Value-Based Medicine

Foundations of Value-Based Medicine

In assessing the value of any intervention, several factors must be taken into account.

By Melissa M. Brown, MD, MN, MBA

My first column laid out the plan to define, discuss and elucidate Value-Based Medicine (VBM). In this second column we'll start with some definitions, maybe not scintillating, but critical just the same.

Value itself is viewed as improvement in length of life and/or quality of life. Just think about it: there is nothing else we as physicians do. It is the core responsibility of any physician to provide diagnoses and interventions that attempt to do one or the other, hopefully both.

VBM is simply the practice of medicine based broadly upon data that identify interventions which best maximize value. It supposes use of a defined, rigorous methodology to quantify value, comparison of such with other interventions (comparative effectiveness) intended to treat similar problems in the most cost-effective manner (cost effectiveness). This gives us the absolutely highest quality of care at the very best cost.

Building Blocks of VBM

The three foundation blocks of VBM are: (1) medical evidence of safety and efficacy of the interventions, (2) value to patients and (3) amalgamation of these data with their related costs. Many physicians are aware of the term evidence-based medicine and appreciate that we turn to our best evidence to determine efficacy of interventions. Evidence refers to the outcomes of various types of scientific studies; our confidence and, thus, the usefulness of these outcomes is related to the degree of scientific rigor and errors found in the studies. And most often, the data only support efficacy of the therapy and rarely or incompletely define and measure quality-of-life improvement. We should demand data on both.

There are five levels of evidence-based interventional evidence: (1) The highest level or “best” medical evidence is thought to be the randomized clinical trial (RCT) with low errors in measurement, followed by (2) the RCT with higher degrees of error, (3) the non-randomized controlled study, (4) the case series and finally, (5) the case report. Non-randomized studies may have potentially confounding (statistically confusing) variables that can allow misleading results. The case series and case report clearly give data of interest and are informative; however, generally less credence is given the information until it is validated.

To obtain the most complete information, ideally, an intervention would be tested on all patients or a population, revealing exact safety and efficacy parameters with certainty. This, of course, is not realistic. As a compromise, companies conduct RCTs that are designed to detect a statistically significant (relevant) difference (typically p < 0.05). At the end of a trial, one of the outcomes is possible: either the intervention is demonstrated to be efficacious or not. However, errors can be made.

Types of Study Errors

If the trial supports use of an intervention, drug or device, yet in reality the intervention doesn't work but still comes to market, there is a big problem. We are depending on an effect that is thought to occur but does not. This is a Type I error and the most egregious sort to make. We think we are treating an illness and we actually are not.

A Type II error is the risk of not detecting a clinically important difference between the treated and non-treated groups. This intervention is the one that never was. This drug study, for instance, would not show the useful drug that actually has a therapeutic effect and thus would not make it to market and become a soldier in our medical armamentarium.

Evaluating Study Errors

How do we use the concept of study errors in ophthalmology practice? If the study is positive, meaning it supports that a difference exists between the studied intervention and a placebo, we look for the Type I error size by evaluating the p-value. A p-value of 0.02 means that a Type I error, showing a positive effect when it truly does not exist, occurs two times out of 100. A p-value of <0.05 often signifies a valid study; however, clinicians are much more likely to have a higher bar to accept the new intervention, device or drug if the p-value is less (0.0001).

An RCT that doesn't demonstrate a difference between the study intervention and placebo is termed a negative trial. It does not, however, mean that there actually is no difference, just that the study did not identify one. Clinicians need to be confident that testing more people would not uncover a difference in treatment effect. The power of the study demonstrates the strength of the numbers tested and is considered robust if the associated Type II error is less than 0.2. By definition, the power of a study equals 1.0 (the Type 2 error), and in general a power of greater than 80% is required for a clinical trial to be a Level 1 RCT. In other words, the negative Level 1 RCT has greater than an 80% chance of detecting a positive result if it exists. The power of a study can be increased if more patients are included in the study.

We will continue to look at evidence, its validity, significance and importance before we then jump to turning evidence into value. OM

“Value-Based Medicine” is a registered trademark of the Center for Value-Based Medicine.

Melissa M. Brown, MD, MN, MBA, is president and CEO of the Center for Value-Based Medicine in Philadelphia. She can be reached via e-mail at mbrown@valuebasedmedicine.com.