¯ 3M Company Page t of 63 FINAL REPORT Epidemiology, 220-3W-05 Medical Department 3M Company St. Paul, MN 55144 Date: October 11, Title: A Longitudinal Analysis of Serum Perfluorooctanesulfonate (PFOS) and Perfluorooctanoat¢ (PFOA) Levels in Relation to Lipid and Hepatic Clinical Chemistry Test Results from Male Employee Participants of the 1994195, 1997 and 2000 Fluorochemical Medical Surveillance Program Study Start Date: Protocol Number (not applicable) IRB Approval J’.uly I, 2001 Exempt X IRB ApprovalDate: (not applicable as these data ar~ from a medical surveillance progam) Principal Investigator. Geary W. Olsen, D.VA4., Ph.D) Co-investigators: Michele M. Buriew, M.SJ Jean M. Bun’is, R.N., M.P.H.I Jeffrey H. Mandel, M~D., M.P.H.I Study Di~’cto~. Jeffrey H. Mandel, M.D., M.P.H. I. 3M Medical Department, 220-3W-05, St. Paul, MN 55144-1000 "(Corrections made from previous version) 3M MN02482163 1799.0001 Page 2 of 63 ABSTRACT The 3M fluor~chemical medical surveillance pro.m"am was conducted in t994/95, 1997 and 2000 at the company’s Antwerp (Belgium) and Decatur (Alabama) manufacturing plants. Althou~ cross-sectional assessments of the data have reported, the opportunity to conduct a longitudinal assessment became possible as a result of a large number of employee participants in the 2000 fluorochemical medical surveillance program. A total of 175 male employees voluntarily participated in the 2000 program and at least one of the two previous program years. A total of 106 (61 percent) of the 175 employees participated in the 1994/95 program and 1 l0 (63 percent) of the 175 parti¢ipatedin the 1997 program. Of these 175 employees, a total of 41 (24 percent) participated in all three years (Antwerp = 21, Decatur = 20), 65 (37 percent) participated in 1994/95 and 2000 (Antwerp = 45, Decatur = 20) and 69 (39 percent) participated in 1997 and 2000 (Antwerp = 34, Decatur - 35). There were insufficient number of female employees to conduct any meanin~ul lon~tudinal ~sessment. Only 14 female employees participated in the 2000 fluorochemical medical surveillance program and at least one of the previous program years. Serum peffluorooctanesulfonate (PFO$) and perfluorooctanoate (PFOA) were assayed in each surv,eillance program year although the method of analysis (high performance liquid ¢h .romatography mass spectrometry) differed sli~tly between years. A different research laboratory was used to assay PFOS and PFOA in each year. The same hospital laboratory analyzed the clinical chemistries for all thre~ surveillance years, These included: cholesterol (m~dl), high density lipoproleins (HDL, 3M MN02482164 1799.0002 3M Company Page 3 or 63 m~dl) and triglycerides (m~dl); alkaline phosphatase (IU/L), gamma glutamyl transferase (GGT, IU/L), aspartate aminotransferase (AST, [U/L), alanine aminotransferase (ALT, I’U/L). total and direct bilirubin (m~dl). Most reference ranges remained relatively constant over dine except for ALT. In each surveillance year, potential confounding factors were also determined. These covariates included age, body mass index, number of alcoholic drinks per day and cigarettes smoked per day. The continuous outcomes of lipid and hepatic clinical chemistry tests were evaluated as repeated measures incorporating the random subject effect fitted to a mixed model by the M!XED procedure in the SAS statistical package. Restricted maximum likelihood estimates of variance parameters were computed. Adjusted regression models were built by introducing all covariates and testing the covariance structure. Significant coefficients were defined when the p value was < .05. There was a positive association between PFOA and serum cholesterol and ¯ triglycerides over time but not with PFOS. This was association was limited to the Antv,’erp employees and, in particular, the 21 Antwerp employees who participated in all three surveillance years. This positive association between PFOA and serum lipids is opposite the inconsistent toxicological evidence that suggested a possible hypolipidemic effect of PFOA in rodents and no effect in primates. Adjusting for potential confounders, there were no temporal changes associated with the fluorochemical tests, PFOS, PFOA and TOF. and the hepatic clinical chemistry, tests. Limitations of this study included the number Of employees with three years of surveillance data (only 24% of the 175 subjects), the inability to analyze temporal changes due to small number~ in female employees, the use of different laboratories and 3M MN02482165 1799.0003 3M Company Pag~ 4 of 63 the associated systematic (experimental error) with each fluorochemical assay for ~he three surveillance pros~am years and the lower levels of serum PFOS and PFOA measured in each pro.