When one thinks of generating superiority data for a new agent, there are several possible meanings of the idea of a “superiority trial”:
1. Head-to-head: A head-to-head prospective comparison of the new drug vs. an existing drug.
2. Sub-analysis: A superiority sub-analysis (or analyses) embedded in a non-inferiority trial.
3. External (Historical) control: An indirect comparison of a new drug with data from patients treated (inadvertently) with ineffective therapy for resistant pathogens
The first meaning (head-to-head) is the one that seems most commonly assumed but is actually unrealistic except under rare circumstances. In such a design, one group of patients would be randomized to some sort of standard of care whereas the second group would receive the new drug (perhaps alone or in a combination). In the case of antibiotics, though, this is exceedingly tricky. The standard of care patients are assumed to be getting therapy that might actually work. Indeed, the trial can’t be designed to deliberately seek superiority in a way that puts patients at risk: it is a requirement of the trial that a patient can NOT be enrolled if thought infected with a resistant organism!
Thus, the design of head-to-head superiority trials makes a demonstration of superiority unlikely, a point noted clearly by the regulators in their publications. The only possible loophole is that if the only available regimen is highly toxic (e.g., colistin-based), then a demonstration of superiority might be possible. However, the success of even one drug here would then create a new tool for future trials that would make it impossible to study the next new drug via a head-to-head superiority trial.
So - what are the alternative superiority designs for antibiotics? There are three possibilities in my view.
1. Embed a superiority trial within the context of a standard non-inferiority trial. In this design - you have to show that, at the very least, your new drug is not inferior to some gold standard treatment. If you achieve superiority in a non-inferiority trial - that is your new drug is not only not inferior - but is actually superior to the gold standard - you can make that claim and get approved on that basis. But if you are non-inferior in the large population, you can carry out a pre-designed analysis on a subpopulation (patients where the activity of the gold standard drug might be equivocal for example) and try and show superiority there. If you can't - you can still be approved based on your non-inferiority to a gold standard antibiotic. Here you have hedged your bets and you are unlikely to lose your drug. If you cannot show non-inferiority - then maybe you deserve to lose it.
2. Historically or externally controlled superiority trials. These are my favorite designs for superiority trials for antibiotics. But they are not without controversy and clear disadvantages. The big gorilla here is the word "historical."
(a) The first such design would use a historical control based on pharmacometric analysis of previous but contemporary trial data. The best example of this is Paul Ambrose's analysis of the tigecycline trial in ventilator associated pneumonia. In that analysis, his group clearly shows that patients who do not achieve therapeutic drug levels are more likely to fail. The results in inadequately treated patients show what happens when you treat resistant infections with antibiotics that are unlikely to work because you cannot achieve high enough levels of drug to kill the infecting bacteria. Obviously, this is the case if the organisms are resistant - that is if the amount of antibiotic required to kill them is beyond what can be achieved in the body. If you can show that your drug is superior in effect to this inadequate therapy - you should be able to achieve approval. The big problem is that the controls are still historical and therefore the patient populations under study might be different.
(b) A second approach would be to carry out a prospective observational study of patients with highly resistant infections being treated with so-called standard of care (like colistin or tigecycline or others). This treatment would have some result in these patients. You then carry out a trial where patients receive your new drug, either in combination with "standard of care" or with some gold standard antibiotic. Only those with infections resistant to the accompanying therapy would be evaluable to see if your new drug was better than standard of care. You compare to the patients in whom you carried out your prospective observational study. In this case, if the inclusion and exclusion criteria in the observational study and your later trial are very similar or the same, you can look for a superior result in your treatment group.
In this discussion, I have ignored the issue of endpoints. What constitutes a response to therapy in these trials? That is too much for this blog - but I have one word for you - pharmacometrics!
DON'T FORGET THE JOHN QUINN MEMORIAL FUND!