Methodology

Sample Sites

Samples for pharmaceutical chemical analysis were collected from eight stream sites (Tualatin River tributaries), two Tualatin River sites upstream and downstream of discharge from an advanced WWTF, and four locations within that treatment facility (table 2, fig. 2). The stream sites were chosen to represent a range of urban, agricultural, and forested land uses in the Tualatin River basin, with an emphasis on the urban areas. Six of the stream sites (sites 1–6) are located in the highly urbanized Fanno Creek drainage, while site 7 on Gales Creek represents more of a forested drainage and site 8 on Dairy Creek has substantial agricultural land use upstream. Sites 1 and 2 in the upper reaches of Fanno Creek, site 5 on Summer Creek, and site 6 on the unnamed creek represent small drainages in largely residential urban areas. The Summer Creek drainage may have a larger fraction of residences with on-site septic systems rather than service from the sanitary sewer system. The remaining two sites farther downstream on Fanno Creek (sites 3 and 4) include increasing amounts of commercial and industrial land use. In addition to point-source discharges, few of which are present in the Fanno Creek drainage, pharmaceutical chemicals may be present in urban streams as a result of accidental discharge, intentional dumping of waste (from a recreational vehicle, for example), sanitary-sewer/storm-sewer cross connections, leaking sewer lines, and failing septic systems, to name a few. Caffeine may be present in streams as a result of people discarding coffee-cup residue on the street, which then can be washed to a storm drain and from there into the nearest creek.

Samples were collected from four locations within Clean Water Service’s Durham WWTF to assess the concentrations of pharmaceuticals that are delivered to the facility and their removal from the waste stream. The Durham WWTF is an advanced treatment facility that uses state-of-the-art treatment technology and attains a higher level of phosphorus removal, for example, than 98 percent of WWTFs in the United States (Clean Water Services, 2002). The facility uses screening, primary settling, enhanced biological treatment, tertiary chemical treatment, chlorination, sand-bed filtration, aeration, and dechlorination processes (fig. 3). Although the treatment is designed primarily to remove solids, oxygen demand, ammonia, and phosphorus, the effluent is near drinking-water quality and typically dilutes many of the regulated chemicals in the Tualatin River’s receiving water. The Durham WWTF serves a customer base of more than 200,000 people and has a dry-weather discharge of about 25 ft³/s (16 Mgal/d). On July 31, 2002, the day of sample collection, the mean facility discharge was 24 ft³/s, but ranged as high as 40 ft³/s and as low as about 12 ft³/s. Samples of influent (at the headworks) and effluent were collected. Additional samples were collected before and after the filtration step, which is downstream of primary, secondary, and tertiary treatment and after chlorination, but prior to final aeration and dechlorination (fig. 3). Samples were collected without attempting to follow a single parcel of influent through the WWTF; the 6-hour residence time and multiple recirculation pathways within the facility make such a sampling strategy difficult to perform. Results from these samples, therefore, are only strictly comparable if the influent waste stream was somewhat invariant in its loading of pharmaceutical chemicals.

To assess the instream effect of WWTF effluent, samples were collected from the Tualatin River 0.6 mi upstream (site 9, Cook Park) and 0.6 mi downstream (site 10, Boones Ferry Road) of the Durham WWTF outfall on the same day and at about the same time that samples were collected from the WWTF. On July 31, 2002, streamflow in the Tualatin River was about 171 ft³/s upstream of the outfall and 198 ft³/s downstream of the outfall. These streamflows were estimated based on measured streamflows at several streamflow-gaging stations upstream and downstream of the WWTF outfall.

Sampling and Processing Methods

Water samples from stream and river sites were collected using standard USGS protocols. Where stream size and depth allowed, the depth- and width-integrating equal-width increment (EWI) water-sampling technique was used. In the EWI method, a sample is collected by lowering and raising a sampler through the water column at a constant specified rate, repeating the process at the center of a set of equally spaced locations in the stream cross-section, and compositing the collected waters (Edwards and Glysson, 1999). EWI samples were collected using either a DH-81 hand-held sampler for wadeable streams or a D-77 sampler for non-wadeable streams with a 3-L Teflon^® sample bottle and a Teflon^® cap/nozzle assembly. The unnamed creek at Walnut Street (site 6, fig. 2) was not deep or large enough to use the EWI method; instead, a grab sample was collected using the protocols and considerations for nonstandard sampling documented by Shelton (1994). Samples were placed on ice for transport back to the USGS Oregon Water Science Center laboratory for later filtration.

Samples of water were collected from four different locations at the Durham WWTF. Because of safety, access, and structural considerations at the facility, nonstandard sampling techniques were used, typically resulting in grab samples of the source water. Samples were placed on ice before transporting them to the laboratory for later filtration.

Special protocols were followed to prevent contamination of the water samples. The sampling crew was careful to avoid contact with or consumption of any products or materials that contain the target analytes in this study. For example, the potential introduction of trace amounts of caffeine into collected samples was minimized through the crew’s abstention from caffeine-containing beverages (such as coffee), and protecting the samples from contact with human exhalations. These types of sample-handling and contamination-avoidance procedures are described by Lewis and Zaugg (2003) in the USGS National Field Manual.

