NPC Archive Item: The idolatry of surrogates and the 10 commandments of the new therapeutics

NOTE – This is an archive post from the NPC and has not been updated since first publication. Therefore, some hyperlinks may no longer be working.
MeReC Rapid Review NPC Logo

13 January 2012

An article in the BMJ notes that surrogate outcomes are often used in clinical trials instead of patient-oriented outcomes such as death, quality of life, or functional capacity. Using blood glucose and other surrogates in type 2 diabetes as examples, the authors argue that over-interpretation of surrogates and ascribing them undue importance can damage patient care. Reviewing this article, Richard Lehman published the lead author’s ‘10 commandments for the new therapeutics’.

Prescribers, prescribing managers and all those involved in the use of medicines should be cautious in accepting claims for medicines’ effectiveness based on surrogate or ‘disease-oriented’ outcomes. This includes the amount of effort given to control surrogate markers of disease rather than to therapies proven to improve patient-oriented outcomes. They should also consider these suggested ‘10 commandments and reflect on whether they should apply them to their own practice.

The use and misuse of surrogate outcomes
The US Institute of Medicine defines surrogates as ‘biomarker[s] intended to substitute for a clinical endpoint [and] expected to predict clinical benefit (or harm…) based on epidemiologic, therapeutic, pathophysiologic, or other scientific evidence’. Examples include blood pressure, blood glucose and blood lipid levels. As the authors of the article discussed here note, surrogate markers are not intrinsically flawed. When interpreted appropriately, they can be helpful in risk stratification and in treatment. However, over-interpretation of surrogates can lead to misinterpretation of the evidence. Moreover, they can be so over-interpreted that control of the surrogate marker becomes an end in itself, which may distort treatment priorities and actually lead to patient harm.

The authors use several examples from the field of type 2 diabetes. Blood glucose is widely used as a surrogate marker in this condition and indeed there is no argument for poor blood glucose control in people with type 2 diabetes. However, as discussed in MeReC Rapid Review 2726, the CONTROL meta-analysis of the four key randomised controlled trials (RCTs) in people with type 2 diabetes (n=27,049) found that more-intensive blood glucose control reduced the relative risk of major cardiovascular (CV) events compared with less-intensive control but produced no significant difference in all-cause mortality, stroke or CV mortality. Furthermore, those RCTs found that reducing HbA1c levels too low may actually increase mortality, particularly when treatment is intensified with insulin. A large observational cohort study by Currie CJ, et al (see MeReC Rapid Review 1017) identified a ‘U-shaped’ relationship between HbA1c levels and mortality in type 2 diabetes. It found that an HbA1c of about 7.5% (59 mmol/mol) was associated with the lowest risk of all-cause mortality, with higher or lower levels associated with greater risk.

Rosiglitazone was licensed largely on the basis of its hypoglycaemic effects, but it was withdrawn from clinical use in September 2010 when it became clear that its use was associated with an increased risk of CV events (See MeReC Rapid Review 1872). Although several newer hypoglycaemic drugs are also effective at reducing HbA1c levels, they all lack robust patient-oriented outcome data, particularly around their CV effects and long-term safety.

Other widely-used surrogates can also be misleading. Although high levels of LDL-cholesterol are associated with an increased risk of CV disease, a large meta-analysis(n=41,778), found no statistically significant reduction in all-cause mortality or CV mortality with intensive statin (e.g. atorvastatin 80 mg/day) dosing compared to moderate or low dosing (e.g. simvastatin 20 mg/day), except in patients with acute coronary syndrome. This was despite a reduction in LDL-cholesterol (see MeReC Rapid Review 2835). It did find a significant reduction in non-fatal MI, a composite of coronary heart disease (CHD) death and non-fatal MI, and a composite of fatal- and non-fatal stroke (excluding TIA); but, importantly, it did not find the widely-assumed linear association between reductions in LDL-cholesterol and reductions in the risk of CHD death or non-fatal MI.

Not all interventions that reduce LDL-cholesterol have been shown to prevent CV disease. For example, there is no published evidence that ezetimibe, alone or added to a statin, reduces the risk of CV disease or mortality compared with an active comparator (see MeReC Rapid Review 1722). Torectrapib was an investigational lipid modifying agent which showed impressive changes in blood lipids compared to placebo when added to atorvastatin: an average 25% relative reduction in LDL-cholesterol, 9% relative reduction in triglycerides and an encouraging 72% relative increase in HDL-cholesterol. Unfortunately, it was also associated with a 25% increased risk of the composite of CHD death, non-fatal MI, stroke and hospitalisation for unstable angina and a 58% increased risk in all-cause mortality in relative terms (hazard ratio [HR]1.25, 95% confidence interval [CI] 1.09 to 1.44 and HR 1.58, 95%CI 1.14 to 2.19 respectively).

The new therapeutics: 10 commandments
Commenting on this article, Richard Lehman outlines the 10 commandments as devised by John Yudkin, the lead author of this article.

  1. Thou shalt treat according to level of risk rather than level of risk factor.
  2. Thou shalt exercise caution when adding drugs to existing polypharmacy.
  3. Thou shalt consider benefits of drugs as proven only by hard endpoint studies.
  4. Thou shalt not bow down to surrogate endpoints, for these are but graven images.
  5. Thou shalt not worship Treatment Targets, for these are but the creations of Committees.
  6. Thou shalt apply a pinch of salt to Relative Risk Reductions, regardless of P values, for the population of their provenance may bear little relationship to thy daily clientele.
  7. Thou shalt honour the Numbers Needed to Treat, for therein rest the clues to patient-relevant information and to treatment costs.
  8. Thou shalt not see detailmen [representatives], nor covet an Educational Symposium in a luxury setting.
  9. Thou shalt share decisions on treatment options with the patient in the light of estimates of the individual’s likely risks and benefits.
  10. Honour the elderly patient, for although this is where the greatest levels of risk reside, so do the greatest hazards of many treatments.

Further information on many of the concepts included in these 10 commandments can be found in the evidence-informed decision-making e-learning section of the NPC website.

Please comment on this rapid review using our feedback form.

Make sure you are signed up to NPC Email updates — the free email alerting system that keeps you up to date with the NPC news and outputs relevant to you.