinicaltrials.gov, and prior meta‐analyses using the algorithm in Supplement 2 in Appendix S1. Essentially, this algorithm required both a ventricular hypertrophy term and one or more of the diuretics of interest as a keyword or in the title or abstract. After elimination of duplicates and articles whose titles indicated obvious irrelevance (eg, titles pertaining to pulmonary hypertension, animals, children, or pregnancy), two of us (SS and GR) conducted an independent evaluation to select articles for review and meta‐analysis. After discussion, a consensus was reached in all instances. Two co‐authors (RA and GR) independently extracted data from the articles pertaining to LVM and SBP in order to eliminate any inaccuracy in data collection.

2.1. Statistical synthesis

Several studies compared the same diuretic with multiple nondiuretics. This raised the issue of double counting, and the number of patients used in the calculations was divided appropriately.37 Thus, if the diuretic was randomized to 30 patients and compared with nondiuretics A, B, and C in three different arms, the number of patients receiving the diuretic in each randomized comparison was counted as 10.

Each network with its number of trials can be diagrammed as shown in Figure 1. The equation for the RASI network is:

$$({\text{CHIP}}_{\mathrm{RASI}}-\text{RASI})-({\text{HCTZ}}_{\mathrm{RASI}}-\text{RASI})={\text{CHIP}}_{\mathrm{RASI}}-{\text{HCTZ}}_{\mathrm{RASI}},$$

where “CHIP_RASI,” “HCTZ_RASI,” and “RASI” represent the reductions in LVM from each of those agents in the RASI network. The analogous equations were used for networks based on BBs and CCBs. The 4th network (the lowest branch in Figure 1) was a direct comparison between indapamide and HCTZ in a trial by Senior et al. These 4 networks [(CHIP_RASI − HCTZ_RASI), (CHIP_BB − HCTZ_BB), (CHIP_CCB − HCTZ_CCB), and (CHIP_SENIOR − HCTZ_SENIOR)] were pooled in a meta‐analysis to give a common effect for the reduction in LVM from CHIP diuretics vs HCTZ.

Figure 1.

Diagram of network analysis with the number of trials in each network. BB, beta‐blockers; CCB, calcium channel blockers; CHIP, chlorthalidone, indapamide, and potassium‐sparing/hydrochlorothiazide diuretics; HCTZ, hydrochlorothiazide; RASI, renin‐angiotensin system inhibitors

Models used were Knapp‐Hartung for meta‐analysis of 5+ trials and DerSimonian‐Laird for meta‐analysis of <5 trials, consistent with recommendations.38 Assessment of heterogeneity across networks was based on tau (the estimate of the SD of the main effect across trials), I ² (the ratio of among trial variability to total variability which is constructed to range from 0% to 100%), and the P value to test whether the observed heterogeneity was statistically significant.39 Meta‐regression was used to examine the overall correlation of reductions in SBP and DBP with reductions in LVM after adjustment for trial characteristics where these were significant predictors of LVM. The strength of evidence was evaluated by GRADE criteria.40 Software was Comprehensive Meta‐Analysis (CMA), version 3.2.00089, March 24, 2014.

3. RESULTS

3.1. Descriptive features

The above criteria yielded 27 articles representing 38 randomized trials with one trial comparing indapamide with HCTZ and 37 comparing diuretics with nondiuretics with a total of 2299 patients (Supplement 3 in Appendix S1 for flow diagram and Supplement 4 in Appendix S1 for references; Table 1 and Figures 2 and 3). This compares favorably with the most recent comprehensive meta‐analysis which identified 24 trials and 1339 patients involving thiazide‐type diuretics.11 The percent men in HCTZ trials and CHIP trials were 47% and 57%, respectively, and the mean age was 56.1 and 54.0, respectively. The risk of bias according to the Cochrane criteria was moderate (Supplement 5 in Appendix S1). Trials of duration 0.5‐1 year contained 90% of patients.

Table 1.

