SEPA

IJnlted States Office of Research andEnwronmental Protection DevelopmentAgency Washington DC 20460

EPA,‘600/R-92/08 1September 1993

Methods for AquaticToxicity IdentificationEvaluations

Phase Ill ToxicityConfirmation Procedures forSamples Exhibiting Acute andChronic Toxicity

c

EPA/600/R-92/081September 1993

Methods for AquaticToxicity Identification Evaluations

Phase I I I Toxicity ConfirmationProcedures for Samples Exhibiting

Acute and Chronic Toxicity--

bYbY

D. I. Mount’T. J. Norberg-King*

With Contributions from:

G. T. Ankley*L. P. Burkhard2E. J. Durban*

M. K. Schubauer-BeriganlM. T. Lukasewyczl

‘AScl Corporation - Contract No. 68-CO-0058*U.S. Environmental Protection Agency

Prev’iis Phase Ill Methodsby 0. I. Mount

EPA-600/3-88/036

National Effluent Toxicii Assessment CenterTechnical Report 02-93

Environmental Research LaboratoryOffice of Research and DevelopmentU.S. Environmental Protection Agency

Duluth, MN 55804

@ Printed on Recycled Paper

Disclaimer

This document has been reviewed in accordance with U.S. EnvironmentalProtection Agency Policy and approved for publication. Mention of trade names orcommercial products does not constitute endorsement or recommendation for use.

Foreword

This Phase III document is the last in a series of guidance documentsintended to aid dischargers and their consultants in conducting aquatic organismtoxicity identification evaluations (TIES). TIES might be required by state or federalagencies as the result of an enforcement action or as a condition of a NationalPollutant Discharge Elimination System (NPDES) permit. These documents shouldaid individuals in overseeing and determining the adequacy of effluent TIES as a partof toxicity reduction evaluations (TREs).

There are two major reasons to require the confirmation procedures. First theeffluent manipulations used in Phase I characterizations (EPA, 1988; EPA, 1991 A;EPA, 1992) and Phase II identifications (EPA, 1989A; EPA, 1993A) might (with someeffluents) create artifacts that might lead to erroneous conclusions about the causeof toxicity. Therefore in Phase III confirmation steps, manipulations of the effluent areavoided and/or are minimized, therefore artifacts are far less likely to occur. Some-times, toxicants will be suspected through other approaches (such as the treatabilityroute) which on their own are not definitive and in these instances, confirmation isnecessary. Secondly, there is the probability that the substances causing toxicitymight change from sample to sample, from season to season or some otherperiodicity. As toxicity is a generic measurement, measuring toxicity cannot revealvariability of the suspect toxicant whereas the Phase III confirmation procedures aredesigned to indicate the presence of variable toxicants. Obviously, this crucialinformation is essential so that remedial action may be taken to remove toxicity.

t

Confirmation, whether using the procedures described in this document orothers, should always be completed because the risk is too great to avoid or eliminatethis step. Especially for discharges where there is little control over the influent or fordischarge operations that are very large or complex, the probability that differentconstituents tiill cause toxicity over time is great. Most of the approaches in Phase IIIare applicable to chronically toxic effluents and acutely toxic effluents.

In this confirmation document, guidance is included when the treatabilityapproach (EPA, 19898; EPA, 1989C) is taken. Use of the treatability approachrequires confirmation as much as or more than the toxicant identification approach(Phase II). The reader is encouraged to use both the acute Phase I characterization(EPA, 1991 A) and the chronic Phase I characterization (EPA, 1992) documents fordetails of quality assurance/quality control (CWQC), health and safety, facilities andequipment, dilution water, sampling and testing. The TIE methods are written asgeneral guidance rather than rigid protocols for conducting TIES and these methodsshould be applicable to other aqueous samples, such as ambient waters, sediment

” elutriate or pore waters, and 1eachate.s

. . .III

Abstract

In 1989, the guidance document for acutely toxic effluents entitled Methodsfor Aquatic Toxicity Identification Evaluations: Phase Ill Toxicity Confirmation Proce-dures was published (EPA, 19890). This new Phase Ill manual and its companiondocuments (EPA, 1991 A; EPA, 1992; EPA, 1993A) are intended to provide guidanceto aid dischargers in confirming the cause of toxicity in industrial and municipaleffluents. The toxicity identification evaluation (TIE) starts with a characterization ofthe effluent toxicity using aquatic organisms to track toxicity; this step is followed byidentifying a suspect toxicant and then confirming the suspect toxicant as the causeof toxicity.

This Phase III confirmation document provides greater detail and moreinsight into the procedures described in the acute Phase III confirmation document(EPA, 1989D). Procedures to confirm that all toxicants have been correctly identifiedare given and specific changes for methods applicable to chronic toxicity are included.Adifficult aspect of confirmation occurs when toxicants are not additive, and thereforethe effects of effluent matrix affecting the toxicants are discussed. The same basictechniques (correlation, symptoms, relative species sensitivity, spiking, and massbalance) are still used to confirm toxicants and case examples are provided toillustrate some of the Phase III procedures. Procedures that describe the techniquesto characterize the acute or chronic toxicity (EPA, 1988) and to identify (EPA, 1989A)toxicants have also been rewritten (EPA, 1991 A; EPA, 1992; EPA, 1993A).

-

iv

c

Contents

PageForeword ....................................................................................................................... iiiAbstract ......................................................................................................................... ivTables ............................................................................................................................ viFigures ......................................................................................................................... .viAcknowledgments ....................................................................................................... .vii

1.

2.

3.

4.

5.

6.

7.

8.

9.

10.

11.

12.

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-l

Correlation Approach ......................................................................................... 2-l2.1 Correlation ................................................................................................ ..2- 12.2 Correlation Problems Caused by Matrix Effects ........................................ .2-4

Symptom Approach.. ........................................................................................ ..3- 1

Species Sensitivity Approach ............................................................................ .4-l

Spiking Approach .............................................................................................. .5- 1

Mass Balance Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-l

Deletion Approach.. ........................................................................................... .7-l

Additional Approaches ...................................................................................... 8-I

Hidden Toxicants .............................................................................................. .9-l

Conclusions ...................................................................................................... 1 o-1

When the Treatability Approach Has Been Used . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11-1

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12-1

, ,

Tables

N u m b e r Page

6-l. Comparison of effluent toxicity and toxicity measured in effluentfraction add-back tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-2

L

Figures

Number

2-l.

2-2.

2-3.

2-4.

2-5.

Page

Correlation of toxic units (TUs) for an effluent and one suspect toxicantin a POTW effluent . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-2

Correlation of toxic units (TUs) for an effluent and one suspect toxicantin a POTW effluent when two toxicants are the cause of toxicity . . . . . . . . . . . . . . . . . 2-2

Correlation of toxic units (TUs) for an effluent and two toxicants in aPOTW effluent . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-3

Correct (top) and incorrect (bottom) plots of toxic units (TUs) for non-additive toxicants . . . . . . . . . . . . . . . ..*........................................................................ 2-4

Correlation of toxic units (TUs) for a POTW effluent and the suspecttoxicant, nickel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-5

vi

6

Acknowledgments

This document presents additional information acquired since the documententitled Methods for Aquatic Toxicity Identification Evaluations: Phase III ToxicityConfirmation Procedures (EPA-60013-88-036; EPA, 1989D) was prepared by DonaldMount and published in 1989. This manual reflects new information, techniques, andsuggestions made since the Phase III confirmation methods for acute toxicity weredeveloped. The suggestions, techniques and cautions contained in this document arebased on a large database generated by the staff of the National Effluent ToxicityAssessment Center (NETAC) at the U.S. Environmental Protection Agency (EPA),Environmental Research Laboratory, Duluth (ERL-D), MN. NETAC staff that providedtechnical support consisted of Penny Juenemann and Shaneen Schmidt (ERL-Dstaff), Joe Amato, Lara Anderson, Steve Baker, Tim Dawson, Nola Englehorn, DougJensen, Correne Jenson, Jim Jenson, Elizabeth Makynen, Phil Monson, GregPeterson, and Jo Thompson (contract staff). Their collective experience has madethis document possible and the contributions are gratefully acknowledged. Thesupport through EPA’s Office of Research and Development (ORD) and Office ofWater made this research possible at ERL-D.

L

Vii

.

Section 11Introduction

The final confirmation phase of a toxicity identifi-cation evaluation (TIE) consists of a group of stepsintended to confirm that the suspect cause(s) of toxicity iscorrectly identified and that all the toxicity is accountedfor. Typically this confirmation step follows experimentsfrom the toxicity characterization step (Phase I) andanalysis and additional experiments conducted in toxicityidentification (Phase II) (EPA, 1991A; EPA, 1992; EPA,1993A). However, there often may be no identifiableboundary between phases. In fact, all three phases mightbe underway concurrently with each effluent sample anddepending on the results of Phase I characterization, thePhase II identification, and Phase III confirmation activi-ties might begin with the first sample evaluated. Phase Illconfirmation procedures should also follow after toxi-cants have been identified by other means or whentreatability approaches are used. Rarely does one step orone test conclusively prove the cause of toxicity in PhaseIII. Rather, all practical approaches are used to providethe weight of evidence that the cause of toxicity has beenidentified. The various approaches that are often usefulin providing that weight of evidence consist of correlation,observation of symptoms, relative species sensitivity,spiking, mass balance estimates and various adjust-ments of water quality.

The approaches described in this document havebeen useful in TIES at ERL-D. While the guidance pro-vided in this manual is based largely on experience withwastewater effluents, in general the methods discussedare applicable to ambient waters (Norberg-King et al.,1991) and sediment pore or elutriate water samples aswell (EPA, 19918). However, specific modifications ofthe TIE techniques might be needed (e.g., sample vol-ume) when evaluating these other types of samples.

Confirmation is important to provide data to provethat the suspect toxicant is the cause of toxicity in aseries of samples and to assure that all other toxicantsare identified that might occur in any sample over time.There may be a tendency to assume that toxicity isalways caused by the same constituents, and if thisassumption carries over into the data interpretation butthe assumption is false, erroneous conclusions might be

reached. That is why the correlation step (Section 2) isaccompanied by other approaches (i.e., Sections 3-9)because each approach aids in revealing any changes inthe toxicant in the confirmation phase of the TIE.

