382. Methodological Deficiencies and Quality of Reporting in Published Clinical Trials of Antibacterial Drugs for the Treatment of Nosocomial Pneumonia
Session: Poster Session: Hospital-acquired and Transplant Infections
Friday, October 30, 2009: 12:00 AM
Room: Poster Hall A
Background: Equivalence and non-inferiority (NI) have been described in multiple clinical trial reports for various antibacterial drugs in treating nosocomial pneumonia (NP). We reviewed reports published over the past 25 years to assess the quality of methodology and reporting.
Methods: Using PubMed® and Web of Science®, we performed a computer-based search of the English-language scientific literature for all comparative clinical efficacy trials in NP published between 1983 and 2008. Retrieved articles were reviewed for trial design, pre-specified hypothesis, sample size calculation, scientific evidence supporting equivalence and NI margins, analysis populations, statistical methodology, and reporting results.
Results: 32 clinical trial reports were retrieved. All were prospective and randomized, but only 7 (21.9%) were blinded. Nineteen reports (59.4%) did not contain a pre-specified hypothesis; 8 (25%) were equivalence and 5 (15.6%) were NI trials. All elements for sample size calculations were provided in 10 (31.3%) reports. None of the NI and equivalence trial reports provided scientific justifications for the margins chosen. The primary analysis populations varied as 19 (59.4%) trials used the evaluable, intent-to-treat (ITT), or per protocol (PP) populations, whereas 13 (40.6%) used ITT and PP variants as co-primary populations. Tests for superiority were employed in 16 of the 19 trials lacking a pre-specified hypothesis, and equivalence was claimed erroneously in 12 of those trials for which non-significant p-values were reported. Results were displayed with p-values only in 12 (37.5%) reports, 95% confidence intervals in 16 (50%) reports, and neither in 4 (12.5%) reports.
Conclusion: Various design, statistical methodology, and reporting deficiencies were observed that could increase the probability of erroneously concluding equivalence or NI. The deficiencies included lack of pre-specified hypotheses, lack of factors for sample size calculations, errors in interpretation of failed superiority trials, and lack of scientific justification for equivalence and NI margins.
Scott Komo, DrPH1, Alfred Sorbello, DO, MPH1, Thamban Valappil, PhD1 and  A. Sorbello, None..
T. Valappil, None..
S. Komo, None., (1)US FDA, Silver Spring, MD