U.S. Environmental Protection Agency, Program Contract Service Center (3803R), 401 M Street, SW, Washington, DC 20460-0001

B -- MULTIVARIATE ANALYSIS OF SEDIMENT TOXICITY DATABASE SOL RFQ-DC-99-00024 DUE 031299 POC Valoree S. Lilley, lilley.valoree@epamail.epa.gov E-MAIL: click here to contact the contracting officer, lilley.valoree@epamail.epa.gov. BACKGROUND: Approximately 10 percent of the sediment underlying our nation's surface water is sufficiently contaminated with toxic pollutants to pose potential risks to fish and to humans and wildlife who eat fish (USEPA 1997). Considering only the upper five centimeters, where many bottom organisms live, this represents about 1.2 billion cubic yards of contaminated sediment. An important step in addressing the contaminated sediments problem is to develop scientifically sound methods that can be used to assess whether and to what extent sediments are "contaminated" or have the potential for posing a threat to the environment. One tool that has been increasingly used is bulk sediment toxicity testing (USEPA 1992). When combined with chemical concentrations measured in the same samples, these tests can be used to identify problematic chemicals. In previous efforts we have compiled a large database that contains paired sediment toxicity and chemistry data. This database has been used to develop logistic regression models that can be applied on a chemical by chemical basis to predict the probability of an adverse effect (Field et al., in press, attachment 1). STATEMENT OF WORKThe purpose of this project is to conduct multivariate analyses that predict the probability of toxicity based on sediment chemistry through use of the database provided by EPA and described below. Specifically, the contractor shall provide: 1. Multivariate concentration-response model(s) that predict bioassay outcomes based on chemical concentrations 2. Confidence limits on point estimates derived using the model (e.g., for 50% probability of response). 3. An evaluation and discussion of model quality. EPA will provide the contractor with the database prior to Task 1, and may provide additional data after the analytical approach (Task 1) has been finalized.Task 1: The contractor shall provide a memo discussing the approach that will be used for modeling and model evaluation. At a minimum the approach should encompass the following modeling variations: Species: Ampelisca only, Rhepoxinius only, combined species Definition of a toxic sample: Statistical significance only (i.e., includes "maybe toxic" and "clearly toxic" samples), statistical significance plus minimum significant difference from control (i.e., includes only "clearly toxic" samples). Because percent control-normalized survival will be provided, this information may also be used directly in the models. Inclusion of covariates: TOC normalized, no normalization. The contractor shall specify how below detection limit values are to be designated in the final database (e.g., as 1/2 detection or as missing values). We expect that dimension reduction techniques (e.g., principal components analysis) may be useful to simplify the chemistry portion of the data. In addition, one column of the database will reduce the PAH concentration information as per the Swartz et al. PAH model (attachement 2). We also expect that multivariate logistic modeling will be an appropriate modeling approach. If an alternative is preferred the contractor shall provide justification for the approach selected. After the memo has been produced, the contractor shall meet (via phone) with EPA to discuss the approach. The contractor will incorporate EPA comments and revise the approach as necessary. Task 2: The contractor shall develop and run the models, and provide a report describing the results. The report shall include: A description of the modeling approaches used to derive the multivariate concentration response curve(s). .Model structure and parameter estimates. Confidence limits on point estimates (e.g., as shown in Field et al. in press). A description of the approach used for model evaluation. 1. A discussion of model quality and performance. Task 3: The contractor shall attend the Sediment Advisory Group meeting to be held in Tampa FL, May 21st 1999. The contractor shall present a 30 minute description of the modeling approach and preliminary results at this meeting. Task 4: The contractor shall incorporate comments received from the EPA project officer after the Sediment Advisory Group meeting and produce a final report. Task 5: The contractor shall provide 16 hours of telephone consulting and review of materials provided by the EPA project officer. We anticipate that consultation will be requested regarding aspects of the chemical-specific models (derived using the approach described in Field et al. In press) and will include model evaluation, model comparison, and evaluation of data handling variations (e.g., classification of toxic samples). The contractor shall invoice by 10 minute intervals for the number of hours spent on this task each month. Each phone call made by a team member to the contractor will be charged a minimum of one interval. The contractor shall keep records of the time spent in providing telephone consultation to team members and submit that cost as a spearate line item. MATERIALS TO BE PROVIDED BY EPA Database Description The database will be provided in dbase or SAS database format. Samples will be formatted in rows with the columns representing chemical concentrations and results of toxicity tests. The database will contain chemistry and toxicity data for 2 amphipod species 10-d survival tests. It contains about 2000 samples for Ampelisca abdita and about 1000 samples for Rhepoxynius abronius. Each data row will represent an individual test sample and will have concentrations for 25-50 chemicals or chemical mixtures (e.g., PAHs), percent total organic carbon (TOC) for most samples and the results of toxicity test. Below detection limit data will be entered as either 1/2 detection limit value or coded as a missing value, as indicated by the contractor. Toxicity test data and analyses that EPA will provide the contractor will include: statistical difference from negative control (Y/N), percent control-normalized survival (mean sample percent survival relative to mean control percent survival), and toxicity code with three classifications (not toxic, maybe toxic, and clearly toxic ) according to the following definitions:not toxic: not significantly different from negative control; maybe toxic: significantly different from negative control and greater than 80% control-normalized survival;clearly toxic: significantly different from negative control and less than 80% control-normalized survival PERIOD OF PERFORMANCE: Date of award thru September 30, 1999. TECHNICAL EVALUATION FACTORSThe award will be made to the entity whose offer provides the greatest value to the goverrnment in terms of price, quality and delivery. The contractor will be evaluated based on the following factors: Key personnel and availability during time period. Demonstrated experience in conducting multivariate statistical analyses. Demonstrate technical comprehension of statement of work. Posted 02/19/99 (W-SN300270). (0050)

Loren Data Corp. http://www.ld.com (SYN# 0022 19990223\B-0005.SOL)