Navigating the shoals in hypertension: discovery and guidanceBMJ 2012; 344 doi: https://doi.org/10.1136/bmj.d8218 (Published 13 January 2012) Cite this as: BMJ 2012;344:d8218
- Morris J Brown, professor of clinical pharmacology1,
- J Kennedy Cruickshank, professor in diabetes, cardiovascular medicine, and nutrition2,
- Thomas M MacDonald, professor of clinical pharmacology3
- 1Clinical Pharmacology Unit, University of Cambridge, Addenbrooke’s Hospital, Cambridge CB2 2QQ
- 2King’s College and St Thomas’ Hospital, King’s Health Partners, London SE1 9NH
- 3Medicines Monitoring Unit and Hypertension Research Centre, Medical Research Institute, Ninewells Hospital and Medical School, Dundee DD1 9SY
- Correspondence to: M J Brown
- Accepted 2 December 2011
The treatment of hypertension is arguably the most evidence based and cost effective medical intervention ever.1 Not only is there a greater choice of drug classes than for other common diseases but there are more long term data that establish their efficacy in reducing risks from strokes, heart attacks, and heart failure. Uniquely, a prospective meta-analysis was planned by the leaders of the long term trials before their completion to avoid the post-hoc selection of questions and answers that can confound meta-analyses and guidelines.2 The drugs concerned are now off-patent and inexpensive. The benefits are evident: treating patients saves money as well as reducing morbidity and mortality.3
However, the population attributable risks of raised blood pressure in those not currently classified as hypertensive are large. The prize for public health is to extend the benefits of blood pressure reduction to all those who would benefit. Two such groups exist. One is people whose blood pressure is below current definitions of hypertension (which are based on proved benefit of treatment in outcome trials) but above the threshold where epidemiological studies show a log-linear increase in risk with blood pressure.4 The other is young patients with stage 1 (mild) hypertension, who paradoxically have higher relative and lifetime risk of complications than older people but are not treated in the UK for reasons of cost effectiveness. If we wait for patients to reach a blood pressure >160/100 mm Hg, or absolute cardiovascular risk >2% a year, before starting treatment we cannot be confident that a return to “normal blood pressure” can normalise risk. In outcome trials the benefits of treatment are limited to those with a clinic blood pressure >140/90 mm Hg and to reductions in blood pressure averaging just 10/6 mm Hg.4
Any changes in treatment should be based on evidence. Outcome trials designed to extend the use of antihypertensive drugs will need to be large and long because the event rate in lower risk groups will be lower. Because of concerns about medicalising a healthy population, these trials should also include, by stratification or randomisation, a prospective look at identifying people at whom future treatment should be targeted. For instance, if people were entered into an outcome trial that used home or ambulatory blood pressure monitoring, the trial could test the prediction that people whose blood pressure is variable will benefit more than those whose blood pressure is steady.5 6 Variability might reflect stiffness of the large arteries, and user friendly equipment now allows us to test whether variability or stiffness is useful in selecting patients and blood pressures for treatment.6 7
Other questions remain about best treatment (box 1). In its 2006 guidelines NICE called for a comparison of the options for managing resistant hypertension, such as spironolactone and other diuretics, and of starting treatment with combinations rather than single drugs to prevent resistance.1 It also wanted studies to predict the best treatment for individual patients.8 9 Several of the shorter term questions are being examined in a British Hypertension Society (BHS) Research Network collaborative programme (PATHWAY) established soon after the questions were identified. The PATHWAY trials build on the ACCELERATE study, which showed that combining drugs from the outset improves blood pressure control and reduces adverse events.10
Box 1: Unanswered questions in hypertension
Should treatment be extended to all those whose blood pressure increases their cardiovascular risk?
Should blood pressure targets be reduced to a similar, or greater, degree than thresholds?
Can the cost-effectiveness of treating more borderline hypertension be enhanced by supporting clinic blood pressures with additional measures of risk?
Does initial combination therapy prevent compensatory haemodynamic or neurohumoral responses that attenuate the efficacy or tolerability of single drugs?
Are potassium sparing diuretics the most effective treatment for resistant hypertension, and do they counter the diabetes risk of higher dose thiazides?
Does the cheap test for plasma renin herald an era of personalised medicine in hypertension?
