F and PCB Levels in Cows' Milk

Over 180 milk, soil, and grass samples, taken from 38 farms across 3 different river systems (River Dee, Trent, and Doe Lea/Rother/Don) in the United ...
1 downloads 0 Views 243KB Size
Research Effects of River Flooding on PCDD/F and PCB Levels in Cows’ Milk, Soil, and Grass I A I N R . L A K E , * ,† CHRISTOPHER D. FOXALL,† ANDREW A. LOVETT,† ALWYN FERNANDES,‡ ALAN DOWDING,§ SHAUN WHITE,‡ AND MARTIN ROSE‡ School of Environmental Sciences, University of East Anglia, Norwich NR4 7TJ, U.K., Central Science Laboratory, Sand Hutton, York YO41 1LZ, U.K., and Food Standards Agency, Aviation House, 125 Kingsway, London WC2B 6NH, U.K.

This paper presents the results of a study examining whether the flooding of pasture by rivers gives rise to higher PCDD/F and PCB concentrations in cows’ milk. Over 180 milk, soil, and grass samples, taken from 38 farms across 3 different river systems (River Dee, Trent, and Doe Lea/Rother/Don) in the United Kingdom, were analyzed for PCDD/Fs and PCBs. The concentrations were compared between flood-prone farms, where the animals had access to pasture that is often flooded, and control farms where the land does not flood. The results indicated that concentrations of PCDD/Fs and PCBs in cows’ milk were higher in samples taken from farms prone to flooding, but only from the river systems flowing through industrial and urban areas. Raised levels of PCDD/F and PCBs were also found in soil and grass from farms prone to flooding providing strong corroborative evidence that the higher concentrations in cows’ milk from such areas is likely to be due to the ingestion of contaminated grass and soil. Overall, the results provide strong evidence that flooding of pastureland can indeed result in elevated concentrations of PCDD/Fs and PCBs in milk from the farms so affected.

Introduction Polychlorinated dibenzo-p-dioxins, polychlorinated dibenzofurans (PCDD/Fs, collectively referred to as “dioxins”), and polychlorinated biphenyls (PCBs) are widely recognized environmental and food contaminants. They mainly originate from anthropogenic activity, and their ubiquitous presence in the environment has caused significant concern due to their persistence (half-lives measured in decades), their bioaccumulation potential in the tissues of animals and humans, and their toxic properties (1). It has been estimated that 95% of human intake of PCDD/ Fs and PCBs is from the consumption of food (2) and there are a variety of pathways through which PCDD/Fs and PCBs can contaminate foodstuffs. It is widely recognized that atmospheric deposition is an important source of this * Corresponding author phone: +44 1603 593744; fax: +44 1603 591327; e-mail: [email protected]. † University of East Anglia. ‡ Central Science Laboratory. § Food Standards Agency. 10.1021/es051433a CCC: $30.25 Published on Web 10/28/2005

 2005 American Chemical Society

contamination, and it has been demonstrated that levels of these contaminants are often higher in milk samples from farms close to urban areas or industrial facilities as opposed to those from farms in rural areas (3, 4). However, transfer through river systems may also be important as suspended particulate and bottom sediment in river systems can serve as a sink for PCDD/Fs and PCBs and then provide a longterm source of release for these pollutants (5). This has been confirmed by studies of PCDD/F and PCB loadings in the sediments of rivers worldwide (5-8). Despite this association with river sediments, little attention has been paid to the possibility that the flooding of land by water carrying PCDD/F and PCB contaminated sediment could be an important source of localized contamination. This would provide another mechanism for transfer to the human food chain. Higher PCDD/F concentrations have been found in cows’ milk and soil from the flood plains of the Rhine Delta as opposed to those from background areas (9), although this study was based upon only 3 moderately polluted sample sites. Elevated levels of PCDD/Fs in cows’ milk from farms near the River Rother in the U.K. have been associated with flooding events washing contaminated river sediment onto pasture grazed by cattle (10). However, this study lacked data on concentrations of PCDD/Fs in the surface soil and was based on only 2 milk samples. This study represents the first controlled investigation into the potential problem of flooded pastures and was designed to investigate the possibility that the over-bank flooding of grazing land might affect the concentration of PCDD/Fs and PCBs in cows’ milk samples. Milk is an important commodity to study because its consumption may account for 30% of general population exposure to PCDD/Fs and PCBs, and is an important foodstuff consumed by young children (11). Furthermore, frequently flooded land is more likely to be devoted to pasture than to arable cropping. To provide supporting evidence for any trends emerging from the milk data, soil and grass samples were also collected.

