Drug Efficacy and Safety
Obviously, a drug (or any medical treatment) should be used only when it will benefit a patient. Benefit takes into account both the drug's ability to produce the desired result (efficacy) and the type and likelihood of adverse effects (safety). Cost is commonly also balanced with benefit (see Economic Analyses in Clinical Decision Making).
Efficacy is the capacity to produce an effect (eg, lower BP). Efficacy can be assessed accurately only in ideal conditions (ie, when patients are selected by proper criteria and strictly adhere to the dosing schedule). Thus, efficacy is measured under expert supervision in a group of patients most likely to have a response to a drug, such as in a controlled clinical trial.
Effectiveness differs from efficacy in that it takes into account how well a drug works in real-world use; often, a drug that is efficacious in clinical trials is not very effective in actual use. For example, a drug may have high efficacy in lowering BP but may have low effectiveness because it causes so many adverse effects that patients stop taking it. Effectiveness also may be lower than efficacy if clinicians inadvertently prescribe the drug inappropriately (eg, giving a fibrinolytic drug to a patient thought to have an ischemic stroke, but who had an unrecognized cerebral hemorrhage on CT scan). Thus, effectiveness tends to be lower than efficacy.
Patient-oriented outcomes, rather than surrogate or intermediate outcomes, should be used to judge efficacy and effectiveness.
Surrogate, or intermediate, outcomes involve things that do not directly involve patients' well-being. They are often such things as physiologic parameters (eg, BP) or test results (eg, concentrations of glucose or cholesterol, tumor size on CT scan) that are thought to predict actual patient-oriented outcomes. For example, clinicians typically presume that lowering BP will prevent the patient-oriented outcome of uncontrolled hypertension (eg, death resulting from MI or stroke). However, it is conceivable that a drug could lower BP but not decrease mortality, perhaps because it has fatal adverse effects. Also, if the surrogate is merely a marker of disease (eg, HbA1c) rather than a cause of disease (eg, elevated BP), an intervention might lower the marker by means that do not affect the underlying disorder. Thus, surrogate outcomes are less desirable measures of efficacy than patient-oriented outcomes.
On the other hand, surrogate outcomes can be much more feasible to use, for example, when patient-oriented outcomes take a long time to appear (eg, kidney failure resulting from uncontrolled hypertension) or are rare. In such cases, clinical trials would need to be very large and run for a long time unless a surrogate outcome (eg, lowered BP) is used. In addition, the main patient-oriented outcomes, death and disability, are dichotomous (ie, yes/no), whereas surrogate outcomes are often continuous, numerical variables (eg, BP, blood glucose). Numerical variables, unlike dichotomous outcomes, may indicate the magnitude of an effect. Thus, use of surrogate outcomes can often provide much more data for analysis than can patient-oriented outcomes, allowing clinical trials to be done using many fewer patients.
However, surrogate outcomes should ideally be proved to correlate with patient-oriented outcomes. There are many studies in which such correlation appeared reasonable but was not actually present. For example, treatment of certain postmenopausal women with estrogen and progesterone resulted in a more favorable lipid profile but failed to achieve the hypothesized corresponding reduction in MI or cardiac death. Similarly, lowering blood glucose to near-normal concentrations in patients with diabetes in the ICU resulted in higher mortality and morbidity (possibly by triggering episodes of hypoglycemia) than did lowering blood glucose to a slightly higher level. Some oral antihyperglycemic drugs lower blood glucose, including HbA1c concentrations, but do not decrease risk of cardiac events. Some antihypertensive drugs decrease BP but do not decrease risk of stroke.