Against this background the recent NICE guidance appeared.3 NICE generally prides itself on a bottom-up process, moving from evidence to conclusions to advice, so guarding against selection of evidence by special pleading.11 However, the 2011 hypertension guidance made headlines not because it answered any of the questions NICE had raised in 2006 but because of the conclusions of two cost effectiveness articles that appeared only in the week of the guidance yet seemed pivotal to its advice.12 13 Most surprising was the conclusion that we overtreat hypertension and that we could reduce the numbers starting antihypertensive treatment by 25% if we used ambulatory blood pressure monitoring instead of clinic blood pressure to diagnose hypertension. The cost-benefit argument in favour of ambulatory monitoring concluded that it would save £10m, but this was true only if the monitoring was not repeated for five years. The combination of a rise, compared to previous guidance, in the blood pressure threshold for treatment and a longer interval before repeat monitoring is not plausible, evidence based, or safe.14
Other recommendations in the guidelines also do not seem to be based on evidence (box 2). These are the relegation of diuretics from first choice to third choice treatment and the change in diuretic drug of choice from bendroflumethiazide to either indapamide (for which there are no data from a primary prevention trial that met its primary outcome) or chlortalidone (for which there is no suitable formulation available in the UK). There are no comparative effectiveness data to underpin this guidance.
Box 2: Guideline assumptions that are in need of evidence
Hypertension is overtreated
Hypertension should be redefined by daytime ambulatory blood pressure monitoring
A daytime average systolic blood pressure <150 mmHg should not be treated, irrespective of clinic or other peak pressures, unless annual risk of cardiovascular morbidity exceeds 2%
People with daytime average systolic blood pressure <135 mm Hg should not receive drug treatment but be rechecked by ambulatory monitoring every five years
Diuretics are less effective than calcium channel blockers at reducing blood pressure variability and should therefore be third choice treatment
Chlortalidone and indapamide are superior to bendroflumethiazide and hydrochlorothiazide
Redefinition of hypertension
NICE’s relegation of clinic blood pressure measurement depends on two unstated assumptions, neither of which is supported by robust evidence. One is that the condition diagnosed by measuring blood pressure in the clinic is the same as that diagnosed by ambulatory measurement. Clinic measured hypertension has been studied in about 500 000 patients in outcome trials; these have shown that drug treatment reduces the risks of stroke, myocardial infarction, heart failure, and overall mortality.2 15 By contrast, no trials have randomised patients to treatment on the basis of hypertension diagnosed by ambulatory monitoring. Indeed, the recent evidence cited by NICE that variability of blood pressure increases risk supports the need to treat patients whose blood pressure is higher on some occasions—that is, people with white coat hypertension.5 16
The second assumption is the equating of a clinic systolic blood pressure of 140 mm Hg with an ambulatory daytime systolic pressure of 135 mm Hg. It is unclear where this ambulatory monitoring threshold came from. Previous UK guidance has recommended, in round numbers, a 10/5 mm Hg difference in thresholds between ambulatory and clinic blood pressure measurements, there being a reported 12/7 mm Hg difference in measured blood pressure when these methods were compared.14 This was based on the Pressioni Arteriose Monitorate e Loro Associazioni (PAMELA) study’s comparison of ambulatory, home, and clinic blood pressure in 2000 participants.17 A decade later, it reported the predictive value of each measurement on cardiovascular deaths and found no significant difference between clinic, home, or ambulatory measurements; indeed, none was individually as informative as their combined risks—that is, a blood pressure that was high only at home or in clinic was higher risk than one that was always normal but lower risk than one that was always high.18 In PAMELA, a daytime mean ambulatory pressure of 135±14 mm Hg corresponded to a mean clinic systolic pressure of 155±22 mm Hg. Thus redefinition of hypertension as a daytime measurement of 135 mm Hg would detect many fewer “people with hypertension” than a clinic pressure of 140 mm Hg. However, half the people who died in PAMELA had a daytime ambulatory pressure below 135 mm Hg, suggesting this threshold is far too high.
