Draft 3/10/2008 Drinking Water Program Health Outcome Based Performance Measures for Chemical Contaminants and Microbial Contaminants Overview EPA has developed new performance measures to link Drinking Water Program activities to public health outcomes. This document details the new measures - a chemical performance measure and a microbial performance measure- that EPA developed based on consideration of several possible approaches, data available to support potential measures, and input from the National Drinking Water Advisory Council (NDWAC). EPA will use the annual avoided new cases of bladder cancer attributable to the implementation of the Stage 1 and Stage 2 Disinfection By-Products Rules (DBPRs) as the health-based Chemical Contaminants Performance Measure. The health-based Microbial Performance Measure will be the national annual avoided cases of endemic Cryptosporidiosis illnesses attributable to the implementation of the Long-Term 2 Enhanced Surface Water Treatment Rule (LT2). This document describes the methodology for the calculation of the baselines, 2014 target estimates, and the actual 2014 out-year target. The measure of bladder cancer cases and Cryptosporidiosis avoided will focus on the year 2014 but will be calculated with data available from the States by 2012. The 2014 target estimate of annual avoided bladder cancer cases attributable to DBPs range from 1,380 to 2,480, with a 95% CI of 460 to 4,460. This estimate is based on data, risk assessment model, benefits assessment model, and assumptions used in the Economic Analysis (EA) for the Stage 2 DBP rule. EPA will work with States to obtain and use available compliance monitoring data to evaluate the target estimate in 2014. The 2014 target estimate of annual avoided endemic cases of Cryptosporidiosis is 231,000 to 964,000, with a 95% confidence interval of 37,000 to 2,307,000. EPA used the risk assessments and benefits assessment models used in the LT2EA to predict the changes in annual cases of Cryptosporidiosis. EPA will use the additional occurrence and treatment change information that will become available as a result of the implementation of LT2 to calculate the number of the cases of Cryptosporidiosis that will be avoided in 2014. This document describes how EPA derived the above 2014 performance measure target estimates, and the planned approach to evaluating these estimates in 2014. 1. Chemical Contaminants Performance Measure The health-based Chemical Contaminants Performance Measure will be the avoided new cases of bladder cancer nationally attributable to the implementation of the Stage 1 and Stage 2 Disinfection By-Products 1 ------- Draft 3/10/2008 Rules (DBPRs). The measure of cases of bladder cancer avoided will focus on the year 2014, but will also consider the expected additional annual cases avoided in the years following 2014 as a result of the implementation of the DBPRs. There are three steps to obtaining this performance measure: 1) Establish the baseline estimate of the number of new bladder cancer cases attributable to DBPs that would be expected to occur in 2014 in the absence of the DBPRs; 2) Calculate a "target" number of the new cases of bladder cancer that will be avoided in 2014 based on projections that can be made in 2008 of expected treatment changes and the attendant reductions of DBPs - specifically, of TTHMs - occurring in finished water that reaches consumers. 3) In 2014, calculate the number of bladder cancer cases that will be avoided that year based on compliance monitoring information that will become available by 2012 as a result of the implementation of the DBPRs. 1.1 Overview of Modeling Approach, Inputs & Uncertainty 1.1.1 Modeling Approach As indicated above, the Chemical Contaminants Performance Measure section will be an estimate of the reduction in the number of new annual cases of bladder cancer that are attributable to the estimated reduction in the national average concentration of TTHMs based on monitoring data obtained in conjunction with the implementation of the Stage 1 and Stage 2 DBP rules. The approach that EPA will use to estimate the reduction in the annual bladder cancer cases attributable to DBPs will be based upon the risk and benefits assessment models that were developed and used by EPA in support of the Economic Analysis (EA) for the Stage 2 DBP rule (EPA, 2005a). It is important to note that the Stage 2 EA models considered the cumulative effect of both the Stage 1 and the Stage 2 rules. The Stage 2 EA model included the following key aspects: 1) An estimate was made of the number of annual bladder cancer cases from all causes and of the fraction of those cases that are attributable to DBPs for conditions prior to the implementation of either the Stage 1 or Stage 2 rules. a. The number of new bladder cancer cases each year from all causes was estimated to be 56,500 based on NCI's SEER (National Cancer Institute's Surveillance Epidemiology and End Results) data updated through 2004. For the purposes of modeling the impact of these rules, it was assumed the annual number of 56,500 would remain constant over the course of the modeling period. b. The estimated fraction of all bladder cancer cases that are attributable to DBPs in drinking water was estimated from three sources: (1) several individual epidemiology 2 ------- Draft 3/10/2008 (mostly case control) studies; (2) a meta-analysis (Villanueva et al., 2003) based largely upon those same epidemiology studies; and (3) a pooled data study (Villanueva et al., 2004) that also derived in part from some of those same studies. A Population Attributable Risk (PAR) value of 15.7% (with 95% confidence intervals of 8.5% to 27.2%) was obtained from the meta-analysis study by Villanueva et al., 2003. EPA considered the results obtained from the meta-analysis study, including the confidence interval noted, to be representative of the best estimates and range of PAR values obtained from the other two approaches considered in the EA. c. Applying the above PAR value to the 56,500 bladder cancer cases from all causes results in an estimate of approximately 8,900 cases attributable to DBPs (with a corresponding 95% confidence interval of approximately 4,800 to 15,400). 