Seasonal trends in toxicants have been observedin publicly owned treatment works (POTW) effluents andsome sediment samples. For example, organophosphatepesticides have been observed to increase in concentra-tions in wastewaters during the late winter and springmonths (Norberg-King et al., 1989). Therefore, the confir-mation steps of Phase III might need to include seasonalsamples. This effort cannot always be pre-determined.The presence of a different toxicant must be consid-ered throughout the TIE, and when samples are collectedover several months the seasonality of a suspect toxicantshould be carefully considered and studied. When reme-dial action requires treatment changes, one must becertain that toxicity from specific toxicant is consistentlypresent and that the suspect toxicant accounts for allthe toxicity. Treatment modifications will not necessarilyresult in removal of all toxicants to acceptable concentra-tions. If toxicity is caused by a variety of toxicants presentat varying intervals, the remedial actions that are practicalmight differ from the remedial action required when toxic-ity is caused by the same constituents consistently.

cTIES conducted at ERL-D have shown that toxi-

cants often are not additive or toxicants are present inratios such that the toxicity contribution by one might bediluted out in the range of the effluent effect concentration(e.g., LC50 or ICp value). Thus, the toxicant present atlower yet toxic concentrations may not be readily dis-cerned. The frequency of occurrence and impact on datainterpretation of either of the above cases was not ad-dressed previously (EPA, 1989D) but are now discussedin Section 2. Toxicants that do not express their toxicitybecause of the presence of other toxicants (either thetoxicants are non-additive or the toxicants occur in dispar-ate ratios) are referred to as hidden toxicants (Section 9).Detection of hidden toxicants is one of the most difficultaspects of confirmation. It is a mistake to search for aconcentration of any chemical present in the effluent at atoxic concentration and to declare any found as the cause

l - l

of toxicity. Matrix effects of the effluent samples makeconclusions such as these subject to error without furtherwork as either the hidden toxicant or the principaltoxicant are likely to be missed using such an ap-proach.

There is a strong tendency to shorten or eliminatethe confirmation steps because by the time Phase IIIconfirmation has been reached, the investigators mightbe convinced of the cause of toxicity and the confirmationsteps seem redundant. However, one cannot expect toconcentrate the effluent on a C,, solid phase extraction(SPE) column and not change a complex mixture such aseffluents, and arrive at some false conclusions about thetoxicants in the earlier phases.

Not all approaches discussed in the followingsections will be applicable to every effluent, and addi-tional approaches might need to be developed during theTIE. The various approaches need not be performed inany particular sequence, and the list of possible ap-proaches will get larger as experience is gained. Toeffectively evaluate effluent samples from one particulardischarger to obtain a correlation, substantial calendartime could be required and any steps for correlationshould be initiated at the beginning stages of Phase III.Judgement must be made as to how many of the ap-

proaches described in Phase III confirmation should beused and how many samples for each should he CXKRpleted. How completely Phase III confirmation is done willdetermine the authenticity of the outcome. The amount ofconfidence in the results of the TIE that is required isdependent at least in part on the significance of thedecision that will be based on the results. For example, ifa suspect toxicant can be removed by pretreatment or bya process substitution, a higher degree of uncertaintymay be acceptable than if an expensive treatment plant isto be built. Such considerations are subjective and cannotbe reduced to a single recommended decision makingprocess with a specified number of samples.

Time and resources might be conserved if identi-fication (Phase II) and confirmation (Phase Ill) can bestarted on the very first effluent sample used in the Phas’eI characterization. However, this is only possible when theresults from the Phase I characterization are definitiveenough to allow the investigators to proceed to identifica-tion and confirmation. In the acute Phase III confirmationdocument (EPA, 1989D), although perhaps not explicitlystated, performing Phase I characterizations on severalsamples before attempting Phases II and III was implied.Initiating the Phase Ill confirmation steps earlier in the TIEis often particularly useful. In addition, many regulatoryagencies have adopted a policy that requires that theprevious TIE approach be modified. For some discharg-ers, action might be required after the first exceedence intoxicity, which means that each effluent sample collectedfor toxicity testing is of equal regulatory concern when thetoxicity is greater than the permit allows. This regulatory

practice was not in place in 1989 when the earlier TIEguidance was available (EPA, 1989D) and at that time wedid not expect that the cause of toxicity in one samplecould be sufficiently deduced as we have been able to do.The importance of confirmation on several samples is notreduced by the importance of conducting confirmationsteps on single samples; rather, the cause of toxicity foreach sample must be confirmed.

In addition to the importance of each sample withtoxicity greater than the allowable amount specified in apermit, a sample that is quite different from the previoussamples must be evaluated to determine if the data pointmust be included in the Phase III correlation final dataanalyses. For each effluent sample, the data points mustbe explainable. If one sample is quite different than othersamples it can cause the correlation to be less useful;however, if it can be shown to have a different toxicant thedata point for that sample can be eliminated from thecorrelation. For example, suppose five consecutivesamples during a Phase III evaluation exhibited toxicitythat correlated well with a suspect toxicant. Then a sixthsample exhibits greater toxicity than previous sampleswhile the measured concentration of the suspect toxicantis much lower than measurements on previous samples.In this sixth sample, the greater toxicity is thought to becaused by a different toxicant. Now in plotting the data forthe correlation (Section 2), the datum point for the sixth%3XV$p, \4ciIl, r;ryJ, & dW;a? +fi +& -j33rrh3 50-t tire ex’rst’rngregression and could render the correlation non-signifi-cant. If however, when the sixth sample is then subjectedto intensive study using Phase I characterization andPhase II identification techniques, and if another toxicantis identified (or even if Phase I only shows that the toxicityhas very different characteristics), datum for the sixthsample can legitimately be excluded from the correlation.This preserves the worth of the data for the previous fivesamples. In confirmation, every effort should be made todetermine why a particular sample shows different re-sponses in the various TIE steps from other samples.

This is not to imply that multiple effluent samplesneed not be subjected to Phase I manipulations, even ifPhase II and/or Phase III are initiated on the first sample.Most effluent samples tend to be representative of theroutine effluent discharge. However, determining what isthe characteristic discharge for each effluent is importantto the final success and completeness of the TIE.

When Phase III is completed, all results that wereobtained during the TIE should be explainable. Unless theresults make sense for all samples (aside from an occa-sional aberrant data point) something has been missed oris wrong. If so, the confirmation is not complete. Manytechniques used in Phase III require keen observationsand extensive or broad knowledge of both chemistry andtoxicology but above all the ability to synthesize small bitsof evidence in a logical sequence is essential. This TIEwork is most effective when scientists interact daily.

l-2

A note of caution. If data obtained on early samplesduring Phase I are to be used for Phase III purposes,quality control will have to be suitable to provide defen-sible data (cf., EPA, 1991A; EPA, 1992; EPA, 1993A). InPhases I and II, the permissibility of using small numbersof animals and replicates, and omitting measurementssuch as pH, DO, and temperature that are required forroutine monitoring tests or single chemical tests wasdiscussed (EPA, 1989E; EPA, 1991 A; EPA, 1992; EPA,1993A). These modifications were made to reduce costand allow more testing, but at this point shortcuts must beavoided because definitive data that constitute the basisfor important decisions are generated in Phase III. ForPhase III testing, the effluent test protocols that triggeredthe TIE (EPA, 199X; EPA, 1993B) should be followed,paying careful attention to test conditions, replicates,quality of test animals, representativeness of the effluentsamples tested, and strict QA/QC analytical proceduresincluding blanks and recovery measurements. Analyticalwork must be selective for the identity of the toxicant andits concentration measurement. When small differencesin toxicity must be detected, concentration intervals shouldbe smaller to obtain partial effects (e.g., use dilutionfactors of 0.60 or 0.65 versus 0.5). Remember, all of thedata from Phases I and II (for either acute or chronictoxicity) are considered preliminary relative to Phase IIIdata. However, if a suspect toxicant is identified andPhases I and II data may be necessary for confirmation,stricter QA/QC can be applied for each of the subsequentPhases I and II techniques so that the data can be used inPhase III.

For samples exhibiting chronic toxicity, modifica-tions or ‘changes to some of the TIE procedures arerequired for confirming the cause of chronic toxicity. Re-member that for confirmation (as well as for Phases I andII), only a single sample of effluent should be used foreach renewal in any chronic test (cf., EPA, 1992; EPA,1993A). This is important because one cannot correlate ameasured concentration of a toxicant with the toxicitymeasured in a test if multiple samples are used for each

-

renewal and the toxicant is not present in some samplesbut other toxicants appear. Even more likely, the ratios ofthe toxicants, when more than one is present, mightchange from sample to sample. In these instances, thereis no valid way to calculate the toxicity of a given toxicant.Overall, considerations for chronic toxicity tests in PhaseIII are not much different than acute toxicity tests in PhaseIII. At present, permit requirements specify the 7-d testand unless data are gathered to show that the 4-d and 7-d tests yield the same results and that the same toxicantsare involved, the 7-d test should be used for confirmation(cf., EPA, 1993A). If the 4-d Ceriodaphnia dubia test hasbeen used instead of the 7-d C. dubia test (see EPA,1992) during Phases I and II, serious consideration shouldbe given to returning to the 7-d test for Phase III.

When identification of the toxicant causingchronic toxicity is desired, and the effluent also exhibitsacute toxicity, it might be possible to use acute toxicity asa surrogate measure to characterize the toxicity in PhaseI and assist in an identification in Phase II. It must bedemonstrated that the cause of the acute toxicity is thesame toxicant as the toxicant causing the chronictoxicity. Yet for confirmation, use of chronic toxicity end-points to confirm the cause of the chronic toxicity isstrongly recommended to avoid misleading the TIE re-sults when using acute toxicity as a surrogate for chronictoxicity. As discussed in the chronic Phase I manual(Section 5.8; EPA, 1992), effect levels for chronic testsshould be calculated using the linear interpolation methodrather than the hypothesis test (EPA, 1992). In order toget more precise estimates of endpoints, test concentra-tion intervals might have to be narrowed (see above).However, when point estimation techniques for other thansurvival endpoints (such as the inhibition concentration(ICp); EPA, 1993B) are used, a point estimate effectconcentration can be estimated. The effect concentrationestimates will also be more accurate when intermediateconcentrations are used (i.e., use dilution factors of 0.6 or0.65).

1-3

Section 2Correlation Approach

2.1 Correlation

The purpose of the correlation approach is toshow whether or not there is a consistent relationshipbetween the concentration of suspect toxicant andeffluent toxicity. For the correlation approach to be useful,the toxicity test results with the effluent must demonstratea wide range of toxicity with several effluent samples toprovide an adequate range of effect concentrations forthe regression analysis. For sediment samples, spatialvariability might be used to perform correlation analyses(EPA, 19916).

