Should we revise our approach to 'optimal medical therapy'? The case of chronic heart failure

The concept of ‘optimal medical therapy’ relies on recommendations of current consensus-derived guidelines, most of which are supported by ‘level of evidence C’ (no scientific evidence).1 However, guidelines recommendations for commonly employed cardiovascular drugs often are based on strong evidence (‘level A’).2 Generally, such drugs are recommended at ‘target doses’ employed in the trials on which their putative efficacy and regulatory approval are based, and are considered ‘optimal medical therapy’. Information about the effectiveness of other doses, lower, or higher, seldom is available. Currently, reports of randomized controlled trials of new drugs in chronic heart failure (CHF) as well as in other conditions must quantitatively define the use of guidelines-recommended ‘optimal medical therapy’, usually as the proportion of patients who took these drugs, regardless of dose. However, in contemporary registries in CHF, for which therapy is particularly complex, large segments of the affected population fail to receive the recommended drugs and substantially less than half receive guidelines-based ‘target doses’. This is apparent from the Table 1, in which doses of drugs for CHF actually employed, from the 12-country 2010 European Survey of the European Society of Cardiology3 are compared with ‘target doses’ of the same drugs from contemporary (2008) published guidelines.4 Rate of use and dosages of the most frequently prescribed renin–angiotensin–aldosterone system blockers and beta-blockers. Modified from McMurray et al.2 Rate of use and dosages of the most frequently prescribed renin–angiotensin–aldosterone system blockers and beta-blockers. Modified from McMurray et al.2 The use of beta-blockers for CHF is an example. In a recent trial—SHIFT—testing the effect of a pure heart rate slowing agent—ivabradine—against placebo in patients with CHF, maximized guidelines-recommended background therapy was mandated, specifically for beta-blockers, which also slow heart rate.5 In SHIFT, 89% of patients received a beta-blocker but only one-fourth reached the ‘target doses’, and only 56% achieved ≥50% of target. These results subsequently were echoed in CIBIS-ELD,6 which evaluated doses of bisoprolol and carvedilol (beta-blockers commonly used in clinical practice) achieved in elderly patients with CHF. The drugs were randomized and then force-titrated to ESC Guidelines targets. At least 50% of the target dose was achieved in only 55% of the patients. Similar results were obtained in community studies7,8 consistent with the data reported in Table 1. This huge ‘underdosing’ might be related to physician inexperience, undesired effects, or patients fears of them, multimorbidity and the attendant polypharmacy, limiting the uptitration which may have been done in the original trials from which the targets were derived, (of note, in the landmark trials many patients, ranging from 22 to 53%, could not reach the protocol-defined beta-blocker target doses in spite of forced-titration). Nonetheless, the target doses recommended in the current guidelines are those predefined as targets in these remote studies, perpetually repeated in guidelines as ‘optimal medical therapy’ in spite of evidence of increasingly diverging clinical practice.2 In fact, if the recommended target doses are appropriate, we might infer that most patients currently are treated inadequately. Alternatively, we might wonder whether the range of doses currently used in clinical practice represents patient needs and drug tolerability in the current era. If guidelines-recommended ‘target doses’ should be reconsidered, how can we define therapeutic recommendations? Perhaps, for drugs affecting easily detectable ‘biomarkers’ related to their pharmacological effects and associated with proved clinical benefit and safety, the concept of ‘target dose’ should be replaced by ‘target effect’. This would imply that once the clinical benefit of a drug is established, a strict and predictable relationship between drug efficacy and safety should be sought relative to an appropriate biomarker, if such exists. Thereafter, the biomarker response could guide dosing. Early drug development often provides a basis for this effort: new drugs commonly are tested in dose-finding studies employing surrogate biomarker endpoints. However, such trials are relatively small, commonly safety-oriented, and feature efficacy endpoints frequently not practically transferable to clinical outcomes. The individual responsiveness to the drugs is variable; lower doses may be better than higher as a function of patient age, individual responsiveness, comorbidities, polypharmacy, etc.9 A biomarker could help in defining optimal dosing for individual patients. Since many drugs used in prevention of major cardiovascular events are intended for lifetime use, defining the lowest effective dose for an individual may have important medical and socio-economical implications. Professional Societies or Regulatory Bodies should decide when a surrogate marker is acceptable to guide therapy and which effect should be taken as the ‘target’. Is this concept practically applicable? Indeed, it is applied systematically to titrate the drug doses in hypertension, diabetes, and dyslipidaemia.10 Besides the beta-blockers, discussed above, statins are examples of drugs having a