2) It was assumed in the EA that the estimated 8,900 cases of bladder cancer attributable to DBPs nationally was directly related to the pre-Stage 1 national average concentration of TTHMs in all public water supplies performing chlorination. That national average concentration (population weighted) was determined to be approximately 38 ug/L. This value was based upon data from the ICR, NRWA, and several states and reflected national averages and populations affected for the following major strata: Large Surface Water Systems: 49 ug/L (-161 million people) Small Surface Water Systems: 83 ug/L (~ 8 million people) Large Disinfected Ground Water Systems (>10K): 15 ug/L (~ 65 million people) Small Disinfected Ground Water Systems (<10K): 17 ug/L (~ 29 million people) 3) It was also assumed in the EA that a reduction in the national average drinking water concentration of TTHMs would result in a proportional (linear) reduction in the attributable cases of bladder cancer cases nationally. So, for example, if the national average TTHM concentration were to decrease by 10% and stabilize at that lower value, then over time the number of new annual bladder cancer cases attributable to DBP would be 10% fewer than the pre-Stage 1 estimate of 8,900 note above. That is, a 10% reduction in the national average DBP concentration would result in 890 bladder cases avoided annually. 4) In the benefits modeling performed for the Stage 2 EA, EPA also included explicit consideration of the "cessation lag" to account for the expected time delay between when the reductions in the TTHM concentrations occur and when the full benefits of that reduction in terms of the estimated annual cases avoided would be expected to occur. So, given the above example of an eventual 890 bladder cancer cases avoided annually with a 10% reduction in the national average TTHM concentrations, the inclusion of cessation lag would result in many fewer than 890 cases per year in the period immediately following the achievement of that 10% reduction. The number will ultimately approach and reach 890 cases avoided per year once there has been a substantial turnover in the population such that the majority of those exposed would have been consuming water with lower TTHM levels for most of their lives. 3 ------- Draft 3/10/2008 a. Because of limited data and substantial uncertainty in quantifying cessation lag, EPA included estimates based on three sets of data with different combinations of carcinogen and cancer end-point: i. Smoking / Bladder Cancer ii. Arsenic (from drinking water) / Bladder Cancer iii. Smoking / Lung Cancer b. These three approaches yielded somewhat different patterns with respect to how quickly the reductions in cancer cases are expected to occur once the TTHM occurrence reductions occur. The arsenic / bladder cancer model showed that most of the cancer risk reduction was achieved fairly quickly (over 80% after 10 years) while the smoking / bladder cancer model showed a considerably slower pattern (about 40% after 10 years), with the smoking / lung cancer model falling between them (about 70% after 10 years). (See Exhibit 1 below) 4 ------- Draft 3/10/2008 Exhibit 1: Comparison of Alternative Cessation Lag Models: Estimates of Percent of Maximum Risk Reduction Achieved Each Year Following Exposure Reduction o E 0% 10 20 30 40 50 60 70 Years After DBP Exposure Reduction Begins 90 100 In applying the Stage 2 EA risk and benefits modeling for the performance measures effort, EPA "streamlined" certain aspects of the risk and benefits models used in the Stage 2 EA. Specifically, EPA: • Included only one of the three bladder cancer PAR estimate approaches, namely that based on the Villanueva et al. (2003) meta-analysis. EPA also included the estimated uncertainty in the PAR value obtained from this data. • Included only two of the three cessation lag models, namely those based on the arsenic / bladder cancer data and the smoking / bladder cancer data. These two models comprise the lower and upper ends of the rate of cessation lag indicated from the three approaches used in the EA. Again, EPA also included the estimated uncertainty in these two cessation lag models. 1.1.2 Inputs With one exception discussed below, the input data for the performance measures modeling will be identical to that used in the modeling performed for the Stage 2 EA. Indeed, for the baseline and target estimates described below, EPA will essentially "re-run" the model and input data as it was used for the Stage 2 EA (modified with respect to the "streamlining" aspects indicated above). 