What Does the Research Really Tell Us? (Part II)

Understanding research studies is one of the keys to delivering good patient care. So much of what we hear about health care—from stories in the media to treatment guidelines issued by health organizations— is based on research study findings. So we need to understand what the data are really saying. Our patients will appreciate our interpretation of these data as well. In Part I of this column, which ran in the May/June 2011 issue of AJNP and is available online at www.webNPonline.com, I discussed the meaning of clinical significance and the difference between relative risk reduction, a term often used but that can be deceptive, and absolute risk reduction. I review these concepts here and then look at a specific research study to show how these concepts are applied. I also discuss other aspects of research studies that nurse practitioners need to keep in mind.

A drug sales representative recently told me about a study showing that the use of one of his company's products in patients with diabetes was associated with a 48% reduction in stroke risk.¹ This result sounded impressive. Who wouldn’t want to reduce his or her chance of stroke by almost half? Well, the answer actually depends on a number of factors. When you hear such information about a product, you need to ask a few more questions: (1) What was the primary end point of the study? (2) How big is that 48% risk reduction? and (3) Does the study apply to my own patient population?

The first thing to note in any study is the primary end point. I’ve read many articles and heard many lectures referring to studies wherein the major arguments made by the writer or speaker are not related to the primary end point of the study. The primary end point is the main effect or outcome that the researchers set out to study, usually based on a certain intervention. If the primary effect or outcome does not occur at a level of statistical significance, then it has not been proven or established or confirmed. Some studies have secondary end points. However, if the primary outcome of the study is notmet, then the secondary outcomes, whether or not they are met, are not relevant. The study alluded to by the drug sales rep had two primary end points. The first primary end point was an acute coronary event (outlined specifically in the study) and the second was stroke. So, in this case, stroke was a primary outcome in this study.

How relevant was the 48% reduction in stroke associated with use of the study drug? To answer this question, you need to know the risk of stroke in patients who did not receive the study drug; that is, you need to put this reduction into context. The 48% reduction in stroke is a relative risk reduction (RRR). The RRR is computed by subtracting the event rate in the drug group from the event rate in the placebo group and then dividing the difference by the event rate in the placebo group.

Remember, RRR tells you nothing about the risk reduction for a given patient. An RRR of 48% tells you that, whatever a patient’s risk is without the drug, the use of the drug will cut this risk by nearly half. To ascertain how relevant this 48% reduction really is for your patient, you need to determine his or her baseline risk—that is, the risk to this patient if he or she didn’t take the drug in question. In a clinical study, the baseline risk is the pretreatment risk or, if there is a placebo group, the risk associated with the placebo treatment. In the study by Colhoun et al,¹ the baseline risk of having a stroke (ie, the risk in the placebo group of having a stroke) was 2.8%. The risk of having a stroke in the drug group was 1.5%. A decrease in the risk of stroke from 2.8% to 1.5% is an RRR of 48%. But, note this: The baseline risk of stroke in this study population is already very low—not quite 3 people out of 100.

In the study itself, 39 of 1410 placebo recipients had a stroke and 21 of 1428 drug recipients had a stroke. Using these numbers, the results can be viewed from another perspective, namely, that you needed to treat 1428 patients with the study drug during the time frame of the study to prevent 18 cases of stroke. Among the 1428 patients treated with this drug, 1410 would not have had a stroke anyway and therefore didn’t derive any stroke-protective benefit from the drug. Most study  participants, even those in the placebo group—in fact, 1371 of 1410—didn’t have a stroke. This doesn’t mean the data aren't true or that the drug is ineffective, just that an almost 50% reduction in risk doesn't mean much when the absolute risk is very small. The absolute risk reduction (ARR) of stroke is calculated by subtracting the rate of events in the drug group from the rate of events in the placebo group. In this study, the ARR is 2.8% – 1.5%, or 1.3%. The time frame of the study is another consideration. How long did patients need to be treated to see this reduction? In this study, median follow-up was about 4 years.

Before you consider prescribing medication to prevent stroke in a given patient, you need to determine his or her baseline stroke risk. To do this, you need to match this person to a similar group of patients who were enrolled in a clinical trial. If you already have a particular trial in mind—say, the study by Colhoun et al—then you need to consider whether the results may apply to any of your patients. Ask yourself: “What were the characteristics of the population studied?” Among participants in the study by Colhoun et al, the mean age was 61 years, 32% were women, 94% were Caucasian, and 100% had type 2 diabetes plus one additional cardiovascular risk factor (eg, hypertension, smoking, microalbuminuria, retinopathy). Participants had an average body mass index of 28 kg/m², a mean low-density lipoprotein cholesterol level of 120 mg/dL, a mean high-density lipoprotein cholesterol (HDL-C) level of 52 mg/dL, and a mean glycosylated hemoglobin (HbA1c) of 7.8%. So, if your practice consists mostly of obese Hispanic women with diabetes and low HDL-C values, then your population is not like this study population, and the results may not apply to your patients. However, if your patient population is similar to the study population in question, then your patients' baseline stroke risk is about 2.8%.

With this information in mind, I deal with my patients on a case-by-case basis. I estimate each patient's baseline risk for stroke (based on how he or she conforms to the typical patient in the study I'm looking at) and then consider the mean reduction in events from baseline with this drug—in this case, about 48%. If I have a slender 45-year-old patient with an HDL-C of 65 mg/dL who has a lower baseline risk of stroke than the study participants did—let's say that her baseline risk is 1.5%—then treating her with this drug would reduce her stroke risk to about 0.80%. By contrast, an elderly obese smoker with poorly controlled hypertension has a much higher baseline risk for stroke. A 48% reduction may bring his 10% stroke risk down to a nearly 5% risk over the next 4 years. You and your patients can now make treatment decisions based on their own risk for an actual event.

Reading research studies from beginning to end, as well as interpreting the data and applying them to the care of our patients, is an important responsibility of NPs. Just reading the abstracts or the conclusions is not enough. Most articles published in major medical journals don’t provide baseline risk or placebo group findings in the abstract.² So, be sure to read beyond the abstract. Become familiar with all the content in research articles and ask the questions for which you need answers: What was the baseline risk? What was the study population like? What was the primary end point? What was the study time frame? With practice, you’ll be able to share with patients how much benefit (or lack thereof) they can expect to receive from a specific drug treatment and then, together, you and the patient can decide whether the drug is appropriate or reasonable to try.