Given the critical importance to NICE of equating its new ambulatory definition of hypertension with the previous clinic definition, the evidence cited is surprisingly circular. The guidance cites a BMJ article, which cites an “international standard” that turns out to be an unreferenced statement in the US joint national committee report from 2004.19 The Lancet study on cost effectiveness of ambulatory monitoring cites both the NICE guidance and the BMJ article.13 The NICE guideline does cite one primary source of data, but this article seems to undermine the call for use of ambulatory monitoring rather than support it: Head et al’s 2010 retrospective comparison of doctors’ and nurses’ readings (both clinic and ambulatory) in 8500 Australians.20 NICE has interpreted this study as showing that the daytime threshold of 135 mm Hg corresponded to a clinic threshold of 140 mm Hg. However, Head et al show only that 135 mm Hg was the average daytime systolic pressure for patients with stage 1 hypertension (defined as systolic 140-159 mm Hg). Their average clinic systolic pressure was 151 mm Hg if measured by a doctor (very similar to that in PAMELA) and 142 mm Hg if measured by a nurse. These figures seem inconsistent with the logic that led to the NICE guidance thresholds.
The white coat effect of doctors is neither new21 nor benign.18 22 Head et al do not suggest the routine introduction of ambulatory monitoring. Indeed, in outcome trials such as the MRC Mild Hypertension Trial, it was the doctors’ readings not the lower readings by nurses, that were used to enter patients.23 Of course, there is scope to supplement clinic blood pressure with other measurements, not least because the PAMELA data suggest additive risk prediction when this is done. It is common to use the same method for monitoring treatment as for making the diagnosis, and this would favour home over ambulatory measurements24; home measurements are also favoured on practical and cost grounds by leaders in primary care.25
But even if the redefinition of hypertension according to an ambulatory threshold were valid, and the headline 135 mm Hg threshold were robust, this is not the figure to be used in most patients. Previous UK guidelines already excluded most people with borderline hypertension from treatment, unless their 10 year cardiovascular risk exceeded 20%. NICE’s preamble acknowledges that this absolute risk requirement has disenfranchised younger patients in the UK. NICE further concedes that lifetime risk assessment might be the more appropriate measure in younger people and would greatly increase the numbers eligible for treatment.14
However, the new NICE guidance disqualifies even more patients than previously by raising the bars for treatment. Anomalies result. A patient might now be told one day that his clinic blood pressure predicts a 1 in 50 risk of myocardial infarction or stroke in the year ahead but discover the next day, after ambulatory monitoring, that he will not be treated for five years because his daytime systolic pressure averaged <135 mm Hg. Another patient at lower risk could have a clinic systolic pressure >180 mm Hg but remain untreated for five years if his daytime pressure averaged <150 mm Hg (figure⇓). We are unaware of any evidence underpinning this higher threshold for “must treat” patients. Although NICE cites Head et al for using this ambulatory value instead of a clinic systolic pressure of 160 mm Hg, as we described above their figure was a mean daytime pressure not a threshold. The corresponding clinic systolic pressure for patients with stage 2 hypertension was 160-179 mm Hg.20 In PAMELA, a daytime systolic pressure of 150 mm Hg equated to a clinic pressure >170 mm Hg.18 Given the risks of hypertension, a threshold of 150 mm Hg seems unjustifiably high for starting treatment. On PAMELA’s data, such guidance would translate into at least five avoidable deaths a year for every 1000 patients with hypertension.
Repositioning and redefinition of diuretics
For many years there has been an almost pantomime conflict between the drug industry’s hope that one drug class was superior to the others and guideline writers’ views that they are all equal. No one treatment is right for all patients, and most patients need a combination of drugs that reduce sodium load and vasodilate resistant arteries.8 The original AB/CD (angiotensin converting enzyme inhibitor or β blockers/calcium channel antagonist or diuretic) algorithm offered a choice at each stage, recognising that a major hurdle in optimising treatment is the 25-30% of patients who have adverse effects from treatment. The AB/CD algorithm avoided random switching with its inevitable loss of control.9
The only drug class that targets a known cause of hypertension is the diuretics.26 27 The new NICE guidance demotes diuretics from first to third choice, even though the evidence for these drugs is among the best of any drug for any indication. The guidelines now recommend chlortalidone, which is available in the UK only as a 50 mg tablet. Patients will need to use a tablet cutter and take half every other day to achieve the 12.5 mg dose used as starting dose in the Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial (ALLHAT) and other studies.28 Ironically, the new NICE guidelines recommend that doctors obtain informed consent before prescribing the most effective diuretic for hypertension—spironolactone—because it is not licensed for hypertension in the UK.29 30 Finally, NICE gives the option of using indapamide, for which there are no data from any primary prevention trial that met its primary end point.