5 ------- Draft 3/10/2008 EPA will use the TTHM occurrence estimates and the changes in occurrence following implementation as they were predicted in the EA to determine the pre-Stage 1 baseline estimate and to determine the target estimate of bladder cancer cases avoided in 2014. However, when the Agency analyzes how well the performance measure meets to 2014 target, EPA will obtain and use compliance monitoring data from the states covering the period from implementation of these rules up through approximately 2012 to calculate the national average TTHM concentration over that period and, from that, estimate the number of bladder cancer cases avoided from the observed reductions in occurrence relative to the pre-Stage 1 baseline. Compliance monitoring data is not routinely reported to EPA. EPA will work with the states to obtain this data. 1.1.3 Uncertainty In implementing the risk and benefits modeling for the Stage 2 EA, EPA included explicit, quantitative consideration of a number of factors that contributed to the uncertainty in the estimates of bladder cancer cases avoided from these rules. These factors will also be included to characterize uncertainty in the baseline and target estimates, as well as in the actual 2014 performance measure. The three main elements of the modeling for which contributions to uncertainty will be considered are: 1) The PAR value 2) The national average TTHM concentrations and the reduction in that average from pre- Stage 1 to post-Stage2 3) Cessation Lag 1.2 Baseline Metrics: In the Stage 2 EA, EPA estimated that, in the years prior to the promulgation and implementation of the Stage 1 rule, there were approximately 56,500 new bladder cancer cases in the US each year. Using the PAR derived from the Villanueva et al. (2003) meta-analysis, EPA also estimated that approximately 8,900 of those new yearly bladder cancer cases are attributable to exposure to DBPs in public drinking water systems. Taking into account the uncertainty in the PAR value (based on uncertainties in the data used to generate it), the 95% confidence interval on that estimate was a baseline range from 4,800 to 15,400 cases. It is important to note that the uncertainty in the attributable cases is consistent with the Stage 2 EA and does not take into account expected changes in the overall population affected, changes in demographics, or changes in individual exposure patterns that may take place overtime. In the Stage 2 EA, EPA assumed that, in the absence of either the Stage 1 or the Stage 2 rules, there would continue to be an additional 8,900 new bladder cancer cases attributable to DBP exposure every year over the approximately 25 year time span that was considered for the benefit - cost analysis. For the purposes of this performance measure effort, EPA is assuming (as was done in the Stage 2 EA) that, in the absence of these two rules, there would continue to be 8,900 new bladder cancer cases attributable to DBPs occurring each year, including 2014, the target year for the measures assessment. 6 ------- Draft 3/10/2008 1.3 Out-Year Targets This section of the document provides EPA's "target estimate" of the number of cases of bladder cancer that will be avoided in the year 2014 resulting from the expected reduction in the occurrence of DBPs in public drinking water systems due to the combined effects of the Stage 1 and Stage 2 rules. The 2014 target estimate is based on data, methods, and assumptions that were used in the Stage 2 EA. Key among these for the 2014 target estimate are: 1) Prior to the promulgation of the Stage 1 rule, the population-weighted national average concentration of TTHMs (the DBP that EPA will use for this performance measure effort) was 38.1 ug/L. 2) Following full implementation of the Stage 1 rule, the population-weighted average TTHM concentration would be 27.7 ug/L. 3) Following full implementation of the Stage 2 rule, the population-weighted average TTHM concentration would be 25.5 ug/L. 4) There are implementation schedules for these rules that affect the timeframe over which these reduced DBP levels are realized. 5) There is a linear relationship between the reduction in TTHMs and the reduction in the number of bladder cancer cases attributable to DBPs. 6) The effect of the change in DBP occurrence with respect to bladder cancer cases is not immediate, but involves a cessation lag. Exhibit 2 below provides a simplified depiction of the change in the expected reduction in the national average TTHM levels following implementation of the Stage 1 and Stage 2 rules. These reductions reflect a short period following promulgation of the rules when no treatment changes are made as systems (and states) perform monitoring and gather other information necessary for compliance. EPA has made a simplifying assumption that once treatment changes begin the national average TTHM levels will decline linearly. TTHM levels will reach the predicted post-rule levels when all systems are required to be in compliance. So, reductions from Stage 1, which was promulgated in 1998, are shown to occur through 2006 when the rule was to be fully implemented and the predicted post-Stage 1 national average TTHM level of 27.7 ug/L was expected to be realized. Further reduction from the Stage 2 rule occur from 2009 and are fully achieved by 2014 where the national average TTHM level is expected to be 25.5 ug/L. 7 ------- Draft 3/10/2008 Exhibit 2: Expected Reduction in the National Average TTHM Levels Following Implementation of the Stage 1 and Stage 2 Rules 40 38 3" 36 ¦» 3, S 34 1 ¦£ ,, 8 32 C 8 30 $28 I ^ 26 n S ii 24 22 20 8 ------- Draft 3/10/2008 Exhibit 3 below converts the foregoing changes in national average TTHM levels to changes in the attributable cases of bladder cancer using the assumption