Materials and Methods Study Area. The study was focused on three river systems in England and Wales illustrated in Figure 1a. Two of these rivers, the Doe Lea/Rother/Don (Figure 1b) and the Trent (Figure 1c), flow through substantial urban and industrial areas (and hence might be classed as “potentially contaminated”). The River Dee (Figure 1d) passes through predominantly rural areas and might therefore be expected to be relatively “uncontaminated”. Once these three river systems had been selected, maps of the dairy farm distribution and flood history were used to identify a number of farms where a significant proportion of the land was subject to regular flooding. For each flood-prone farm, a nearby farm whose land was not subject to flooding was selected as a control. By referring to the location of other potential aerial sources of PCDD/Fs and PCBs in the area, such as industrial facilities and motorways, control farms were selected that would be expected to be subject to similar levels of aerial deposition of these contaminants as the neighboring flood-prone farms. Thirty eight farms satisfied these criteria and were selected to take part in the monitoring program. It was decided that rather than clustering samples in relatively few “uncontaminated” or “potentially contaminated” sections it would be more informative to use sites encompassing both the upper and lower reaches of each drainage system. Of the 38 farms selected, 34 were grouped into pairs of flood-prone VOL. 39, NO. 23, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

9033

FIGURE 1. The three river systems: (a) locations; (b) sampling sites on the Doe Lea/Rother/Don river system; (c) sampling sites on the Trent River system; and (d) sampling sites on the Dee river system. and control farms. Five pairs were located on the Doe Lea/ Rother/Don river system, seven were on the Trent, and the remaining five were on the Dee. Sampling and Sample Analysis. To take account of seasonal variations in feeding regimes for dairy cattle, the sampling program was designed as a three-phase process. The first batch of samples (milk, soil, and grass) was collected in October 1998 when the majority of the dairy herds were still grazing exclusively on pastureland and had not spent any significant time indoors. The second phase (March 1999) was timed so that samples (milk only) were taken while the 9034

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 23, 2005

cattle were still predominantly housed indoors. The final sampling phase (milk, soil, and grass) was carried out in August 1999 by which time the cattle had been feeding outdoors for at least three months. In total, 114 milk, 79 soil, and 79 grass samples were collected during the three phases. PCDD/F and PCB concentrations were analyzed according to methods fully UKAS accredited to the ISO 17025 standard (12). All analyses were based on the seventeen 2,3,7,8-Cl substituted PCDD/F congeners, four nonortho PCBs (77, 81, 126, and 129) and twenty-one ortho congeners (18, 28, 31, 47, 49, 51, 52, 99, 101, 105, 114, 118, 123, 128, 138, 153, 156,

FIGURE 2. Distribution of total TEQ for milk, soil, and grass concentrations: (a) milk; (b) soil; and (c) grass.

TABLE 1. Median TEQ Concentrations (ng/kg Fat) of Milk Samples on Flood-prone and Control Sites Stratified by River System and Sampling Datea October 1998

March 1999

August 1999

river system

flood-prone

control

flood-prone

control

flood-prone

control

total

Doe Lea/Rother/Don Trent Dee

5.46 (6) 4.67 (7) 1.73 (4)

4.01 (6) 3.02 (8) 1.97 (6)

4.45 (6) 3.97 (7) 1.70 (5)

3.79 (5) 2.47 (8) 1.91 (6)

2.04 (6) 2.42 (6) 1.09 (5)

2.13 (6) 1.49 (8) 1.05 (6)

3.63 (35) 2.67 (44) 1.58 (32)

Total

4.29 (17) 2.85 (20) 3.60 (37)

a

3.29 (18) 2.66 (19) 2.90 (37)

1.70 (17) 1.46 (20) 1.56 (37)

Figures in parentheses are the number of samples.

157, 167, 180, and 189). TEQ values were calculated using WHO-TEFs and are reported as upper-bound total TEQs incorporating PCDD/F, ortho and nonortho PCB concentrations (13). One hundred and eighty three samples (milk 111, soil 56, grass 16) were analyzed for PCDD/Fs and PCBs and the sample design ensured that the sample types were evenly spread across river systems and between flood-prone and control sites for all of the time periods.

Results Figure 2 presents the distributions of milk, soil, and grass TEQ concentrations. The milk data (Figure 2a) demonstrate large variations in TEQ concentrations (0.71-14.08 ng/kg fat). The median value is less than the mean indicating that the distribution has a long right tail with a majority of lower values and a few higher ones. The mean value (3.10 ng/kg fat) is greater than that reported by the latest U.K. Total Diet Study (0.9 ng/kg fat) (14). This is expected as two of the three river systems were deliberately selected because their catchments contained substantial urban and industrial areas. The soil (Figure 2b) and grass data (Figure 2c) demonstrate similar distributions although the soil data have a much greater range (2.48-746.36 ng/kg dry weight) whereas the grass data have a lower range of values (0.11-1.31 ng/kg dry weight). To determine the impact of flooding on TEQ concentrations in milk, soil, and grass the data were analyzed in two stages. To obtain an overview of these data the first stage stratified the samples by sampling period, and river system, and then compared the values between flood-prone and

