Mutation of p53 is the most common genetic change in human cancer, causing complex effects including not only loss of wild-type function but also gain of novel oncogenic functions (GOF). It is increasingly likely that p53-hotspot mutations may confer different types and magnitudes of GOF, but the evidences are mainly supported by cellular and transgenic animal models. Here we combine large-scale cancer genomic data to characterize the prognostic significance of different p53 mutations in human cancers. Unexpectedly, only mutations on the Arg248 and Arg282 positions displayed significant association with shorter patient survival, but such association was not evident for other hotspot GOF mutations. Gene set enrichment analysis on these mutations revealed higher activity of drug-metabolizing enzymes, including the CYP3A4 cytochrome P450. Ectopic expression of p53 mutant R282W in H1299 and SaOS2 cells significantly upregulated CYP3A4 mRNA and protein levels, and cancer cell lines bearing mortality-associated p53 mutations display higher CYP3A4 expression and resistance to several CYP3A4-metabolized chemotherapeutic drugs. Our results suggest that p53 mutations have unequal GOF activities in human cancers, and future evaluation of p53 as a cancer biomarker should consider which mutation is present in the tumor, rather than having comparison between wild-type and mutant genotypes.