of a linear relationship as noted above. Exhibit 3: Expected Reduction in the Annual New Bladder Cancers Cases Attributable to DBPs Following Implementation of the Stage 1 and Stage 2 Rules Q 0 8000 +¦» 0) 3 •*-» £ ¦c •*-» SE 1 <3 8 s >- 4000 ci ~o i§ 23 3000 75 2000 3 1995 2000 2005 2010 2015 2020 9 ------- Draft 3/10/2008 Exhibit 4 below displays the number of bladder cancer cases that would be avoided each year if the effect of those reductions were immediate. As indicated in this exhibit, the number of bladder cancer cases that would be avoided in 2014 if there were no cessation lag would be -2,930. The 95% CI for this estimate is -1,590 to -5,060. Exhibit 4: Annual Bladder Cancer Cases Avoided without Cessation Lag 1995 2000 2005 2010 2015 2020 10 ------- Draft 3/10/2008 As indicated in Section 1, above, EPA is using the cessation lag functions based on the arsenic/bladder cancer and the smoking/bladder cancer data as upper and lower bounds, respectively, of the rate at which the potential cases avoided are actually realized. Exhibit 5 below shows the results of the application of those cessation lag functions to the cases avoided, which indicate that in 2014 the annual number of bladder cancer cases avoided fall in a range of-1,380 to -2,480. Exhibit 5: Annual Bladder Cases Avoided with Cessation Lag 3500 1995 2000 2005 2010 2015 2020 Therefore, the estimate of bladder cancer cases avoided in 2014 resulting from the implementation of the Stage 1 and Stage 2 ranges from -1,380 to -2,480. Taking into account quantifiable uncertainties in the PAR value, in the Pre-Stage 1 national average TTHM level, in the reduction in TTHM from these rules, and in the cessation lag functions, the overall 95% CI for the number of cases avoided in 2014 is -460 to -4,460. (Note that, as indicated previously, the 95% CI for the number of cases avoid in 2014 without consideration of the cessation lag is -1,590 to -5,060.) Compliance monitoring data is not routinely reported to EPA. EPA will work with the states to obtain this data. 11 ------- Draft 3/10/2008 1.4 Methodology for Evaluating Out-Year Targets The approach that EPA will use to evaluate the 2014 target estimate will, use the same modeling assumptions as used to develop the target estimate above, with one key exception. In addition to using the estimates of the national average TTHM levels expected as a result of the implementation of these rules, EPA will also work with the states to obtain and use available compliance monitoring data to calculate the national average in 2014. EPA can use this national average to compute the reduction from the Pre-Stage 1 national average and to estimate the reduction in attributable bladder cancer cases (using the same assumptions used to generate the target estimate). There are some known limitations that EPA will need to address in the 2014 target evaluation. First, it is unlikely that EPA will be able to access all of the compliance data for all public water systems. EPA will assess how representative the data received are and to make any adjustments needed to use the data to calculate the national average. Second, it is most likely that in 2014, the compliance data that will be available will be reflective of monitoring completed in earlier years, perhaps no more recently than 2012. EPA will therefore need to account for any additional TTHM reductions that might occur between when the obtained monitoring data were collected and 2014. Finally, it is recognized that as result of the IDSE component of the Stage 2 rule, many of the monitoring locations for systems (especially the larger systems) will have changed from those used to estimate the pre-Stage 1 TTHM levels to reflect the highest levels in the system. Those changes will result in higher TTHM levels than observed previously. TTHM reductions resulting from the Stage 1 and Stage 2 rules may be an underestimate. To some extent, this effect is accounted for in the target estimates provided above, which are based on an approach that EPA used in the Stage 2 EA to account for the effect of the IDSE. EPA will consider this issue in examining the Post Stage 2 monitoring data when calculating the 2014 target. 1.5 Additional Out-Year Analyses In addition to targeting and evaluating the number of bladder cancer cases avoided in 2014 specifically, EPA will also consider the cumulative number of bladder cancer cases avoided from promulgation of the Stage 1 and Stage 2 Rules. EPA will develop two cumulative estimates based on the compliance monitoring data that will also be used for the 2014 estimate. The first cumulative estimate will address the bladder cancer cases avoided beginning with the promulgation of the Stage 1 rule in 1998 through the 2014 performance measure year. Using the same procedures as those derived from the Stage 2 EA to obtain the target 2014 estimates provided in Section 1.3, EPA currently estimates that the cumulative cases avoided through 2014 will range from approximately 8,500 (using the smoking/bladder cancer cessation lag model) to 17,300 (using the 12 ------- Draft 3/10/2008 arsenic/bladder cancer cessation lag model). The 95% confidence interval for this estimate is 2,800 to 31,200. The second cumulative estimate will address the bladder cancer cases avoided from 1998 through the year 2025, reflecting a 20 year period following promulgation of the Stage 2 rule in 2006. Using the same procedures form the Stage 2 EA as those used for the 2014 estimates provided in Section 1.3, EPA currently estimates that the cumulative cases avoided through 2025 will range from approximately 28,200 (using the smoking/bladder cancer cessation lag model) to 47,400 (using the arsenic/bladder cancer cessation lag model). The 95% confidence interval for this estimate is 9,300 to 85,200. 