Design of clinical trials of antibacterial agents for community-acquired bacterial pneumonia

Abstract

Standards for the conduct of clinical trials of antibacterial agents for community-acquired bacterial pneumonia (CABP) have changed dramatically in recent years. A draft guidance from the US FDA on the conduct of such trials was issued in March 2009. However, the guidance has already faced substantial criticism during the open public comment period, resulting in uncertainty regarding the appropriate design of such studies from a regulatory perspective. Controversies regarding the magnitude of the treatment effect associated with antibacterial therapy versus placebo/no therapy, the appropriate timing, nature and noninferiority margin for the primary efficacy end point, and other clinical and statistical issues have complicated efforts to reach consensus on appropriate trial design of antibacterial therapy for CABP. It is critical that studies of new drugs for CABP are designed to ensure that they are feasible to conduct and that their results are scientifically valid, statistically rigorous and clinically meaningful. Based on 3 years of active dialog between clinical, statistical, and regulatory experts, this article proposes an approach to enable a balance of clinical trial feasibility with appropriate scientific, statistical and clinical rigor.

Figure 1. Improvement in clinical response in patients with community-acquired bacterial pneumonia treated with sulfonamide antibacterial agents versus standard background medical therapy without antibacterial agents

(A) Percent of clinically responding patients by day post-presentation to the hospital in three cohort studies, including one from 1937 (no antibacterial agents) [23] and two from 1939 (sulfapyridine #1 and 2) [28,29]. The pre-antibiotic study did not include dead patients in the analysis; by contrast both sulfonamide cohorts incorporated dead patients into the analysis. (B) Point estimates (open circle) and 95% CI (error bars) of the difference in clinical response rates between patients treated with sulfonamide (average of both studies through day 4, and then sulfapyridine #1 alone from day 5–7) versus no antibacterial therapy.

Figure 2. Determination of non-inferiority margins

(A) Difference in efficacy between a comparator drug, an experimental drug, and placebo. The M1 margin is the lower bound of the 95% CI of the difference in efficacy between the comparator drug versus placebo based on previous studies. The M2 margin is the pre-planned noninferiority margin for the current study, and is often set to preserve 50% of the effect of the comparator drug (i.e., it is set to be 50% of the magnitude of M1). (B) The power curve for a clinical trial with a noninferiority margin of -15%, an α of 0.05, and an expected 80% success rate in the comparator arm. The power curve shows the probability of demonstrating noninferiority as a function of the true difference in efficacy between the treatment arms. In the example shown, the trial has a power of 95%, or a probability of 0.95 for demonstrating the noninferiority of the experimental drug if both the experimental and comparator drugs yield a success rate of 80% (a difference of 0%). The probability of a positive trial (i.e., establishing noninferiority of the experimental to the comparator drug) is graphed on the vertical axis against the true difference in success rate between the experimental and comparator drugs (i.e., experimental drug success rate - comparator drug success rate) on the horizontal axis. For example, if the experimental drug is associated with a true absolute difference of -5, -10 or -15% relative to the comparator drug, the trial has a 67, 26 or 5% chance of achieving a positive outcome (i.e., establishing noninferiority), respectively.