Reanalysis of data from ten years of koala chlamydia vaccine trials reveals inherent bias and a lack of protective efficacy

Koalas within Australia are facing multiple threats from habitat loss, climate change and pervasive chlamydial disease. A safe and effective vaccine to protect koalas against chlamydia would be a welcome veterinary intervention to prevent this devastating disease which causes blindness, renal and reproductive pathology, infertility and mortality. However, over a decade of vaccine development and research trials has produced conflicting results, which vary from no observable indication of protective efficacy to a 64% reduction in chlamydia associated mortality. Given this discrepancy we conducted an independent analysis of data underpinning the highly cited publication demonstrating vaccine efficacy that was the basis for a permit for limited use in koalas. In doing so, we identified a substantial error in the way the data was encoded for analysis, which resulted in survival from birth to time of vaccination incorrectly being attributed to vaccination. After correction for this error, we show using the exact methods of the original authors that there was no significant difference in the survival or disease rate between vaccinated and unvaccinated koalas. Furthermore, we identified a substantial source of bias whereby animals with disease at their first health assessment were not excluded from the study and as those animals were generally not vaccinated, they were almost six times more likely to be included in the control group. Reanalysis including only koalas that were healthy on initial assessment and for which follow-up health assessments were performed, showed that vaccinated individuals were equally likely to contract disease or to die with disease; 18 of 143 vaccinated koalas (12.6%) developed disease and 3 (2.1%) died, compared to 63 of 439 control animals (14.4%) that developed disease and 11 (2.5%) that died. This study highlights the importance of randomised, observer blinded, placebo-controlled studies in veterinary settings to uphold scientific rigor and prevent stratification bias.