Pharmaceutical Chemical Detections and Concentrations

Reconnaissance Results

This reconnaissance of pharmaceutical chemicals in streams of the Tualatin River basin was conducted in late July 2002. Stream samples were collected between July 25 and July 30, and the Tualatin River and the Durham WWTF were sampled on July 31 (table 2). Late July in western Oregon typically is warm and dry, and streams in the Tualatin River basin at that time are composed predominantly of groundwater baseflow. In the case of the Tualatin River, additional flow comes from upstream reservoir releases and the treated effluent from two advanced WWTFs. Prior to sample collection, no measurable rain had occurred since July 8, although a trace was recorded on July 26 (USGS data: see site 452359122454500, for example, at http://or.water.usgs.gov/grapher/).

A wide variety of sources can deliver pharmaceutical chemicals to streams. During dry weather, such sources might include failing septic or other on-site waste-treatment systems, leaking sewer lines, permitted and accidental discharges, illicit and unpermitted dumping, sanitary-sewer/storm-sewer cross connections, and unmanaged or poorly managed pet and livestock wastes. Stormwater sources are not included in this reconnaissance, which is important to remember because the storm-sewer system in the study area generally discharges to the stream network rather than to a treatment facility. The dry-weather sampling increased the chances of detecting pharmaceuticals from dry-weather sources because dilution by rainfall was avoided. The absence of rain also meant that the Durham WWTF was handling its normal load of domestic and other municipal/industrial sewage; therefore, the effect of WWTF effluent on the Tualatin River was representative of summer low-flow conditions and the relative proportion of treated effluent in the river was somewhat maximized, thus making its source load easier to detect.

Of the 21 target analytes included in this study, six were not detected in any sample, and only six were detected in any of the stream or river samples (tables 4 and 5). Stream sample detections included cotinine, caffeine, acetaminophen, carbamazepine, codeine, and sulfamethoxazole, in decreasing order of detection frequency. Sulfamethoxazole and carbamazepine primarily were found in the Tualatin River rather than in the smaller drainages, and the main source of the two compounds in the river probably was treated WWTF effluent; none of the smaller creek sites have WWTF discharges upstream of their sampling locations. Fifteen of the 21 target analytes were detected in WWTF influent; in contrast, only five were detected in treated effluent.

The most commonly detected compound was cotinine, which was found in every urban stream sample, both samples from the Tualatin River, and all WWTF samples. Cotinine is a metabolite of nicotine, which is an ingredient of tobacco-containing products; therefore, cotinine is delivered to municipal WWTFs on a consistent basis through the sanitary sewer. Given the widespread outdoor use of nicotine-containing products, it is not surprising that cotinine was detected in all urban stream samples. Similarly, caffeine was the second most-detected compound, and it also was found in all urban stream samples. Caffeine is present in high concentrations in coffee-based beverages, soft drinks, and energy drinks, and the propensity of many people to discard the remains of those beverages on streets and parking lots may account for the prevalence of caffeine in urban stream samples. Wash-off from streets and parking lots typically is conveyed to the nearest stream through the storm-sewer system with minimal opportunity for degradation, and little of the storm-sewer network has been retrofitted to deliver stormwater to wetlands or other naturalized treatment facilities. Note that if wash-off from streets and parking lots was instrumental in delivering these compounds to streams, then the dry-weather sampling performed in this study may have been affected by previous stormwater delivery processes. Despite the universal detection of caffeine in urban stream samples in this study, none of those samples included detections of 1,7-dimethylxanthine, which is the primary metabolite of caffeine (Guerreiro and others, 2008) and is present in high concentrations in untreated wastewater. The absence of this metabolite in urban stream samples reinforces the suggestion that the source of caffeine to these urban streams does not include processing through the human body, and therefore does not come from treatment, septic, or sewer-related sources.