Similarly, clinically relevant adverse effects are patient-oriented outcomes; examples include the following:
Surrogate adverse effects (eg, alteration of concentrations of serum markers) are often used but, as with surrogate efficacy outcomes, should ideally correlate with patient-oriented adverse effects. Clinical trials that are carefully designed to prove efficacy can still have difficulty identifying adverse effects if the time needed to develop an adverse effect is longer than the time needed for benefit to occur or if the adverse effect is rare. For example, cyclooxygenase-2 (COX-2) inhibitors relieve pain quickly, and thus their efficacy can be shown in a comparatively brief study. However, the increased incidence of MI caused by some COX-2 inhibitors occurred over a longer period of time and was not apparent in shorter, smaller trials. For this reason, and because clinical trials may exclude certain subgroups and high-risk patients, adverse effects may not be fully known until a drug has been in widespread clinical use for years (see Drug Development).
Many drug adverse effects are dose-related.
Whether a drug is indicated depends on the balance of its benefits and harms. In making such judgments, clinicians often consider factors that are somewhat subjective, such as personal experience, anecdotes, peer practices, and expert opinions.
The number needed to treat (NNT) is a less subjective accounting of the likely benefits of a drug (or any other intervention). NNT is the number of patients who need to be treated for one patient to benefit. For example, consider a drug that decreases mortality of a certain disease from 10% to 5%, an absolute risk reduction of 5% (1 in 20). That means that of 100 patients, 90 would live even without treatment, and thus would not benefit from the drug. Also, 5 of the 100 patients will die even though they take the drug and thus also do not benefit. Only 5 of the 100 patients (1 in 20) benefit from taking the drug; thus, 20 need to be treated for 1 to benefit, and the NNT is 20. NNT can be simply calculated as the inverse of the absolute risk reduction; if the absolute risk reduction is 5% (0.05), the NNT = 1/0.05 = 20. NNT can be calculated for adverse effects also, in which case it is sometimes called the number needed to harm (NNH).
Importantly, NNT is based on changes in absolute risk; it cannot be calculated from changes in relative risk. Relative risk is the proportional difference between two risk levels. For example, a drug that decreases mortality from 10% to 5% decreases absolute mortality by 5% but decreases relative mortality by 50% (ie, a 5% death rate indicates 50% fewer deaths than a 10% death rate). Most often, benefits are reported in the literature as relative risk reductions because these make a drug look more effective than the absolute risk reductions (in the previous example, a 50% reduction in mortality sounds much better than a 5% reduction). In contrast, adverse effects are usually reported as absolute risk increases because they make a drug appear safer. For example, if a drug increases the incidence of bleeding from 0.1% to 1%, the increase is more likely to be reported as 0.9% than 1000%.
When balancing NNT against NNH, it is important to weigh the magnitude of specific benefits and harms. For example, a drug that causes many more harms than benefits may be worth prescribing if those harms are minor (eg, reversible, mild) and the benefits are major (eg, preventing mortality or morbidity). In all cases, patient-oriented outcomes are best used.
Genetic profiling is increasingly being used to identify subgroups of patients that are more susceptible to the benefits and adverse effects of some drugs. For example, breast cancers can be analyzed for the HER2 genetic marker that predicts response to particular chemotherapy drugs. Patients with HIV/AIDS can be tested for the allele HLA-B*57:01, which predicts hypersensitivity to abacavir, reducing the incidence of hypersensitivity reactions and thus increasing NNH. Genetic variations in various drug-metabolizing enzymes help predict how patients respond to drugs (see Pharmacogenetics) and also often affect the probability of benefit, harm, or both.
One goal in drug development is to have a large difference between the dose that is efficacious and the dose that causes adverse effects. A large difference is called a wide therapeutic index, therapeutic ratio, or therapeutic window. If the therapeutic index is narrow (eg, < 2), factors that are usually clinically inconsequential (eg, food-drug interactions, drug-drug interactions, small errors in dosing) can have harmful clinical effects. For example, warfarin has a narrow therapeutic index and interacts with many drugs and foods. Insufficient anticoagulation increases the risk of complications resulting from the disorder being treated by anticoagulation (eg, increased risk of stroke in atrial fibrillation), whereas excessive anticoagulation increases risk of bleeding.