Characteristics of trials in this analysis. All patients had a diagnosis of hypertension. Unless otherwise noted, data represent means

Percentage of patients in trials of diuretics
HCTZ trials	38%
CHIP trials	62%
All trials	100%
Percentage of patients in trials of CHIP diuretics
Chlorthalidone trials	39%
Indapamide trials	39%
PSD/HCTZ trials	22%
All trials of CHIP diuretics	100%
Percentage of statistical weight in double‐blind trials of CHIP diuretics
Chlorthalidone trials	31%
Indapamide trials	54%
PSD/HCTZ trials	14%
All trials of CHIP diuretics	100%
Agent in the nondiuretic arm: CCBs 41% RASIs 38% Beta‐blockers 22%
Trial characteristics
Men	58%
Age	56.1
CHF explicitly excluded	54%
Patients in double‐blind trials	82%
Trial duration	0.62 y (90% of patients were in trials of duration 0.5‐1 y.)
Background antihypertensive medications continued throughout the trial	46%
Titration of dose to achieve a target BP	62%
Add‐on therapy to achieve a target BP	34%a
Patient censored if target BP is not achieved	19%

BP, blood pressure; CCBs, calcium channel blockers (all dihydropyridines); CHF, congestive heart failure; HCTZ, hydrochlorothiazide; PSD, potassium‐sparing diuretic; RASIs, renin‐angiotensin system inhibitors (13 trials used an angiotensin‐converting enzyme inhibitor, and one trial used an angiotensin receptor blocker).

^aFor example, in the trial by Fagard et al, patients were randomized to either triamterene/HCTZ or trandolapril. Amlodipine was added if the blood pressure was not controlled.

Figure 2.

Relationship of reduction in systolic blood pressure (SBP) to reduction in left ventricular mass (LVM)

Figure 3.

Reduction in left ventricular mass from diuretics relative to nondiuretics. CHIP, chlorthalidone, indapamide, and potassium‐sparing/hydrochlorothiazide; HCTZ, hydrochlorothiazide

A 1% decline in SBP predicted a 1% decline in LVM (Figure 2). DBP also predicted LVM (P = 0.029) but, when included in a model with SBP, DBP was no longer statistically significant (P = 0.187), while SBP continued to predict LVM (P = 0.0001). Individual trials comparing diuretics with nondiuretics are shown in Figure 3.

3.2. Results for heterogeneity

The difference between CHIP diuretics and HCTZ in reducing LVM varied substantially across trials; that is, there was heterogeneity in the CHIP‐HCTZ difference. The ratio of tau to the main effect was 1.8‐1, indicating that about 29% of trials would be opposite to the main effect. The I ² was 81%, and the P for heterogeneity was highly significant statistically: 0.001. This made interpretation of the results uncertain when all trials were analyzed. However, double‐blind trials and trials with no background antihypertensive medications had no detectable heterogeneity, and therefore, analyses were limited to these trials.

3.3. Publication bias and absolute levels of nondiuretic comparators

Among double‐blind trials, there was no detectable publication bias. The mean decline in the nondiuretic comparators for HCTZ was −7.24%, SE 1.28%, and the mean decline in the nondiuretic comparators for CHIP diuretics was −10.22%, SE 2.21%. The difference between the two is 2.98% (95% CI −2.03%, 7.99%), P = 0.243, providing no evidence that the two sets of nondiuretic comparators differ from one another.

3.4. CHIP diuretics vs HCTZ in reducing LVM

In double‐blind trials, chlorthalidone, indapamide, and all CHIP diuretics combined were each significantly superior to HCTZ in reducing LVM (Figure 4). PSD/HCTZ tended to be superior to HCTZ, but this was not statistically significant. However, the statistical weight (inverse of the variance) from PSD/HCTZ trials was only 14% of the total weight of all CHIP diuretic trials, compared to 31% for chlorthalidone and 54% from indapamide (Table 1). This is reflected in the substantially wider confidence limits around the PSD/HCTZ‐HCTZ estimate. Among trials with no background antihypertensive medications, CHIPs were also superior to HCTZ. Among trials with the longest duration of 0.5‐1.0 year, these relationships persisted.

Figure 4.