Most of the target analytes in this study (15 of 21) were not detected in stream or river samples. Only three compounds (acetaminophen, caffeine, cotinine) were detected in urban stream samples, and the great majority of those detections were at concentrations less than 0.04 µg/L with half of those detections at less than 0.01 µg/L (table 4). For perspective, 0.01 µg/L is nearly identical to half a drop of the pure compound diluted into a 660,000-gal Olympic-sized swimming pool, and 660,000 gal is the total amount of water used if you were to flush a modern 1.6-gal water-saving toilet 25 times a day for the next 45 years. With only a couple of exceptions, the detected concentrations were less than about 4 to 5 times the analytical method detection limit. One sample did show a substantially higher concentration of caffeine (0.314 µg/L) than any other stream sample, perhaps indicating a nearby source. Only two compounds (acetaminophen, carbamazepine) were detected in Gales Creek and only one (caffeine) was detected in Dairy Creek, and all three were estimated at concentrations less than their method detection limits. Even the Tualatin River samples, which were affected by one or more upstream WWTF sources, showed only four compound detections (carbamazepine, codeine, cotinine, sulfamethoxazole), and two (codeine and sulfamethoxazole) of the four were estimated at concentrations less than the method detection limit. In general, this reconnaissance of the target pharmaceutical chemicals indicates that these compounds either are not present in Tualatin River basin streams or are present only at trace concentrations that are many orders of magnitude below their usage or pharmacological dosing levels. For many of these compounds, it is not yet known whether these trace concentrations have ecological effects.

The effects of upstream land use were not clearly defined with the limited number of samples collected in this study. The samples collected from the Gales (forested) and Dairy (agricultural) Creek sites, however, certainly had fewer of the target compounds detected and all such detections were estimated at concentrations less than their method detection limits. The target analyte list, however, did not include the most commonly used antibiotics used in veterinary medicine, which might partially explain the paucity of detections in the agricultural drainage. The biggest difference between these two samples and the urban stream samples was the consistent presence of caffeine and cotinine in the urban samples. Caffeine and cotinine, therefore, may be good indicators for sources associated with urban areas. The Gales and Dairy Creek sites have little upstream urban land use (2 and 7 percent, respectively).

Comparison to Other Data

Samples were collected by the USGS in 1999–2000 from 139 streams across the United States, including several from Oregon streams (Barnes and others, 2002; Kolpin and others, 2002), in a reconnaissance of pharmaceutical chemicals, hormones, and organic wastewater contaminants. Results from that investigation provide a useful comparison to this study. The list of pharmaceutical compounds included in the national reconnaissance was similar to the target analyte list in this study, largely because the same analytical method was used. Seventeen of the target compounds in this study were included in the national reconnaissance, and of the six compounds that were detected in stream samples in this study, only carbamazepine was not included in the national reconnaissance. Results for the five remaining detected compounds are compared to results from the national study in figure 4. The concentrations detected in this study are similar to and consistent with those from the national study. Maximum concentrations detected in this study, however, were at least an order of magnitude less than the maxima from the national study. This could be due to chance, given that the national study included about 10 times more samples, but it also may indicate that samples from the Tualatin River basin have fewer or weaker upstream sources of the target pharmaceuticals.

In both the national reconnaissance and this study, a sample was collected from the Tualatin River at Boones Ferry Road (site 10). The sample in the national study was collected on June 3, 2000, was preceded by only 3 days of dry weather, and showed detections for caffeine, codeine, 1,7-dimethylxanthine, sulfamethoxazole, and trimethoprim. In this study, only codeine, sulfamethoxazole, and cotinine were detected (carbamazepine also was detected but was not included in the national study). Concentrations of the detected compounds in the two samples were of similar ranges, but neither consistently higher nor lower. Because both samples were collected during the low-flow summer period, upstream sources of these compounds were likely to be similar. Although storm-related sources may have been more prevalent in the sample from the national study, such sources were not specifically assessed in this study and their importance remains unknown. In any case, the results indicate that the low concentrations and few detections obtained for the Tualatin River sample in this study are not an anomaly.

Depending on the chemical properties of the target pharmaceuticals, these compounds might be more likely to be found in stream sediment rather than in the overlying water. Nilsen and others (2007) determined that sediment samples collected from streams in the lower Columbia River drainage network do contain trace concentrations of pharmaceutical chemicals. Four of the samples in that study were collected from either Fanno Creek near its confluence with the Tualatin River, or from the Tualatin River near the outfall of the Durham WWTF. Those samples included detections of caffeine, diphenhydramine, thiabendazole, diltiazem, miconazole, trimethoprim, dehydronifedipine, carbamazepine, cotinine, fluoxetine, and codeine, in descending order of detected concentration or frequency. Detected concentrations ranged from about 1 to about 70 ng/g of sediment. In contrast, only four of these compounds (caffeine, carbamazepine, cotinine, and codeine) were detected in filtered water samples from the same general sampling sites in this study.