13 ------- Draft 3/10/2008 2. Microbial Performance Measure The health-based Microbial Performance Measure will be the annual cases avoided nationally of endemic Cryptosporidiosis illnesses attributable to the implementation of the Long-Term 2 Enhanced Surface Water Treatment Rule (LT2). This measure of annual endemic Cryptosporidiosis cases avoided will focus on the year 2014, but will also address the expected annual cases avoided in the years following 2014 as a result of the implementation of LT2. There are three main steps involved in obtaining this performance measure: 1) Establish the baseline of the number of annual cases of endemic Cryptosporidiosis that would be expected to occur in 2014 in the absence of the LT2; 2) Calculate a "target" number of the cases of Cryptosporidiosis that will be avoided in 2014 based on projections that can be made in 2008 of treatment changes and the attendant reductions in Cryptosporidium occurring in finished water that reaches consumers; 3) In 2014, calculate the number of the cases of Cryptosporidiosis that will be avoided that year based on the additional occurrence and treatment change information that will become available as a result of the implementation of LT2. 2.1 Overview of Modeling Approach, Inputs & Uncertainty 2.1.1 Modeling Approach Ideally, the health-based microbial performance measure would involve a more direct determination of the number of cases of Cryptosporidiosis avoided nationally in 2014 based on the reported incidence of cases prior to and subsequent to the implementation of LT2. However, limitations in available public health data preclude using such an approach. It is therefore necessary to estimate the number of avoided cases of Cryptosporidiosis in 2014 using other data that are, and will become, available on the occurrence of Cryptosporidium in source waters, together with information on the effectiveness of treatment to determine the levels in finished water, dose-response relationships, and morbidity to predict the probability of illnesses occurring in an exposed population given those expected levels of Cryptosporidium in finished drinking water. EPA developed and implemented a risk and benefit assessment models in support of the final LT2 rule, and that model is described in detail in Chapters 4 and 5 (as well as in several appendices) of the Economic Analysis for the Long Term 2 Enhanced Surface Water Treatment Rule (USEPA, 2005b), referred to here as the LT2EA. That model is designed to predict the changes in annual cases of Cryptosporidiosis due to expected treatment changes in response to source water compliance monitoring carried out by affected surface water systems in the US. EPA used the benefits and risk modeling approach from the LT2EA (with some modifications) for this microbial performance measure effort. 14 ------- Draft 3/10/2008 The main elements of the LT2EA model are as follows: 1. Based on representative source water monitoring data, estimates are made of the baseline occurrence of Cryptosporidium in the influent water at surface water systems in the US. 2. Based on information available on existing treatment methods in use in the US, and on the effectiveness of those treatment methods, the model predicts the reduction in Cryptosporidium levels and the resulting occurrence in finished water. 3. Using water consumption information and dose-response relationships that together describe the probability of infection and illness as a result of consumption of Cryptosporidium in drinking water, the model predicts the number of cases of Cryptosporidiosis that occur annually in the US attributable to public drinking water. 4. Using the same source water monitoring data as referred to in element (1) above together with the requirements of the LT2 rule to implement treatment to achieve additional removal as a function of the observed source water occurrence, new estimates of the finished water occurrence of Cryptosporidium are made to reflect the lower, post-LT2 implementation finished water occurrence. 5. Using the same water consumption and dose-response relationships as indicated in (3) above together with the lower finished water occurrence estimates indicated in (4) above, the model predicts a lower number of annual cases of Cryptosporidiosis in the US from drinking water consumption, with the "cases avoided" being simply the difference between the pre-LT2 and post-LT2 estimates. The LT2EA model does not consider conditions at each public water supply, but rather it is constructed as a Monte Carlo simulation in which a large number of surface water systems and populations served by them are created and characterized with respect to source water occurrence, treatment effectiveness, finished water occurrence, exposure and dose-response relationships. With respect to element (1) of the LT2EA model, the representative source water occurrence monitoring data used was obtained from: • The Information Collection Rule (ICR), which focused on all large filtered and unfiltered systems for an 18 month period in 1997-1998. • The ICR Supplemental Surveys (ICRSS), which involved a representative sample of 40 large and 40 medium systems (mainly filtered) for a 12 month period in 1999 - 2000. EPA used the occurrence information