The effluent effect concentration (i.e., LC50 orICp) data and the measured toxicant concentration datamust be transformed to toxic units (TUs) for the regres-sion analysis to evaluate whether or not a linear relation-ship exists. Effluent TUs are obtained by dividing 100%by the effect concentration expressed in percent of theeffluent (cf., EPA, 199lA; EPA, 1992). The suspect toxi-cant concentration is converted to TUs by dividing themeasured toxicant concentration by the LC50 or ICp forthat toxicant (data to make this comparison might have tobe generated; EPA, 1993A). If more than one toxicant ispresent, the concentration of each one is divided by therespective LC50 or ICp value and the TUs can then besummed (cf., discussion below for non-additive toxicants).

Most of the effluents we have tested have exhib-ited a wide range of toxicity with several different samplesand therefore the data can be used in the correlationapproach. Typically for the correlations that .we haveconducted, the data used are from toxicity tests withoutany manipulations and from chemical measurements onthe effluent samples for the concentrations of the suspecttoxicant. However for effluents where ammonia was thecause of the toxicity, the effluent toxicity. results have notvaried in toxicity enough, nor have the ammonia concen-trations fluctuated enough to use the data in a correlation.Also, when the effect concentration is greater than 1 OO%,this information is not useful since the data point cannotbe included in the regression analysis. However, whensamples are marginally toxic or when the suspect toxicantconcentrations do not vary enough from sample to sample(i.e., ammonia is cause of toxicity), changes in toxicity canbe induced by sample manipulation (cf., EPA, 1993A) andthis toxicity data can be used to develop a different type ofcorrelation. For example, the toxicity of a given amount of

total ammonia can be changed by over an order ofmagnitude by altering the pH of aliquots of the effluentwithin an acceptable physiological range (e.g, pH 6 to 9).For some metals and some species, the toxicity can alsobe changed by adjusting the pH and using dilution watersof varying hardness. This type of data is useful in thecorrelation step as providing additional weight of evi-dence. Therefore, the idea of minimal manipulation(s)and any risk of creating artifactual toxicity are off set by theutility of the data.

An example of the regression from an effluentfrom a POTW in which the suspect toxicant was diazinonis given in Figure 2-l. The independent variable (x-axis) isthe TUs of diazinon and the dependent variable (y-axis) isthe effluent TUs. The solid line is the observed regressionline obtained from the data points, and the dashed line isthe expected or theoretical regression line. If there is 1 .OTU of the toxicant in 100% effluent, then the effluentshould have 1 .O TU (i.e., the LC50 =lOO%). Likewise for2.0 TUs of suspect toxicant, the effluent TUs should be2.0, et cetera. Thus, the expected line has a slope of oneand an intercept of zero. In Figure 2-1, the intercept (0.19)is not significantly different from zero and the slope is veryclose to 1 (1.05). The P value is 0.63 which, while nothigh, indicates that the majority of the effluent toxicity isexplained by the concentration of the toxicant. As the r2becomes lower, less confidence can be placed on slopeand intercept. In a small data set such as this, one datumpoint that had 5.0 TUs for the effluent toxicity lowered ther2 value substantially. As discussed in Section 1, if anintensive effort had been expended on that sixth sampleand another toxicant had been found, this particulardatum point could have been excluded and the P valuewould have been higher.

In another POTW effluent, diazinon was also thesuspect toxicant. For these data (Figure 2-2) the slope is1.38, the intercept is 1.24 and the r2 value is only 0.15,which all indicate poor fit for diazinon as the only toxicant.The low r2 value indicates a large amount of scatter,therefore little can be inferred from the slope and theintercept. Based on this correlation, we returned to PhaseII analytical procedures and identified two other organo-phosphates (chlorfenvinphos (CVP) and malathion). Tox-icity data indicated that CVP was present at toxic

2-l

1 2 3 4 5 6TUs of Suspect Toxicant

Figure 2-1. Correlation of toxic units (TUs) for an effluent and onesuspect toxicant in POTW effluent.

00

00

00

1.24

Figure 2-2.

1 2 3 4 5 6TUs of Suspect Toxicarlt(s)

Correlation of toxic units iTUs) for an effluent and cnesuspect toxicanr in a POTW effluent when two toxicantsare the cause of toxicity.

concentrations while malathion was no?. After testingeach compound both separately and as a mixture, thetoxicity from all three chemicals was determined to beadditive, so a new correlation was begun with analyticalmeasurements made for all three chemicals. CVP anddiazinon have nearly identical LC50 values for the spe-cies (C. dub@ used in this TIE. Malathion is about one-

.fourth as toxic as CVP or diazinon. Since the measuredconcentrations of malathion were lower than its toxicity, itwas not included in the regression analysis. In a newcorrelation with data for the TUs summed for CVP anddiazinon versus the effluent TUs, the data show a muchbetter fit to the expected slope and intercept and a high r2value (Figure Z-3). Malathion TUs could also have beenincluded in the regression (although its contribution totoxicity was minimal) because it was additive with othertoxicants. This type of situation is discussed below.

In addition to slope and intercept, some judge-ment of the scatter about the regression line must bemade. This can be done statistically, but when the samplesize is large, the scatter can be very large and yet notnegate the relationship. A-suggested approach to avoidthe effect of sample size on the significance of scatter isto set a lower limit on r2. This value (often expressed aspercent) provides the measure of how much of the ob-served effluent toxicity is correlated to the measuredtoxicant. It is not dependent on choosing the correct effectconcentration of the toxicant. The specific choice of theminimum value of r2 should be made based upon theconsequences of the decision. It is important to recognizethat experimental error makes an r2 value greater th2r-r0.80 or 0.85 difficult to obtain. Therefore, where minimalchance of an incorrect decision is required, an r2 value ofnearly 0.80 may be used. Where an increased risk of anincorrect decision (i.e., a lesser amount of the toxicityaccounted for) is acceptable, a lower value such as 0.60may be used.

Since <I .O TU cannot be directly measured in theeffluent, such values are, of necessity, excluded from theregression. (This comment is exclusive of the use ofconcentrates such as the C,, SPE fractions’ where TUs of~1 .O are possible.} However in some instances, when theTUs based on chemical analyses are ~1 .O TU and efflu-ent effect values are ~1 .O TU, the data support the validityof the regression provided a suspect toxicant has beenfound in several previous samples. In the correlation forthe effluent toxicity depicted in Figure 2-2, toxicity waspresent in a different fraction (Phase II non-polar organicidentification) than where the pesticides were identified. Aspecific toxicant was not identified in that fraction andtoxicity was not always measurable in that fraction. How-ever, this additional toxicity may have decreased the r2value.

Correlation might be more definitive when two ormore toxicants are present. For example, suppose threetoxicants are involved. If each toxicant has the sameLC50 and each is strictly additive with the ratio of theirconcentrations remaining the same, the slope will be theexpected but the intercept will be positive if all toxicants

ITUs can be calculated from toxicity tests with the fractions, theconcentrate or the HPLC fractions as described in Phase II (EPA,’1993A).

2-2

.

slope = 0.82y-intercept = 0.46

4H

0” ” ” ” ” ”0 1s 2 3 4 5 6

TUs of Suspect Toxicant

Figure 2-3. Correl?tion of toxic units (TUs) for an effluent and twotoxicants in a POTW effluent.

are not identified. If the relative amounts (ratios) of eachtoxicant vary from sample to sample, the slope, interceptand r2 will be different from the expected if only onetoxicant is identified. If the toxicity of one of the toxicantsis substantially different, and if the ratios of the threetoxicants vary from sample to sample, then the slope,intercept, and r2 value will all be different from expected ifall are not identified. Much can be learned from studyingthe interrelationship of slope, intercept and the r2 value.For example, a high r2 value and an intercept near zerowith a slope larger than 1 can be caused by using aneffect concentration for the toxicant that is not appropriatefor the toxicant in the effluent matrix (e.g., suspect toxi-cant is more toxic in effluent matrix than in single chemi-cal test). This error causes the toxicant TUs to be too fewrelative to the effluent TUs (Figure 2-4) (cf., discussionbelow on non-additive toxicants). If toxicant concentra-tions and effluent toxicity show a wide distribution, asignificant correlation will be easier to demonstrate thanfor a narrow range.

Great care must be taken to understand whetheror not toxicants are additive or if the TUs for each toxicantare so different that only one toxicant determines theeffect level. For either situation, the resulting data willhave to be interpreted as though the toxicants are non-additive. For example, suppose the ratio of TUs is sodisparate that at the effluent effect concentration, thetoxicant with fewer TUs is always present at a fraction of aTU (e.g., 0.25 of a TU). Whether the two toxicants areadditive or not is irrelevant because the major toxicant willset the effluent effect concentration. While 0.25 TUs of

the minor toxicant appear to be relatively unimportant inview of experimental variability, this affects the regres-sion. If in one sample the effect concentration is 25% andthe 4 to 1 ratio of toxicants occurs, there are 4 TUs of themajor toxicant and 1 TU of the minor toxicant. If thetoxicant concentrations are summed, 5 TUs will be plottedagainst 4 effluent TUs, and this results in a 25% error.When secondary toxicants are present in concentrationsthat will not contribute to the effect concentration of theeffluent, they should not be included in the correlationdata set. Obviously if an effluent had several toxicants indissimilar ratios, the error of including the minor TUs in acorrelation plot could be large and may negate the corre-lation significance. The investigator should evaluate thedata in regression plots to consider the significance of thecontribution of the secondary toxicant especially if thetoxicants appear to be additive.

Unfortunately the minimum fraction of a TU that isdetectable will depend on the precision of the laboratoryperforming the testing. And of course the precision of thetesting is not only dependent on the quality of the work,but the inherent precision of measuring specific toxicantTUs. That is, the toxicity measurement for some chemi-cals is more precise than for some other chemicals. Ingeneral, a chemical such as NaCl whose toxicity is gener-ally not affected by pH, alkalinity, hardness, total organiccarbon (TOC), suspended solids or solubility, can bemeasured more precisely than a chemical whose toxicityis affected by these factors, such as lead or copper.Therefore, each laboratory must determine which frac-tional value of a TU at the effect concentration isunmeasurable, thus indicating which TUs contributed bythe minor toxicant should be deleted from the correlationdata set.

Clearly, if two or more toxicants are strictly non-additive, then only the major one (the one present in themost TUs) should be included in the correlation data set.Since additivity might be easier to measure than theminimum measurable contribution of a fraction of a TU, itmay be preferable to first determine if additivity occurs. Ifsubstances appear to be partially additive, then verycareful work is required to properly add TUs.