Percent reduction in left ventricular mass from chlorthalidone, indapamide, and potassium‐sparing/hydrochlorothiazide (CHIP) diuretics relative to hydrochlorothiazide (HCTZ) among trials where there was no detectable heterogeneity, that is, no detectable difference in effect across trials. “No background meds” refers to the absence of antihypertensive medications at baseline

Among double‐blind trials, HCTZ reduced LVM by −7.29 (95% CI −10.40, −4.18). All CHIP diuretics combined were superior to HCTZ by −7.66 (95% CI −12.22, −3.10). Consequently, CHIP diuretics reduced LVM more than HCTZ by 105% (−7.66/−7.29); that is, CHIP diuretics were about twice as effective as HCTZ in reducing LVM. The minimum superiority of CHIP diuretics was 30% (−3.10/−10.40), and the maximum superiority was 292% (−12.22/−4.18).

3.5. CHIP diuretics vs HCTZ in reducing LVM, SBP, and DBP

Complete information on SBP and DBP that included SDs or SEs was present in 89% of patients in trials without background antihypertensives and in 66% of patients in double‐blind trials (Figure 5). When limiting the analysis of LVM to patients having complete information on SBP and DBP, CHIP diuretics again surpassed HCTZ. With reductions in SBP and DBP as outcomes, there was no detectable difference between CHIP diuretics and HCTZ. There was some evidence of heterogeneity for the SBP and DBP comparisons for double‐blind trials, but this did not achieve statistical significance.

Figure 5.

Percent reduction in left ventricular mass (LVM), systolic blood pressure, and diastolic blood pressure from chlorthalidone, indapamide, and potassium‐sparing/hydrochlorothiazide (CHIP) diuretics vs hydrochlorothiazide (HCTZ). The analysis of LVM is limited to trials with all necessary information on blood pressure

4. DISCUSSION

These findings confirm the a priori hypothesis. Among the highest quality trials, double‐blind trials, chlorthalidone, indapamide, and all CHIP diuretics combined were superior to HCTZ for reducing LVM. Results were both statistically and clinically significant, with an effect on LVM from CHIP diuretics which was twice the effect on LVM from HCTZ. The benefit of PSD/HCTZ for reducing LVM relative to HCTZ was not statistically significant, but the weight from PSD/HCTZ trials was very small relative to the weights from chlorthalidone and indapamide trials, and the trend favored PSD/HCTZ over HCTZ with a magnitude similar to the other two CHIP diuretics. The advantage of CHIP diuretics persisted among trials of long duration. The overall strength of the evidence that CHIP diuretics are superior to HCTZ is high based on GRADE criteria (Table 2).

Table 2.

Strength of evidence that CHIP diuretics are superior to HCTZ in reducing LVM based on GRADE criteria40

Feature	Findings	Strength of evidence
Design	Results are from a network meta‐analysis of randomized trials	Intermediatea
Reasons to downgrade the quality of the evidence
Risk of bias of individual trials	Moderate to low (all trials were double blind for conclusions)	Unchanged
Inconsistency	No detectable heterogeneity among double‐blind trials.	Unchanged
Indirectness	Mean age in the study (56) is nearly identical to the mean age of U.S. hypertensives (55). There were no restrictions as to comorbidities.	Unchanged
Imprecision	CHIP diuretics vs HCTZ for reducing LVM: −7.7 (95% CI −12.2, −3.1)	Unchanged
Publication bias	Not detectable	Unchanged
Reasons to upgrade the quality of the evidence
Magnitude of effect	Effect on LVM from CHIP diuretics is twice that of HCTZ.	Upgrade strength
Plausibility	Compatible with prior research (see BACKGROUND)	Upgrade strength
Probability that null hypothesis is correctb	(a) Observed P < 0.001. (b) Compared to head‐to‐head comparisons, network meta‐analysis has reduced power to detect differences. Thus, the observed CHIP‐HCTZ difference likely underestimates the true effect	Upgrade strength
Final GRADE		Highc

^aGRADE starts with a rating of HIGH for randomized trials and LOW for observational studies. GRADE does not specify what strength of evidence to assign to network meta‐analysis. A network meta‐analysis of randomized trials is inferior to a head‐to‐head meta‐analysis of randomized trials because of indirectness of comparisons but superior to a meta‐analysis of observational studies because it preserves randomization.41 Therefore, the strength of evidence based on study design was classified as intermediate.

^bGRADE criteria do not address this feature.

^cGRADE criteria characterize the strength of evidence as “High” when an observational design is upgraded twice. In this network analysis of randomized trials, the study design is superior to an observational design and has been upgraded three times, more than justifying a high strength of evidence.