NICE’s demotion of diuretics rests on the assumption that calcium channel blockers achieve a greater reduction in blood pressure variability. But variability has never been tested prospectively nor used as an entry criterion in any clinical trial of antihypertensive drugs. And the guidance seems contradictory since NICE invokes increased variability first as the indication for not treating 25% of patients with hypertension and then as the reason for treating most of the remainder with one class of drugs rather than another. In reality, the latest meta-analysis comparing calcium channel blockers with diuretics, which includes data from the international trialists’ collaboration, shows no difference in their effect on variability (P M Rothwell, personal communication).2 16 The conventional outcome data cited by NICE shows diuretics to be superior to calcium channel blockers in preventing heart failure and equivalent for other major outcomes, with no significant difference in cost effectiveness.3 NICE selects one outcome trial in which low dose thiazide plus angiotensin converting enzyme (ACE) inhibitor was less effective than high dose calcium channel blocker plus ACE inhibitor at reducing blood pressure and preventing cardiovascular complications.31 Since this was not a trial of first line treatment (97% were previously treated), and its results run counter to other trials comparing calcium channel blockers and diuretics, it cannot alone justify this big change.
After more than 50 years of use, NICE has redefined diuretics not according to their mechanism of action (on the Na+-Cl− cotransporter) or evidence from prospective outcome trials but according to their chemical structure. Thiazide and thiazide-like drugs are separated because they have “differential effects on carbonic anhydrase.” Once again, the underlying facts are questionable, since hydrochlorothiazide, but not bendroflumethiazide, inhibits carbonic anhydrase, whereas NICE wishes to band these drugs together.
Meanwhile, NICE’s (uncharacteristically) selective data review omitted the whole question of whether potassium sparing diuretics are desirable additives and, specifically, overlooked co-amilozide, which was as effective as the top dose of calcium channel blocker in one trial, and much better than placebo in another.32 33 34 The 44% reduction in coronary events with co-amilozide in the MRC Older Adults trial was remarkable.32 Co-amilozide is cheap and could replace bendroflumethiazide as first choice for diuretic.34 However, increasing the dose of bendroflumethiazide could be just as effective. Whenever maximal, equihypotensive doses of thiazide and calcium channel blockers have been compared in morbidity and mortality trials, there is no difference in the primary outcomes.2 28 33 The lower efficacy of adding bendroflumethiazide to a β-blocker in reducing secondary outcomes in a trial of 18 000 patients, is more plausibly related to the submaximal (1.25-2.5 mg) dose than to the choice of diuretic.35 36
The success of antihypertensive therapy does not blunt the need for further research on a condition that remains a major global cause of serious morbidity and mortality.37 Why NICE’s consultation exercise did not forestall the criticisms raised here is unclear. Several of the points were made during stakeholder discussions, and the absence of outcome trials based on ambulatory measurements is well known.38
What should doctors do now? The British Hypertension Society agreed unanimously at its 2011 annual meeting that only robust research is a basis for substantive changes to practice. Let this NICE guidance be the catalyst. We need to know “beyond all reasonable doubt” that we are doing the right thing—and that requires prospective clinical trials.
Cite this as: BMJ 2012;344:d8218
Contributors and sources: This article arose from presentations and discussion of the NICE 2011 guidance at the annual open meeting of the British Hypertension Society, recorded as a BHS webcast (www.bhsoc.org/stream/index.html ). The views expressed are those of the authors and are based both on extensive literature research and on personal experience as clinical trialists, membership of NICE and BHS guideline committees, and of the Blood Pressure Lowering Treatment Triallists Collaboration.2 All authors contributed equally to the writing of the article.
Competing interests: All authors have completed the ICJME unified disclosure form at www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author) and declare no support from any organisation for the submitted work; no financial relationships with any organisations that might have an interest in the submitted work in the previous three years; The authors are past or present president, vice president, or executive members of the British Hypertension Society, and chief or principal investigator of the BHS’s PATHWAY programme of trials. MJB is a member of the BHS guideline committee.
Provenance and peer review: Not commissioned; externally peer reviewed.