Interpreting negative virus results from highly treated water

Highly treated waters often contain viruses at concentration levels too low to detect. This paper describes a method for drawing inferences about the concentration of viruses in treated water when very few or none are detected in collected samples. The approach uses a volume-varying version of the negative binomial distribution to construct one-sided confidence intervals about the mean virus concentration. Information on the presence of viruses prior to treatment (i.e., when densities are high enough to measure) is used to estimate parameters for the model. Samples are provided based on measurements from three southern California wastewater-treatment plants. Based on 93,679 L of tertiary treated sewage effluent, confidence intervals for virus concentrations for Pomona, Calif., ranged from 0 to 2.68.10(-5) plaque forming units (PFU) per liter at the 50% confidence level to 0 to 5.83.10(-4) PFU/L at the 95% confidence level. These values are considerably above the risk-based public health goal that is being used to regulate viruses in drinking water (2.22.10(-7) PFU/L). Direct measurements to demonstrate treatment to drinking water quality are therefore impracticable.