7% per cm; and for fish with 4% learn more lipid, the rate was 2.1% per cm. Coho with high filet % lipid exhibited higher PCB concentrations even at small lengths, but PCB concentrations appeared to increase at a slower rate in these fish as length increased. While these interactions improved the fit of the model, they represent only minor changes in the primary relationships among PCB concentrations and time, body length, % lipid, and season that were suggested by the original main effects model

described previously. Exploratory plots and GAM models suggested patterns for chinook similar to coho with a rapid decline in filet PCB concentrations until the mid to late 1980s, then a slower decline to the 2010; increases in PCB concentrations as both body length and % lipid in filets increased; and higher PCB concentrations Obeticholic Acid molecular weight in filets from fish collected in the fall than in the summer. We fit the same set of models

that we fit for coho, and estimated the point of intersection of piecewise linear trends to be 1985, one year later than for coho. The two models for chinook with lowest AIC included the same predictors as the two best-fitting models for coho: predictors for the model with minimum AIC were piecewise linear time trends, fish body length, % filet lipid, and season collected (Table 4). The model including the additional predictor of location fit slightly worse. The estimated rate of decrease in PCB concentration was − 16.7% per year for 1976–1985 (95% CI: − 18.2% to − 15.2%) and − 4.0% per year for 1986–2010 (95% CI: − 4.4% to − 3.6%; Table 5 and Fig. 3). PCB concentration increased by 2.3% per cm of length (95% CI: 2.1% to 2.5%) and by 10.2% for each 1% increase in % lipid (95% CI: 8.9% to 11.6%). For chinook at a given length and % lipid content, PCB concentrations were 80.6% larger for fish caught in the fall than the summer (95% CI: 67.7% to 94.5%). As with coho, we also examined models that included condition as a predictor using a smaller dataset containing only records with condition. Similar to our findings

with coho, models with minimum AIC were the same as those for the larger dataset; models including condition fit substantially worse. We examined models with all combinations of 2-way interactions among the predictor variables in the model just described; among those models, the one with minimum AIC included 2-way interactions between chinook body Clostridium perfringens alpha toxin length and the two time trends, between length and season, and between length and % lipid. The interactions between body length and the time trends suggested that larger chinook exhibited slower declines than smaller fish in the early time period (− 17.7% for a 60 cm fish vs − 13.3% for a 100 cm fish), but more rapid declines in the later time period (− 3.5% for a 60 cm fish vs − 5.3% for a 100 cm fish). The interaction between chinook body length and season caught was due primarily to differences in filet PCB concentrations for smaller fish between the two seasons.