We hypothesized that observer bias could markedly effect the results of all pertussis vaccine efficacy trials including those with double-blinding. Specifically, a less efficacious vaccine which prevents typical disease, but not mild disease could be determined to be more efficacious than it is, if the study personnel have a preconceived opinion as to what clinical pertussis is. This could lead to an a priori dismissal of mild cases as being other respiratory illnesses, and therefore, cultures and acute-phase serum not obtained and proper follow up not carried out. We carried out a pertussis vaccine efficacy trial in Southern Germany from May 1991 until December 1994 in 227 physician offices. In a randomized, double-blind fashion, infants received 4 doses of either DTaP or DTP. Controls (unblinded) received DT. For this analysis we separated study investigators into 3 groups based upon their rate of investigation of possible cases:group 1 - high study rate (cough reporting rate [CR] ≥ 20% or evaluation rate [ER] ≥ 10%); group 2 - intermediate study rate (CR ≥ 10% or ER ≥ 5%) or group 3 - low study rate (CR < 10% or ER < 5%). Using ≥ 7 days of cough and laboratory confirmation the efficacy of DTaP vaccine was: group 1 40% (95% CI = -3-65), group 2 78% (95% CI = 65-86) and group 3 75% (95% CI = 63-87). With ≥ 21 days of cough and laboratory confirmation the DTaP vaccine efficacies were 69%, 86%, and 84% respectively. Our results indicate that observer bias can inflate efficacy; evidence of this bias is present in 2 other recent trials.

This study was supported by Wyeth-Lederle Vaccines and Pediatrics.