The consistent decline in tPCBs across all data groups supports the conclusion that tPCBs in Channel Catfish have decreased from 1989 to 2021 on the Ohio River. The observed decline in a top predator particularly susceptible to lipophilic contaminant bioaccumulation is potentially a positive indicator for the Ohio River aquatic community. The steepest declines were observed in the earliest time periods which was expected as these samples were taken closer to when PCB production was first banned and light weight congeners with shorter half-lives were more present in the environment. The observed decreasing trends across all data groups are likely due to a combination of factors.
While these compounds are resilient and long-lasting in the environment, they are also mobile and naturally degrade over time (ORSANCO 2006). The cessation of PCB production and subsequent decades of degradation and downstream mobility likely contributed to decreasing environmental PCB exposure to Channel Catfish. Unfortunately, few contemporary aqueous and bed-sediment collections of PCBs from the Ohio River exist across the study period to support this claim. ORSANCO completed longitudinal collections of aqueous PCB concentrations on the Ohio River between 1997 and 2004. Every sample collected during this period exceeded criteria protective of human health (ORSANCO 2006). Given the magnitude of the exceedances, longer congener half-lives of highly chlorinated PCBs, and the resource intensive nature of the collections few aqueous collections have been obtained since. The observed decline in Channel Catfish tPCBs warrants future analyses of aqueous or bed-sediment PCB concentrations on the Ohio River to more confidently determine if environmental exposure has similarly decreased.
Aside from decreased environmental exposure, a decline in lipid content of Ohio River fishes may partially explain the observed tPCB trends. Decreasing lipid content has been observed elsewhere in freshwater fishes (USEPA 2017) and anecdotally in other Ohio River species during preliminary investigation of the ORSANCO fish tissue contaminant dataset. Due to the lipophilic nature of PCBs, fish with lower lipid contents tend to contain proportionally lower tPCBs in their tissues (USEPA 2017). A recent study by Holm et al. (2022) proposed a potential link between declining fatty-acids in marine food-chains and increasing water temperatures. Inclusion of potential contributing factors of climate change on decreasing tPCBs in fish tissue were beyond the scope of this research as paired water temperature data were not collected at the time of composites. ORSANCO routinely collects surface water temperatures at fixed locations along the Ohio River. Thomas et al (2019) analyzed these data from a similar time frame as this study and determined temperatures have been increasing since the 1990s. It is possible the observed temporal decline in lipid content may partially reflect a decreased capacity of Channel Catfish, and the larger Ohio River aquatic community to retain and accumulate lipophilic contaminants. Future surveys and analyses would be necessary to determine whether this is the case and to what degree the observed tPCBs decline is attributable to declining lipid content rather than a decrease in environmental exposure.
Seasonal variation of fish tissue lipid content can further complicate interpretation of temporal tPCBs trends. It is well established in the literature that lipid concentrations in fish tissue can fluctuate throughout the year relative to spawning cycles, (Butcher et al. 1997; USEPA 2017; Louzeau et al. 2001; Montano et al. 2022), prey availability, and other environmental factors (Nelson and McPherson 1987; Moore and Potter 1972; Newsome and Leduc 1975; Medford and Mackey 1978; Pierce et al. 1980). Fish tissue trend analyses can minimize this potential variation by restricting sample composites to a particular time of year. Seasonal restriction of the ORSANCO dataset was not a viable option given only sporadic dates of collection were recorded for some composites or individual specimens. To maintain optimal annual sample sizes within each data group, the inclusion of lipid predictor variables in the multiple regressions and the lipid-normalization techniques were employed to minimize the effect of seasonal lipid variation in lieu of further subsetting or omission of composite data. The resulting lipid-normalized data generated the most conservative rates of tPCBs decline relative to the other two methods, which is similar to past findings when a declining lipid content trend was also observed (USEPA 2017).
This study demonstrates how historical fish contaminant data can be utilized to establish general trends relative to the fate of legacy contaminants in the absence of direct environmental measures. These data were not collected with the intention of long term trends analyses, but with caution, they can still provide insight concerning long term trends when biases in the dataset are apropriately addressed. Environmental regulatory agencies could use this framework as impetus to revisit impaired waterway listings to determine if those sites are still reflective of an impaired condition.
One shortcoming of using multiple regressions to minimize inherent biases is the resulting residuals no longer possessed concentration units (i.e., mg/kg). This can limit the usefulness of such trends to the general public that consume Ohio River fishes. The comparison of the proportion of samples designated as DNE (%DNE) across the study period for various datasets allowed for the corroboration of observed trends, retention of tPCBs, and provides a means to disseminate meaningful results. Since 2010, there have been three Channel Catfish composites that were classified as DNE; one in 2010, 2013, and 2019. It is not surprising that these three samples were collected from the upper reach (upstream of river mile 203.9; Thomas et al. 2011) of the Ohio River. While this stretch is approximately one fifth of the length of the river, the records from this area contain 57% of DNE samples in the dataset. Figure 5 shows a potential explanation of why higher concentrations could be observed in the upper end of the Ohio River. The observed decreases in %DNE has regulatory and human health implications particularly relevant to the upper portion of the Ohio River where fish consumption advisories have been historically more restrictive.
