Multiple Treatment Meta-Analysis of Products Evaluated for Control of Fire Blight in the Eastern United States

Abstract

The aim of this analysis was to estimate the effect sizes and consistency of products evaluated for fire blight control in the eastern United States over the last decade. Because only 3% of the 69 studies published from 2000 to 2008 explicitly presented a measure of within-study variability, a method for estimating the least significant difference (LSD) and, hence the sampling variance, for studies with at least two significant mean separations in the presented mean multiple comparisons was developed. Lin's concordance analysis indicated that the estimated LSD was an accurate predictor of the actual LSD based on 35 studies in a calibration evaluation (ρ(c) = 0.997). Separate multi-treatment random-effects meta-analyses were performed for three control categories: antibiotics, biological control, and plant defense-activating products and mean log response ratios relative to the nontreated controls ([Formula: see text]) were computed for each treatment and then back-transformed to obtain the mean percent disease control. None of the products evaluated performed as well as streptomycin, the standard product for fire blight control, for which the mean disease control was 68.6%. As a group, experimental antibiotics provided the best fire blight control with mean effect sizes ranging from 59.7 to 61.7%. Among the biological controls, the best control was noted for treatments combining the antibiotic streptomycin with a product based on Pantoea agglomerans (55.0% mean disease reduction) or Bacillus subtilis (53.9%). Mean disease control was 31.9, 25.7, and 22.6%, respectively, for products based on B. subtilis, Pantoea agglomerans, and Pseudomonas fluorescens without an antibiotic, suggesting that the higher efficacy of the combination treatments was due to the antibiotic. Among the plant defense-activating products, prohexadione calcium had the highest and most consistent effect size (50.7% control), while other products provided modest mean disease control of between 6.1 and 25.8%. Percent control values were significantly moderated by study location and cultivar used in the study, and were smaller, but more variable, when products were tested under high disease intensity compared with low disease intensity. Results indicate that wide-scale use of biological control and plant defense-activating products in the eastern United States is likely to remain low.