Some very unusual decisions are required inaccepting data into the correlation database when toxi-cants are strictly non-additive. For example, consider zincand ammonia in the same effluent sample; we have fcundthem to be strictly non-additive. Also consider that insome samples zinc and ammonia occur in TU ratios of 3to 1 and in other samples the ratio is 1 to 2. In theregression for the 3 to 1 ratio samples, only zinc TUsshould be plotted. In the regression for the 1 to 2 ratiosamples, only ammonia TUs should be plotted. For thisparticular example, 3 TUs for the first sample and 2 TUsfor the second sample would be used if the data isinterpreted correctly (i.e., plotting total TUs) or 4 and 3TUs would be used respectively, if the data is interpreted

4Correct Plot

I I I I1 2 3 4

Calculated TUs

Incorrect Plot

1 I I I1 2 3 4

Calculated TUs

Figure 2-4. Correct (top) and incorrect (bottom) plots of toxic units(TUs) for non-additive toxicants.

incorrectly. The slopes for both plots would be 1 but anegative intercept instead of an intercept of 0 would beobtained for the incorrect plot. The more similar the TUsof each toxicant are to each other, the greater the error inthe correlation will be.

2.2 Correlation Problems Caused by MatrixEffectsCorrelation becomes much more difficult when

the toxicants interact with the other effluent constituentsin ways that change their toxicity and we refer to thesechanges as matrix effecti. There are numerous matrixeffects and all of them will not be discussed here; instead

2-4

.a framework is provided to aid in designing tests or testconditions to validly incorporate matrix effects in such amanner that useable correlation data can be obtained.

Matrix effects generally fit into one of two catego-ries. One category is when the toxicants change form insome manner which exhibit a different toxicity. A verycommon example is ammonia which changes from NH, toNH,+ as pH decreases. NH,+ is so much less toxic thanNH, that it is often considered nontoxic? Another exampleis HCN whose most toxic form is as un-dissociated HCN,a form predominating at low pH values. As pH increasesthe equilibrium shifts to more H+ and CN-. If metals arepresent, metal-cyanide complexes form which are oftenless toxic than HCN but metal-cyanide complexes mightvary in toxicity depending on the metal. For example, iron-cyanide complexes are much less toxic than some of theother metal complexes. Metal-cyanide complexes mightalso photodecompose in sunlight releasing HCN or H+and CN-, depending on PH.

A second category of matrix effects involves suchphysical changes as sorption or binding in some mannerso as to make the toxicant unavailable to the organism.For example, non-polar organics sorb onto suspendedsolids, and some metals, such as copper, also sorb o%osuspended solids. The presence of organic matter onsuspended solids might increase the sorbtive capacity.Predictably, changes in water chemistry often change thesorption/solution equilibrium and thereby, change the por-tion of total toxicant that is available to the organism.

To further complicate matters, biological charac-teristics of the test organisms might change the availabil-ity of the same toxicant form. For example a non-polarorganic sorbed on suspended solids such as bacterialcells, might be unavailable to a fish but readily available todaphnids because cells might be ingested and digestedby daphnids. The uptake route then is through the diges-tive tract but the toxicant has entered the body none-the-less.

From the above discussion, it is obvious that onemethod of correlation will not be applicable for all toxi-cants. A temptation may be to remove the toxicant fromthe effluent and then use the effluent as a diluent tomeasure toxicity. However, because effluents are so com-plex and undefined, there is virtually no way to removeone or a few constituents and still be certain other charac-teristics have not been changed. For example, zeoliteremoves ammonia but it also removes some metals andnon-polar organics; the C,, resin removes metals as wellas non-polar organics; Ion exchange columns removeionized constituents, but non-polar organics also are re-tained by the columns. Toxicant removal procedures haveutility but require very complicated simultaneous testingof the effluent and proper blanks (cf., EPA, 1992; EPA,

*See specific discussion in Section 3, Phase II (EPA, 1993A).

.

5 01993A) is necessary to properly interpret results (cf.,Section 9 on hidden toxicants).

In Phase III, quantitative comparisons are beingmade between toxicity and concentrations of toxicantsrather than qualitative comparisons as in Phases I and II(EPA, 199lA; EPA, 1992; EPA, 1993A). In the correlationapproach, such comparisons are the essence of thetechnique. Therefore even small changes in form or avail-ability might be unacceptable. This means that manipula-tions and changes must be minimized when effluenttoxicity and toxicant concentrations are to be compared.

Solvent extraction, so commonly used for organicanalyses, is likely to extract biologically unavailable or-ganics as well as soluble forms. The total measuredconcentration may be larger than the true exposure con-centration. Use of the C, SPE column also is not freefrom problems as the C,, SPE column is a finer filter thanthe glass fiber filters commonly used for pre-columnfiltration. Therefore solids are likely to be physically re-tained on the upper part of the column. When the columnis eluted with methanol, the methanol extracts toxicantfrom the solids (which might not be biologically available)as well as elutes the C,, sorbent itself. For Phases I andII, this might be unimportant, but for the Phase III correla-tion step where careful quantitative comparison is neces-sary, the effect might be unacceptable. Such problemsprobably reach a maximum when working with samplessuch as highly organic sediment pore water (with highorganic characteristics) where much of the chemical mightbe biologically unavailable.

The central problem for either type of matrixeffect is the difficulty of analytically measuring the biologi-cally available portion of the specific toxic form. A correla-tion for a POTW effluent where for nickel was suspectedof causing the toxicity is shown in Figure 2-5. DuringPhase I, the acute toxicity was removed with EDTAadditions, and in Phase II the nickel was measured attoxic concentrations to C. dubia. The toxicity correlatedvery well with total nickel concentration (r2 = 0.89 and aslope of 1 .17) and it appeared that only nickel seems tobe involved. But the intercept of -12.34 is quite differentfrom the expected zero. Such an intercept would beexpected if there were a relatively fixed amount of nickelwhich was not biologically available in all samples. In thisexample, because all other confirmation data corrobo-rated nickel as the toxicant, a constant concentration ofnontoxic nickel was thought to provide the explanation forthe unexpected intercept value. However, there is noobvious reason to think that the quantity, or even thepercentage of total toxicant, is the same across samplesfor other toxicants, or for nickel in other matrices.

For the effluent samples that lose their toxicity ina short time, the nontoxic effluent can be used for thesuspect toxicant tests as a diluent in parallel testsusing a standard dilution water to elucidate matrix effectson toxicity. Toxicity test results with quite different toxicitywould reflect matrix effects. If toxicity is persistent, devel-

40

10

y-intercept = -12.34

0 1010 20 30 40TUs of Suspect Toxicant (Nickel)

Figure 2-5. Correlation of toxic units (TUs) for a POTW effluent andthe suspect toxicant, nickel.

oping two separate correlations using pure cheplical addi-tions on two different effluent samples, each with sub-stantially different toxicant concentrations, might be useful.If the toxicity test results indicate that the biologicallyunavailable portion changes with measured concentra-tions, the slope should be different than one. This ap-proach requires careful work and the investigator mustconsider incorporating equilibrium time experiments (cf.,EPA, 1993A).

Metals can be especially difficult toxicants toimplicate using correlation because the toxicity of metalsis typically very matrix dependent. When the knowledgeof these characteristics is extensive for a chemical, as it iswith ammonia (see Phase II), testing can be tailored tothe chemical and a very powerful correlation obtained.The large amount of available information on ammoniadoes not exist for most metals. In these instances, thelogic pattern should to be reversed where the approachhas to become: if x is the toxicant, what are the matrixeffects?. These can be found by pure chemical testingcombined with Phases I or II manipulations. Once anadequate understanding of matrix effects is obtained, theinformation can be used to answer the question: Is theeffluent toxicant behavior consistent with the matrix ef-fects for the suspect toxicant?

Matrix effects will have varying impacts on toxi-cant behavior that also depends on the effluent effectconcentration. For effluents which have effect concentra-tions in the ~10% range, the test solutions will moreclosely resemble the diluent water matrix than the efflu-ent. If the effluent has effect concentrations in the 50% to100% range, the matrix effects of the test solution willmost likely resemble those of the effluent, not of thedilution water. Since effluent TUs are calculated from

2-5-

responses occurring in the dilution near the effect con-centration, the matrix characteristics of that concentrationare of the most concern for correlation. Thus the impor-tance of the effluent matrix effects diminishes as thetoxicity of the effluent is greater (i.e., matrix at effect levelis more like dilution water).

One can safely say that the difficulty of simulatingthe matrix effects with a simulated effluent is quite largeso that the choice is clearly to use the actual effluentwhen possible. An important reason for this choice is thatso few matrix effects have been studied extensively, andbeyond pH and hardness little data exists. Even then theinterrelationship between pH, alkalinity and hardness wereoften ignored.

The above discussion does not provide all of theoptions on how to handle matrix effects. However, it

.

should provide convincing evidence that more than thecorrelation step alone is necessary to provide adequateconfirmation!

In summary, the TIE research experience hasrevealed two major areas of potential problems in usingthe correlation approach. The lack of additivity for toxi-cants found in effluents requires careful analysis whencalculating TUs for regression purposes. Secondly, whenthere are matrix effects, correlation becomes difficult be-cause the effluent matrix might change from sample tosample and because there are no analyses specific forthe toxic forms. For such effluents, other confirmationtechniques should be used more extensively to bettersupport the overall confirmatory efforts.

c

2-6

Section 3Symptom Approach

Different chemicals may produce similar or verydifferent symptoms in a test species. Probably no symp-tom of intoxication is unique to only one chemical. There-fore, while similar symptoms observed between twosamples means the toxicant could be the same ordifferent, different symptoms means the toxicant isdefinitely different, or there are multiple toxicants in thetwo samples. By observing the symptoms displayed bythe test organisms in the effluent and comparing them tothe symptoms displayed by test organisms exposed tothe suspect toxicants, failure to display the same symp-toms means the suspect toxicant is probably not thetrue one or the only one.

Behavior of most test species is difficult to putinto words so that a clear image of behavior is obtained.Behavioral and morphological changes of 30-d old fatheadminnows (Pimephales promelas) were used as diagnos-tic endpoints in 96 h flow-through single chemical tests.Organic chemicals of various modes of action were testedand video recordings were used to monitor the behav-ioral response (Drummond et al., 1986; Drummond andRussom, 1990). Substances within a single chemicalclassification did not necessarily cause the same type ofresponse (Drummond and Russom, 1990). Therefore, itis difficult to predict chemical classification using behav-ioral monitoring alone.

This type of behavioral monitoring data does notexist for the cladocerans or the newly hatched fatheadminnows or other species that are most frequently usedin the TIE process. However, noting various symptoms isuseful in the TIE. This is done by simply exposing the testspecies to the suspect toxicant and observing howthey react. By the time confirmation is initiated, toxicitytests with the suspect toxicants will have been conductedusing pure compounds and symptoms may have beenobserved. It is important to note the symptoms observedduring all testing because such characteristics can bevery helpful in confirmatory work.

