NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

National Clinical Guideline Centre (UK). Hypertension: The Clinical Management of Primary Hypertension in Adults: Update of Clinical Guidelines 18 and 34 [Internet]. London: Royal College of Physicians (UK); 2011 Aug. (NICE Clinical Guidelines, No. 127.)

Cover of Hypertension

Hypertension: The Clinical Management of Primary Hypertension in Adults: Update of Clinical Guidelines 18 and 34 [Internet].

Show details

42004 Methods

4.1.1. Review methods

The aim of reviewing was to identify and synthesise relevant published and unpublished evidence to allow recommendations to be evidence-based wherever possible.630 The search was carried out using the electronic databases MEDLINE, EMBASE and CENTRAL, attempting to locate systematic reviews and meta-analyses, and original randomised trials using a combination of subject heading and free text searches. We made extensive use of high quality recent review articles and bibliographies, as well as contact with subject area experts. New searches were concentrated in areas of importance to the guideline development process, for which existing systematic reviews were unable to provide valid or up to date answers. The expert knowledge and experience of group members also backed up the search of the literature.

Electronic searches used a sensitive search strategy based on a combination of text and index terms to locate randomised controlled trials of treatments relevant to the guideline. If data necessary for our analyses were not reported, we wrote to authors or sponsoring agencies. We are grateful to investigators and sponsors who provided unpublished information to aid our work.

We assessed the quality of relevant studies retrieved and their ability to provide valid answers to the clinical questions addressed by the group. Assessment of study quality concentrated on internal validity (the extent to which the study measured what it intended to measure), external validity (the extent to which study findings could be generalised to other treatment settings) and construct validity (the extent to which measurement corresponded to theoretical understanding of a disease).139

Table 8. Quality Criteria for Randomised Controlled Trials.

Table 8

Quality Criteria for Randomised Controlled Trials.

Once data had been abstracted from individual papers and their quality assessed, the information was synthesised. Individual trials often have an insufficient sample size to identify significant outcomes with confidence81, so where appropriate, the results of randomised studies were combined using meta-analytic techniques 175. Questions were answered using the best evidence available. When considering the effect of an intervention, if this could be addressed by the best study design then weaker designs were not reviewed. Where studies were of poor quality, or contained patient groups considered likely to have different responses, the effects of inclusion or exclusion were examined in sensitivity analyses. No trials that met our inclusion criteria were excluded from the primary analyses. However, where data on relevant outcomes were not available, these studies could not be included, thus leading to the potential for publication bias.

Review criteria

Scoping work revealed a vast number of trials of pharmaceutical interventions. Recent work suggests that study size is a useful proxy for study quality.189,224 Consequently to achieve the task in the timescale provided we reviewed only those pharmaceutical studies which enrolled 200 or more patients. Since the prime motivation for treatment in hypertension, an asymptomatic condition, is the prevention of mortality and morbidity, we reviewed those studies with a planned follow-up of at least a year since such studies are likely to have been designed to inform about these endpoints. Few non-pharmacological studies directly address cardiovascular endpoints or feature substantial durations of follow-up. Consequently in these areas we evaluated blood pressure reduction as a proxy endpoint and included trials with a follow-up of 8 weeks follow-up or more, which compared a group receiving a lifestyle intervention with a control group who received no treatment, usual treatment, sham therapy or a placebo.

Statistical methods

Pharmacological interventions

The outcomes analyzed were: all cause mortality, fatal and non-fatal myocardial infarction, fatal and non-fatal stroke. We did not consider the following endpoints: renal disease (rare in non-diabetic patients); heart failure (inconsistently reported in trials); cardiovascular events (a concatenation of myocardial infarction and stroke). For each trial, the risk ratios comparing the risk of each outcome in the active treatment and control groups - or, for head-to-head trials, in the different treatment groups - were calculated. Results of trials were combined in a meta-analysis using the DerSimonian and Laird random effects model175, to estimate an overall pooled risk ratio (RR) and its 95% confidence interval (95%CI). This model assumes that there are different effects of treatment in different populations, which are clustered about a mean effect; the pooled RR gives the best estimate of this mean effect. In the placebo-controlled trials reported in this guideline, a RR less than 1 favours treatment and a RR greater than 1 favours control. If the 95%CI include 1, there is no statistically significant difference between the treatments being compared.

