Wednesday, 29 April 2015

EQUIVALENCE vs NON-INFERIORITY vs SUPERIORITY TRIALS

In clinical research studies, what do the terms EQUIVALENCE, NON-INFERIORITY, and SUPERIORITY trials refer to, and why even have different types of clinical trials?


Medical research studies or clinical trials (in relation to cancer) are initiated for various reasons. Often what governs how a clinical trial will be conducted is dependent on what they aim to find out, the type of intervention, and the current treatment standard as well as state of knowledge about a particular disease.

Most commonly, clinical trials (in relation to cancer) may want to establish or look at:
  1. Causes of cancer by investigating genetics, lifestyle, work environment and diet (or combination)
  2. Prevention of cancer by studying specific lifestyle, diet, and / or drug factors to reduce risk
  3. Diagnostics to examine new laboratory tests or scanning equipment that can diagnose cancer earlier, more precisely, or in a easier and more patient friendly way
  4. Therapies to investigate new drugs / types of treatment, combinations, or methods of administering a drug
Clinical trials that investigate new therapies (see point 4 above), generally aim to find out if a new treatment or procedure is safe, and whether or not it has side effects and if these outweigh the health benefits. In addition, clinical studies aim to find out if a new therapy is better than the current standard of treatment, or if it improves the quality of a patient's life.

Perhaps the most burning question for people involved in drug development (as well as patients), when they are entering clinical trials for a new drug / treatment:

Is the new therapy "better" than the established standard treatment?

Diagram explaining the difference between Non-inferiority trial, equivalence trial and superiority trial

So, how do you go about answering the above question?


Up until the recent past, most clinical trials compared new drugs to placebo (which is called a superiority trial). However, once a proven effective treatment for a disease exists, the clinically important question is whether or not a new treatment is "better" than the old one. Also, it would become unethical to conduct placebo controlled trials in life threatening diseases such as cancer for which therapies already exist. 

Ideally, new treatments should be progressively better than old treatments. Unfortunately, it becomes increasingly difficult to demonstrate statistically significant superiority of new treatments when it almost takes a "quantum leap" to achieve clinical benefit above and beyond what standard therapies achieve (exceptions do exist, e.g. Antibody Drug Conjugates (ADC)). As such, drugs are often more precisely defined and targeted to a particular sub-set of a given disease (based on genetics or a tumor's molecular composition - i.e. rationally designed drugs). Hence personalised medicine...

Nevertheless, new treatments may still be desirable, even when they do not have a superior treatment effect. For example if a new therapy is safer, is cheaper, has fewer side effects, or is more user friendly or easier to store (i.e. without the need for refrigeration in hot climates), it may be useful to licence a drug despite it not being "better" than the original. 
Furthermore, it is considered prudent to have alternative options for patients who become intolerant to standard treatments. 
This school of thought has led to several drugs being granted a licence following clinical trials that demonstrate they are "as good as" or "not significantly worse than" the established standard therapy (i.e. without a superiority trial ever taking place - and as such never having been compared to placebo). These 'active control' trials include Equivalence and Non-inferiority trials.

In order to demonstrate that a new treatment is 'equivalent' to the active control, an infinite sample size would normally be required (for exact equivalence to be determined). Therefore, during the trial design phase a margin is calculated / selected. Then, once the data from the trial has been analysed and a Confidence Interval (CI) is calculated for the difference between two test statistics and it falls within the negative and positive margin the two treatments are deemed equivalent. These type of trials are often conducted to demonstrate that a generic version of a drug is functionally / biologically comparable to the original treatment (a recent example is the Phase III GATE
study sponsored by Synthon in which they demonstrate that the efficacy and safety profile of Synthon’s Glatiramer Acetate (GTR) is equivalent to Copaxone® of Teva Pharmaceutical Industries Ltd). 

Thus randomised interventional type clinical trials (phase II, and phase III) can essentially be categorised into three types of "comparisons":
  1. Superiority
  2. Non-inferiority
  3. Equivalence
While superiority trials are designed to demonstrate that one treatment is better than another, a non-inferiority trial is designed to establish whether or not a new treatment is ‘not unacceptably worse’ than the current medically accepted standard (within a given geographic region).