The intensity of exposure concentrations mightchange the symptoms observed with the suspect toxicantin the effluent. Therefore, it is important to comparesymptoms at concentrations that require about the sameperiod of onset. This can be done by comparing symp-

toms at exposure concentrations that have similar TUs. Inthis way both the unknown (sample) and the knowntoxicants (pure compound) can be set at the same toxicitylevel.

Observations of the organisms should not bedelayed until the normal length of the test has elapsed.With some toxicants, the test organisms will show distinc-tive symptoms soon after the exposure begins, whereaslater, symptoms are often more generalized and lesshelpful. For some other toxicants, a sequence of differentsymptom types are displayed by the test organism overthe exposure period and the sequence may be moredefinitive for a given chemical than the individual symp-toms. In few cases will the symptoms be unique enoughto specifically identify the toxicant, but symptoms differentfrom those caused by the pure suspect toxicant areconvincing evidence that the suspect toxicant is not thetrue or only one.

A second caution is needed regarding mixtures oftoxicants. Mixtures of toxicants can produce symptoms intest animals different from the symptoms of the individualtoxicants comprising the mixture. When more than onetoxicant is involved, the investigator must not only includeall the toxicants, but include them in the same ratio asmeasured in the effluent. Often the toxicant of the mixtureat the highest concentration relative to its effect concen-tration will cause most of the symptoms. As for singletoxicants, the mixture concentration causing the sameendpoint in a similar exposure period should be com-pared. Spiking effluent with the suspect toxicants andcomparing the results of the spiked effluent sample andthe unspiked effluent sample toxicity tests, both near theireffect concentrations, is a good approach to take (Sec-tion 5).

Symptoms caused by the toxicant might bequite different among different species of organisms;therefore the use of two or more species provides in-creased definitiveness of the observations. For both spe-cies, the researcher must compare symptoms atconcentrations that are equitoxic. The greater the differ-ence in sensitivity, the more important this becomes, Thechemical concentration is unimportant; the important con-sideration is that equitoxic concentrations are compared.

3-l

Suppose, for example, species A and B have LC50values for a suspect toxicant of 1 and 80 mg/l. Thenconcentrations of 2 and 160 mg/l may be used to com-pare symptoms of species A and 9, respectively. If theonset of symptoms is rapid, then perhaps 1.25 and 100mg/l (1.25xLC50) should be tried. Since symptoms varywith the exposure intensity, using various multiples of theLC50 (i.e., 0.5, 1, 2x) can add additional confirmationdata, if the same set of symptoms are seen in both series.If more than one toxicant is involved, and the ratio of thetwo species’ LC50 values for toxicant A is markedlydifferent than for toxicant 9, C, D, . . . . then the definitive-ness of using symptoms is even greater.

For acute toxicity, time-to-mortality at equitoxicconcentrations can be used as a symptom type of test.

Some chemicals cause mortality quickly and some causemortality slowly. If for two effluent samples, toxicity isexpressed quickly for one and for the other very slowly,the toxicants are probably not the same.

In chronic testing, use of symptoms is also appli-cable. For example, adult mortality, number of young/female, death of young at birth, growth retardation, abor-tion, or time to onset of symptoms, all can also bemonitored and such observations may be useful. Theshape of the dose response curve may also be a determi-nant in assisting in confirmation. Some chemicals showan all or none type of response (diazinon) while others(i.e.. NaCI) display a relatively flat concentration-responseslope for chronic toxicity.

-

3-2

Section 4Species Sensitivity Approach

The effect concentrations can be compared forthe effluent of concern and the suspect toxicants, usingspecies of different sensitivities. If the suspect toxicantis the true one(s), the effect levels of effluent sampleswith different toxicity to one species will have the sameratio as for a second species of different sensitivity. Alsothe ratio for each species should be the same as forknown concentrations of the pure toxicant. The sameratio of effect values for two species implies the sametoxicant in both samples of effluent. Obtaining the sameeffluent toxicity ratio among various effluent samples foreach species as is obtained by exposure to comparableconcentrations of known toxicants, implies that the sus-pect toxicants are the actual ones present. However, ifother effluent characteristics affect toxicity and if theyvary, the ratios could also be affected.

The common notion that goldfish are resistant tomost toxicants and trout are sensitive to most toxicants isnot readily substantiated (AQUIRE, 1992). Many speciesare more sensitive to certain groups of toxicants thantrout. Of course, there are generalizations that can bemade. For example, sunfish (Centrarchids), frequentlyare much more resistant to metals than goldfish, min-nows, and daphnids (AQUIRE, 1992). Daphnids tend tobe more resistant to chlorinated hydrocarbon insecticidesthan many fish species and more sensitive to organo-phosphate insecticides (AQUIRE, 1992). These differ-ences must always be verified for the suspect toxicants;generalities can only be used as an initial guide tospecies selection. Sensitivity differences of 1 O-i 00x mayoccur in some chemical groups and not in others. Ifseveral toxicants are involved, interpreting the resultsand designing the ancillary experiments is more difficult.If successful, the power of the result for multiple toxicantsis much greater than for a single toxicant. The differencein sensitivity between Ceriodaphnia and fathead min-nows has, on several occasions, revealed either a changein the suspect toxicants present in a series of effluentsamples, or the presence of other toxicants in addition tothose suspected.

-

_

Comparison of sensitivity among species hasanother very important use. Some species may evidencetoxicity from an effluent constituent that the TIE testspecies did not. If this happens, then the above compari-son will be confused, but at least there will be a warningthat the suspect toxicant may not be the cause of toxicity.In order to determine what is happening, the investigatorshould step back to Phase II, and possibly step back toPhase I to characterize the additional toxicant and thenidentify the toxicant using the new species. A secondPhase III effort might be necessary for this toxicant andspecies. It is important not to assume that tb residentspecies have the same sensitivity as the TIE test species.Especially for freshwater discharges into saltwater thisconcern is critical when a saltwater organism triggeredthe TIE, because at present the techniques and proce-dures described in Phases I and II are most likely to bedone using freshwater organisms especially since theeffluent is freshwater. If the concern is for marine organ-isms and their protection cannot be assumed (cf., Section8, Phase I; EPA, 1991A), confirmation must be conductedwith marine organisms.

In chronic testing, chemical and physical condi-tions might differ more among tests on different speciesbecause food must be provided during the test period anddifferent foods are used for each species. For example,the final pH of fathead minnow 7-d tests might be lowerthan in acute fathead minnow tests and both are likely tobe lower than in Ceriodaphnia chronic tests due to greaterrespiration rates for fish than cladocerans and food in fishtests. If the investigation was to confirm ammonia toxicity,this pH difference could result in confusing results byshowing the Ceriodaphnia to be more sensitive than thefathead minnows when the reverse should be true (cf.,EPA, 1993A; Phase II). The above example illustratesreasons to maintain careful quality control in Phase IIIwork.

4-l

Section 5Spiking Approach

In spiking experiments, the concentration of thesuspect toxicant is increased in the effluent sampleand then toxicity is measured to see whether toxicity isincreased in proportion to the increase in concentration.While not conclusive, if toxicity increases proportionallyto an increase in concentration, considerable confidenceis gained about the true toxicant( Two principles formthe basis for this added confidence. To get a proportionalincrease in toxicity from the addition of the suspecttoxicant when it is in fact not the true toxicant, both thetrue and suspect toxicants would to have 1) very similartoxicity and 2) to be strictly additive. The probability ofboth of these coinciding by chance is small.

Removing the suspect toxicants from the effluentwithout removing other constituents or in some wayaltering the effluent is usually not possible. The inabilityto do this makes the task of establishing the true toxicityof the suspect toxicants in the effluent difficult. For manytoxicants, effluent characteristics, such as TOC, sus-pended solids, or hardness, affect the toxicity of a givenconcentration, Some characteristics, such as hardness,can be duplicated in a dilution water, but certainly notTOC or suspended solids because there are many typesof TOC and suspended solids, and generic measure-ments do not distinguish among the different types. Forexample, effluent TOC occurs as both dissolved andsuspended solids. In POTW effluents, the source of theTOC is likely to be largely from biological sources, bothplant and animal (e.g., bacteria) and bacteria are likely tomake up a large component of suspended solids. If therehave been recent storms, oily materials from stormwaterrunoff might be high. Simulating TOCs from such variablesources is next to impossible because TOC is not solelythe result of man-made organic chemicals. For sus-pended solids, shape, porosity, surface-to-volume ratio,charge and organic content (all or any), will impact sorp-tion characteristics. None of these qualities are mea-sured by the standard methods for measuring suspendedsolids nor can they be reproduced in a simulated effluent.

In a simple system, such as reconstituted softwater, it is reasonable to expect that for most chemicals adoubling of the chemical concentration will double thetoxicity, at least in the effect concentration range. If thesolubility of the toxicant is being approached or there are

effects from water characteristics such as suspendedsolids, then the toxicity might not double or conceivablycould more than double. For example, if a chemical with alarge n-0ctanoVwater partition coefficient (log P) is largelysorbed on solids, doubling the total concentration mightmore than double the toxicity because the added chemi-cal might remain in solution. Another important issue isthat equilibrium might not be established during the entiretest period and is probably unlikely to occur before thetest organisms are added. For example, in our TIE re-search, we found various surfactants sorb to solids andcan be removed by filtration (Ankley et al., 1996). In theseexperiments, however, filtration failed to remove surfac-tants immediately after they were spiked in an effluent butsurfactants were removed after a few days equilibriumtime. Other chemicals are likely to show similar behaviorin regard to equilibrium time.

If several toxicants are involved, then their inter-action (additivity, independent action, synergism) must bemeasured or otherwise included in the confirmation pro-cess (cf., Section 2). Since ratios might be as importantas concentration, the best way to spike when multipletoxicants are involved is to increase each toxicant by thesame number of TUs (e.g., by doubling each). In this waythe ratios of the toxicities remain constant.

