Guidelines for treating risk factors should include tools for shared decision makingBMJ 2016; 353 doi: https://doi.org/10.1136/bmj.i3147 (Published 14 June 2016) Cite this as: BMJ 2016;353:i3147
- John S Yudkin, emeritus professor of medicine1,
- Jayne Kavanagh, lead of medical ethics and law unit2,
- James P McCormack, professor3
- 1Division of Medicine, University College London, London, UK
- 2Academic Centre of Medical Education, UCL Medical School, Royal Free Campus, London, UK
- 3Faculty of Pharmaceutical Sciences, University of British Columbia, Vancouver, BC, Canada
- Correspondence to J S Yudkin
In a recently published case report titled “The tyranny of guidelines,” Sarosi recounts the story of an 86 year old man living on his farm in Wisconsin and caring for his 92 year old brother with early dementia.1 Six years earlier he had been started on an angiotensin converting enzyme inhibitor and metformin after a health check, with other oral drugs subsequently added. But, when his family practice was taken over by a large organisation, he was given a copy of the American Diabetes Association guidelines and started on insulin because his haemoglobin A1c concentration was 8.5%; his antihypertensive dose was also doubled because his blood pressure was 154/92 mm Hg. Three weeks later he was admitted to hospital hypotensive and hypoglycaemic, with a hip fracture and a stroke. Both he and his brother subsequently needed residential care. The author pointed out that the guidelines stated: “Older adults who are functional and cognitively intact and have significant life expectancy should receive diabetes care with goals similar to those developed for younger adults” (HbA1c 7% and blood pressure <140/90 mm Hg).2
Importance of individual benefit
The clinicians might claim that they were only following guidelines. But, when linked to quality measures and reimbursement, guidelines can morph into orders. These guidelines suggest a target HbA1c below 8% and blood pressure <140/90 mm Hg in elderly patients unless their health status is “very complex/poor … (long term care … end-stage chronic illnesses or moderate-to-severe cognitive impairment)” with limited remaining life expectancy.2 And though the American Diabetes Association and the European Association for the Study of Diabetes recommend that “where possible, such decisions should be made with the patient, reflecting his or her preferences, needs, and values,”3 they provide no tools for quantifying the harm associated with a particular risk factor or information comparing likely benefits and risks of treatments.4 We argue that such tools, based on outcomes relevant to patients and likely gains in healthy life expectancy, are vital—not just for shared decision making but also better to inform clinicians, guideline committees, and comparative effectiveness agencies. In this man’s case, an outcome model would have estimated that the changes to his treatment would have extended his healthy life expectancy by no more than five weeks.5 6
The UK National Institute for Health and Care Excellence (NICE) updated its guidelines for management of adults with type 2 diabetes in December 2015.7 The NICE targets are similar to, but in some cases more aggressive than, the US and European guidelines. They recommend HbA1c of 7% (or 6.5% if it can be achieved by a single drug that does not cause hypoglycaemia) and blood pressure <140/80 mm Hg, although these targets can be relaxed in “people who are older or frail … or with a reduced life expectancy.”
NICE also states that “patients should have the opportunity to make informed decisions about their care and treatment, in partnership with their healthcare professionals.” But although the guidelines provide a useful decision aid for assessing risks of treatment—giving rates of adverse effects, including hypoglycaemia, with pictograms8—the only information on possible benefits is provided separately in the “decision aid user guide for healthcare professionals” and in the form of relative risk reductions.9 No information is provided to allow clinicians to quantify the risks associated with different levels of risk factors, or likely benefits with their reduction.
