Statistical tests for clustering of fatal and nonfatal overdoses during this interval identified seven heroin-related deaths that occurred on March 1 to 2 as a statistically significant cluster. An extension of the analysis for clustering to a 15-month period identified two additional clusters, one of fatal overdoses and one of nonfatal ones. When all victims of fatal overdose in cluster intervals were combined and compared with all other heroin-related deaths, no significant differences were noted for levels of morphine or ethanol in blood. However, bile morphine concentrations of cluster decedents were significantly lower than those of noncluster decedents, suggesting that these decedents were less tolerant to the effects of narcotics than the comparison group. Heroin concentration in street-level heroin samples collected during clusters did not differ from those collected during comparison intervals. These data conflict with the traditional explanation of overdose clusters, which attributes these events to unusually potent street-level heroin. (Author abstract)