A New Paradigm for Clinical Trials in Antibiotherapy?

Recommending or actively prescribing antibiotics is a fundamental daily activity in the field of infectious diseases. Our primary objective is to offer the most effective agent to optimize patient outcomes. However, numerous issues, not limited to toxicity and tolerability, cost and risk for emergence of antimicrobial resistance, frequently influence our choice of therapeutic agents. It is widely accepted that experimental clinical trials provide the strongest evidence to guide the choice of a given therapy. Numerous large trials (1-3) have been performed in some areas of infectious diseases, such as in viral hepatitis, HIV infection, vaccines and tuberculosis, that can help guide infectious disease specialists. However, it is ironic that for bacterial infections, which are undoubtedly the most common infectious diseases encountered by generalists and infectious diseases specialists, there is a paucity of evidence from well-performed randomized controlled trials. We performed a brief exploratory review of the published literature to gain a sample of the contemporary body of evidence that may be used to help guide antibiotic prescribing practices. We searched PubMed using the terms “antibiotics” and “infections” and limited our review to human clinical trials published in the English language with links to full text articles. We arbitrarily chose to review 30 consecutive, recently published trials investigating systemic antibiotic therapies for bacterial infections (4-33). As shown in Table 1, many of the studies were relatively small, with a median sample size of 218 patients (interquartile range 109 to 434 patients); the trials did not identify a superior treatment strategy. Few studies focused on severe infections, and none included mortality as a primary outcome measure. Remarkably, eight of 30 trials investigated various different regimens for treating Helicobacter pylori infections (Table 1). These trials represent only a small and, potentially biased, sample of current evidence. While they may be reflective, they are by no means proof of depth of the literature base. Nonetheless, it is our anecdotal experience clinically and from interhospital rounds, meetings and committees, that most decisions and recommendations for infectious diseases related to antibacterial therapies are based on laboratory, clinical experience and, perhaps most frequently, arbitrary personal preference. Conduct of antibiotherapy trials has been challenged by a number of factors, the most important of which is the immense cost associated with the conduct of clinical trials. Public research funding bodies rarely support comparative antibiotic clinical trials, and there is understandably little interest from industry to fund the evaluation of older off-patent agents. Pharmaceutical industry-supported trials are typically designed to fulfill regulatory requirements for new agents. These studies, therefore, generally aim to demonstrate safety and noninferiority of the new agent compared with standard therapies. It is well recognized that once agents are approved and available for one indication, they are rapidly used by clinicians for a range of other infections in patients who were not included in the trials (34). However, frequently, we are left with ongoing questions regarding the optimal management of these untested indications. As clinicians, we recognize that many agents, even if suboptimal, successfully treat mild to moderate infections. We are grateful to have even a relatively small amount of clinical trial literature to support our options. However, what we often really want to know, and are asked to provide expertise in consultation, is whether one agent will improve a patient’s chance of survival over another. However, is it realistic for us to expect clinical trials to demonstrate superiority of one antibacterial agent over another? We can argue that for a serious life-threatening infection, a demonstrated absolute reduction (as low as 1%) in the risk of death would provide support for use of one agent over another. A small benefit, such as the one previously described, has been used as justification for the use of certain agents in other disciplines (35). However, if a 1% mortality difference is chosen to be clinically significant (α=0.05, two-tailed, β=0.1), then at control group mortality rates of 10%, 25% and 50%, approximately 40,000, 80,000 and 100,000 patients would need to be enrolled, respectively! Because of these challenges, we have often turned to observational studies to gain insight into optimal treatments. Increasingly large and complex databases have enabled the assessment of treatments and outcomes on a large scale (36,37). These observational designs are particularly useful when mortality is the outcome because this may be reliably determined using vital statistics data. However, at the risk of oversimplification, the Achilles’ heel of observational studies is that observed differences in treatment outcomes may be falsely attributed to an uneven distribution of confounding variables. Multivariable regression and other statistical methodologies (such as the use of instrumental variables) may permit adjustment for unmeasured confounders, and may be particularly attractive given the well-recognized regional variation in patterns of medical practice. These have been used to evaluate drivers of antimicrobial overuse (38) and the impact of antimicrobials on outcomes in chronic obstructive pulmonary dis