Article Text

Download PDFPDF

The choice of diuretic in hypertension: saving the baby from the bathwater
  1. Morris J Brown
  1. Correspondence to Professor Morris J Brown, Clinical Pharmacology Unit, University of Cambridge, Addenbrookes Hospital, Cambridge CB2 2QQ, UK; morris.brown{at}

Statistics from


There has been a growing realisation that almost 2 decades of diuretic prescribing in hypertension have not been evidence based. A recent meta-analysis confirmed that low-dose thiazides—hydrochlorothiazide (HCTZ) 25 mg in most of the world—achieve only small falls in ambulatory blood pressure (BP),1 and the 2011 revision of the National Institute for Health and Clinical Excellence's (NICE's) guidance for hypertension concedes lack of outcome data to support HCTZ 25 mg or bendroflumethiazide 2.5 mg.2 NICE does a U-turn on previous UK and all other international guidelines by dropping diuretic from first to third choice in most patients (unless deemed at ‘high risk of heart failure’), and NICE seems to jump out of the frying pan into the fire by recommending formulations which are either not available in the UK (chlortalidone 12.5–25 mg) or have been used in only one outcome trial in hypertension, which was too small to achieve its primary outcome benefit (indapamide 1.25–2.5 mg). How did this mess come about, and what should now be done?


Sodium (Na+) retention causes hypertension; diuretics developed to reverse Na+ retention

The thiazides were developed in the late 1950s as effective diuretics. In the next decade, they entered use as antihypertensive drugs, and over the next 3 decades, numerous outcome trials demonstrated their efficacy, alone and in combination, in preventing strokes and heart failure, being superior to placebo and non-inferior or equivalent to other antihypertensive classes. The doses used in such trials were those shown to be effectively natriuretic, for example, HCTZ 50 mg in the Multiple Risk Factor Intervention Trial (MRFIT), Medical Research Council (mild hypertension and elderly trials) and International Nifedipine GITS Study: Intervention as a Goal in Hypertension Treatment (INSIGHT).3–5 Lower doses were effective in combination with renin–angiotensin system (RAS) blockers, and such combinations are indeed effective natriuretic treatments, as is apparent from their widespread use in heart failure. Over the same period, numerous studies investigating the pathogenesis of hypertension demonstrated a variable contribution from salt. This was not too surprising given that pressure is force/area and that the main determinant of force in a column of fluid is the weight of its principal content, namely, the amount of salty water. The clincher at the level of molecular proof was the series of elegant demonstrations that ‘gain-of-function’ mutations in channels transporting Na+ from kidney into blood were the sole and sufficient cause of severe hypertension and strokes.6–9 These rare monogenic syndromes explain <1% of all hypertension. However, the common feature in their phenotype—a low plasma renin concentration—was invaluable in demonstrating that the more common-or-garden syndrome of low-renin hypertension was likely to be due to Na+ retention. This demonstration, and entry of a rapid, cheap plasma renin mass assay into clinical practice, has had two practical clinical outcomes. One has been the discovery that small, curable Conn's adenomas are more frequent than hitherto realised—about 100-fold more so than the 0.01%–0.03% incidence cited by NICE; a suppressed plasma renin despite diuretic treatment and transient thiazide-induced hypokalaemia are often the triggers to diagnosis (figure 1).10 The second outcome is the recognition, overall and in individual patients, that resistant hypertension is commonly due to the iatrogenic underdosing of diuretics.11 12

Figure 1

Rapid diagnosis of Conn's adenoma. A 37-year-old man with blood pressure (BP) 152/99 on three drugs was found to have a plasma renin 5 mU/L (normal 5–68) and became transiently hypokalaemic (K+=3.1 mmol/L) when started on bendroflumethiazide 2.5 mg. The positron emission tomography (PET)–CT, shown here, showed lateralisation to a 5 mm adenoma in the right adrenal, whose laparoscopic removal cured the hypertension. Standardised uptake value (SUV) was measured between 35 and 45 min after injection of 11C-metomidate synthesised in the Wolfson Brain Imaging Centre, Cambridge.