The fact that two or more toxicants fail to showadditivity is useful evidence in confirmation. Interpretingspiking data might require a very high level of compe-tence in both toxicology and chemistry; otherwise thedata could be very misleading. Using more than onespecies of differing sensitivity is effective in adding confi-dence to the results. When matrix effects are compli-cated, other types of spiking can be done to reduce theeffects of the effluent matrix characteristics. If a methodexists for removing the toxicants from the effluent, suchas the C,, SPE procedures (EPA, 1993A), the extracts ormethanol fractions can be spiked with pure chemicals inaddition to spiking effluent, using the same principles asdescribed for effluents. The advantage in this approach isthat matrix characteristics such as suspended solids andTOC will be absent or much reduced and will not affectspiking experiments as much. The disadvantage is thatproof that the extracts or fractions contain the true toxi-cants must be generated. Some approaches for doing

5-1

,

this are given in Section 6. The use of the spiking ap-proach is especially applicable to fractions from the C,,

I SPE column or the high performance liquid chromatogra-phy (HPLC) column used for the isolation of non-polarorganics. In these procedures, the constituents are sepa-rated from much of the TOC, suspended solids andhardness, so that spiked additions might be strictly addi-tive where they might not be in the effluent. Suggestionsand precautions about ratios and all other previouslydiscussed concerns apply here too. In addition, concernsabout the methanol percentages in the toxicity tests, theamount of SPE or HPLC eluate required for the toxicitytests and the issue of toxicity enhancement by methanolmust be considered in order to generate the appropriatetoxicity data. Spiking the methanol fractions with suspecttoxicants, however, does not provide the same confi-dence about the cause of toxicity in the effluent as spikingthe effluent directly. The mass balance approach de-scribed in Section 6 could be coupled with spiking theeffluent with a portion of the fractions to make the datamore relevant to whole effluent toxicity.

For chronic testing spiking a portion of the metha-nol fractions, such as C,, SPE methanol fractions intodilution water to mimic the effluent, requires some specialconsiderations as discussed in the chronic Phase I (EPA,1992) and the new Phase II (EPA, 1993A). For any testspecies, the effects of the methanol at the effluent spikingconcentration for the test species must either be essen-tially non-existent or clearly established so that properinterpretation is applied. The use of spiking for chronictoxicants of the methanol fractions is not as easy as thespiking for acute toxicants due to the limitations in thequantity of methanol that would be added with eachfraction for the toxicity test. If the chronic toxicity effectlevel is around or ~25% effluent and the highest fractiontested is 4x higher than the chronic effect level, add-backtests can be conducted similar to the acute add-backs butthe quantity of methanol required for the testing andanalysis must be considered (cf., Section 2; EPA, 1993A).As discussed in Phase II, once a suspect toxicant has

been tentatively identified, the steps of confirmation shouldbe started although sample volumes of methanol eluatesmight limit the amount of testing (see Phase II, Secticn 2;EPA, 1993A) with chronically toxic samples. Spiking ofappropriate levels for chronic toxicity for single chemicals(or mixtures) is limited as sublethal data are not asplentiful as acute data. The acute toxicity of some chemi-cals might be altered by methanol (i.e., surfactants). Thepossibility that this is occurring must be checked and acorrection applied if warranted. Spiking fractions also hasapplicability for hidden toxicants; refer to Section 9 forfurther details.

Spiking can also be done effectively when thesuspect toxicant of concern can be removed. However,since other toxicants might also be removed, the datamust be carefully interpreted. Ammonia is a good ex-ample (cf., Phase II; EPA, 1993A) to use with this tech-nique where one toxicant can be removed. Ammonia canbe removed from the effluent by passing samples overthe zeolite resin, after which the concentration can berestored in the post-zeolite effluent by the addition ofammonia. If toxicity is also restored, then it is likely thatthere is sufficient ammonia to cause the toxicity observed.However, it cannot be concluded from these data alone,that ammonia is the cause of toxicity because the zeolitecan also remove substances other than ammonia. An-other substance which is non-additive with ammonia yetpresent at a lesser or the same number of TUs couldcause the initial effluent toxicity but not be discernable bythis removal technique. This is an example of a hiddentoxicant (see Section 9). For acute toxicity, zinc couldbehave exactly this way because it is non-additive withammonia yet zinc is also removed by zeolite. Using otherammonia removal methods, such as high pH stripping,followed by spiking to the initial ammonia concentrationwill enhance confidence that a hidden toxicant is notpresent. Other examples involving the C,, SPE columnand various ion exchange resins would be approachedand interpreted similarly.

5-2

Section 6Mass Balance Approach

This approach is applicable only to those situa-tions in which the toxicant can be removed from theeffluent and recovered in subsequent manipulation steps.The objective is to account for all toxicity to assure thatsmall amounts of toxicity are not being lost. This concernis partly covered by the correlation approach (Section 2);however, a totally different toxicant present at a smallconcentration could appear as experimental variability inthe correlation and go unnoticed.

The mass balance concept is best described byillustration for acutely toxic effluents and the C,, SPEfractions. As described in Phase II (Section 2.2.7; EPA,1993A) for acutely toxic effluents, the effluent has beenpassed over a C,, SPE column which is then eluted withthe methanol/water fractions. After the toxicity tests onthe individual fractions are completed, add-back testscan be initiated to determine whether all of the toxicity inthe original sample was accounted for in the SPE frac-tions. For this step, there are three separate tests (withdilutions and replicates to calculate effect endpoints) thatmust be conducted which consist of the all-fraction test,the toxic-fraction test, and the nontoxic-fraction test. As-suming a complete recovery of all non-polar organicsfrom the SPE column, this should yield a solution of non-polar organic compounds equal to the original sampleconcentrations, In the mass balance approach, theseadd-back tests are conducted using an aliquot of theeffluent that has passed through the C,, SPE column(post-SPE column nontoxic effluent) or an aliquot ofdilution water. Each toxic fraction is added back to thepost-SPE column effluent, so that each is present atoriginal effluent concentrations (i.e., lx effluent concen-tration). For example for acutely toxic effluents, the toxic-fraction test solution is prepared using methanolconcentrations as described in Phase II (i.e., Section2.2.7; EPA, 1993A) and for each fraction where toxicitywas observed in the fraction toxicity test, 30 ul of each isadded to the same IO ml of nontoxic post-C,, SPEcolumn effluent (or dilution water). A portion of each ofthe remaining fractions where toxicity was not demon-strated are now added to a second post-SPE columnaliquot at effluent concentrations for the nontoxic-fractiontest. Finally portions of all the fractions (e.g., n= 8 foracutelycolumn

toxic’ effluents) are added to a third post-SPEaliquot at effluent concentrations for the all-frac-

tion test. If all the toxicity is exhibited in the toxic-fractiontest, then the all-fraction test results and the toxic-fractiontest results should be the same as in the unalteredeffluent. Results from the nontoxic-fraction test shouldindicate that no toxicity is present. This mass balance (oradd-back) approach allows the researcher to ascertainwhether or not the toxicity in the toxic-fraction test equalsthe effluent toxicity. Small amounts of toxicity can beundetectable in the toxic-fractions when tested separatelyor the toxicant might not have been eluted from the C,,SPE columns. Unless mass balance experipents areconducted, such loss of toxicity might not be detected. Inthe effluent example discussed in Section 2, the toxicitywas contained usually in the 75%, 80%, and 85% frac-tions and occasionally in the 70% fraction! The r2-value,slope, and intercept were all close to the expected valuesif two toxicants (diazinon and CVP) were causing theeffluent toxicity (Figure 2-3). However, in Table 6-l theresults of mass balance tests indicate that toxicity fromthe all-fraction test was greater than the toxicity of thetoxic-fraction test. While this difference is small, it didseem to be real and was attributed to a small amount ofanother toxicant in the 70% fraction. In 11 of 12 samples,the results from the all-fraction tests indicate there wasgreater toxicity than was found in the toxic-fraction tests.On the few occasions when the 70% fraction was toxic, itdid not contain any of the three suspect toxicants. Withoutthe mass balance data, consistent presence of the addi-tional toxicant would not have been discovered.

At the stage where the toxic-fractions have beenidentified, the test of the fractions in a mass-balance testis highly desirable. For chronic toxicity testing, the amountof eluate available might be limited following the fractiontoxicity tests. Using eluate for the add-back tests might bea trade-off between tracking toxicity and having sufficienteluate to concentrate for further analysis. This limits theadd-back tests broad applicability for chronic toxicity TI Esunless the effluent is toxic enough that at 4x the chroniceffect level, the methanol concentrations do not exceed

3During development of the non-polar organic procedures, variouselution profiles were used that included the 70% methanol/Waterfraction.

6-l

Table 6-1. Comparison of Eff Lent Toxicity and Toxicity Measured inEffluent Fraction Add-back Tests

Sample Effluent

12/03/87 1.18

01/l 2/88 2.00

01/13/88 1.93

02/03/88-P cl .oo

02/03/88-II 2.00

03/03/88-la 1.15

03/03/88-II 1.33

03/23/88-l 3.70

03/23/88-H 2.86

04128188 2.27

05/ 17188 2.27

05/l 7188 2.27

Toxic Units (TUs)All-fractions

1.64

2.94

2.86

1.15

1.75

1.06

1.52

3.03

2.86

1.72

2.04

1.67

Toxic-fractions

1.43

3.13

2.53

<I .oo

1.64

cl .oo

1.13

2.86

2.44

1.64

2.00

1.59

Mean 2.13 2.18 2.00

aValues excluded from mean calculations due to less than values.

the organisms tolerance. For chronically toxic samples,the all-fraction add-back test with C. dubia is not possibledue to high methanol concentrations in test cups unlesschronic toxicity is below 25% and add-backs are doneusing 25% effluent as the high test concentration (cf.,Phase II; EPA, 1993A). The data from the individualmethanol/water tests may be summed; however this ap-proach must be considered more tentative than add-backtests (see below).

A deficiency in the above approach to massbalance is that there can be some toxicity in the post-SPEcolumn effluent which has not been removed by the C,,SPE but which is not present in concentrations highenough to detect. The above mass balance approachalone will not identify this. However, if the add-back testsdescribed above are repeated using a standard dilutionwater, residual toxicity in the post-SPE column effluentshould cause the toxic-fraction test and all-fraction test toshow more toxicity when added to the post-SPE columneffluent than when added to dilution water. A confoundingeffect of this approach is that if the toxicity is changed bymatrix effects (suspended solids or TOC), then the toxic-ity will be different in the clean water test. Matrix effectscan be discerned, in part, by a third spiking experimentwhere a portion of all of the fractions and a portion of eachtoxic-fraction test are spiked into whole filtered effluent(which has not passed through the C,, SPE column). Ifthe addback tests in dilution water indicates greater toxic-ity than the addback tests with the post-SPE columneffluent, and the same type of addback test experimentwith filtered effluent (i.e., 1 lrn filter) indicate that thefractions are exactly additive, then matrix effects areindicated.