~’am year among these employees compared with ~ose that cause effects in laboratory animals. 3M MN02482166 1799.0004 3M Company Page 5 of 63 kNTRODUCTION The 3M fluorochemical medical surveillance program is conducted on a routine basis at the company’s Antwerp (Bel~urn) and Decatur (Alabaraa) manufacturing plants. Employee pa.nicipation is voluntary. Prior to 1994, 0nly total organic fluorine was measured and no specific fluorochemical analytes were measured. Serum perfluorooctanesu[fonate (PFOS) and perfluorooctanoate (PFOA) have been routinely assayed since 1994/95 rather than total organic fluorine as the analytical capabilities have improved. Cross-sectional analyses of the 1994/95 and 1997 medical surveillance program data and the 2000 data in relation to Antwerp and Decatur employees’ serum PFOS levels have been reported elsewhere (Olsen et al, 1999a, 1999b, 2001). In the 1994/1995 medical surveillance program, a total of 178 male employees participated (Antwerp = 88; Decatur -- 90) and 149 male employees participated in the 199"] program (Antwerp = 65; Decatur = 8~t). For these two program years, there were too few female participants to include in the data analysis (Olsen et al 1998). In the 2000 fluorochemical medical surveillance program, there were considerably more participants: 421 males (Antwerp = 206; Decatur = 215) and 97 females (Antwerp -- dg; Decatur = 48). It was suspected that the increased voluntary participation in 2000 was due to increased employee awareness of the persistence and prevalence of PFOS in human tissue and the environment and the company’s May 16. 2000 phase out announcement that it would cease the production o(pcrfluorooctanyl chemistry in certain rcpcllent~ and surfactants by the end of 2000. Regardless of the surveillance year, there have been several consistent differences between the Antwerp and Decatur male employee populations. The Antwerp male 3M MN02482167 1799.0005 3M Company Page 6 of 63 employee population has been si~ificantly younger than Decatur, has had lower Body Mass Indices (BMI) and higher self-reported daily consumption of alcohol. In addition, the Antwerp male employee population clinical chemistry profiles were different for several tests including lower mean alkaline phosphatas¢ and triglyceride values and higher total bilirubin and I-IDL values than the D~catur male employee population. Analyses of workers’ lipid and hepatic clinical chemistry r~sults have not been associated with hypolipidemic effects and PFOS as reported in rodents (3M Company 2000; Haughom and Spydevold 1992; Ikeda et al 1987; Pastoor et al 1987; Seacat et al 2001a; Sohlenius et al 1993) and primates (Seacat et al 2001b). In the 2000 medical surveillance program, statistical analyses also examined the relation between PFOA and a calculated total organic fluorine indc.x (TOF) to clinical chemistries, hematology, thyroid hormones and urinalyses (Olsen et al 2001). A positive association was observed between triglycerides andPFOA; however, this association was opposite the data that have inconsistently reported a h.vpolipidcmia effect of PFOA in rodents (Haughorn and Spyclevold 1992; Pastoor et al 1987) and no effect in primates (Butenhoffet al 2001). Furthermore, this positive association between PFOA and triglycerides has not been observed at the 3M Cottage Grove manufacturing plant (Olsen et al 20(30) where employees’ serum levels have, historically, been much higher than those measured among Antwerp and Decatur employees (Olsen et al 1999; 2001a; Olsen et al 2001b). The inability to, assess temporal changes in cross-sectional studies is a wellknown limitation of this design. The large participation of employees in the 2000 fluorochemical medical surveillance who may have participated in the 199ad95 and/or 199"/surveillance programs at these two manufacturing sites allowed for an opportunity 3M MN02482168 1799.0006 3M Company Page 7 ot’63 to conduct a longitudinal analysis among the male employee population. Altogether, a total of 175 employees (Antwerp = 100; Decatur = 75) who participated in 2000 had also participated in at least one previous fluorochemical medical surveillance exam since 1994~95. Therefore, the purpose of this analysis was to conduct a longitudinal assessment of this 6 year time period regarding