Applying the findings of clinical trials to individual patients
ACP J Club. 1995 Mar-April;122:A12. doi:10.7326/ACPJC-1995-122-2-A12
Related Content in the Archives
• Letter: Applying the findings of clinical trials to individual patients
In this editorial, we consider how the findings of studies can be applied to individual patients. In a previous editorial, we described an approach to interpreting the results of randomized trials (1). We suggested that clinicians should begin by considering the validity of the trial. The key issues here are whether patients were in fact randomly assigned to the different treatments and then analyzed in the groups to which they were allocated, and whether follow-up was complete.
After confirming the validity of the study, clinicians should consider the magnitude and precision of the treatment effect, focusing particularly on the extent to which the treatment reduced the risk for an adverse outcome. This relative risk reduction (RRR) and the confidence interval around it tells us the extent to which we might expect the base-line risk (that is, the risk in untreated patients) to decrease when we administer the intervention. If an intervention reduces the risk for death by half, the RRR is 50%.
In applying the results of a valid study to a patient, clinicians should consider their patient's baseline risk and the magnitude of the benefit that the patient might expect if given the treatment. A useful way of thinking about this magnitude of the benefit is to consider the number of patients we need to treat to prevent a single adverse outcome, the number needed to treat (NNT) (2). Theclinician can calculate the NNT by considering how many of 100 untreated patients would have the adverse event and how many of 100 treated patients would have the event and then dividing 100 by the difference. For example, if 10 untreated patients would die and the treatment is associated with a RRR of 50%, 5 treated patients would also die. The difference is 5 patients, and the NNT is 100/5, or 20.
Thus, to optimally use clinical trial results, clinicians not only need to know the RRR associated with treatment but also must estimate their patient's baseline risk for the adverse outcome that the treatment prevents. Where can clinicians get this information? They may use their intuition about their patient's risk, but intuition is unlikely to provide as accurate an estimate as do carefully done studies. The most readily accessible published estimate is the overall risk from the patients who participated in the trial. But this information will be of limited use because patients arrive with widely varying risks; thus, the overall risk from the study may be a poor estimate for an individual patient. For example, in our previous editorial (1) we used the example of 2 patients with myocardial infarction, a 44-year-old man with no signs of heart failure and a 2% risk for death in the next 30 days, and a 72-year old man with heart failure and a 50% risk for death.
Information with these details can come from large trials if the authors have provided information about risk in different subgroups of patients. Further, clinicians may look to other randomized trials, meta-analyses of trials, or formal studies of prognosis to provide complementary information. In the rest of this editorial, we describe an example of our approach using an article about therapy from this issue of ACP Journal Club.
Doval and colleagues (3, 4) conducted a trial of low-dose amiodarone in patients with heart failure. The study was randomized, and patients were analyzed in the group to which they were allocated. 12 of 256 controls and 7 of 260 treated patients were lost to follow-up. The loss to follow-up undermines the validity of the study but not to the extent that we must reject the article. 87 of the treated patients and 106 of the control patients died, yielding a RRR of 19% (95% CI, -1% to 36%). Thus, although our best estimate of the treatment effect is an RRR of about 20%, the results are compatible with no treatment effect at all. Mindful of the relatively wide CI and the loss to follow-up, the cautious clinician may choose to wait for another trial before using amiodarone for patients with heart failure. A more aggressive practitioner could argue that the results provide more support for treatment than we have for many of our widely used interventions and that administration of amiodarone is therefore appropriate. We take this latter stance to show the treatment implications of the study in more detail.
Although the mortality in the control group was approximately 40% over a mean follow-up of 13 months, the mortality was 25% in the New York Heart Association class II control patients, 40% in class III patients, and 55% in class IV patients. How do these results correspond to those of other recent large trials of heart failure? We kept track of these studies as they were published, but had we not, we would use search strategies described in a recent ACP Journal Club editorial to find the relevant articles (5). In 1 large study of class IV patients, the mortality was almost 50% after more than 1 year of follow-up (6), a finding consistent with the results of the amiodarone trial. 4 recent large studies that included approximately half class II patients and half class III patients showed a mortality of approximately 10% to 15% per year in patients treated with angiotensin-converting enzyme (ACE) inhibitors (7-10). Because the mortality in these trials is appreciably lower than that in the amiodarone study, clinicians must decide which results are more likely to apply to their patients. The amiodarone study was done in Argentina and included approximately 500 patients. The other studies that argued for the use of the latter data came from North America and Europe and included thousands of patients combined. In addition, data from the Framingham study, in which patients who survived for at least 90 days after the diagnosis of heart failure were followed outside of the randomized trial setting, showed a 1-year mortality of approximately 20% (11).
Using the gradient of risk between class II and class III of approximately 1.5 from the amiodarone study, reasonable estimates of the baseline risk for death at 1 year in class II, class III, and class IV patients are 10%, 15%, and 50%, respectively. Given the best estimate of a 20% RRR for amiodarone from the randomized trial, the numbers of patients we must treat to prevent a death are 50, 33, and 10 in class II, class III, and class IV, respectively. These figures suggest that we should aggressively treat patients in all 3 groups, but particularly those inclass IV.
Clinicians may find the prospect of examining multiple studies to determine baseline risk intimidating, if not completely unrealistic. We empathize, but the alternative of assuming similar baseline risk in patients with differing age and disease severity is likely to bemisleading. In the future, therefore, we will askour commentators to help with the work for our readers. For appropriate studies, we will ask the commentators to provide estimates of risk in terms of low-, medium-, and high-risk patients, to explain how to identify these differing risk groups, and to determine what is the associated NNT for each of them. We hope that this innovation will make the results reported in ACP Journal Club even more easily applicable so that patient care is optimized.
Gordon H. Guyatt, MD, MSc
Roman Z. Jaeschke, MD, MSc
Deborah J. Cook, MD, MSc
1. Guyatt GH, Cook DJ, Jaeschke RZ.How should clinicians use the results of randomized trials? ACP J Club. 1995 Jan-Feb;122:A10-1.
5. McKibbon KA, Walker-Dilks CJ.Beyond ACP Journal Club: how to harness MEDLINE for therapy problems. ACP J Club. 1994 Jul-Aug;121:A10-2.
6. The CONSENSUS Trial Study Group. Effects of enalapril on mortality in severe congestive heart failure. Results of the Cooperative North Scandinavian Enalapril Survival Study (CONSENSUS). N Engl J Med. 1987;316:1429-35.