Some post-SPE column effluent samples developfungal or bacterial growth or perhaps a precipitate formsafter the effluent passes through the column. For thefungal type of growth, this is thought to occur when somemethanol bleeds into the effluent as it passes through thecolumn and more rinsing will not eliminate this problem.Some effluents consistently develop this type of growth inthe post-column effluent while others exhibit this patternin only an occasional sample. To alleviate this problem,conditioning the column with acetonitrile has helped (cf.,the acute Phase I (EPA, 1991A) and chronic Phase I(EPA, 1992) for details). When methanol fractions arespiked into the effluent this problem might or might not beenhanced; we have found this to be an effluent-specificoccurrence.

Caution is warranted in situations where toxicityis contained in more than one SPE fraction. The re-searcher should not necessarily expect the toxicity ex-pressed by each individual fraction that is tested separatelyto add up to the total effluent toxicity. First, toxicants maynot be additive and second, some toxicity which cannotbe detected in individual fractions may add to the wholetoxicity. For example, any one C,, SPE fraction may notshow toxicity but may contain some of the toxicant that isin the adjacent toxic-fraction. In this case, the toxicity ofthe toxic-fraction test would be less than expected. If thishappens in more than one pair of fractions, the sum of thetoxicity from the toxic-fraction test will be less than theeffluent toxicity or all-fraction test. These concerns areespecially important when several toxicants are involvedand one or more occur in more than one fraction.

For effluents where the C SPE column is notused, but where the toxicants can%e removed from thesample, the same objectives should be achievable, butthe methods will be different. For example, if an effluentappears to contain a volatile toxicant, the mass balancecould be done on the trap and on the purged sample.Since we have not yet done mass balance on samplessuch as these we have no experience from which to offeradditional guidance or advice.

Some of the mass balance process begins inPhase II, and there is a subtle difference in the purpose ofmass balances in Phases II and III. In Phase II, usuallyonly a few samples are used and mass balances arenecessary to determine the need for more identification inthose few samples. The mass balance is useful in earlystages of Phase II as well before toxicants are identifiedat all, because it allows the investigator to decide if thetoxicants present at 2x or 4x whole effluent concentra-tions are also expressing toxicity at lower concentrations.

In Phase III as many samples are tested, themass balance approach can provide information overtime with many samples whether or not the suspecttoxicants consistently account for all or the majority of thetoxicity. As illustrated above, the power of the mass

6-2

balance approach to detect small degrees of toxicity isbetter than for the correlation approach.

from the sample does not remove biologically non-avail-able portions. An example of this situation may be the

When a portion of the toxicant is not biologicallyalternative solvent extraction procedures which may re-

available and therefore does not contribute to toxicity,move a bound toxicant sorbed on suspended solidswith the solvent and is now toxic, yet it was not toxic in the

care must be taken to assure that removal of the toxicant unaltered sample.

6-3-

Section 7Deletion Approach

In some situations, particularly for industrial dis-charges, keeping the suspect toxicants out of the wastestream influent or effluent for short periods of time andalso conducting toxicity tests on the wastewater simulta-neously may be practical. When this approach can beused, it offers the most convincing evidence obtainablethat the suspect toxicants are the true ones. Care mustbe taken however, that other substances are not deletedor that some characteristic such as pH does not changealso. If a researcher can be certain that all changes areknown, then this approach is definitive. Changes in the

toxicants with time are as much of a concern here as inany other approach. These can be handled by the ap-proaches outlined in earlier sections and the deletionapproach need not be done repeatedly; however, if itwere practical to do so, it would certainly be effective. Ifsome samples do not contain one or more suspect toxi-cants, these effluent samples can be used to the advan-tage in confirmation in much the same way as intentionaldeletions described in this section can be used to confirmtoxicity.

*

7-1

-

_

Section 8Additional Approaches

-

This section mentions only a few of many stepsthat can be used to further confirm the cause of toxicity.The steps mentioned are mostly those that we have usedand found helpful and practical.

The pH is one of the most important effluentcharacteristics that changes toxicity. The pH of POTWeffluents, sediment pore or elutriate waters, and ambientwaters will almost always rise when they are exposed toair, especially in the small test volumes used in TIE work.Commonly, pH in an effluent sample at 25°C will .risefrom 7.1-7.3 to 8.3-8.5 during a 24 h period. That pHchange is enough to increase ammonia toxicity (basedon total ammonia) about three fold. Such pH changescan destroy work for some purposes, but by regulatingthese pH changes, the pH fluctuations can be used togreat advantage for other purposes.

Phase II (EPA, 1993A) describes the use of pHchange to identify ammonia toxicity. The toxicity of somemetals, hydrogen cyanide and hydrogen sulfide amongothers, is altered by pH change. Other characteristics,such as hardness, can also be varied to see if thechanges in toxicity follow a predictable pattern. Thetoxicity of some metals could be approached in this way.Not all equilibria are as rapid as the ammonia equilibrium,so the amount of time for equilibria to occur should becontrolled and standardized (cf., Phase II; EPA, 1993A).Various time . periods may have to elapse before theexpected changes occur and this may differ with eacheffluent. With the improved methods of pH control de-scribed in the Phase I documents (EPA, 1991; EPA,1992), much more use can be made of pH manipulation.

Often chemicals in effluent samples may not bebiologically available, and if they are not, then they arenot likely to cause toxicity. They may be made biologi-cally available through some manipulation in Phase I andsubsequently identified in Phase II. Through confirma-tion, the toxicity due to such a toxicant will becomeapparent when the correlation indicates a poor fit (cf.,

Section 2). For many toxicants, biological availability canbe demonstrated by measuring body uptake. If the con-stituent of concern enters the body from the effluent, it iscertainly biologically available. Exposure to pure com-pounds may be necessary to establish which particularorgan should be evaluated for the toxicant. In acute metalexposures using fish, most metals concentrate first in thegills while non-polar organics concentrate in fatty tissuessuch as the liver. When a chemical is metabolized by theorganism, a residue measurement for that compound isnot a valid measure of the lethal body burden because itis unknown whether the metabolite is more of less toxicthan the parent compound. If the suspect toxicant has aknown mode of action, such as the acetylcholinesteraseinhibition produced by organophosphate pesticides, thisexposure effect can be measured to assess if toxic effectsconform with the predicted effect. The use of enzymeblockers such as piperonyl butoxide (PBO) is also an aidin confirming toxicity caused by specific classes of toxi-cants (cf., Phase II: EPA, 1993A).

As additional steps are needed for confirming thecause of toxicity, combinations of various Phase I andPhase II procedures should always be used wheneverpractical. When several results are combined and allresults are indicating the same type of toxicant, the dataare more conclusive than when only one procedure yieldspredicted results.

Total dissolved solids (TDS) are a common prob-lem in certain areas of the country and for certain indus-tries. TDS will not cause toxicity from osmotic stress (thiscan easily be shown because their toxicity is not related toosmotic pressure) but rather TDS acts as a set of specifictoxicants. For toxicity caused by TDS, the ratios andconcentrations of the major cations and anions can bemeasured analytically. A similar mix of these major ionscan be added to a dilution water to see if the expectedtoxicity is present. By testing various mixtures, the re-searcher can ascertain which of the TDS componentscontribute most to the toxicity.

8-7

Section 9Hidden Toxicants

In the previous section, references were made tothe problem of hidden toxicants. Essentially there are twosituations which may produce the problem of hiddentoxicants. The first situation occurs when disparate ratiosof TUs of two toxicants are present in the effluent sample.Since the effect concentration is measured by diluting theeffluent, when disparate ratios occur, the TUs of thetoxicant present in fewer TUs in 100% effluent are so lowat the effect diluent, that its contribution if any, is notmeasurable. This problem exists whether the toxicantsare additive or non-additive. This situation generally willnot be encountered in effluents that have very slighttoxicity (i.e., effect concentration 75% to 100%) becauselittle or no dilution is required to achieve the effect con-centration. For those toxicants present in disparate ratiosin effluents with marginal toxicity, the chemical present atthe low levels may be nontoxic even in 100% effluent.

The second situation where hidden toxicantoccurs is when the toxicants are non-additive or partiallyadditive in the effluent sample. These toxicants mayoccur at approximately equal TUs or at disparate ratios ofTUs, as long as those present at lesser TUs are presentat 1 TU in the 100% effluent (cf., discussion of performingcorrelation on these types of toxicants, contained inSection 2).

If confirmation is being conducted for both acuteand chronic toxicity or if acute toxicity is being used as asurrogate for chronic toxicity, the acute to chronic ratiomust also be considered. For example, consider aneffluent with toxicants A and B for which the acute-to-chronic ratios are 3 and 12, respectively and the TUs foracute toxicity are 2 and 1 in an effluent sample for A andB, respectively. By definition, 1 acute TU (TU,) for toxi-cant A equals 3 chronic TUs (TUJ and for B, 1 TU, = 12TU,. In this example, the acute toxicity of the effluent willbe determined by A and the chronic toxicity will bedetermined by B. If in another situation, the acute-to-chronic ratios for two compounds were similar, then oneof the toxicants would determine the effect concentrationfor both acute and chronic toxicity. These examplesillustrate the importance of acute-to-chronic ratios fornon-additive toxicants. Acute-to-chronic ratios have spe-cial importance for additive toxicants when acute toxicityis being used as a surrogate measure for chronic toxicity.

If acute toxicity is being used as a surrogate it must bedemonstrated that the cause of the acute toxicity is thesame as the chronic toxicity. When acute toxicity is usedas a surrogate for chronic toxicity in Phases I and II,interpretation of the results can easily be biased andthese considerations are important.

When a toxicant can be removed from the efflu-ent and recovered, the identification of the presence of ahidden toxicant is more readily known. For example, theuse of the C,, SPE column may remove hidden toxicants.The toxicant is recovered in the eluate a&measuredboth analytically and toxicologically. This type of hiddentoxicant may be observed if ammonia is present at con-centrations that could cause toxicity. For example, in aneffluent sample ammonia is present at 3 TUs. Ammoniawill not be removed by the C,, SPE column and yet anadditional 1.5 TU of a non-polar organic toxicant is evi-dent when the C,, SPE eluate test is conducted. If thedischarger applied remedial treatment they would be ableto remove the ammonia toxicity yet the effluent would stillbe toxic. The same concept of hidden toxicants can befound when toxicants are removed by sublation which isfollowed by recovery and concentration of toxicity (cf.,Phase I; EPA, 1991 A; EPA, 1992). For example, sublationcan separate some surfactants, resin or fatty acids, andpolymers from such constituents as metals and ammonia.Hydrogen sulfide can be removed by a purge and trapmethod, thereby separating it from other effluent constitu-ents.