from the ICR and ICRSS sources to derive four separate characterizations of the distribution of the baseline occurrence of Cryptosporidium in the source waters used by surface water systems in the US. EPA derived four different distributions: 1. Unfiltered system occurrence from the ICR data 2. Filtered system occurrence from the ICR data 15 ------- Draft 3/10/2008 3. Filtered system occurrence from the ICRSS large system data 4. Filtered system occurrence from the ICRSS medium system data Filtered and unfiltered systems are treated separately both because of recognized differences in occurrence of Cryptosporidium in their source waters and because of different requirements in the LT2 rule for these two types of systems. Exhibit 6 depicts the four cumulative source water occurrence distributions. These are lognormal distributions of the mean occurrence at plants across the sampling period and reflect the variability of Cryptosporidium occurrence from plant to plant. The distributions shown in this exhibit reflect the central tendency of 1,000 such distributions generated for each of the four sets that captured uncertainty in the occurrence; the LT2EA model used all 1,000 such distributions to characterize uncertainty in the results. Exhibit 6: Lognormal Distributions for Source Water Occurrence 100% - jj! scft _ra ~_ ra Q> ~_ £ 13 23% " [ft " ICR Unfitted ICR Filtered SSM iliih IIIIiim -m ¦" T 1e-005 0.0001 0.001 0.01 0.1 1 Plant Mean Cryptosporidium Concentration (Total oocysts/L) r 10 As shown in Exhibit 6, the occurrence levels of Cryptosporidium in the source waters used by unfiltered systems are generally lower, by approximately an order of magnitude, than those for filtered systems. (Note there were some data from the ICRSS sets on unfiltered systems, but too few to support the 16 ------- Draft 3/10/2008 derivation of national distributions.) The three distributions for filtered systems show similar central tendency (median) values, but are notably different in the tails, especially the upper tail which drives the risk and benefits estimates. In particular, the ICR data shows substantially higher probability of high counts (>1 oocyst/L) of Cryptosporidium occurrence than either of the two ICRSS data sets, and of those two the ICRSS medium system data showed slightly greater probability of occurrence in the upper tail than did the ICRSS large system data. For all of the subsequent steps of the modeling conducted for the LT2EA, EPA used the three different filtered water baseline occurrence distributions as three separate and equally valid representations of baseline occurrence in the source waters used by filtered systems, using the differing results from them to further characterize the uncertainty in the estimates. (Only the distribution from the ICR data for unfiltered systems was used to represent occurrence in the source waters for unfiltered systems; however, as noted above, there were 1,000 such distributions used in the model to capture some of the uncertainty for the unfiltered system estimates). As in the LT2EA, a range of different assumptions were made when predicting the efficiency of Cryptosporidium removal between source water and finished water depending upon different characteristics of the systems (LT2EA model element (2)). For all unfiltered systems it was assumed that there was no removal of Cryptosporidium between source water and finished water. For small, filtered systems (serving < 10,000 people), a range of removal efficiencies from 2-log to 4-log was used, with the mostly likely removal being around 3-log. For large filtered systems, the range of removal efficiencies used was from 2-log to 5-log, with the most likely value being about 3.25-log removal. The treatment effectiveness ranges used for in the model were incorporated into the LT2EA model as triangular distributions as shown in Exhibit 7 to capture both variability and uncertainty in the estimates of finished water occurrence. 17 ------- Draft 3/10/2008 Exhibit 7: Pre-LT2 Treatment Assumptions Small Systems, Standard Estimate Large Systems, Standard Estimate Small Systems, with 0.5 Log Credit Large Systems, with 0.5 Log Credit 18 ------- Draft 3/10/2008 Exhibit 8 provides an indication of the finished water occurrence resulting from the integration of the source water occurrence distributions with the treatment effectiveness assumptions. As indicated in Exhibit 8, the unfiltered systems, for which there is no change between source water and finished water occurrence, have finished water levels that are approximately 2 to 3 logs higher than the finished water levels in filtered systems. As with the three estimates of source water occurrence for filtered systems, the finished water occurrence distributions also show a greater probability of higher levels from the ICR data than from the two ICRSS data sets. 19 ------- Draft 3/10/2008 Exhibit 8: Pre-LT2 Finished Water Occurrence Distributions for Small and Large Systems Flnl Hi e d Wa tar Cry pto * por HI un (Ceo y»U! L) Rnlrtiiri HiiirOip-h ^irlrflum EQHfibriL] 20 ------- Draft 3/10/2008 For element (3) of the LT2EA model, EPA used the SDWIS database to determine the number of people served by surface water systems. EPA used these data and the information that EPA has developed on daily drinking water ingestion patterns to characterize exposure, the probability of individuals ingesting a Cryptosporidium oocyst on a given day and over the course of a year. This exposure information was integrated with a number of different