Finally, we assessed the tolerability of the interventions by comparing the rate of overall withdrawal (percentage of patients who withdrew each year) in each treatment arm of a trial and calculating the difference in these rates (called the ‘incident risk difference’). These incident risk differences were combined in a meta-analysis using the DerSimonian and Laird random effects model175, to estimate an overall pooled incident risk difference and its 95% confidence interval.

We assessed heterogeneity between trials using a chi-squared statistic (Q). This assesses whether the trials are sufficiently similar to be validly combined. Although the test for heterogeneity is weak, it is usually assumed that if it gives p-values greater than 0.10, there is no significant heterogeneity and it is valid to discuss the combined findings.

We also assessed whether the effect in individual trials was related to the size of the trial; any such trend might indicate publication bias, e.g. where small trials were published only if they showed a positive effect. Again, this test for systematic variation in the magnitude of the estimated effect with the size of the trial is weak, but it is usually assumed that if it gives a p-value greater than 0.10, there is unlikely to be any such bias.

Lifestyle interventions

None of the studies identified were designed to quantify significant changes in rates of death or cardiovascular events, so we analysed the surrogate endpoint of reduced blood pressure. For each trial, the difference in the final value mean blood pressure in the treatment and control groups - or, for head-to-head trials, in the different treatment groups - was calculated. Change scores from baseline were used where complete data for final values was unavailable. These mean differences were weighted according to the precision of each trial (which depends largely on its size, with larger trials getting more weight) and combined in a meta-analysis using the DerSimonian and Laird random effects model175, to estimate an overall pooled weighted mean difference and its 95% confidence interval. While most of the trials were of parallel design (two or more groups received the various interventions at the same time), some were of crossover design (all participants received both active treatment and control interventions, but in a random order). Crossover trials have about four times greater precision than parallel trials of the same size, so we used methods have been developed recently to combine the parallel and crossover trials in the same meta-analysis.147,193 Heterogeneity and the potential for publication bias were assessed in the same way as for pharmaceutical trials.

The mean percentage achieving a reduction of 10mmHg or more in systolic blood pressure was then estimated from the cumulative normal distribution637 and confidence intervals were estimated using the delta method.51

Finally, we assessed the tolerability of the interventions by comparing the proportion of withdrawals (% of patients who withdrew) in each treatment arm of a trial and calculating the difference in these proportion (called the ‘risk difference’). These risk differences were combined in a meta-analysis using the DerSimonian and Laird random effects model,175 to estimate an overall pooled risk difference and its 95% confidence interval.

4.1.2. Group process

The guideline development group was run using the principles of small group work and was led by a trained facilitator. The group underwent initial exercises to set its own rules to determine how it wanted to function and received brief training on reviewing methods, economic analysis and grading methodology. Additional training was provided in the group as the need arose in subsequent meetings. Findings, expressed as narratives, statements of evidence and recommendations, were reached by informal consensus. There was no obligation to force an agreement where none existed after discussion: dissensions were recorded in the guideline narrative.471

4.1.3. Evidence statements and recommendations

The guideline development group process produces summary statements of the evidence concerning available treatments and healthcare and from these makes its recommendations. Evidence statements and recommendations are commonly graded in guidelines reflecting the quality of the study designs on which they are based. An established scheme adapted from the Agency for Health Care Policy and Research (AHCPR) Classification is shown in Table 9 and Table 10.14

Table 9. AHCPR derived categories of evidence.

Table 9

AHCPR derived categories of evidence.

Table 10. AHCPR derived strengths of recommendations.

Table 10

AHCPR derived strengths of recommendations.

