Participant characteristics. The subgroup of 320 participants with complete data consisted of 203 males 44.9 ± 11.47 years of age (mean ± SD; median, 48 years) and 127 females 47.3 ± 9.24 years of age (median 48 years), with an overall mean age of 45.8 ± 10.7 years (median, 48 years). Among males, the age range was 20–75 years, and in females 21–70 years. The median and mean serum concentrations (in picograms WHO TEQ per gram lipid) of DLCs in these participants are shown in Supplemental Material, Table S2 (http://dx.doi.org/10.1289/ehp.1205739). Data on mean and median serum concentrations of PCDDs, PCDFs, DL-PCB congeners, and the most abundant NDL-PCB congeners from samples with concentrations > LOD are shown in Supplemental Material, Table S3. The median concentrations of individual congeners with concentrations > LOD correlated with median concentrations overlapping with TCDD > LOD (r = 0.998). Thus, we assumed that parameters calculated from samples overlapping with TCDD > LOD well represent those from samples > LOD.
For males and females, the mean (± SD) of volume of the thyroid gland were 11.56 ± 4.42 mL (median, 10.20) and 9.49 ± 4.75 mL (median, 8.35), respectively. Mean (± SD) serum concentrations of FT4 for males and females were 16.93 ± 2.65 pmol/L (median, 16.7) and 15.72 ± 3.22 pmol/L (median, 15.39), respectively.
Identification of the index congener.
There is general agreement that an index compound should be the most well-studied member of its class and that it should provide the largest body of acceptable scientific data (U.S. EPA 2000
). At the same time, an index chemical should be potent with regard to the expected end point. We used multiple regression with backward elimination to query the selection of TCDD as the index congener in concurrence with other PCDD or PCDF congeners. We created four models (A–D) for this purpose. When we entered thyroid volume as the dependent variable and concentrations of the seven most toxic PCDD congeners [TCDD, 1,2,3,7,8-pentachlorinated CDD (PeCDD), 1,2,3,4,7,8-HxCDD, 1,2,3,6,7,8-HxCDD, 1,2,3,7,8,9-HxCDD, 1,2,3,4,6,7,8-heptachlorinated CDD (HpCDD), and octachlorinated CDD (OCDD)] as independent variables, cross-tabulation for samples > LOD reduced the number of individuals to an insufficient 25. If we omitted the two HxCDD congeners with relatively low concentrations (1,2,3,4,7,8-HxCDD and 1,2,3,7,8,9-HxCDD), the study population increased to 62 individuals. Model A showed that with respect to thyroid volume reduction, TCDD was the most potent congener [see Supplemental Material, Table S4 (http://dx.doi.org/10.1289/ehp.1205739)]. In model D, with FT4
as the end point of interest, multiple regression eliminated four PCDF congeners when they were combined with TCDD (see Supplemental Material, Table S4). However, multiple regression did not confirm the role of TCDD with FT4
as the dependent variable and PCDD congeners as the independent variable (see Model B in Supplemental Material, Table S4) or with thyroid volume as the dependent variable and PCDF congeners as the independent variable (see Model C in Supplemental Material, Table S4).
Assessment of REPs for PCDDs, PCDFs, and DL-PCBs.
Data in show that PCDDs were associated with a decrease in both thyroid volume and FT4
level. The association between thyroid volume and dioxins decreased with the increasing number of chlorine substitutes in the compound, except for 1,2,3,7,8,9-HxCDD. The PCDFs were associated with a decrease in thyroid volume in a similar manner except for two compounds (1,2,3,4,7,8-HxCDF and OCDF). With respect to FT4
, we observed a mixed response: There was a negative association with 2,3,7,8-tetraCDF (TCDF), 2,3,4,7,8-PeCDF, 1,2,3,4,6,7,8-HpCDF, and OCDF and a positive association with 1,2,3,7,8-PeCDF and the three HxDF congeners (1,2,3,4,7,8-HxCDF, 1,2,3,6,7,8-HxCDF, and 2,3,4,6,7,8-HxCDF). The DL-PCBs were related to an increase in both thyroid volume and FT4
serum level, except for the non-ortho
-substituted congener PCB 81 for both thyroid volume and FT4
and the mono-ortho
-substituted congener PCB 105 for FT4
. Of all the congeners, TCDD was most strongly associated with a decrease of thyroid volume and FT4
level. NDL-PCBs were associated with slight changes, compared with TCDD, appearing as increases with the most abundant PCB congeners [see Supplemental Material, Table S5 (http://dx.doi.org/10.1289/ehp.1205739)]. To comply with the assumption that congeners have a similar mode of action (U.S. EPA 2000
), we calculated the REPs only for those acting in the same direction as the index chemical. Thus, congeners associated with an increase of thyroid volume or FT4
level were not further analyzed.
The calculated REPs of PCDD, PCDF, and DL-PCB congeners.