approaches to describing the dose-response relationship between oocyst ingestion and the probability of becoming infected, the probability of becoming ill, and the probability of that illness resulting in mortality. These alternative approaches to characterizing the dose- response relationships (which notably include a primary model and six alternatives for estimating the probability of infection given ingestion) are described in detail in the LT2EA and are not elaborated on here. For the purposes of the performance measures effort, EPA will only use the primary dose-response model approach. Also, for the purposes of the performance measure effort, EPA will limit the end-point to Cryptosporidiosis illnesses due to drinking water exposures (including additional cases that occur as secondary spread related to the drinking exposure cases). LT2EA model elements (4) and (5) essentially entail conducting a 'second pass' of the simulation model where, based upon the source water occurrence information and the binning and treatment requirements of the LT2 rule, there is an additional 2 or 3 log removal for all unfiltered systems and an additional 1 to 2.5 log removal for those filtered systems having >0.075 oocysts/L in the source water. Except for the inclusion of these additional log reduction amounts, all other aspects of the simulated water systems and populations served are kept the same. As a result, the model produces pre-LT2 and post-LT2 estimates of the annual cases of endemic Cryptosporidiosis, with the difference between those being the "annual cases avoided" attributable to the LT2 rule. 2.1.2 Inputs With the exception discussed here, the inputs to the LT2EA model for the performance measures effort will be the same as those used to support the final LT2 rule. The exception will be the inclusion of approximately one year of compliance monitoring data that will be available in early 2008 on Cryptosporidium levels in source waters for large systems. EPA will use that data as a fourth data set to be modeled along with the ICR, ICRSSL, and ICRSSM data sets to characterize the baseline and to develop the target estimate. (Note: This new baseline data is currently being analyzed and is not included in this draft. It is expected to be available by the beginning of March). Further, when the evaluation of the target estimate is carried out in 2014, EPA will use the more complete set of compliance monitoring data from all size systems that will be available at that time to serve as the basis for determining the actual binning of systems for treatment change determinations, as well as to provide input to the risk and benefits calculations. The use of the compliance monitoring data in the performance measures effort is discussed further in Section 2.2. 2.1.3 Uncertainty In implementing the risk and benefits modeling for the LT2EA, EPA also quantified the uncertainty in the estimates of the number of endemic cases of Cryptosporidiosis avoided from the LT2 rule associated with uncertainty in several of the inputs. EPA will similarly characterize uncertainty in the baseline and target estimates, as well as in the actual 2014 performance measure. 21 ------- Draft 3/10/2008 The main elements of the modeling for which contributions to uncertainty will be considered are: 1) The source water occurrence of Cryptosporidium (as characterized by the use of the alternative source water occurrence data sets) 2) The effectiveness of existing treatment 3) The dose-response relationship parameters 2.2 Baseline Metrics: In the LT2 EA, EPA developed three estimates of the baseline cases of endemic Cryptosporidiosis occurring annually in the US. These were based on the ICR, ICRSSL and ICRSSM data. For the purpose of the performance measure, EPA is using those three estimates as alternative baselines. (A fourth baseline measure based on large system LT2 compliance monitoring data is currently being analyzed and is not included in the this draft.) EPA will use these alternative estimates to capture the recognized variability in Cryptosporidium occurrence over time and the resulting uncertainty that this, and other factors, contribute to the estimate of the baseline cases of endemic Cryptosporidiosis for a "typical" year or for any specific year (such as 2014, the target year for the performance measures effort). Exhibit 9 summarizes these three estimates of baseline annual cases of Cryptosporidiosis due to drinking water: Exhibit 9: Estimated Baseline Annual Cases of Cryptosporidiosis Due to Drinking Water Filtered Unfiltered Total 95% Confidence Interval ICR 491,091 501,706 992,797 147,826 to 2,390,920 ICRSSL 147,185 146,449 293,634 45,028 to 714,508 ICRSSM 257,985 257,342 515,327 76,177 to 1,308,597 These three estimates of the baseline cases of Cryptosporidiosis in 2014 due to drinking water consumption are EPA's estimates of what would occur in the absence of the LT2 rule. These range from approximately 290,000 to over 990,000 cases, with 95% confidence bounds that range from approximately 45,000 to 2,390,000 cases per year. 2.3 Out-Year Targets The LT2 rule compliance monitoring schedule requires that all systems that are required to install treatment to comply with the rule do so before 2014. Although the rule offers some systems extensions to 22 ------- Draft 3/10/2008 the deadlines, these extensions end in 2014 for large systems and extend only to 2015 for small systems. For the purpose of the 2014 target estimates used in this performance measure, it is assumed that all systems will be in compliance by 2014. The estimated cases of endemic Cryptosporidiosis avoided in 2014 are summarized in Exhibit 10. Exhibit 10: Estimated Endemic Cryptosporidiosis Cases Avoided in 2014 Filtered Unfiltered Total 95% Confidence Interval ICR 464,069 500,291 964,360 144,890 to 2,307,013 ICRSSL 84,609 146,121 230,730 37,334 to 541,300 ICRSSM 198,426 256,744 455,170 69,755 to 1,135,677 Using the lowest and highest of these estimates, the 2014 target estimate of avoided endemic cases of Cryptosporidiosis is 231,000 to 964,000, with a 95% confidence interval of 37,000 to 2,300,000. 