Welcome to this feature from iForumRx from iForumRx - the evidence that informs ambulatory care pharmacy practice. Hi I'm Christine Pruschak at the University of Maryland School of Pharmacy. Thank you for your interest in this presentation. Today I will be discussing interpretation of non-inferiority trials By the end of this presentation, I hope you will be able to understand how to interpret non-inferiority trials and differentiate them from other study types. Before we get into the specifics of a non-inferiority trial, it is important to understand the various trial designs and how they compare to each other. There are three major trial designs used in research-- superiority, equivalence, and non-inferiority. Superiority is the historical design that most people think of when reviewing literature. Its goal is to demonstrate that the study treatment is superior to control. This is the only way to prove that an intervention has benefits over another treatment or placebo. Contrary to what some people may think, a trial that fails to show superiority does not indicate that the two treatments are actually equal. For that reason, two other study designs are used. Investigators use an equivalence design to demonstrate that the compared treatments are actually equal. And finally, non-inferiority trials seek to demonstrate that the treatment is not inferior to or worse than the control. Two things to consider when reviewing the various designs are the null and alternate hypotheses. In a superiority design, the null hypothesis is that treatment x is no more effective than treatment y for a given condition. By disproving the null hypothesis in this trial, one can state that the new treatment is better than the control. In an equivalent study where the goal is to demonstrate equal treatments, the null hypothesis is that treatment x is either worse or better than treatment y for a given condition by greater than delta. Delta represents the equivalence margin or the amount of difference that is acceptable by the investigators to state that the two treatments are the same. Similar to equivalency studies, non-inferiority trials incorporate delta into its hypotheses. However, since the goal is to demonstrate that the treatment is not inferior or worse than the control, the hypotheses vary slightly. The null hypothesis is that treatment x is worse than treatment y by greater than delta, while the alternate is that treatment x is not worse than treatment y by greater than delta. Delta is a non-inferiority margin. Therefore, if one rejects the null hypothesis, they are able to state that the new treatment is not any worse than the control group. Now that we know the differences between the trials, it is important to answer the question of why non-inferiority trials are preferred in certain situations. The first answer is cost. Non-inferiority trials often require far fewer patients than its counterpart of superiority trials. With this, there are fewer resources needed for completion, which reduces cost. This is also directly related to the second advantage of convenience. The third reason non-inferiority trials are completed is if placebo is considered to be unethical. If there is a current treatment proven to be superior to placebo and has become the standard of care, it is unethical to create a trial comparing the new treatment to placebo. In addition, it may be too costly or inconvenient to do an active comparative superiority trial. Lastly, the most common reason for completing a non-inferiority trial is when the new treatment may be no worse than the current treatment, but has other advantages, such as improved safety, greater convenience, improved adherence, or lower cost. Next, we will review the considerations one has to make when reviewing non-inferiority trials. First, the reader must consider how investigators chose delta. In 2009, a review of published non-inferiority trials showed that few truly explain the development of delta. Delta is often based on previous trials, with the active control compared to placebo. It is usually derived from a lower bound of the confidence interval or a specified fraction of the active control effect. As a reader, once you compare the trial design to those trials that were able to distinguish efficacy of the active control, the design should be compared closely in terms of inclusion criteria, methods of diagnosis, and concomitant treatments. If they are not similar, the delta may not be a representation of what is considered to be non-inferior. The next consideration for the reader is the active comparator dosing. If this dosing is suboptimal, there may not be a fair comparison of the two treatments. For example, if the control group dosing is too low, it may be an ineffective dose. This may bias the results towards showing non-inferiority. The second design feature to be cautious about is blinding. While it is considered to be the gold standard in superiority trials, blinding will not always remove bias in a non-inferiority trial. In a non-inferiority trial, the goal is to prove that the new treatment is no worse. Therefore, the investigator may lean towards all results being similar. Of note, this is most problematic with subjective parameters, such as the severity of depression