For a patient resembling the man we described above, a conversation about his glycaemic control using the information in the NICE guideline might go as follows. With the information in the user guide, he would be told that getting his HbA1c down by 1% might reduce his risk of a non-fatal myocardial infarction by 13% (a relative benefit), but the healthcare professional could not give any indication of his baseline risk or the possible gains in healthy life expectancy (box 1). He might be told, based on the guidelines, that it was important to improve his glycaemic control to prevent blindness or renal failure. However, he would not be told that the evidence for this benefit is extrapolated from a 20%-25% reduction in surrogate endpoints (retinal photocoagulation and proteinuria changes) and that there is virtually no evidence for glucose control actually reducing risks of blindness and end stage renal disease.10 He would also not have been told his lifetime risks of these events are at most 1-2%.11
Box 1: Likely benefits of starting insulin in patient described*
Reduction of 1%
Fatal event—no effect
Non-fatal event–13% reduction in relative risk
Absolute risk reduction—3.7% at 10 years
25% reduction in relative risk (from surrogate endpoints)
Absolute risk reduction 2.3% at 10 years
End stage renal failure
25% reduction in relative risk (from surrogate endpoints)
Absolute risk reduction 0.03% at 10 years
Life expectancy gain
About 5 weeks
*Estimates are derived from the UKPDS outcomes model 25
Fully informed decisions
Fully informed decision making needs fully informed clinicians as well as patients. Absolute risk reductions and numbers needed to treat are now more widely disseminated, but when benefits are expressed as likely gains in healthy life expectancy these are, at best, moderate.12 Moreover, in people with a 10 year cardiovascular risk of 40% who start statins (relative risk reduction about 25%) nine out of 10 will not benefit in that timespan. In people at much lower risk, or in those treated with less effective strategies (like glucose lowering),13 the interventions are more for the benefit of population health than individual health. Many clinicians are unaware of this when they write prescriptions,2 14 15 In one study, physicians presented with three “grey cases” grossly overestimated the probable benefits of intensifying glucose control (median around 7 years versus around 5 weeks from an outcomes model) and blood pressure lowering (about 7 years versus about 10 weeks) 14 Better tools to estimate likely harms, benefits, and risks might also help the NICE guideline development group assess the likely additional benefits from reducing target HbA1c to 6.5%.6 16
In its professional guidance, the UK General Medical Council (GMC) advises doctors that their role is to outline the benefits, risks, and burdens of a treatment or procedure in clear and understandable fashion, but it is the patient who weighs up the information, together with other relevant issues, and makes the decision.17 A recent UK Supreme Court judgment concerning the information provided to Nadine Montgomery about the benefits and risks of a caesarean section for her and her baby has now configured legal obligations with ethical guidance.18 The Montgomery judgment highlighted the importance of shared decision making that is properly informed; this may be interpreted as establishing a new legal requirement that information about the potential harms and benefits of a proposed course of action should be communicated accurately.19
In the context of treating risk factors, two other considerations often apply—the patient has no symptoms and the treatment has the potential to be lifelong. Although many patients will still defer to their clinician on decisions about such treatment, guideline developers should surely include, or at least provide some directions to, shared decision making tools that could be used (box 2). Because people differ widely in their willingness to take preventive medication,23 24 and their responses to different formats for explaining benefits,21 such tools need to show estimates of benefit in a variety of formats—absolute risk reductions (ARR), numbers needed to treat (NNT), and gains in healthy life expectancy.6 12 21 22 Furthermore, the multiplicity of complications of diabetes, which benefit to different degrees from improved glycaemic control, make it difficult to express benefit as absolute risk reduction or number needed to treat.13 Because of this, a model derived summary measure, such as gains in healthy life years, may be more relevant for intensive glucose lowering.6 However, the observation that in people with type 2 diabetes the effect of glucose lowering on such summary measures may be quantified in weeks or months 6 12 should also influence the agenda of cost effectiveness agencies.
Box 2: Tools for estimating benefits of interventions to lower cardiovascular risk factors20
Absolute cardiovascular disease risk/benefit calculator (http://chd.bestsciencemedicine.com/calc2.html)
Mayo Clinic heart disease risk calculator (http://www.mayoclinic.org/diseases-conditions/heart-disease/in-depth/heart-disease-risk/itt-20084942)
Mayo Clinic diabetes decision aid (https://diabetesdecisionaid.mayoclinic.org/index.php/site/compare?PHPSESSID=k3sgf2rju1t2bpo8738bf95354)
Healthy Living for People with Diabetes web-based self-management programme21
London School of Economics statin ranking tool* (http://www.lse.ac.uk/IPA/ResearchAndEngagement/ProjectArchive/VisualisingData/StatinRankingTool.aspx).
*Although this is not a diabetes decision aid, it shows how it is possible to integrate comparative treatment rankings from network meta-analysis with patient preferences in decision making.
Use of summary measures might permit a more patient specific assessment of the value of glucose lowering medication. Token statements such as “patients should have the opportunity to make informed decisions about their care and treatment” do not make for patient centred guidelines. If guideline writers with all their expertise and resources can’t come up with specific tools or approaches, individual users are unlikely to be able to fill in the knowledge translation blanks left by these guidelines.
Guidelines on preventive treatment are generally based on population data
Strict adherence to recommendations may not benefit individuals
Guideline writers should provide guidance to help the clinician and patient consider not just the risks of treatments but also the likelihood of benefit for that individual, expressed in different formats
Contributor and sources JSY conceived the concept of individual versus population health benefit. JK introduced the discussion of the relevance of the recent Supreme Court judgment. The three authors have discussed several previous drafts of the manuscript. JSY is a physician with an interest in diabetes, who has conducted research over the past 25 years on the benefits and hazards of risk factor reduction and in particular on the problems of a “glucocentric” approach to type 2 diabetes. More recently he has explored the potential benefits and hazards of overmedicalisation with regard to intensive treatment of diabetes and prediabetes. JK is a medical ethicist and a physician working in sexual health. JM is a pharmacist with an interest in outcomes based research and in communication for shared decision making.
Competing interests: I have read and understood BMJ policy on declaration of interests and have no relevant interests to declare
Provenance and peer review: Not commissioned; externally peer reviewed.