2.4 Methodology for Evaluating Out-Year Targets EPA intends to use the same basic modeling procedures for the 2014 evaluation of the target estimates with one key difference. When estimating the reduction of endemic cases of Cryptosporidiosis in the target estimate as presented above, EPA will use as fourth estimate of Cryptosporidium occurrence in the source waters of the systems affected by the LT2 rule. This fourth estimates will be derived from the actual compliance monitoring data obtained by the systems and reported to the states as required by the LT2 rule. The compliance monitoring required under LT2 determines the occurrence of Cryptosporidium in source water which then determines how the systems are binned and, therefore, what additional log removal will need to be achieved through treatment or other mitigation strategies. When the evaluation of those estimates is carried out in 2014, EPA plans to use only this compliance monitoring data and the estimates of additional log removal implied by them to characterize treatment changes in affected systems. It is recognized, however, that there will still be year-to-year variability in the occurrence of Cryptosporidium in source waters due to variability in weather patterns. EPA will therefore model the estimates of baseline cases and cases avoided resulting from the added treatment using the compliance monitoring data and the three original sets of baseline occurrence data from the EA to generate a range of estimates to capture what the avoided cases in 2014 might be. As with the target estimates, EPA will present the evaluation of the target in 2014 as a high and low end value, derived from the four baseline occurrence data sets. EPA will also include a confidence interval based on the quantifiable uncertainties included in the modeling. 23 ------- Draft 3/10/2008 2.5 Additional Out-Year Analyses In addition to targeting and evaluating the number of Cryptosporidiosis illness cases avoided in 2014 specifically, EPA will also consider the cumulative number of cases avoided from promulgation of the LT2 rule. Two cumulative estimates will be developed based on the compliance monitoring data that EPA will obtain and use for the 2014 estimate. The first cumulative estimate will address the total Cryptosporidiosis cases avoided from promulgation of the LT2 rule in 2005 through the 2014 performance measure year. Using the same procedures as those derived from the LT2 EA to obtain the 2014 target estimates provided in Section 1.3, EPA currently estimates that the cumulative cases avoided through 2014 will range from approximately 902,000 to 3,762,000 as summarized for the three data sets in Exhibit 11. Exhibit 11: Estimated Cumulative Endemic Cryptosporidiosis Cases Avoided through 2014 Filtered Unfiltered Total 95% Confidence Interval ICR 1,785,000 1,977,000 3,762,000 565,000 to 8,997,000 ICRSSL 325,000 577,000 902,000 146,000 to 2,115,000 ICRSSM 762,000 1,013,000 1,775,000 272,000 to 4,429,000 The second cumulative estimate will address the Cryptosporidiosis cases avoided from 2005 through the year 2025, reflecting a 20 year period following promulgation of the LT2 rule. Using the same procedures as those derived from the LT2 EA for the 2014 estimates provided in Section 1.3, EPA currently estimates that the cumulative cases avoided through 2025 will range from approximately 902,000 to 3,762,000 as summarized for the three data sets in Exhibit 12. 24 ------- Draft 3/10/2008 Exhibit 12: Estimated Cumulative Endemic Cryptosporidiosis Cases Avoided through 2025 Filtered Unfiltered Total 95% Confidence Interval ICR 6,889,000 7,480,000 14,369,000 2,158,000 to 34,371,000 ICRSSL 1,256,000 2,184,000 3,440,000 557,000 to 8,069,000 ICRSSM 2,945,000 3,837,000 6,782,000 1,040,000 to 16,922,000 Summary: EPA will use the methodologies outlined in the Stage 2 and LT2 rules to develop chemical and microbial performance measures, respectively, to link Drinking Water Programs to public health outcomes. The 2014 target estimate of annual avoided bladder cancer cases attributable to DBPs range from 1,380 to 2,480, with a 95% CI of 460 to 4,460. This estimate is based on data, risk assessment model, benefits assessment model, and assumptions used in the Economic Analysis (EA) for the Stage 2 DBP rule. EPA will work with States to obtain and use available compliance monitoring data to evaluate the target estimate in 2014. The 2014 target estimate of annual avoided endemic cases of Cryptosporidiosis is 231,000 to 964,000, with a 95% confidence interval of 37,000 to 2,307,000. EPA used the risk assessments and benefits assessment models used in the LT2EA to predict the changes in annual cases of Cryptosporidiosis. EPA will use the additional occurrence and treatment change information that will become available as a result of the implementation of LT2 to calculate the number of the cases of Cryptosporidiosis that will be avoided in 2014. 25 ------- |