The host specificity of Leptinotarsa undecimlineata Stål, a candidate biocontrol agent for Solanum torvum Sw. (turkey berry) a woody pasture weed in Vanuatu was investigated. In no-choice tests, adult L. undecimlineata beetles fed and oviposited on eggplant (Solanum melongena L.) plants, which also supported development of first instar L. undecimlineata larvae through to adult. A systematic review of field host records indicated that L. undecimlineata has never been reported to attack S. melongena throughout most of its native range in Mexico, the Antilles and Colombia, although minor and infrequent spillover attack has been observed in Central America. Choice oviposition, ‘multiple generation’ (continuation) tests, and the relative performance risk score approach (updated to use Firth’s bias reduced logistic regressions) were used to investigate the potential risk posed to S. melongena should L. undecimlineata be released in Vanuatu. The relative risk score approach was inconclusive, indicating a low but not zero risk of S. melongena being a persistent field host for both choice and no-choice tests. In contrast, the multiple generation tests demonstrated that L. undecimlineata cannot persist on S. melongena. Adult feeding was lower and significantly fewer eggs were laid on S. melongena compared to S. torvum in no-choice adult feeding and oviposition tests. The combination of tests used provided an explanation why only minor and infrequent spillover attack on S. melongena has been reported in parts of the native range of L. undecimlineata. We conclude that L. undecimlineata poses a very low risk to S. melongena should it be introduced into Vanuatu. However, the relative performance risk score approach requires further analyses using larger datasets to investigate the potential for threshold scores to be used to reliably predict risk.