Though collection methods remained consistent throughout the study period, other attributes of the samples varied and potentially introduced biases within the dataset. In order to address each bias, they must first be identified and their effect on resulting tPCBs understood to more accurately evaluate temporal trends in tPCBs. Potential biases identified in ORSANCO’s long-term fish contaminant dataset largely fall into three categories: analytical, physiological, and spatial. Advancements in analytical methods led to several different methods employed to detect and enumerate tPCBs in the composite samples. Additionally, ORSANCO contracted analytical services with multiple laboratories throughout the study. While each PCB detection method is standardized, the means of quantification and final reported tPCBs and how they are calculated and reported can vary between laboratories (Butcher et al. 1997). Similar to Butcher et al. (1997) the changes in laboratories and analytical methods precluded the direct temporal comparison of resulting contaminant concentrations. Split composite samples with which to compare intra-method results from a single laboratory or inter-laboratory results using a single method were absent from the dataset. Without the ability to directly compare or quantify potential variation across analytical laboratories and methods, the dataset was subdivided and analyses performed separately on the previously defined data groups. Of all the species in the ORSANCO dataset, the ubiquity of Channel Catfish in the Ohio River allowed for PCB analysis at a wide spatial and temporal scale, across the defined data groups.
The necessity to proceed with Channel Catfish also required addressing two potential physiological biases. A length bias exists because the premise for the collections was rooted in the collection of samples that were most likely to be caught and consumed by the general public (i.e. anglers, commercial fishermen, etc.), leading to the exclusion of extremely small or large specimens. Even with a focus on averaged size fish, length within the dataset still varied annually, likely due to normal population dynamics. It is important to note that length is being used as a surrogate for age of the fish and age is extremely important as it represents duration of environmental exposure to PCBs.
Limited metadata had implications for lipid standardization, as majority of the records lacked a specific date of collection. Without collection time more resolute than year it was difficult to account for any potential seasonal variations in sample lipid content. During the spawning period, feeding intensity declines and egg production increases, therefore it is important to consider seasonal fluctuations in regards to changing tPCBs (Butcher et al. 1997; USEPA 2017; Louzeau et al. 2001; Montano et al. 2022). Feeding is the primary mechanism with which PCBs enter the food web (as opposed to assimilation of dissolved contaminants in surrounding water), especially for PCB congeners with more than four chlorine atoms (Loizeau et al. 2001). Different habitat preferences and prey sources (e.g., animal versus detritus) impact the amount and variation of PCB congeners introduced into the food web. Environmental factors such as water temperature and ecosystem productivity influence the rates of consumption, growth, respiration, and reproduction of fishes on all trophic levels, which directly impacts the uptake of PCBs and the degree to which bioaccumulation occurs. Fishes’ physiological responses to environmental factors (e.g., growth rates of predator/prey and lipid content/availability) will vary from each of the Ohio River’s navigational pools as they are all unique in their size and composition of drainage areas, in-stream physical habitats, nutrient loading, water quality parameters, and the presence/absence of tributaries of varying orders (Thomas et al. 2019). The aforementioned biases inherent to unknown factors pertaining to seasonal variation of lipid content were accounted for by including lipid content as a predictor variable in the multiple regression equations as well as standardizing results by lipid content.
Spatial biases also needed to be addressed in these analyses given the relative distribution of potential PCB sources along the Ohio River (Fig. 4). According to a recent study on the Hudson River, it was concluded that declines in fish tissue levels of total PCB homologue equivalent were generally attributable to declines in sediment and water exposure concentrations (USEPA 2017). While it is clear that tPCBs in fish tissue are decreasing, it is difficult to quantify how much of that decrease can be attributed to decreasing environmental exposure, confounding factors such as seasonal variation of lipid content observed in fish tissue samples, rising water temperature affecting rate of uptake, or declining lipid content observed in fish tissue samples between 1989 and 2021. Other studies from heavily contaminated large river systems have shown decreases in tPCBs being roughly double the decrease in lipid-normalized PCB concentrations (USEPA 2017); one study suggests that the disparity between these two measurements is likely due to inherent biases related to seasonal fluctuation of lipid content. This study addressed these disparities using three different approaches: changes in observed tPCBs, changes in multiple regression modeled residuals over time, and changes in lipid-normalized concentrations over time (Table 3). Percent annual changes in tPCBs were the highest percent changes of the three different approaches. This is likely due to the fact that this raw value does not account for biases related to length of fish collected, lipid content, or sample location. The only biases accounted for in this approach are differences in laboratory and analytical method. Residuals extracted from the optimum multiple regression models represent the attempt to remove as much bias as possible, allowing the software to incorporate the relationships of the predictor variables to observed tPCBs. The use of residuals extracted from the models to represent tPCBs resulted in a loss of concentration units with which to quantify temporal changes in concentration. The calculation of mean annual rates of percentage change provided a means to make such comparisons. Lipid-normalized concentrations were the lowest percent changes of the three different approaches. Lipid content of samples from the Ohio River showed a declining trend over the span of the study, which is similar to what was observed in the Hudson River study which concluded that lipid-normalized PCB concentrations represent the most conservative approach with regard to rates of decline (USEPA 2017).
These findings suggest that tPCBs are likely decreasing on the Ohio River. The rate of decrease is difficult to quantify, however agreement across all three approaches and similarities to the findings of similar studies, lends confidence to this conclusion. While it is difficult to quantify rates of decline, PCB content in fish tissue has decreased. Fish tissue records classified as “Do Not Eat” have almost disappeared from modern collections. While attributing this directly to decreasing environmental exposure or other confounding environmental factors remains elusive, our analyses indicate that Channel Catfish on the main stem Ohio River presently contain far fewer tPCBs than in the past.