The Weak Correlation Between Antibiotic Action and Mortality

At the recent FDA workshop on narrow-spectrum antibiotic development[1], the concept of using bacteriological response (BR) as an efficacy endpoint was flat-out rejected. The reason given is the purported lack of correlation between BR and survival/mortality which FDA considers the most important clinical endpoint[2]. FDA felt so strongly about this, that BR was not even considered a ‘surrogate endpoint’.

We know for a fact and can all agree that bacteria cause disease, and that antibiotics kill bacteria or prevent their growth, first and foremost, and by so doing improve outcomes. Survival is the expected outcome for all but the most severe infections. Certainly, there are many bacterial infections that our immune system can take care of all by itself, but it is the immunosuppressed patient who has a hard time fending off even trivial bacterial infections without antibiotics. If antibiotics fail to protect such patients, the outcome is uniformly dismal.

Immunity, antibiotics, surgery, amputation, fire are the lines of defense, in that order…

In the immune-compromised patient, bacterial persistence and mortality are tightly linked but in the normal host, only BR and CR are most directly correlated to antibiotic action. Our antibiotic trials are conducted in the immunologically intact host in whom with a high likelihood survival is expected. Study entry criteria make sure that high-risk patients are excluded; those who die during the 30-day observation period may succumb because of the infection but a myriad of other factors come into play: age, pre-existing conditions, trauma, surgery, complications, superinfections, you name it. Imagine an HAP/VAP patient in the modern day ICU: intensive care can support vital organ functions with mechanical ventilation, dialysis, pressors and artificial coma, making ‘30-day survival’ not a true measure of antibiotic efficacy; local medical practices and health care resources may have as much of an impact as the lowly antibiotic being tested.

Not very long ago regulators insisted that the efficacy of a new AIDS antiviral be measured by its effect on mortality or on the prevention of opportunistic infections. It took years for FDA to accept VL as a surrogate endpoint. The recent flurry of HCV trials would not have been possible without acceptance of VL as a sensitive and specific response measure. Please note that for both HCV (and HBV) trials VL was an accepted surrogate efficacy measure long before it was validated on the basis of data showing that VL and progression to liver cirrhosis, the hard clinical endpoint, are indeed correlated. Nonetheless, the analogy to HIV led to early acceptance of VL as a reasonable outcome measure.

For bacterial infections, we can test for culture-negativity at some infection sites but there are no straight-forward quantitative tests correlating bacterial disappearance rates and antibacterial effectiveness. We do know though for a fact that unchecked bacterial multiplication will not get better on their own. Hence, we really do not need a ‘bacterial load’ test and follow its course because cure is always predicated on bacterial eradication.

Even if we cannot distinguish antibiotic action from immunologic effect, why would we want to add more and unnecessary noise to the evaluation of antibiotic efficacy?

Antibiotics are not fever medicines, or pain medicines or anti-inflammatory medicines. The only thing they do is …kill bacteria. Secondary benefits are derived from this basic direct antibacterial effect. Yes, cough diminishes, fever comes down, WBC, interleukin and interferon levels normalize, and so do the PCT, ESR and CRP values. But the first necessary and sufficient step is bacterial eradication. Hence, let us not dismiss BR so fast; it is and should remain the primary read-out of antibiotic action. Mortality and other efficacy measures of lower sensitivity should take a back seat to BR.

Abbreviations:
HAP/VAP   hospital-acquired / ventilator-associated pneumonia
BR / CR     bacteriological / clinical response
VL     viral load
PCT  procalcitonin test
ESR  sed rate
CRP  C-reactive protein

References:
[1] FDA Briefing Document.  Developing Antibacterial Therapies Targeting a Single Bacterial Species.  Meeting of the Antimicrobial Drugs Advisory Committee (AMDAC). April 13, 2017
[2] E Durante-Mangoni. Colistin and Rifampicin Compared With Colistin Alone for the Treatment of Serious Infections Due to Extensively Drug-Resistant Acinetobacter baumannii: A Multicenter, Randomized Clinical Trial. CID 2013; 57: 349

Print Friendly

Leave a Reply

Your email address will not be published.