Hereditary nonpolyposis colorectal cancer (HNPCC) is widely considered to be a syndrome of defective mismatch repair (MMR). A major concern with genetic diagnosis of HNPCC is the variable, often low, percentage of pathogenic germline mutations that can be detected in MMR genes using common screening methods. The variable percentage of mutation detected is in part related to the sensitivity of conventional screening methods and may also depend on the heterogeneous genetics of HNPCC. Thus, identification of phenotypic criteria predictive of germline mutations in MMR genes may be helpful in efficient HNPCC genetic testing. Clinical diagnostic criteria, initially developed for HNPCC (e.g., Amsterdam I and II, or Bethesda criteria), can be used to clinically select patient candidates that carry germline mutations in MMR genes. More useful criteria were previously developed by analyzing families with germline MMR mutations. Using a complementary approach based on tumor microsatellite instability analysis, we confirm that the Amsterdam criteria are significantly better than the Bethesda criteria in predicting families with MSI-H tumors (P = 0.0227). Our results also suggest that a cutoff at < 50 years' mean age at diagnosis of HNPCC-related cancers (especially colorectal and endometrial cancer) may be an additional tool for the identification of families with defective MMR. Recent advances in MMR mutation screening are expected to improve detection of pathogenic MMR mutations in these families. Conversely, the high proportion of MSS tumors observed in our series of families with advanced age at cancer diagnosis is consistent with the low percentage of MMR mutations detected by previous studies in families with this phenotype. These families probably carry mutations in other genes that may or may not be related to MMR. Additional studies are necessary to clarify the molecular basis for HNPCC in families with MSS tumors.