Translation of nonclinical to clinical safety findings for 27 biotherapeutics.

Human adverse drug reactions (ADRs), and in vivo nonclinical adverse and nonadverse findings, were identified in 27 biotherapeutic programs and placed into organ categories to determine translation. The sensitivity of detecting human ADRs was 30.8% with a positive predictive value (PPV) of 53.3% for nonclinical adverse findings; sensitivity increased to 67.3% and PPV fell to 35.0% when including nonadverse findings. Nonclinical findings were associated with a greater likelihood of a human ADR in that organ category, especially for adverse findings [positive likelihood ratio (LR+) >10 (lower 95% confidence interval [CI] of >5)]. The specificity and negative predictive value (NPV) were very high (>85%). A lack of nonclinical findings in an organ category was associated with a lower likelihood of a human ADR in that organ category. About 40-50% of human ADRs and nonclinical adverse findings, and about 30% of nonclinical nonadverse findings, were attributed to pharmacology. Slightly more than half of the human ADRs with a translating nonclinical finding had findings in animals that could be considered very similar. Overall, 38% of nonclinical findings translated to a human ADR at the organ category level. When nonclinical findings did not translate to humans, the cause was usually higher exposures or longer dosing in animals. All programs with human ADRs attributed to immunogenicity also had nonclinical adverse or nonadverse findings related to immunogenicity. Overall, nonclinical adverse and nonadverse findings were useful in predicting human ADRs, especially at an organ category level, and the majority of human ADRs were predicted by nonclinical toxicity studies.