Pubmed search result for the term non-inferiority trial and superiority trial split by year

Selecting appropriate margins


Since the introduction of non-inferiority trials in the mid-1990s it has been debated whether such trials should be performed. Given that the design of a non-inferiority trial is a complex process and rests upon assumptions which can be difficult to verify.
The challenges faced throughout the design of a non-inferiority trial compared with superiority trials include the selection of an appropriate and clinically relevant margin, the primary population for analysis, and the choice of comparator treatment (e.g. are comparisons of a new treatment with a medically accepted treatment used in another geographic region clinically relevant and / or appropriate?).

Similarly, the use of inaccurate or unreasonably generous active control event rates or non-inferiority margins, can skew the result in favor of the new treatment. 
In addition, operational complications (insufficient data to support a selected margin or a calculated margin that leads to an impractical sample size) may hinder the process of settling on an appropriate margin.
In addition, the selection of appropriate non-inferiority margins for the appraisal of non-inferiority trials, is also paramount to safeguard against a phenomenon called "bio-creep" (bio-creep occurs when a new therapy is slightly worse than the preceding drugs, yet is granted non-inferiority and becomes the control arm in future trials. This can ultimately lead to an active therapy being no better than a placebo).  
However, bio-creep / techno-creep (in case of medical devices) can be avoided by selecting the most effective treatment in class as the control for non-inferiority trials, even in circumstances where this is not the treatment most commonly used.

Finally, as has been noted before, the difference established by statistical means can potentially include values which are not necessarily acceptable or useful in the clinic, so this is an issue that most definitely warrants close attention during the design stages of a non-inferiority trial.

Despite these reservations, non-inferiority trials have been embraced by drug developers as can be seen from continued and increasing popularity (based on a pubmed search performed in April 2015).

Diagram is an easy graphical representation of non-inferiority margins and the 6 possible outcomes or result of a non-inferiority clinical trial

What drives the conceptual design of a Non-inferiority clinical trial and how is it guided by scientific principles?


Unlike superiority trials that aim for statistically significant Hazard Ratios of less than 1, Non-inferiority trials "aim" for Hazard Ratios of more than 1. As such, these numbers together with the confidence intervals, epitomize the scientific limits of what constitutes "not significantly worse than" and how we quantify the degree of non-inferiority. Keeping in mind that the upper limit of the Confidence Interval is not of primary concern in non-inferiority trials (in contrast to equivalence trials where both the lower as well as the upper bound CI are required to fall within the margins as shown in figure 1).

Are there any concerns about Non-inferiority trials?


Given the complexity of non-inferiority trials, and the fact that it is designed to determine whether or not the new treatment is "not inferior to an unacceptable extent", a poorly designed or conducted clinical trial could potentially incorrectly demonstrate non-inferiority (which is why some advocate the use of the Per Protocol (PP) population for analysis). However, rather than focussing on Per Protocol populations (instead of Intention To Treat (ITT) populations), it would seem more appropriate to adequately design, conduct and monitor a trial. Regulators in the EU and US have already highlighted this particular issue and both expect ITT as well as PP analyses to be performed (with both results supporting non-inferiority).

Non-inferiority trials - the road to licencing clinically meaningful alternatives


All three clinical trial types (Equivalence, Non-inferiority, and Superiority trials) have a role to play when evaluating the efficacy and clinical relevance of new drugs (or medical devices). While the superiority trial type has enjoyed great popularity for decades, non-inferiority trial types are fast catching up since their inception in the mid nineties. Non-inferiority trial is a valid approach (when conducted and monitored appropriately) for establishing whether or not a new drug is a clinically meaningful alternative for a medically accepted standard of care. This is especially true in cases where a new treatment possesses an additional feature (safety, ease of use, less side effects, lower cost (e.g. generics)) for which physicians, patients and insurers are willing to sacrifice a small margin of drug activity compared to the already approved version of a drug. The "sacrifice" or margin however, will have to be incredibly small for a drug to be relevant in a clinical setting (especially in serious disease settings, such as cancer or life threatening infections).

No comments:

Post a Comment