Standard USGS procedures for handling and filtering samples containing organic compounds were followed prior to the shipment of samples to the USGS National Water Quality Laboratory (NWQL) in Denver, Colorado. Samples were filtered using a stainless steel or aluminum 142-mm or 293-mm diameter plate-filter assembly with a 0.7-micron pore size glass-fiber filter. A valveless-piston variable-speed pump with Teflon^® tubing was used, and sufficient filtrate from each sample was collected to fill two 1-L amber glass bottles that had been cleaned and baked at 350°C. Standard protocols for sample processing are documented in section 5.2.2 of the USGS National Field Manual (Wilde and others, 2004). All filtered samples were shipped on ice to the USGS NWQL for analysis.

The use of filtered water samples in this study, a requirement of the laboratory analysis method, means that the results reflect only the dissolved fraction of the target compounds’ total mass in the collected water samples. Although many of the target compounds are fairly water soluble (some even are ionized under neutral pH conditions, which makes them extremely water soluble), a fraction of their mass in a whole-water sample may be associated with suspended particulate material. As a result, the measured concentrations from a filtered sample may be lower than the concentration that would be measured in a whole-water sample.

Laboratory Analysis

Filtered water samples were analyzed for a suite of 18 pharmaceuticals and 3 metabolites by the Methods Research and Development group at the USGS NWQL using a relatively new analytical method (Cahill and others, 2004; Furlong and others, 2008). In that method, the target analytes and an added performance surrogate first were removed from filtered 1-L water samples by passing the water through a solid-phase extraction (SPE) cartridge. The analytes then were eluted from the SPE cartridge using small volumes of methanol and acidified methanol, concentrated to near dryness, reconstituted into formate buffer with an internal standard, and filtered. The concentrates were analyzed using high-performance liquid chromatography interfaced with a mass spectrometer using electrospray ionization operated in the positive-ion mode (HPLC–ESI–MS). Selected-ion monitoring MS was used to improve sensitivity and specificity. For more details on the specific materials, extraction procedures, and instrument conditions, see Cahill and others (2004) or Furlong and others (2008).

The analytical method used in this study was still under development at the time of sample collection and analysis. For that reason, the lists of target analytes in the published method papers (Cahill and others, 2004; Furlong and others, 2008) are slightly different from the target analytes in this study. In the most recent version of the analytical method documented by Furlong and others (2008), 11 pharmaceuticals and 3 metabolites were analyzed, all of which were included in this study. Seven additional compounds (cimetidine, fluoxetine, gemfibrozil, ibuprofen, metformin, miconazole, and ranitidine) were included in this study, six of which also were included by Cahill and others (2004). Of the 21 compounds analyzed in this study, only miconazole was omitted by both of the published method papers. Method detection limits and spike recovery information for all of the target analytes are included later in this report with the results and the quality assurance data. Because this method incorporates a mass spectrometer as a detector, the qualitative identification of a compound can be verified, if not reliably quantified, at concentrations less than the method detection limit (Childress and others, 1999; Bonn, 2008b); such detections are reported as estimated concentrations only.

Quality Assurance

The quality assurance program for this study included field duplicate samples and equipment blanks, and laboratory blanks and spikes. The field equipment blank was composed of certified organic blank water (EM Science, Universal Blank Water, lot #42044: purity verification data available upon request) that was filtered and handled using the same procedures as those used for the stream samples. Duplicate water samples were collected at five sites (table 2) to test the reproducibility of field and laboratory procedures. In addition, as part of the laboratory method, each set of analyzed samples included a laboratory blank sample and a spiked blank sample.

Spike recovery results from this study, and from 67 additional laboratory spike samples analyzed during method development in 2002, demonstrated that the analytical method produced mixed results for this set of 21 target analytes (table 3). For 11 of the analytes, the spike recovery was consistently greater than 60 percent, and results were deemed reliable enough to report without qualification. Mean spike recoveries for eight other analytes were between 20 and 60 percent; this less reliable recovery requires that quantified concentrations for these compounds be reported only as estimates. Spike recoveries for the final two compounds on the target analyte list, metformin and miconazole, were poor (6 percent or less); any detections of these compounds were reported without quantification.

Blank samples from the field and the laboratory were, with one exception, devoid of the target analytes. A low-level concentration of fluoxetine (0.0039 µg/L) was detected in the equipment blank sample, reducing the reliability of any low-level detections for that compound; fluoxetine results already were qualified as estimates based on low spike recovery results. No other target compounds were detected in the blank samples. Table 3 lists the spike recovery data, blank results, and interpreted reporting guidance for each compound (report without qualification, report as estimate, etc.).

Duplicate sample results showed good agreement, providing assurance that the detection frequency and the quantified concentrations were reliable. For the five duplicate water samples included in this study, a target analyte was detected in 15 instances in which a paired duplicate result also was available. In 12 of the 15 instances, the analyte was detected in both samples; in only 3 instances was a compound detected in only 1 of the paired samples. As might be expected, the percent relative difference in the quantified concentrations of paired results was higher for the lowest concentrations. For concentrations less than 0.1 µg/L, the relative percent difference was about 19 percent (10 comparisons), and that value decreased to about 5 percent for concentrations greater than 0.1 µg/L (2 comparisons). Concentrations closer to the detection limit, therefore, are expected to be more uncertain than those that are well above the detection limit.

Matrix interferences can cause quality assurance issues that are difficult to quantify. The influent sample from the Durham WWTF, although filtered, had a complex organic chemical signal that could not completely be cleaned up or avoided during the sample extraction and concentration process. These interfering chemicals were incompletely separated from the target analytes during the chromatographic procedure, thus decreasing signal-to-noise ratios, increasing detection limits, and decreasing the certainty of compound identification and quantitation. Although results from the Durham influent sample were double-checked by re-running the analysis, all those results have a greater uncertainty due to matrix interferences.