The aqueous concentrations of pharmaceutical compounds in this study cannot be compared directly to the sediment-associated concentrations detected by Nilsen and others (2007) because the samples were not collected at the same time or from exactly the same location. Even if collected concurrently from the same site, water samples would reflect conditions at that time, whereas sediment samples (collected at 1–2 cm depth by Nilsen) represent time-integrated processes of accumulation and degradation. Further research is necessary to quantify the concentrations and proportions of these compounds in various media (water, sediment, fish tissue), but it is instructive to make rough comparisons using available data, in the interest of learning more about instream processes. For example, caffeine was detected in water and in sediments from Fanno Creek near its mouth. Aqueous concentrations from this study were roughly 0.02 µg/L (identical to 
0.02 ng/mL or 0.02 ng/g), and the sediment-associated concentration measured by Nilsen and others (2007) was about 57 ng/g. Given the fact that caffeine is not charged under neutral pH conditions, and that its octanol-water partition coefficient (K_ow) is about 1.0, nearly equal concentrations of caffeine in water and in the organic component of the sediments would be expected. These data, however, indicate that the phases are not in equilibrium—the aqueous concentration is lower than expected based on the sediment-associated concentration. Perhaps the caffeine in the aqueous phase is readily degraded or the upstream sources are sufficiently variable that the phases are never in equilibrium. In contrast to caffeine, many pharmaceutical chemicals (such as ibuprofen) have acid-base properties that greatly increase their aqueous solubility under environmentally relevant pH conditions and decrease the relevance of hydrophobic partitioning to sedimentary organic matter (Wells, 2006). When considering the fate of pharmaceutical chemicals in the environment, it is critical to know how their chemical properties affect their aqueous solubility, bioaccumulation, bio- and photo-degradation, and partitioning to sediments and the atmosphere. More research and sampling is needed to fully understand the presence and fate of pharmaceutical chemicals in Tualatin River basin streams, sediments, and biota.

Effects of Wastewater Treatment

The highest concentrations of any target analyte in this study were found in the influent sample from the Durham WWTF. Five compounds were detected at concentrations greater than 1 µg/L. Caffeine and acetaminophen had the highest concentrations (30 and 29 µg/L, respectively), followed closely by 1,7-dimethylxanthine (24 µg/L), ibuprofen (7.6 µg/L), and cotinine (1.5 µg/L) (table 4). The presence of these particular pharmaceuticals and metabolites in high concentrations in wastewater is entirely consistent with typical usage rates. Although caffeine is an active ingredient in some medications, the adult per capita consumption rate of caffeine through beverages has been estimated as 320 mg/d in the United States; no other pharmaceutical in this study has such a high consumption rate (Wilkison and others, 2006). Furthermore, a national usage survey performed in 1998–99 reported that acetaminophen and ibuprofen were the two most commonly used pharmaceuticals, with caffeine in ninth place, not including the consumption of caffeine in beverages (Kaufman and others, 2002). Despite the high concentrations of these compounds in WWTF influent, 3 of these 5, and 10 of the 15 target compounds detected in WWTF influent, were not detected in samples of the treated WWTF effluent. Only carbamazepine, cotinine, ibuprofen, metformin, and sulfamethoxazole were detected in the effluent (table 4).

The influent and effluent detection data from this study, although from a limited number of samples, can be used to calculate an apparent removal rate of these compounds due to treatment processes occurring in the Durham WWTF, assuming that the compound concentrations detected in influent and effluent are typical. Because WWTFs are efficient at removing particulate material, and because this study analyzed only filtered water samples, these calculated removal rates may be biased low, depending on the water solubility of the target compound. For those compounds that were detected in the influent but not in the effluent, a lower limit on the apparent removal rate can be estimated based on the analytical method detection limit for each compound. For those that were detected in influent and effluent, a percent removal can be computed directly (table 6). Most of the compounds detected in the influent were removed with fairly high efficiency (> 90 percent). Ibuprofen and sulfamethoxazole were mostly but incompletely removed (> 75 percent), but only a small fraction (< 20 percent) of the antiepileptic drug carbamazepine was removed. Results for ibuprofen were somewhat inconsistent because it was detected in only one of two duplicate effluent samples. Results from samples collected before and after filtration in the WWTF showed that filtration was not responsible for the removal of these compounds from the waste stream; most of the removal occurred prior to that treatment step.