Evolution of the low-dose diuretic mythology

So how did we come to the routine prescribing of bendroflumethiazide 2.5 mg—one-quarter of the dose used in the Medical Research Council Mild Hypertension Trial, one-sixth indeed of that used at the start? If the basis of its antihypertensive action is natriuresis, and the dose required for this and prevention of stroke is 10 mg, what changed? It was a concern about side effects, combined with a Ponzi-style pharmacology promising all boom and no bust—all the BP reduction with none of the side effects. Sure, the two can in theory be dissociated if the latter are due to an off-target action of the drug with a right-shifted dose–response relative to that at the primary site of action, in this case, the Na+-Cl co-transporter (NCCT) in the distal tubule. But no second site for the thiazides has been securely demonstrated.13 In contrast, the BP lowering and the common metabolic problems of hypokalaemia and hyperuricaemia are reproduced by single mutations of NCCT in Gitelman's syndrome—while mutations of a kinase which activates NCCT cause hypertension and hyperkalaemia.14–16 Bizarrely, a self-fulfilling defence for reducing thiazide dose has been that its antihypertensive action must be due to an extrarenal vasodilator action because the plasma concentration of low-dose thiazide is lower than its Ki for NCCT.13

In retrospect, the idea that low-dose thiazides achieve maximal fall in BP is a shocking medical myth due to a mixture of underpowered dose comparisons, a collective amnesia or even rewriting of older studies showing the opposite (but true) result, and a sloppy habit of guideline self-reinforcement through quoting reviews that reached the desired conclusion. In 1990, a study of 257 patients sounded large enough to change practice.17 But these were divided into five groups of 50-odd, with a power to differentiate five doses of bendrofluazide only if the SD of BP differences was <7 mm Hg. The fact that the eventual results, showing no significant difference, had SDs of twice this size was masked by reporting SEs of 2 mm Hg, and the point differences in BP response to 10 versus 1.25 mg were in fact almost 5 mm Hg. A second study, supporting use of bendrofluazide 1.25 mg, had just 30 patients per dose.18 For HCTZ, there were no new data. The American guideline, JNC7, cites two reviews in journal supplements and one original paper.19 This is Materson's oft-cited 1993 comparison of six antihypertensives, which did not include dose–response comparisons.20 There is a post hoc analysis of patients controlled at each dose of drug, which no more shows HCTZ 25 mg to be maximal than it shows captopril 25 mg to be a maximal dose of ACE inhibition!

My suspicion that all was not well in the kingdom of Bendmark arose from analysis of the BPs in the International Nifedipine GITS Study: Intervention as a Goal in Hypertension Treatment, where patients were randomised to nifedipine 30 mg or HCTZ 25 mg (+ amiloride 2.5 mg), with dose doubling in patients whose systolic BP did not fall below 140 mm Hg.5 The effect of dose doubling was identical in the two arms (figure 2). Subsequently, a prospective, randomised, crossover comparison of several diuretics and doses showed that doubling of bendroflumethiazide from 2.5 to 5 mg achieved a 5 mm Hg fall, the same as achieved by switching from bendroflumethiazide to spironolactone (figure 3).21

Figure 2

Dose–response to hydrochlorothiazide (HCTZ) and nifedipine. Hypertensive patients in the International Nifedipine GITS Study: Intervention as a Goal in Hypertension Treatment (INSIGHT) study were randomly assigned to low-dose co-amilozide (HCTZ 25 mg + amiloride 2.5 mg) or nifedipine 30 mg. After 2 weeks, the dose was doubled to the high dose unless systolic blood pressure had fallen to <140 mm Hg. After a further 2 weeks, a second drug was added unless systolic blood pressure was <140 mm Hg on high-dose monotreatment. The figure shows the systolic blood pressure for patients whose systolic blood pressure was controlled on low or high dose of monotreatment, and illustrates the identical effect of dose doubling for each drug class.5

Figure 3

Double-blind crossover of diuretics and doses. Fifty-one patients with low-renin hypertension rotated in random order through 5-week cycles on each drug and dose. Low and high doses of each drug were never consecutive. Switching from bendroflumethiazide 2.5 mg to 5 mg or to spironolactone 50 mg was similarly effective.21

Do thiazides cause diabetes?

So in recommending a reduction in dose from that shown to prevent stroke, what is the adverse event which is so severe as to justify a reduction in protection from stroke? A possible answer is glucose intolerance.22 But if so, what is the evidence for differential dose–response of thiazides on BP and glucose tolerance? That thiazides increase the incidence of so-called new-onset diabetes has been a secondary finding of several outcome trials—although the incidence may have been exaggerated as none followed the requirement to obtain two elevated glucose values for diagnosis.5 22–24 Whether the diabetes is sustained if the diuretic is withdrawn is uncertain, and its importance must be in some doubt given the efficacy of adequate doses in preventing cardiovascular complications.5 23 25 The mechanism of the glucose intolerance is also uncertain. Is it due to reduced blood flow, hence delivery of glucose, to muscle—a haemodynamic effect of diuretics likely to be shared by β blockade, but not by ‘newer’ vasodilating antihypertensives? Or is the glucose intolerance related to fall in K+?26 27