These novel results are consistent with the finding that, in a systematic review and head‐to‐head meta‐analysis of all available double‐blind trials, CHIP diuretics are superior to renin‐angiotensin system inhibitors in reducing LVH.42 Moreover, this is the first time to our knowledge that indapamide has been shown to be more effective in reducing a cardiovascular outcome relative to HCTZ. Based on pharmacology and cost, it has been suggested that indapamide might be the best choice when using a thiazide diuretic,43 and that viewpoint is now strengthened based on indapamide's effect on a cardiovascular outcome. These results are compatible with previous research showing that triamterene/HCTZ12 and chlorthalidone18 are each superior to HCTZ for reducing CVEs and CHF.

Greater blood pressure reductions and alterations in serum sodium and potassium explain the superiority of CHIP diuretics over HCTZ. High sodium and reduced potassium may lead to LVH via reduced availability of nitric oxide and increased dimethyl arginine which alter vascular smooth muscle.20

For reducing cardiovascular events, HCTZ has been inferior to enalapril, amlodipine, and chlorthalidone, in spite of similar reductions in office SBP.18, 44, 45 Likewise, in the present study for reducing LVM, HCTZ was inferior to CHIP diuretics, in spite of similar reductions in office SBP and DBP. Explanations for this pattern may pertain to HCTZ's less than 24‐hour duration of action,46 which might have become apparent had blood pressure been measured by 24‐hour monitoring. The importance of 24‐hour duration of action is highlighted by the observation that treatment of nighttime blood pressure by evening dosing may be more effective relative to the usual daytime dosing in reducing cardiovascular risk.47 On the other hand, the superiority of CHIP diuretics over HCTZ for reducing cardiovascular risk may be unrelated to blood pressure. For example, unlike HCTZ, chlorthalidone decreases platelet aggregation and vascular permeability;48 indapamide relative to HCTZ more effectively scavenges oxygen radicals and inhibits platelet aggregation;14 and PSD/HCTZ combinations relative to HCTZ more effectively reduce sodium (Supplement 1 in Appendix S1),29, 30, 31 and preserve potassium.35, 36

Potential adverse effects pertain primarily to serum electrolytes. Chlorthalidone, indapamide, and HCTZ have similar effects on serum potassium.12, 14 Relative to HCTZ, the PSD/HCTZ combination does not lead to hyperkalemia but does minimize the hypokalemia from HCTZ. Chlorthalidone and PSD/HCTZ combinations may lead to greater reductions in serum sodium than HCTZ, but this is partly why these two CHIP diuretics are superior in reducing LVM. Thus, obtaining serum electrolytes are helpful in guiding decisions about the choice of diuretics. Where the optimal antihypertensive effect is needed, indapamide and chlorthalidone may be the best choices.

Limitations of the study are assessed in Table 2. The primary limitation of the study stems from a network analysis which is inferior to a head‐to‐head analysis. In addition, there was relatively less statistical power for the PSD/HCTZ vs HCTZ comparison. Some trials were of moderate quality. As in head‐to‐head comparisons, network analyses may be affected by differences in trial characteristics between the two agents being compared (in this case CHIPs and HCTZ). In these data, for the seven frequently occurring trial characteristics, CHIPs and HCTZ agreed on average by 81% (SD 11%); this limits the potential for bias. Owing to limitations in the number of trials and patients, it was not possible to compare CHIPs and HCTZ with respect to the percent reduction in LVM per percent reduction in SBP and DBP.

5. CONCLUSION

With 49 million U.S. prescriptions annually,49 HCTZ is used many times more often than its alternative diuretics. On the other hand, these novel findings demonstrate that its alternatives are far more effective for reducing LVH in patients with hypertension.

CONFLICT OF INTEREST

None should be declared.

Supporting information

 

Roush GC, Abdelfattah R, Song S, Ernst ME, Sica DA, Kostis JB. Hydrochlorothiazide vs chlorthalidone, indapamide, and potassium‐sparing/hydrochlorothiazide diuretics for reducing left ventricular hypertrophy: A systematic review and meta‐analysis. J Clin Hypertens. 2018;20:1507‐1515. 10.1111/jch.13386