An Assessment of Two-Step Linear Regression and Multifactor Probit Analysis as Alternatives to Acute to Chronic Ratios in the Estimation of Chronic Response from Acute Toxicity Data to Derive Water Quality Guidelines Andrew R Slaughter, Carolyn G Palmer, and Wilhelmine J Muller* Unilever Centre for Environmental Water Quality (UCEWQ), Institute for Water Research, Rhodes University, PO Box 94, Grahamstown 6140, South Africa (Received 5 December 2005; Accepted 11 April 2006) ABSTRACT In aquatic ecotoxicology, acute to chronic ratios (ACRs) are often used to predict chronic responses from available acute data to derive water quality guidelines, despite many problems associated with this method. This paper explores the comparative protectiveness and accuracy of predicted guideline values derived from the ACR, linear regression analysis (LRA), and multifactor probit analysis (MPA) extrapolation methods applied to acute toxicity data for aquatic macroinvertebrates. Although the authors of the LRA and MPA methods advocate the use of extrapolated lethal effects in the 0.01% to 10% lethal concentration (LC0.01–LC10) range to predict safe chronic exposure levels to toxicants, the use of an extrapolated LC50 value divided by a safety factor of 5 was in addition explored here because of higher statistical confidence surrounding the LC50 value. The LRA LC50/5 method was found to compare most favorably with available experimental chronic toxicity data and was therefore most likely to be sufficiently protective, although further validation with the use of additional species is needed. Values derived by the ACR method were the least protective. It is suggested that there is an argument for the replacement of ACRs in developing water quality guidelines by the LRA LC50/5 method. Keywords: Freshwater shrimp Ecotoxicology Extrapolation Water quality guidelines INTRODUCTION Water quality guidelines are values for toxicants expressed as concentration units that are recommended to support and maintain a designated water use (ANZECC and ARMCANZ 2000) and their derivation is dependant on toxicity test data (Warne 2001). Water quality guidelines compiled for South African (DWAF 1996) and Australian/New Zealand (AN- ZECC and ARMCANZ 2000) freshwater systems have highlighted the preferential use of chronic over acute toxicity data. Chronic toxicity data are important for setting water quality guidelines; however, there are many more acute than chronic toxicity data. This is because chronic toxicity tests are comparatively more difficult and costly to perform, sublethal endpoints are difficult to measure, and few chronic test protocols available are validated. However, water quality guidelines have been derived despite a lack of chronic toxicity data. Both the South African Water Quality Guidelines (DWAF 1996) and the Australian/New Zealand (ANZECC and ARMCANZ 2000) trigger values use acute to chronic ratios (ACRs) if there are insufficient chronic toxicity data available. However the use of ACRs have been criticized (Heger et al. 1995; Lange et al. 1998; Warne 1998; Crane and Newman 2000; Roex et al. 2000; Isnard et al. 2001; Hunt et al. 2002). Because of the uncertainty surrounding the accuracy and protectiveness of chronic data estimated through the use of ACRs, the ANZECC and ARMCANZ (2000) trigger values assign reliability indices of medium reliability or low reliability to guidelines derived with ACRs. Various alternative methods of predicting chronic response from acute toxicity data have been proposed. Two-step linear regression analysis (LRA; Mayer et al. 1994) and multifactor probit analysis (MPA; Lee et al. 1995) are 2 methods that have a good theoretical basis and are explored in this study. Although the MPA and LRA methods have been validated with fish chronic toxicity data (Mayer et al. 1994; Lee et al. 1995), no further published research on the accuracy or validity of these methods are available. The aims of this research were to investigate the comparative accuracy and protectiveness of the use of the ACR, LRA, and MPA acute to chronic extrapolation techniques in the derivation of water quality guidelines. These acute to chronic extrapolation methods were applied to indigenous southern African macro- invertebrate acute toxicity data and benchmarked against equivalent experimentally obtained chronic toxicity data. METHODS All extrapolations and chronic toxicity testing were done with Na 2 SO 4 and NaCl as toxicants because the extrapola- tions and chronic testing were part of a larger salinity guideline study (Slaughter 2005). Acute toxicity data used to derive extrapolated chronic data Acute toxicity data were obtained from the Unilever Centre for Environmental Water Quality (UCEWQ; Gra- hamstown, South Africa) toxicity database (Scherman et al. 2002; Palmer et al. 2004). All acute and short-term chronic toxicity test data contained on the database were originally produced for studies unrelated to the one described here. The * To whom correspondence may be addressed: nikite@iwr.ru.ac.za. Integrated Environmental Assessment and Management — Volume 3, Number 2—pp. 193–202 Ó 2007 SETAC 193 Original Research Review