A statistical analysis of injury outcome and biomechanical response was performed using data from 28 left side impact tests employing Heidelberg-type sleds and post-mortem human subjects, with the objective of advancing the development of thoracic injury criteria for lateral impact. Injuries were scored by the test centers according to AIS 90. Rib fractures accounted for the maximum AIS score in each case. Curvature data from chest band gauges were used for calculation of contours depicting the shape of the thorax at 1 ms time intervals following impact. Thoracic deformations were deduced from the contours. Risk factors studied included maxima of curvature, deflection, rib and spinal accelerations, the Thoracic Trauma Index (TTI), and Average Spinal Acceleration (ASA). Subject age at death was found to have a significant effect on injury outcome. Consequently, age was used as a confounder variable in logistic regressions for the prediction of dichotomous outcomes P(AIS≥3) and P(AIS≥4). Stepwise backward logistic regression indicated that subject age and maximum normalized curvature relative to initial curvature are the only surviving independent variables among all considered. Separate logistic regressions employing age and a single risk factor confirmed that age combined with maximum curvature difference yields the greatest statistical significance, and the highest-ranking goodness of fit. Results also showed that by employing the logarithm of curvature difference in the logit, goodness-of-fit can be improved, and the usual problem of a poor fit at low values of risk factor is eliminated. Thoracic deflection was found to be the second highest-ranking injury correlate for side impact, ranked above TTI and ASA in its ability to predict accurately and reliably the extent of side impact thoracic injury.