In a recent crossover trial, we found a significant rise in the 2 h glucose on oral glucose tolerance test after just 2 weeks of treatment with HCTZ 25 mg, with little further rise when HCTZ was increased to 50 mg (figure 4). This action of HCTZ was not seen on an equihypotensive dose of amiloride, namely 20 mg daily, and overall there was a negative correlation between changes in K+ and glucose.28

Figure 4

Hydrochlorothiazide (HCTZ)-induced hyperglycaemia. A double-blind crossover comparison of 4-week treatment with either HCTZ 25–50 mg or amiloride 10–20 mg, each force titrated after 2 weeks, was undertaken in 38 patients with hypertension. Both doses of HCTZ, but not amiloride, impaired the 2 h glucose after a 75 g oral glucose tolerance test (p=0.006).28

Whether there is truly a lack of dose–response for the glucose intolerance is uncertain. Indeed, this is the point—that as a justification for limiting the usual dose of thiazide diuretic, the glucose intolerance is as poorly evidenced as the BP reduction. It is intriguing that, in the Anglo-Scandinavian Cardiac Outcomes Trial (ASCOT), there was a 34% excess of diabetes in the arm randomised to atenolol and bendroflumethiazide 1.25 mg, whose BP was less well controlled than the arm randomised to amlodipine + perindopril, whereas in the Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial (ALLHAT), there was only 18% excess of diabetes in the arm receiving chlortalidone 25 mg (+ atenolol in 40% of patients), whose BP control equalled that in the amlodipine arm.23 24 Hypertension, which increases RR of diabetes 2.5-fold, is more important a risk factor than choice of drugs, and it is likely that, as for other complications of hypertension, it is BP control which is the paramount determinant of outcome.25 29 Clearly, a 34% excess risk of diabetes does not justify poorer BP control and may even be the consequence of this.

Alternatives to low-dose thiazide in current practice; beware turning myths into legends

What should be done now? Is it bendroflumethiazide 2.5 mg, bendroflumethiazide itself or all thiazides which should be dropped as first choice? The latter seems ill advised. In the only double-blind head-to-head comparison of outcomes on diuretic versus calcium channel blockade (CCB)—in which there was no difference except for marginal superiority for diuretic in preventing heart failure—the diuretic base was HCTZ.5 And although this was the study which demonstrated a clear dose–response, a third of 3100 patients with BP 173/99 mm Hg were controlled on 25 mg.5 Yet NICE recommends ‘offering a thiazide-like diuretic, such as chlortalidone (12.5–25.0 mg once daily) or indapamide (1.5 mg modified-release once daily or 2.5 10 mg once daily) in preference to a conventional thiazide diuretic’. What are the objections?

Thiazide-like diuretics

First, there has been no comparison of thiazide with ‘thiazide-like’ diuretics other than short-term BP.30 31 Chlortalidone is twice as potent as HCTZ, but potency is different from efficacy. Indapamide has not been studied long term in typical hypertensives. It was administered to 1930 hypertensives, aged >80, in the HYpertension in the Very Elderly Trial (HYVET) where it failed, despite addition of perindopril, to achieve a significant reduction in the primary end point of stroke.32 In the other trial cited by NICE, a poststroke study published only in outline by the Chinese Medical Journal, indapamide reduced BP by 5/2 mm Hg and stroke by 29%.33 Chlortalidone has the impressive record, by contrast, of coming first equal in the largest-ever hypertension outcome trial, the Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial, even in competition with amlodipine.23 But patients started at a dose of 12.5 mg, titrated in two stages to 25 mg, whereas only a 50 mg tablet is available in the UK.

De novo combination of low-dose thiazide with RAS blockade

The second weakness of the recommended switch in diuretic prescribing is that the recommendation stems from trials in which diuretic was introduced as first-line treatment. Nowadays, and especially after issue of the new guidance, most patients will receive diuretic in combination treatment with an angiotensin receptor blocker (or ACE inhibitor). Here there is much more evidence, short and long term, for HCTZ, even low dose, than for the alternatives.34 In combination with a RAS blocker, the BP in most patients becomes salt sensitive; the consequences are that there is a steeper dose–response for diuretic than when used as monotreatment and that even the lowest doses are more effective. In the Losartan Intervention For Endpoint reduction in Hypertension Study (LIFE), for instance, 90% of patients received HCTZ at an initial dose of 12.5 mg.35 In the Valsartan Antihypertensive Long-term Use Evaluation trial (VALUE), where the angiotensin receptor blocker was predictably less effective than CCB as sole treatment of a population aged >55, addition of low-dose HCTZ almost abolished the gap.36