Two grading schemes were used when developing this guideline, the one above and a new scheme called GREG (Guideline Recommendation and Evidence Grading).392 The new scheme seeks to address a number of problems, by extending grading from treatment to include diagnosis, prognosis and cost, and to handle the subtleties of clinical evidence more sensitively (Table 11).

Table 11. GREG scheme for assessing evidence and writing recommendations.

Table 11

GREG scheme for assessing evidence and writing recommendations.

Use of the two schemes was evaluated in this and another guideline being developed contemporaneously. Both groups consistently favoured the new scheme and so the guideline is presented using the new grading scheme. The evaluation of the two schemes will be reported separately.

The key point of note is that any assessment of evidence quality is ultimately a subjective process. How bad does a trial have to be before it is flawed or how sparse do the findings have to be before we lose confidence in the findings? The purpose of an evidence grading scheme is to characterise the robustness of outcomes from studies, and the random and systematic biases that pertain to them.

Similarly recommendation grading must credibly assimilate evidence and health service context to credibly advise lines of care for average patients. Clinicians must use their judgement and awareness of patients’ circumstances and values when considering recommendations from guidelines.

4.1.4. Costs and consequences

Approaches to cost-effectiveness have assisted in reaching recommendations in a series of primary care evidence-based guidelines.188,393 This guideline involves a systematic appraisal of effectiveness, compliance, quality-of-life, safety and health service resource use and costs of a medical intervention provided in the British health care setting. Using the most current, pertinent and complete data available, the economic analysis attempts a robust presentation showing the possible bounds of cost-effectiveness that may result.

The guiding principle behind economic analysis is that it is desirable to use limited healthcare resources to maximise health improvements in the population. Well defined but narrow notions of health improvement may not reflect all aspects of value to patients, carers, clinicians or society. For example, evidence may lead the guideline group to recommend targeting additional resources to certain patient groups when unequal access to care is apparent. The group process allows discussion of what should be included in the definition of ‘improved health’ and more broadly of other concepts of value to society such as fairness, justice, dignity or minimum standards of care.

  • The range of values used to generate cost-effectiveness estimates reflects the available evidence and the concerns of the guideline development group. Recommendations are graded reflecting the certainty with which the costs and consequences of a medical intervention can be assessed. This practice reflects the desire of group members to have simple, understandable and robust information based on good data.
  • It is not generally helpful to present an additional systematic review of previous economic analyses that have adopted a variety of differing perspectives, analytic techniques and baseline data. However, the economic literature is reviewed to compare guideline findings with representative published economic analyses and to interpret any differences in findings when these occurred. A commentary is included when the group feel this aids understanding.

4.2. 2006 methods

4.2.1. Clinical evidence Methodological introduction

Study inclusion and reporting criteria

A systematic search of the literature was performed on EMBASE and MEDLINE for randomised controlled trials comparing any combination of antihypertensive drugs from among the following five classes of drugs:

  • ACE inhibitors (ACEi)
  • angiotensin-II receptor antagonists (ARB)
  • beta-receptor blockers (BB)
  • calcium-channel blockers (CCB)
  • thiazide-type diuretics (TD).

Placebo-controlled studies were not included because the main aim of this rapid partial update was to make recommendations regarding the optimal sequencing of drug treatment for hypertension, for which head-to-head studies are required, and because sufficient placebo-controlled studies of the main drug classes had been considered in the original NICE guideline. However, placebo-controlled studies were sought for isolated systolic hypertension because of a lack of comparator studies.

The cut-off date for evidence to be considered in the previous guideline was July 2004, so this update only searched for English-language titles published after that date. Papers published up to and including 19 December 2005 were considered – this constitutes the cut-off for evidence for this rapid update.

Studies were excluded due to:

  • inadequate or no randomisation
  • inadequate study power, defined as a sample size of less than 200 patients, or having a follow-up period of less than 12 months
  • having an exclusive diabetic or paediatric patient population, unrepresentative of the general UK hypertensive population
  • stroke, myocardial infarction, and mortality outcomes not being reported.