Many other studies have evaluated the removal rates of pharmaceuticals from wastewater. Ternes (1998) investigated the occurrence and removal of a suite of pharmaceuticals in German WWTFs, and determined that more than 60 percent of the pharmaceutical residues were removed. Many were not detected in the treated effluent, some such as ibuprofen (90 percent removal) were mostly removed through treatment, and others such as carbamazepine (7 percent removal) were only partially removed. Those results for ibuprofen and carbamazepine mirror the results from this study. A 2004 survey of pharmaceuticals in South Korean WWTFs by Han and others (2006) showed similar results for ibuprofen (78 percent removal), but quite different results for carbamazepine (91 percent removal) and acetaminophen (9 percent removal). The type of treatment (primary versus secondary, trickling filter versus activated sludge, with or without nitrification and/or denitrification processes, chlorination versus UV disinfection) and the details of the treatment processes (residence time, etc.) can have an effect on the removal of pharmaceuticals from wastewater (Jones and others, 2005; Phillips and others, 2005).

Samples from the Tualatin River upstream and downstream of the Durham WWTF outfall provide an independent means of assessing the effect of WWTF effluent on the river. Three of the five target analytes detected in WWTF effluent (carbamazepine, cotinine, and sulfamethoxazole) also were detected in the Tualatin River upstream and downstream of the Durham outfall. For those compounds, a mass balance of sorts can be performed. Because the approximate flow in the river upstream of the outfall (171 ft³/s), the flow rate of the effluent (about 24 ft³/s), and the flow in Fanno Creek (< 5 ft³/s) are known, an expected downstream concentration can be computed. In the case of carbamazepine, the in-river concentration increased from 0.038 µg/L upstream to 0.046 µg/L downstream (table 4), which is consistent with an expected downstream concentration of about 0.052 µg/L based on a mass balance. Similarly consistent results were obtained for cotinine, with upstream and downstream measured concentrations of 0.0098 and 0.014 µg/L, versus an expected downstream concentration of 0.016 µg/L based on a mass balance. In-river concentrations of sulfamethoxazole were essentially the same upstream and downstream of the Durham WWTF outfall (0.015 and 0.014 µg/L), whereas the mass balance produced an expected downstream concentration of 0.017 µg/L; given the uncertainty involved in the analyses, these three concentrations are essentially identical. Ibuprofen was not detected in the river samples, perhaps providing evidence to question the detection of ibuprofen in treated effluent, where only one of two duplicate effluent samples showed a detection (1.77 µg/L). If ibuprofen had been present in treated effluent at a concentration of 1.77 µg/L, then the in-river downstream concentration should have been easily detected at an estimated concentration of 0.2 µg/L, which is 10 times higher than the method detection limit. The laboratory analytical method used in this study was not, perhaps, the best method for quantifying ibuprofen concentrations; new analytical methods for ibuprofen and other acidic pharmaceutical chemicals are under development.

In-river concentrations of carbamazepine and cotinine upstream of the Durham WWTF outfall (Tualatin River at Cook Park, site 9) can be shown to be consistent with the presence of a large WWTF source upstream of that location. The Rock Creek WWTF discharges treated effluent to the Tualatin River about 28 mi upstream of Cook Park, and the travel time from the WWTF outfall to Cook Park was about 6–8 days under the flow conditions that occurred in late July 2002. For all of the carbamazepine in the Tualatin River at Cook Park to have come from the Rock Creek WWTF, the effluent flow rate from that facility would have to be about 41 ft³/s, assuming its carbamazepine concentration was identical to that in Durham WWTF effluent and assuming no instream degradation during the travel time to Cook Park. The measured Rock Creek WWTF effluent flow rate for that time period was about 39 ft³/s. A similar calculation for cotinine only required that the Rock Creek WWTF effluent flow rate be at least 34 ft³/s, and sources of cotinine from other urban streams such as Rock Creek certainly were present. It is likely, therefore, that concentrations of carbamazepine and cotinine in the Tualatin River upstream of the Durham WWTF outfall primarily are from the Rock Creek WWTF. Similar calculations for sulfamethoxazole reveal that the Rock Creek WWTF probably is an important source of that compound to the river, but it may not be the only source upstream of Cook Park unless the concentration of sulfamethoxazole in Rock Creek WWTF effluent was higher than the concentration in Durham WWTF effluent.