TABLE 2. Median TEQ Concentrations (ng/kg Dry Weight) of Soil Samples on Flood-prone and Control Sites Stratified by River System and Sampling Datea October 1998 river system flood-prone Doe Lea/ Rother/Don 57.57 (3) Trent 24.41 (4) Dee 3.45 (2) Total a

control

August 1999 flood-prone

10.91 (3) 14.56 (6) 4.03 (3) 16.11 (7) 4.27 (3) 3.65 (5)

control

total

11.06 (6) 5.66 (8) 4.62 (6)

12.86 (18) 7.60 (22) 4.34 (16)

23.00 (11) 4.27 (7) 10.58 (18) 6.73 (20) 7.80 (18) 6.99 (38)

Figures in parentheses are the number of samples.

control sites. These trends were then investigated statistically using a multivariate method. The results are now presented followed by a discussion of these at the end of the paper.

Data Stratification To obtain an overview of the data, tables were created in which the median TEQ values for milk, soil, and grass samples were stratified by type of site, river system, and sampling period. Median TEQ values were used, owing to the relatively low numbers of observations in most subgroups and a skewed distribution of values. The results are presented in Table 1 for milk, Table 2 for soil, and Table 3 for grass. Considering the milk data presented in Table 1, the highest TEQ concentrations were occurring on the Doe Lea/Rother/ Don followed by the Trent and the Dee, a trend which accords VOL. 39, NO. 23, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

9035

TABLE 3. Median TEQ Concentrations (ng/kg Dry Weight) of Grass Samples on Flood-prone and Control Sites Stratified by River Systema August 1999 river system

flood-prone

control

total

Doe Lea/Rother/Don Trent Dee

0.31 (3) 0.31 (4) 0.13 (2)

0.22 (2) 0.29 (3) 0.22 (2)

0.30 (5) 0.29 (7) 0.13 (4)

Total

0.26 (9) 0.29 (7) 0.28 (16)

a

Figures in parentheses are the number of samples.

with the pattern anticipated in the experimental design. In terms of sampling date, the median values were lower in each successive time period. Comparison of the values for flood-prone and control farms within each subgroup indicates a consistent tendency for median TEQ concentrations to be higher in milk samples taken from flood-prone sites as opposed to control sites. However, this is only evident on the Trent and the Doe Lea/Rother/Don and not on the Dee. This is an important finding as the River Dee was the only system designated as “uncontaminated”, with no heavy industry upstream of any of the sampling locations. Another interesting observation is that milk samples from flood-prone farms have higher TEQ concentrations than those from control farms in March 1999 even though the cattle were still predominantly housed indoors at this time. TEQs are internationally accepted measures of PCDD/F and PCB toxicity. However, to enhance the robustness of the findings the subgroup analysis was repeated using another measure of contamination known as ΣICES7 which is based upon the summed absolute concentrations of 7 PCB congeners (28, 52, 101, 118, 138, 153, and 180). This demonstrated that concentrations from flood-prone sites were again found to be consistently higher than for control sites in the Trent and the Doe Lea/Rother/Don systems, but not on the Dee. The soil TEQ concentrations are presented in Table 2 and indicate a pattern almost identical to that observed in the milk data with highest levels occurring in the Doe Lea/Rother/ Don, followed by the Trent and the Dee, and with slightly lower TEQs in the later time period (although this effect appears less pronounced than in the milk data). Concentrations from flood-prone farms are consistently higher than those from control sites along the two contaminated river systems but again this effect is not apparent on the River Dee. In fact, a closer examination of Tables 1 and 2 suggests a consistent trend of lower TEQ levels on flood-prone locations in comparison to control sites on the River Dee for both soil and milk samples (apart from milk TEQ in August 1999 when the values are nearly identical). Table 3 presents the results for the grass samples. These results have to be interpreted with caution due to the very small number of samples in each of the subgroups. Nonetheless the results suggest that grass concentrations are lowest on the River Dee and that flood-prone sites have higher TEQ on the Doe Lea/Rother/Don and the Trent.

Multivariate Regression A multivariate regression analysis was applied to the milk and soil TEQ values. No such analysis was performed on the grass data due to the small number of samples (16). For both the milk and soil data the natural logarithm of the TEQ value was used as the dependent variable to account for their skewed distributions. The first multivariate regression analysis was performed on the milk data and related the natural logarithm of the milk TEQ to 7 binary (1/0) independent variables. The results 9036

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 23, 2005

TABLE 4. Results of the Multivariate Regression for Natural Logarithm Milk TEQa independent variable

regression coefficient

standard error

DoeLRD 0.48 0.12 Trent 0.28 0.11 October98 0.65 0.090 March99 0.63 0.11 Flood 0.33 0.10 March99flood -0.090 0.16 Deeflood -0.49 0.16 model deviance ) 15.1 d.f. ) 103

p