Sex and age were included as confounders (confounders 1 and 2) in all analyses. To assess the effect of confounding by other DLC congeners identified in the exposure mixture on β coefficients, we computed BMCs for thyroid volume decrease related to the serum concentration of individual congeners and entered the various combinations of congener confounders. We set both p0
(the background risk at zero concentration) and benchmark response (BMR) at 0.1, which translates to an increase in risk of 200% (Crump 1995
). Based on the Akaike information criterion, we used these two regression models: f
) = a1
, and f
) = a1
We observed that the BMC and the BMC lower confidence limit (BMCL) for TCDD were slightly influenced by the presence of other congeners in the exposure mixture [see Supplemental Material, Table S1, confounders 3–6 (http://dx.doi.org/10.1289/ehp.1205739)]. In addition, when TCDD was entered as a confounder in combination with other congeners (e.g., with the second most potent congener, 1,2,3,7,8-PeCDD), we obtained similar results. Neither of these adjustments for PCB congeners affected the BMC and BMCL value of TCDD. Therefore, in , we present REPs that were derived after adjusting only for sex and age.
The REPs in were calculated as the relation of the individual congener βi, BMCi, or BMCLi of to the βTCDD, BMCTCDD, or BMCLTCDD, respectively, of the index chemical. The REPs calculated using β coefficient, BMC, and BMCL data correlated strongly between themselves (all r-values were > 0.903, p < 0.0001). Moreover, we observed a strong correlation between the REPs calculated from the largely independent thyroid volume and FT4 data. The Spearman correlations (rS) for REPs were derived from thyroid volume and FT4 data using the βi/βTCDD (rS = 0.81, p = 0.015), BMC (rS = 0.786, p = 0.021), and BMCL (rS = 0.857, p = 0.007) approaches.
As shown in , the β coefficient–derived REP data for thyroid volume and FT4
column in ) correlated significantly with the WHO TEF values (Van den Berg et al. 2006
) (thyroid volume, rS
= 0.693, p
= 0.009; FT4
= 0.616, p
= 0.033), The best fit is logREP = 0.566, logTEF = –0.229 for thyroid volume and logREP = 0.363, logTEF = –0.399 FT4
. According to our estimates, the potencies of congeners above the central axis are greater than the TEFs, and vice versa. The BMC- and BMCL-derived REP data correlated less significantly with the WHO TEF values (data not shown).
To show our REPs in a broader context, we included in the minimum, maximum, and median values published for in vivo
REPs in the REP2004
database (see Table 8 of Haws et al. 2006
). Our REPs for all PCDD congeners studied and thyroid volume outcome [note that data on 1,2,3,6,7,8-HxCDD were not included by Haws et al. (2006)
], irrespective of the method of derivation, are between the maximum and minimum values estimated by other researchers, except for those of OCDD. Our REPs for 1,2,3,4,7,8-HxCDD and 1,2,3,7,8,9-HxCDD, where FT4
is the outcome, were higher than the published maximum estimates (Haws et al. 2006
). Of the three REP values (βi
, BMCL, and BMC) for 1,2,3,4,6,7,8-HpCDD, the βi
ratio (0.029) is smaller than the published maximum estimate (0.035) (Haws et al. 2006
). For PCDF congeners associated with thyroid volume, the REPs were close to the maximum values determined by other investigators, except for 2,3,4,7,8-PeCDF, which is higher than the minimum reported value of 0.0065 (Haws et al. 2006
). We calculated REPs for four PCDF congeners with FT4
as an outcome; for two of them (2,3,7,8-TCDF and 1,2,3,4,6,7,8-HpCDF), values were unavailable for comparison. However, the REP for 2,3,4,7,8-PeCDF fits within the range published by Haws et al. (2006)
, whereas the REP for OCDF is an outlier with regard to TEFs.
When analyzing the relative magnitude of thyroid effects of PCB congeners, we included both DL-PCB and NDL-PCB congeners. shows plotted β values for PCB congeners for thyroid volume against those for FT4 serum level shown in [see also Supplemental Material, Table S5 (http://dx.doi.org/10.1289/ehp.1205739)]. The β-coefficients for the three non-ortho-substituted PCB congeners (PCB 81, PCB 126, and PCB 169) were not plotted with regard to a high proportion of samples with concentrations < LOD. The mono-ortho-substituted PCBs (congeners 105, 156, 167, 189, 157, 123, and 114; TEFs = 0.00003) are distributed along the line of best fit (y = 0.461x – 0.003; R2 = 0.797; p = 0.001). When we included β-coefficients for both PCB and TCDD (coordinates –1.101 for thyroid volume and –0.508 for FT4), we obtained the equation y = 0.459x – 0.003 (R2 = 0.999; p = 0.001). The slopes of these two equations were not statistically different, meaning that the lower end of the PCB best fit has a value similar to that of TCDD. This analysis suggests continuity between a dioxin-like and a non-dioxin-like effect. This conforms with the four orders of magnitude difference between TEFs for TCDD and most DL-PCBs.
Figure 2 Plot of regression coefficients [β; listed in Table 1; see also Supplemental Material, Table S5 (http://dx.doi.org/10.1289/ehp.1205739)] for thyroid volume vs. PCB congener concentration (x-axis) against those for FT4 serum (more ...)