Advancing the developmental and reproductive assessment in medical device preclinical safety evaluations

Currently, a standardized developmental and reproductive toxicity (DART) test of chemicals present in and released from medical devices does not exist. The objective of this study is to investigate if selected non-commercial and commercial in silico models, individually or in combination, can be used as a first step in addressing DART of medical device chemicals. A total of 478 polymeric material chemicals (268 DART and 210 non-DART chemicals) commonly used in medical devices were obtained from the European Chemicals Agency Database. Parameters, including sensitivity, specificity, and accuracy, were calculated and compared. Median sensitivity, specificity, and accuracy of the individual non-commercial or commercial in silico models were 0.529, 0.601, and 0.535, respectively. Median sensitivity, specificity, and accuracy of all possible combinations of two or three in silico models were 0.659, 0.467, and 0.573, respectively. Combining Leadscope and Percepta models resulted in the sensitivity of 0.524, specificity of 0.845, and best performing accuracy of 0.766. Integration of different in silico models may act as a useful first step for screening medical device-related chemicals with potential DART toxicity in addition to the effort in replacing, reducing, and refining in vivo DART testing in medical device preclinical safety assessments.