The following outcomes were recorded for each study, where available:

  • mortality from any cause
  • stroke (ischaemic or haemorrhagic)
  • myocardial infarction (including, where reported, silent MI)
  • new-onset diabetes mellitus
  • vascular procedures (including both coronary and carotid artery procedures)
  • incidence of unstable angina (or angina episodes requiring hospitalisation)
  • study drug withdrawal.
Interpretation and analysis of results

All outcomes, with the exception of study drug withdrawal, vascular procedures and unstable angina, were entered into a meta-analysis for each drug combination using RevMan 4.2 software (©The Nordic Cochrane Centre). The overall effect size was reported as the relative risk (RR) with 95% confidence intervals in each case.

A p-value less than 0.05 was considered statistically significant for overall effect. Forest plots for each comparison are included in Appendix A.

In recording the outcomes, stroke was considered to be synonymous with ‘cerebrovascular event’. Reports of ‘cardiovascular events’ or other composite outcomes other than those listed above were not considered.

Sensitivity analyses were performed based on the inclusion and exclusion of silent myocardial infarction and the inclusion and exclusion of secondary prevention studies. Additional subgroup analyses were performed to identify the source of any significant heterogeneity in study results (defined as an I2 statistic greater than 50%).

Where the heterogeneity has I2 greater than 50%, the trials are reported individually in the evidence statements.

The following outcomes were not subject to meta-analysis due to potential variability or subjectivity in diagnosis or treatment protocols, and were reported as a narrative only:

  • unstable angina
  • revascularisation procedures
  • study drug withdrawal.

Following consultation on the draft guideline, heart failure as an outcome was included in the meta-analysis. Because of inconsistency in definition of heart failure in the trials, this was analysed using a random effects model.

Secondary analyses

In addition to results in general hypertensive populations, the following subgroups were also considered separately:

  • those patients with isolated systolic hypertension (ISH)
  • black people of African and Caribbean descent younger patients (defined as under 55 years).

For ISH, due to the lack of evidence comparing different antihypertensive drugs, the results from placebo-controlled trials were also considered. These results included pre-defined subgroup analyses from trials in general hypertensive populations as well as one trial comprising only ISH patients. The results were entered into a meta-analysis according to the same procedure specified above. The definition of ISH varied slightly between studies: permitting a diastolic blood pressure up to 95 mmHg in one study (SYST-EUR43,124,555) and 90 mmHg in the others (SHEP483,536,537,606, SHEP-P281,484,485).

No trials comprising only non-white patients were found, although two pre-defined subgroup analyses from trials in general hypertensive populations were found (ALLHAT589–591, LIFE154,176,222,369,370,507,618,619). Results involving placebo comparisons in non-white populations were not considered.

Evidence on younger patients was extremely sparse, and evidence consideration was therefore extended to include papers pre-dating July 2004 and in which blood pressure lowering effect was the main outcome measure.

4.2.2. Cost-effectiveness evidence

The GDG drafted recommendations on the basis of the clinical evidence. A health economic analysis was then conducted to balance the clinical outcomes and to test the cost effectiveness of different initial antihypertensive medications.

See ‘Appendix I: Cost-effectiveness analysis – pharmacological treatment (updated 2011)’ for full methods – note that analysis was updated as part of the 2011 update.

Copyright © 2011, National Clinical Guideline Centre.

Apart from any fair dealing for the purposes of research or private study, criticism or review, as permitted under the Copyright, Designs and Patents Act, 1988, no part of this publication may be reproduced, stored or transmitted in any form or by any means, without the prior written permission of the publisher or, in the case of reprographic reproduction, in accordance with the terms of licences issued by the Copyright Licensing Agency in the UK. Enquiries concerning reproduction outside the terms stated here should be sent to the publisher at the UK address printed on this page.

The use of registered names, trademarks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant laws and regulations and therefore for general use.

The rights of National Clinical Guideline Centre to be identified as Author of this work have been asserted by them in accordance with the Copyright, Designs and Patents Act, 1988.

Bookshelf ID: NBK83290


Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...