Specific blockers of toxicity such as EDTA formetals and PBO for organophosphates are also useful inestablishing the cause of toxicity. The more specific theblocker, the more definitive are the results, However,present knowledge does not allow us to be certain thatcompounds such as EDTA do not also affect the toxicityof other chemicals. Use of two specific blockers such asEDTA and sodium thiosulfate for copper, allows moredefinitive conclusions (cf., Phase I; EPA, 1992).

Manipulating characteristics such as pH is usefulbut can easily mislead thinking. For example, if the efflu-ent has ammonia toxicity, the toxicity due to ammoniashould disappear if the pH is lowered appropriately. Theseresults do not allow a conclusion that there are no hidden

9-l

toxicants. If, however, the pH is lowered so as to eliminateammonia toxicity but the effluent toxicity exists or evenincreases, then the likelihood of a hidden toxicant is high.Unfortunately a complication to this rationale is that thetoxicity expressed at the lower pH may be totally artifac-tual due to mechanisms of pH adjustments.

The best approach to find hidden toxicants is tofirst use, those methods that alter the effluent the least,can remove and recover removed hidden toxicants, andare most specific for a few toxicants. This advice is mostapplicable where the effort is to try to find out if somespecified type of toxicant is a hidden one, e.g., is there anon-polar organic as a hidden toxicant.

If, however, the search is for any type of hiddentoxicant then every conceivable technique should be usedthat would help to distinguish a hidden toxicant from thesuspect toxicant( Hidden toxicants are very hard to findwhen ammonia is the primary toxicant. Various tests usedto identify ammonia as the toxicant, i.e., use of the zeoliteresin, graduated pH tests and air-stripping (EPA, 1993A),all have a reasonable probability of changing the toxicityof many other potential toxicants. For instance, it is knownthat zeolite removes some non-polar organics and met-als. Air-stripping (at pH 11) could also remove or destroymany other chemicals as it often must be done for aextended period of time to achieve good ammonia re-moval. The graduated pH test results might also implicatea metal as a toxicant (EPA, 1993A). If these tests wereconducted in Phase II (E?A, 1993A) and the resultsconsistently indicated ammonia toxicity, these .data indi-cate that there are no hidden toxicants. The requiredcharacteristics for a hidden toxicant to behave exactly asammonia are very specific and obtaining results like thosedescribed above for a toxicant other than ammonia isunlikely.

,

If the hidden toxicant is additive with the suspecttoxicant but occurs in a disparate ratio, the confirmationeffort must first emphasize confirming the cause of toxic-ity (or remove the toxicity) of the primary toxicant. Thentoxicity from the hidden toxicant should be measurable.The probability a hidden toxicant that has additive toxicitywill not express its toxicity using several Phase I or PhaseII techniques is less than the probability that a non-additive toxicant will express its toxicity using several ofthe same techniques.

If the remedial action for a primary toxicant isspecific and easy, such as a product substitution, thesearch for hidden toxicants perhaps should be done afterthe remedial action has reduced or eliminated the primarytoxicant from the effluent. The remedial action (especiallyif it is treatment) may also eliminate the hidden toxicant(What must be avoided if at all possible, is to carry outexpensive remedial action only to find that the effluent isstill toxic.

The problem of hidden toxicants is a major rea-son a researcher should not accept the presence of toxicconcentrations of suspect toxicant as sufficient confirma-tion (cf., Section 1). The presence of biologically unavail-able forms (cf., Section 8) is a compelling reason not to doso.

A thorough confirmation is resources well spentin most instances. Non-additivity and disparate ratioscomplicated by non-availability occur too frequently to by-pass confirmation. Seasonal changes or changes withouta pattern, in effluent toxicants are further reasons toperform the confirmation over a period of time to assurethat the entire suite of toxicants has been found.

9-2

Section IOConclusions

Often the most laborious and difficult part of theTIE is developing data to adequately establish the causeof toxicity. In our experience, frequently the suspectcause of toxicity is found without difficulty but developinga convincing case to prove that the suspect cause is thetrue toxicant is the challenge.

Especially for POTW plants, this confirmationphase must be performed over a considerable period oftime to be certain that the cause of toxicity is not chang-

ing. TIES on POTWs and some industrial categories arenot likely to be a one time event but will have to berepeated as long as the inputs to the plant change. Ourcurrent wastewater treatment plants were not designed toremove specific chemicals, so there is no reason toexpect that they will remove everything which they re-ceive. Especially where the control over the influent is notcomplete, as is the case with POTW plants, a solid casemust be developed to assure that the cause of toxicity isnot changing.

10-l ,

Section 11When the Treatability Approach Has Been Used

As discussed in Phase I, two main approachesmay be used to remove a toxicity problem--toxicant iden-tification and source control or treatability. Phases I and IIinvolve the first approach while treatability proceduresaccompanied by toxicity testing are used in the second(EPA, 19898; EPA 1989C).

In the second approach, treatment methods arevaried to determine which will remove toxicity withoutidentifying the specific toxicants. The treatability approachrequires as much confirmation as the toxicant identifica-tion approach. Since the treatability approach shouldremove toxicity, the confirmation procedures are some-what different.

Repeat samples should be tested to ensure thattoxicity has been successfully removed. This should bedone over a sufficient length of time to assure that therange of conditions are included during the confirmationphase. Such events as seasonal changes, production

changes, storms, and intermittent operations all shouldbe included during the confirmation phase. Toxicity shouldbe consistently removed or appropriately reduced, asrequired. Either acute or chronic toxicity removal can beconfirmed this way.

One must be absolutely sure that the toxicity toresident species has been successfully removed. As hasbeen pointed out in Phases I and II, the effluent constitu-ents producing toxicity to one species may not be thesame for other species. Toxicity by a given treatmentmethod may remove all toxicity for one species but not foranother. The species of concern must be tested in theeffluent from the treatment method selected. If chronictoxicity is the concern, this testing may be more difficultbecause chronic testing methods may not be available forresident species. In selected cases, symptoms may besubstituted for the usual endpoints of chronic tests buttheir use would be case specific.

11-l

Section 1212References

Ankley, G.T., G.S. Peterson, M .T. Lukasewycz, and D.A.Jensen. 1990. Characterization of Surfactants in Toxic-ity Identification Evaluations. Chemosphere 21(1-2):3-12.

AQUIRE. 1992. Aquatic Toxicity information RetrievalDatabase and Technical Support Document. Environ-mental Research Laboratory, Duluth, MN.

Drummond, R.A., C.L. Russom, D.L. Geiger, and D.L.DeFoe. 1986. Behavioral and Morphological Changesin Fathead Minnow (Pimephales promelas) as Diag-nostic Endpoints for Screening Chemicals According toMode of Action. Aquatic Toxicology and EnvironmentalFate: Nineth Volume, ASTM STP 921. T.M. Poston andR. Purdy, Eds., American Society for Testing and Ma-terials, Philadelphia, PA. pp. 415-435.

Drummond, R.A. and C.L. Russom. 1990. BehavioralToxicity Syndromes: A Promising Tool for AssessingToxicity Mechanisms in Juvenile Fathead Minnows.Environ. Toxicol. Chem. 9: 37-46.

EPA. 1988. Methods for Aquatic Toxicity IdentificationEvaluations: Phase I Toxicity Characterization Proce-dures. EPA-600/3-88-034. Environmental ResearchLaboratory, Duluth, MN.

EPA. 1989A. Methods for Aquatic Toxicity IdentificationEvaluations: Phase II Toxicity Identification Procedures.EPA-600/3-88-035. Environmental Research Labora-tory, Duluth, MN.

EPA. 1989B. Toxicity Reduction Evaluation Protocol forMunicipal Wastewater Treatment Plants. EPA/600/2-88/062. Water Engineering Research Laboratory, Cin-cinnati, OH.

EPA. 1989C. Generalized Methodology for ConductingIndustrial Toxicity Reduction Evaluations (TREs). EPA/600/2-881070. Water Engineering Research Labora-tory, Cincinnati, OH.

EPA. 1989D. Methods for Aquatic Toxicity IdentificationEvaluations: Phase III Toxicity Confirmation Proce-dures. EPA-600/3-88-036. Environmental ResearchLaboratory, Duluth, MN.

EPA. 1989E. Short-Term Methods for Estimating theChronic Toxicity of Effluents and Receiving Waters toFreshwater Organisms. Second Edition. EPA/600/4-89/001 and Supplement EPA1600/4-89/001 A. Environmen-tal Monitoring and Support Laboratory, Cincinnati, OH.

EPA. 1991A. Methods for Aquatic Toxicity IdentificationEvaluations: Phase I Toxicity Characterization Proce-dures. Second Edition. EPA/600/6-91/003. Environmen-tal Research Laboratory, Duluth, MN.

EPA. 1991 B. Sediment Toxicity Identification Esaluation:Phase I (Characterization), Phase *II, Identification andPhase III (Confirmation) Modifications of Effluent Pro-cedures. EPA/600/6-91/007. Environmental ResearchLaboratory, Duluth, MN.

EPA. 1991 C. Methods for Measuring the Acute Toxicity ofEffluents to Freshwater and Marine Organisms. FourthEdition. EPA/600/4-90/027. Environmental Monitoringand Support Laboratory, Cincinnati, OH.

EPA. 1992. Toxicity Identification Evaluation: Character-ization of Chronically Toxic Effluents, Phase I. EPA/600/6-91/005F. Environmental Research Laboratory,Duluth, MN.

EPA. 1993A. Methods for Aquatic Toxicity IdentificationEvaluations: Phase II Toxicity Identification Proceduresfor Samples Exhibiting Acute and Chronic Toxicity.EPA-600/R-92/080. Environmental Research Labora-tory, Duluth, MN.

EPA. 1993B. Short-term Methods for Estimating theChronic Toxicity of Effluents and Receiving Waters toFreshwater Organisms. Third Edition. EPA-600/4-911022. Environmental Monitoring and Support Labora-tory, Cincinnati, OH.

Norberg-King, T.J., M. Lukasewycz, and J. Jenson. 1989.Results of Diazinon Levels in POTW Effluents in theUnited States. NETAC Technical Report 14-89. US.Environmental Protection Agency, Environmental Re-search Laboratory, Duluth, MN.

Norberg-King, T.J., E.J. Durhan, G.T. Ankley, and E.Robert. 1991. Application of Toxicity Identification Evalu-ation Procedures to the Ambient Waters of the ColusaBasin Drain, California. Environ. Toxicol. Chem. IO:891 -900.

12-l *U.S.COV&RNMENT PRINTING OFFICE:1993 -7SO-002/80298