or cognitive impairment, as they are influenced by investigators' beliefs and interpretation. One way to reduce this kind of bias is to blind the investigator to the trial design. The last trial component to consider when reviewing non-inferiority trials is analysis population. Intention to treat analyzes all randomized patients, whether they received treatment or not. This is considered a conservative approach in superiority trials. In non-inferiority trials, it may sway the results either way, depending on compliance in each group. If there is poor compliance in the control group, the results may favor non-inferiority, or the opposite is true for poor compliance in the treatment group. In a per protocol analysis, results are only based on those patients that follow the treatment protocol without major protocol violations. In this analysis, there is a higher likelihood of finding a treatment difference in either direction. Therefore, the recommendation for non-inferiority trials is to review the results from both the intention to treat and per protocol populations. If they both show non-inferiority, the reader can safely interpret the results as truly being non-inferior. Let's take a look at an example together. A fictional study that compares the efficacy of Terpaban, a fictional target specific anti-coagulant, with the fictional standard of care Defaultarin which has been used to prevent thrombotic events for many years. In this fictional non-inferiority trial, Terpaban 1500 milligrams daily was compared to Defaultarin to prevent stroke in patients with atrial fibrillation. Due to its very significant variability in into patient response over time, the does of Defaultarin factor in must be adjusted based on lab monitoring using standard Defaultarin concentrations, or SDC. The primary efficacy and point was time to first stroke or ischemic event. In order to meet non-inferiority, the upper boundary of the one-sided, 97.5% confidence interval for the hazard ratio of the primary endpoint must not exceed 1.38, based on data from research with other anti-coagulants compared to Defaultarin in a similar population. Results were reviewed as per protocol, as well as intention to treat. As you can see from this design, non-inferiority is a preferred design, as Terpaban offers a significant advantage over Defaultarin, since it does not require routine blood monitoring. We should also consider three of the major components when reviewing a non-inferiority trial. Delta, controlled treatment, and population. Delta was determined based on trials in a similar patient population, so it seems appropriate. The dosage of medication in the control group, Defaultarin, was based on a target SDC. If the results show that patients spent less than 50% of time in the therapeutic range, it should raise a red flag to the reader. However, if the stated time and therapeutic range is consistent with well-managed Defaultarin therapy, the comparison is likely a valid one. And finally, the study results were analyzed using both intention to treat and per protocol. Results show that in a per protocol population, Defaultarin has a 1.5% per year rate of stroke or thrombosis, while Terpaban shows a 1.18% per year rate. The hazard ratio was 0.79 with a confidence interval of 0.52 to 1.29. In the graph below and non-inferiority margin is noted in red at 1.38. A hazard ratio to the left one favors Terpaban, while Defaultarin is favored if the hazard ratio is to the right of one. However, when reviewing the results, it is important to note the confidence interval and where it lies in relation to the non-inferiority margin or delta. If the confidence interval lies to the left of the margin, Terpaban would be considered to be non-inferior. When reviewing the intention to treat population, you see very similar results. Again, the confidence interval is to the left of the non-inferiority margin, showing that Terpaban is not inferior to Defaultarin in preventing stroke and thrombosis. Now that we have reviewed the results of the trial, let's review alternate results. If the confidence interval was completely to the left of one, the results favor terpaban and indicate that Terpaban is superior to Defaultarin. If the confidence interval is completely to the right of the non-inferiority margin, Defaultarin would be considered much better in terms of interpretation and Terpaban is inferior to Defaultarin. If the confidence interval is between the equivalency margin, Terpaban is considered to be equivalent to Defaultarin. And finally, if the confidence interval crosses the delta in any way, results cannot be interpreted, and the trial is considered to be inconclusive. In conclusion, non-inferiority trials show that a treatment is no worse than the control. This type of trial design is growing in use due to its many advantages. As a reader, it is important to consider design specifics and population when interpreting results. In addition, all confidence intervals must be compared to delta. [MUSIC PLAYING] Thank you for your time and interest in this presentation. iForum RX is a free, interactive, web-based design to inform ambulatory care pharmacy specialists, pharmacy residents, and student pharmacists about high-quality, practice-changing evidence.