The ecotoxicological risk of carbamazepine to selected aquatic species has been studied by several research groups. Han and others (2006) studied the effects of several pharmaceutical chemicals on daphnia magna, a common planktonic crustacean, and determined that the typical concentrations of pharmaceuticals downstream of South Korean WWTFs were low enough that no appreciable risk existed. Indeed, the LC50 (concentration lethal to 50 percent of test organisms) for carbamazepine cited in that research is 111 mg/L, about 2,400 times higher than the concentration detected in Tualatin River samples. A safety factor of 1,000 typically is applied in their analyses, indicating that the concentrations found in this study may not pose a significant ecological risk to this particular test organism. Related research by Kim and others (2007) showed little risk by carbamazepine to daphnia magna or to a test fish species. They reported potential concern for effects from acetaminophen and sulfamethoxazole, but the exposure concentrations they used were more than 60 times higher than those found in Tualatin River samples; the lower concentrations measured in this study would push the risk below their hazard thresholds. Another study by Oetken and others (2005) investigated the effect of carbamazepine on an oligochaete, a midge, and a freshwater snail species. The aquatic testing in that study indicated that the concentrations found in the Tualatin River would not pose a threat to these species. In general, insufficient data exist to completely ascertain the ecological risk of pharmaceutical chemicals to species that reside in the Tualatin River and its tributaries, but limited data for compounds such as carbamazepine, acetaminophen, and sulfamethoxazole suggest that the acute ecological risk under normal low-flow conditions is low. Although pharmaceutical chemical concentrations normally might be low in the Tualatin River, accidental releases or spills of sewage could greatly increase their instream concentrations and their associated ecological risk. In addition, low-level chronic exposure to those pharmaceutical chemicals that can act as endocrine disrupters could lead to a harmful ecological effect, but such an effect cannot yet be assessed.

Use of Pharmaceutical Chemicals as Tracers of Human-Related Stream Contamination

Although the analysis of samples for pharmaceutical chemicals is expensive, the cost might be justified if one or more of the compounds proved to be a good tracer for specific sources of stream contamination. Results from previous studies indicate that caffeine and selected pharmaceuticals may be useful as confirmatory evidence of contamination, if not sufficient as primary evidence (Seiler and others, 1999; Verstraeten and others, 2005). Caffeine also has been proposed as an ideal tracer for certain types of sewage-related contamination of streams because of its high concentration in human waste and the absence of substantive natural sources (Ferreira, 2005; Wu and others, 2008). Unless the sewage-related stream contamination is large, however, the use of caffeine as a source indicator is limited because trace concentrations of caffeine appear to be common in urban streams. Indeed, caffeine was universally present in urban stream samples in this study at concentrations generally greater than 0.01 µg/L. That concentration is fairly easy to achieve with only a few spills of coffee, although other sources also may be present. For example, if coffee has a caffeine concentration of 350 mg/L (Seiler and others, 1999), and a stream such as Fanno Creek has a flow of about 4 ft³/s, then a caffeine concentration of 0.01 µg/L in the stream represents a spill of only 1 ounce of coffee residue that is completely mixed into the stream every 2.5 hours. A 1-ounce coffee spill need only mix into 5 minutes worth of the stream’s flow to achieve the highest concentration measured in this study (0.314 µg/L), a dilution level that is certainly possible, although perhaps unlikely. To use caffeine as an effective tracer for a sewer leak, the leak would have to be large enough to produce caffeine concentrations higher than a few tenths of a microgram per liter. At that level, a different indicator such as coliform bacteria or an optical brightener might prove to be less ambiguous, faster to analyze, and more cost-effective.

Although caffeine might be used only as a general tracer for mixed sources of human-related stream contamination, or as a tracer for large sewer leaks, compounds like acetaminophen, carbamazepine, and sulfamethoxazole might be useful as tracers under some circumstances. These compounds (a common analgesic, an antiepileptic drug, and an antibiotic, respectively) are unlikely to be detected in streams unless a human-related or veterinary source is present. Such a source could be a leaking sewer line, a failing septic system, illicit or unpermitted dumping, or a storm-sewer/sanitary-sewer cross connection. In the presence of one of these sources, the instream concentration of acetaminophen ought to be easily measurable, given its high concentration in municipal wastewater (29 μg/L; table 4). Instream concentrations of carbamazepine and sulfamethoxazole might be lower than that of acetaminophen, at least in untreated sewage, but these compounds are not completely removed by wastewater treatment, so they could be used as tracers of either raw or treated wastewater sources. Carbamazepine has been proposed previously as a marker for human wastewater (Clara and others, 2004). In separate research, Glassmeyer and others (2005) suggested that pharmaceutical chemicals such as carbamazepine, diphenhydramine, and caffeine might be good indicators of the presence of human wastewater. Guo and Krasner (2009) recently determined that carbamazepine and primidone may be used as indicators of upstream wastewater sources. The data derived from this Tualatin River basin reconnaissance show that several of these pharmaceutical compounds might make good tracers for human-related contamination of streams under some circumstances, but that additional research and sampling are needed to further evaluate their potential use as tracers.