I want to discuss the idea of superiority trials for antibiotics and some of the issues that many experienced researchers fail to consider when thinking about this topic. I am grateful to George Drusano for his input here.
Before I get into this topic, I want to say that I know I haven’t written a blog since April. That is because I feel that if we do not have an important pull incentive such as is being contemplated in the PASTEUR Act in front of the US Congress – all else is futile including questions of antibiotic trial design . . .
There are many reasons why antibiotics don’t work and, most of the time, it is not related to resistance. Consideration of this question will be important in the design of clinical trials for antibacterials as Contrafect just learned the hard way.
Yes, resistance to antibiotic treatment is an important concern, but it is hard to design trials around this issue for several reasons. Resistance to gold standard comparator antibiotics remains uncommon. Even when this occurs, in a double blinded randomized trial, we usually exclude those patients whose infection is resistant to a comparator antibiotic such that they can be treated with something more likely to be effective. There are some occasional exceptions to this, but this problem makes trial design challenging.
Some experts believe, based on some data, that so called bactericidal antibiotics like beta-lactams are generally superior to bacteriostatic antibiotics like tetracyclines. Others believe that this is mainly an issue of choosing an appropriate dose for the indication being studied. Still others believe that this is only true for certain situations like nosocomial pneumonia or serious infections occurring in severely immunocompromised patients. But proving this in the context of a clinical trial remains very challenging.
Of course, other obvious problems exist. We try and exclude patients with non-bacterial infections from study when possible since antibiotics will not be effective – but we do not always succeed. Including these patients tends to drive the result to the “nul” or -no difference. This would be a big problem in a superiority trial. We also try and exclude patients who are not likely to respond even to effective antibiotic treatment such as those unlikely to survive long enough to provide sufficient time on study and those who are severely immunocompromised where antibiotic treatment may be less effective. But this still leaves many patients with susceptible infections who will respond poorly to antibiotic treatment. Why and how?
Let’s look at one example - serious infections caused by Staphylococcus aureus. Disseminated infections withS.aureus tend to involve multiple foci of infection including abscesses of various sizes. Because there may be multiple such foci, drainage is not possible. Antibiotics may not penetrate well into these areas and may not work well in the local environment of the abscess where the pH may be low. Left-sided staphylococcal endocarditis still carries a 30% mortality rate in spite of appropriate therapy. I personally think this is more likely caused by mechanical problems and multiple abscesses rather than any issue with the lack of killing activity of the antibiotic.
Sometimes, patients slip into a septic state where a systemic inflammatory response is triggered. While this might not be immediately fatal, even with antibiotic therapy, the inflammatory response could progress and might ultimately lead to organ failure and death. Again, protected sites of infection might be the problem here. Yet these patients will be entered into clinical trials and the antibiotic itself will be destined to fail without other interventions such as surgery.
Animal models in the antibiotic world are incredibly helpful in predicting efficacy and efficacious dosing. In vitro pharmacodynamic models using the hollow fiber system can also be useful and are frequently more “conservative” than traditional animal models since you can study more long term effects including that of slowly emerging resistance. But these models cannot be used, in general, to predict superiority as it is studied in the context of a clinical trial.
What I am trying to argue here is that superiority trials for antibiotics will be confounded by infections where the antibiotic is not the issue. Most antibiotics will work well in the absence of such intractable problems. Therefore, even in the absence of resistance, it will be challenging to show superiority clinically even with an antibiotic that might look superior in vitro or in various model sysems.