K+ sparing and mineralocorticoid receptor blockade

Curiously, NICE ignores the question of whether the efficacy and safety of thiazides are enhanced by routine combination with a K+-sparing diuretic, as supported by outcome trials and a number of reviews.5 26 37 38 Co-amilozide (HCTZ/amiloride 25/2.5 or 50/5 mg) has been successful in two outcome trials, reducing stroke and coronary heart disease by 31% and 44% against placebo39 and reducing BP and events by the same amount as high-dose CCB.5 It is as cheap as indapamide, at 4p per day. Pending further research, co-amilozide should be the first-line recommendation for patients aged >55 as an alternative to CCB.

Arguably, however, the principal and certainly most interesting elephant in the diuretic room is mineralocorticoid (aldosterone) blockade, using spironolactone or eplerenone. Although thiazides are conventionally considered more effective diuretics than the K+-sparing diuretics, because more Na+ is reabsorbed through NCCT than through the epithelial Na+ channel, the latter is at least as important in the Na+ retention of hypertension—as evidenced by the sites of molecular defects in the monogenic syndromes.6–9 In our diuretic comparison, the rise in plasma renin, one of the best measures of natriuretic response, was twice as much on spironolactone or amiloride as on high-dose bendroflumethiazide.21 Spironolactone and eplerenone (shortly to join spironolactone as off patent) are highly effective in reducing cardiovascular complications of impaired myocardial function, but have never been tested for outcomes in hypertension.40–42 Since β blockers perform similarly well (to mineralocorticoid blockade) in heart failure, but underperform in hypertension, efficacy in the latter cannot be taken for granted. Practical problems with gynaecomastia and hyperkalaemia will be an issue. But given the primary role of Na+ retention in hypertension, the likely contribution of aldosterone to the Na+ retention and the growing evidence of damaging extrarenal consequences of mineralocorticoid stimulation, I would place a high punt on an outcome trial with spironolactone or eplerenone being positive. (Whether a funding body would share this view is yet to be tested. But interestingly, spironolactone's lack of a licence for hypertension means that, under recent regulation, the bet would be for real as a positive result would now permit a premium formulation, maybe spironolactone + chlortalidone, to be licensed and marketed.)

Calcium channel blockade

Despite conceding ‘uncertainty about the relative cost-effectiveness’ of the two classes, with no CIs given for the minuscule difference shown, NICE recommends CCB rather than diuretic as first choice and demotes diuretic to third line. Within hypertension, the evidence for superiority over diuretic does not hold up. Several head-to-head comparisons showing no difference in outcome—if anything, there is more heart failure on CCB—should not be written off by one subsequent study where a CCB plus RAS blocker was superior to CCB plus HCTZ.5 25 43 In the Avoiding Cardiovascular Events through Combination Therapy in Patients Living with Systolic Hypertension (ACCOMPLISH) study, the doses were not matched for clinic BP reduction, and since 97% patients swapped from previous treatment to a new combination, the intervention studied was neither first-line treatment nor usual practice nor indeed likely to be balanced across the two arms.

What is remarkable about CCBs is their lack of any serious contraindication or complication, with no biochemical monitoring required before or after initiation. In patients with established hypertension, where biochemical screening is required to exclude secondary causes, this is not a deal breaker. Currently, we do not treat a systolic BP <140 mm Hg. But the risk of BP rises log-linearly from 115 mm Hg,44 and the much higher prevalence of such BPs means that most strokes arise within the range of systolic BP 115–140 mm Hg. It is here that wider use of CCBs should be investigated.45 So the danger of reviving yesterday's arguments about one class over another in hypertension is that we miss future potential for massively enlarging the benefits of antihypertensive treatment.

Conclusion: the need for further evidence

A decade ago, when there were questions about preferred treatment in hypertension, the discussions centred around the appropriate trial to answer the question. Now that the drugs in question are old, off patent and cheap, there is a temptation to assume that they are no longer worthy of further study. But the opposite is the case. Whereas industry studies sometimes put up straw men as comparators of new treatment, the academic, specialist community can decide with peer review what studies are most important. With NICE finding that treatment of high BP is unique in being cost-effective and cost-saving, the question of how best to reverse the Na+ retention of hypertension should not be one for reviews, committee rooms or stakeholders, but for clinical investigation. Let the good not be the enemy of the excellent!


View Abstract


  • Competing interests None.

  • Provenance and peer review Commissioned; internally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.