definable target effect related to outcomes. Actually both have many pharmacological effects (some desirable, some not). We do not know the precise dose at which the drug exerts each effect maximally, nor the relative contribution of each pharmacological effect to the net clinical effect of the drug. Thus, we do not truly know the ‘optimal dose’ for any individual patient. However, for beta-blockers in CHF, we know the relationship between the drug-induced change in heart rate and some important clinical outcomes. Hence, heart rate is a trustworthy metric of the effect of beta-blockade, as recently shown by a well-conducted meta-analysis of randomized controlled trials11 and other studies8 in CHF demonstrating that the dose of beta-blocker has little influence on mortality while heart rate change induced by the drug is directly associated with mortality benefit. Statins may be the drug group in which the effects of individual approved drugs differ most widely. Within this group, multiple doses of several drugs have been compared head to head. The results have defined different doses of drugs that can achieve well-defined surrogate biomarker outcomes (usually LDL cholesterol) which correlate well with clinical benefit. LDL cholesterol is, indeed, a desirable target biomarker since it is a modifiable risk factor (i.e. its change relates directly to alteration in clinical outcomes), not just a risk marker, of atherothrombosis. However, most statin trials employed different fixed doses that were not modified according to the level of LDL achieved. A pragmatic biomarker-based approach to dosing has been applied with several other groups of drugs, including anti-hypertensive, antidiabetic, and anticoagulant agents. Indeed, several trials in hypertension and diabetes have shown the benefits of treating patients to target blood pressures and HbA1c concentrations. The COURAGE trial has set the standard for ‘optimal medical therapy’ by treating patients effectively to multiple biomarker targets and averting major adverse cardiac events,12 though the impact on clinical practice has been less than expected.13 The principle of target effect cannot be applied for drugs with primary effects not including suitable biomarkers. For instance, for modulators of the renin–angiotensin–aldosterone system, blood pressure might be the appropriate basis for dose selection for hypertension but not for heart failure. In patients with the latter condition blood pressure could be monitored for safety, to prevent undesired hypotension, but not for efficacy, which may be related to pharmacological effects other than on blood pressure. Antiplatelet agents have been extensively investigated in search of the lowest dose that optimally balances antithrombotic and pro-haemorrhagic effects. However, defining the best dose individually is difficult because of the lack of availability, to date, of a quick, reliable and inexpensive measure that can be applied at the bedside.14 Ultimately, pharmacogenetics and perhaps other and newer approaches will enable optimized individualized dosing. Genetic variations associated with variations in responses to drugs have been demonstrated for cardiovascular drugs including diuretics, ACE inhibitors, and beta-blockers.15 Observational data also can help to clarify drug and dose effects, firstly by recording drug doses used in clinical practice. These may result from beliefs and behaviours of physicians in prescribing drugs or from beliefs of patients on their effectiveness and safety in taking them. Secondly, long-term observational registries can be designed to clarify the reasons why certain drugs are most commonly employed at far lower than the guidelines-recommended targets. Thirdly, these observational studies can assess data on drug effectiveness using analysis plans designed to minimize the (inevitable) uncertainty deriving from non-randomized studies Finally, the writers of clinical practice guidelines (increasingly perceived as mandatory for all patients, with potential legal consequences if they are abrogated) should consider the appropriateness of automatically confirming historically accepted standards, and should take cognizance of the practical applicability of their guidelines, not focusing only on the <50% of affected patients able to reach the standard target doses. It is important to recognize that, as each new therapy is added to the existing list for any condition, we have no information about the continuing benefit of therapies tested before more current additions. In practice, the Guidelines should discuss whether or not the drug doses defined in placebo-controlled trials are still to be recommended in a new therapeutic context, and the level of recommendations should be graduated accordingly. Perhaps the presently recommended target doses might best be considered as thresholds beyond which drugs have not been tested, rather than targets that must be achieved. L.T., A.P.M., and J.S.B. equally contributed to the preparation of this text. Conflict of interest: L.T. is consultant or committee member for Boston Scientific, Cardiorentis, Servier, S. Jude Medical, Vifor Pharma. A.P.M. is consultant or committee member for Novartis, Cardiorentis, LonStar Heart, Servier, Amgen, Sanofi, Vifor Pharma. J.B. is consultant or committee member for Astellas, Novartis, Celladon, Cardiorentis, Biotronik, Biomarin, Pfizer, Roche-Genentech, Somahlution, Servier, and Salix.