Refining the uncertainties and expansion of wastewater-based epidemiology for assessing population exposure to chemicals

Abstract

Systematic sampling and analysis of wastewater is an increasingly used tool tocomplement more traditional techniques for assessing consumption of licit and illicitchemical substances in the population. The use of wastewater sampling andanalysis contributes to a broader field that is referred to as wastewater-basedepidemiology (WBE). Both spatial and temporal analysis can be conductedquantitatively, quickly and cost effectively using the WBE approach.In brief, per capita exposure to a given chemical within a population is estimated bymeasuring the concentration of that chemical or a metabolite in a representativewastewater sample multiplied by the volume of wastewater represented by thesample, divided by the population size from which the sample originates andcorrecting by factors such as the excretion factor of the metabolite or chemical,molecular weight change and potential stability of the chemical within the sewer. Ithas previously been determined that the population size is the largest uncertainty forWBE estimates.The aim of this thesis it to therefore identify useful markers that allow populationestimation for a given wastewater sample and apply this technique including toassess per capita exposure/release of a group of chemicals that have not beenexamined in previous WBE studies.The approach for this thesis was to systematically collect samples on a day when thepopulation is well defined. For this we collected samples on the 2011 Census Day inAustralia from 10 wastewater catchments ranging in size from approximately 1,500to 500,000 people. By providing catchment maps to the Australian Bureau ofStatistics, the accurate population size for each catchment was determined.The most obvious choice of a potentially useful population markers are endogenouschemicals such as creatinine. Therefore, in Chapter 2, creatinine was assessed as apopulation marker. It was found that there was no correlation between the mass loadof creatinine in wastewater and the population. Using laboratory-scale sewerreactors with conditions representative of both gravity sewers and rising mains, itwas found that creatinine, while stable in collected samples, is unstable under sewerconditions. We therefore conclude that creatinine is not suitable for predictingdifferences in population size particularly when different sewer systems arecompared.In Chapter 3, a method was developed to quantify 96 chemicals in wastewaterinfluent and applied to the census wastewater samples to identify potentialpopulation size markers. Thirteen chemicals including acesulfame, caffeine, andpharmaceuticals and personal care products were detected in all samples and foundto have a good correlation (R2 > 0.8) between mass load and population size. ABayesian inference model was developed which incorporated these potentialpopulation size markers to provide a population size estimate. The model wasvalidated using a leave-one-out approach for all sites and comparing the populationsize estimate from the model with the accurate census population data. It was shownthat for small catchments, the uncertainty of the estimate as measured by the widthof the posterior was 1.1 to 2.4 times narrower than the width of the posterior usingonly the WWTP operator population size estimates. For large WWTP catchments,the width of the posterior using the population size model was between 5 and 40times narrower than the WWTP operator population size estimates. Additionally, itwas found that the posterior width of the model was improved with addition of morechemicals in the model.In Chapter 4, using laboratory-scale sewer reactors, the impact of in-sewerdegradation of the population markers identified in Chapter 3 was evaluated. It wasfound that five of the fourteen markers were stable under all conditions over the 12hour study period. Those which were unstable ranged from little degradation over thestudy period and only under certain conditions to rapid degradation regardless ofsewer conditions. Additionally we assessed whether or not the degradation of thesechemicals impacted the population size estimation model by excluding the unstablecompounds from the model. We found that the uncertainty of the estimate did notdecrease through exclusion of the unstable chemicals.In Chapter 5 we assessed whether WBE can be expanded to chemicals other thanthose previously assessed (i.e. illicit drugs, alcohol and tobacco). For this a methodwas developed to analyse organophosphorous flame retardants (PFRs) inwastewater influent. Using the samples collected during census, it was estimatedthat 2.1 mg person-1 day-1 enter Australian wastewater. In addition we found a goodcorrelation (r2 ≥ 0.87) between the population size and mass load for each of the fourmain contributors (TBOEP, TCIPP, TDCIPP and TCEP).Overall, this thesis demonstrates that the uncertainty of WBE estimates can beimproved through identifying population markers, and developing and calibrating apopulation model. Additionally, we found that WBE is not limited to assessingexposure/consumption of the chemicals previously assessed (i.e. alcohol, illicit drugsand tobacco) and can be expanded to other chemical groups.