the relationship of PFOS, PFOA and TOF to the medical surveillance data collected on these 175 Antwerp and Decatur male employees. METFIODS Dam Collection Data were compiled from the 1994/95, 1997 and 2000 fluorocherr~cal medical surveillance prob.’am databases. A total of 175 male employees participated in the 2000 program and at least one of the two previous program years. A total of 106 (61 percent) of the 175 employees participated in the 1994/95 program and 1 I0 (63 percent) of the 175 participated in the 1997 program. Of these 175 employees, a total of 41 (24 percent) participated in all tl’u’ce years (Antwerp -- 21, Decatur = 20), 65 (37 pcn:ent) participated in 1994/95 and 2000 (Antwerp = 45, Decatur = 20) and 69 (39 percent) participated in 1997 and 2000 (Antwerp = 34, Decatur = 35). For purposes of brevity, these three subpopulations will hat, after b¢ referred to as sabcohorts A, B and C. Demo.m’aphic d~ta (age, BMI, alcoholic drinks per day and cigarettes per day) were recorded for each employee in each surveillance year. A standard set of clinical chemistries and hematology, data was also obtained for each employee. Given results from previous toxicological studies, the longitudinal analyses focused on lipid 3M MN02482169 1799.0007 3M Company Page 8 of 63 [cholesterol (m~dl), high d~nsity lipoproteins (I-IDL, m~dl) and triglyceddes (m~dl)] and hepatic [alkaline phosphatas¢ (I’U/L), gamma glutamyl transferase (GGT, IU/L), aspartate aminotransferase (AST, I’U/L), alanine aminotransferase (ALT, IU/L), total and direct bilirubin (mg/dl)] clinical chemistries that were measured in each program year by the same laboratory (Allina Laboratories, St. Paul, MN). Reference ranges were relatively constant over time, althou~ for ALT the range declined from 20-65 IU!L in 1994/95 to 1-40 IU/L in 1997 and 2000. Fluorochemical Analyses PFOS and PFOA were assayed in 1994/9.5, 1997 and 2000. However, the method of analysis differed slightly for each year. In 1994/95, the method used te~’abutylan,h-nonium to ion-pair with PFOS and PFOA in the serum (.]’ohnson et al 1996). The ion-pairs were then extracted with e~hyl acetate. The abstraction product was then analyzed using high-performance liquid chromatograph-thermospray mass spectrometry. In 1997, the serum samples were analyzed by liquid chromatography/mass spectrometry, using selected ion monitoring in the negative-ion mode (Anderson and Mulvanna 1997a; 1997b). In 2000, sere samples were extracted using an ion-pairing extraction procedm’ (Hansen et al, 2001). Only in 2000 were the extracts quantitatively analyzed for PFOS and PFOA as well as the other analytes: PFHS~ (perfluorohrxanesulfonate), PFOSAA (Nethyl perfluorooctanes,ulfonamidoacetate), M570 (N-methyl perfluorooctanesulfonamidoacetate), PFOSA (perfluorooctanesulfonateamide) and M556 (perfluorooctanesulfonamidoacctate). High-p~rformance liquid chromatography/electrospray tandem mass spectrometry (I-IPLC/ESMSMS) was the 3M MN02482170 1799.0008 3 M Compnny Page g of 63 technique used in 2000. The samples were evaluated versus an extracted curve from a human serum matrix. Analyses were conducted at different laboratories in the three surveillance yea~s. For purposes of this longitudinal analysis, a total organic fluorine index (TOF) was determined by calculating the percent of PFOS and PFOA that was attributed to organic fluorine (64.’7 and 69.0 percent, respectively) multiplied by the ppm. measured for each of these two fluorochemicals and then summed to produce the TOF. Data Analysis Briefly, mixed models can be used in the analysis of repeated measures data which are simply data sets with multiple measurements of a response variable on the same subject over time. Detailed explanation of these models is provided elsewhere (Littell 1996; 2000). Mixed models contain factor effects which are considered both fixed and random. An effect is fixed if the levels in the studyrepresent all possible levels of the factor, or at least all levels about which inference is to be made. Factor effects are random if ~he levels of the factor that are used in the study represent only a random sample of a lar~er set of potential levels. The focus of the standard linear model is to model the mean of y by using the fixed-effects parameters ~). That is, y=X~+~ where y represents a vector or" observed data, ~ is an unknown vector of fixed effect~ parameters wi~ kno~n design matrix X, and ~ is an unknown random error vector modeling the statistical "noise’ around Xj~. The residual errors ~ are assumed to be 3M MN02482171 1799.0009 3M Company Page I0 of 63 independent and identically distributed Gaussian random variables with mean 0 and v~riance o-. A generalized standard linear model is a mixed model which is: y =X# + Zy+ ¯ where y is an unknown vector of random-~ffects parameters with known design matrix Z, and ~ is an unknown random error vector whose el~-nents am no longer required to b~ independent and homogeneous. If y + E ar~ assumed to b~ Gaussian random variables that are uncorrelated and have expectations 0 and variances G and R, respectively, then the variance ofy is: V = ~GZ’ + R The variance of the data, y, can be modeled by specifying the structure of Z, G and R. 7fbe model matrix Z is designed in the same fashion as X, the model mat.fix for the fixedeffects pararr~ters. For the matrices G and R, a covariance structure must be selected in using mixed models. Since observations on different subjects ar~ assumed to be independent, the structure refers to the covariance pattern of repeated measurements on the same subject.. For most of these structures, the covariance between two observations on the same subject depends only on the length of the time interval between measurements and the variaace is constant over time. Numerous covariance sUructures exist. Common examples include the following. Simple covariance structure (SIM) specifies that the observations are It is independent, even on the same subject, and have homogeneous variance. usually not realistic for most repeated measures data becaus~ it specifies that observations on the same subject are.independent. Compound symmetric (CS, otherwise 3M MN02482172 1799.0010 3 M Company P~$e [ ! of 63 k~own as variance components) structure specifies that observations on the same subject have homogeneous covariance and homogeneous.variance. Correlations between ~o observations are equal for all pairs of observations on the same subject. Autore~ressive order I (A.R(l))covariance structure specifies homogeneous variance but that covariances between observations on the same subject are not equal, but decrease toward zero ~vith increasing time interval between measurement~ (lag). Its limitation is that observations on the same subject far apart in time would b~ essentially independent. Autoregressive with random effect for subject (AR + RE) covariance strucmr~ specifies homogeneous variance plus the covariance between observations on. the same subject arises from two sources: I) any two observations share a common contribution because they ar~ on the same subject; and 2) the covariance between observations decreases exponentially with lag but only to the common contribution (not to independence). Toeplitz (’rOEP) structure specifies that covarianc¢ depends only on lag but not as a mathematical function with a small number of parameters. The ’unsU’uctured’ structure (UN) specifies no patterns in the covariance matrix and is therefore completely general. The above structures are appropriat~ if equal spacing (of data) is assum~cl in a time series analyses. In situations where unequally spaced longitudinal measurements exist, spatial covaxiance structm-es can be used. In the present analyses, equal spacing was assumed ~ven there ~,’ere approximately 3 years between each medical surveillance program examinations. Akaike’s inform,at.ion criterion (AIC) and Schwarz’s Bayesian criterion (SBC) are indices of relative goodness-of-fit that were used to compare models with the same fixed effects, but different covariance structures. SBC penalizes models more severely for the 3M MN02482173 1799.0011 3M Company Page 12 of 63 number of estimated parameters than AIC and thus the two criteria did not always agree on the choice of-"oest’ model. SBC was preferred. In the present study, the continuous outcomes of lipid and hepatic clinical chemistry tests wer~ evaluated as repeated measures incorporating the random subject effect fitted to a mixed model by the MIXED procedure in the SAS statistical package (Littell et al 1996). Restricted maximum likelihood estimates (REML) of variance parameters wer~ computed. Adjusted regression models were built by introducing all covadates (s¢ below) and testing the covarianc¢ structure. Based on goodness-of-fit tests described above, AR+RE, was routinely considered the best covariance structure chosen for the mixed models. Covariates included PFOS (or PFOA or TOF), years of observation, the-interaction term of PFOS and yearn of observation, age, body mass index (BIVflS, cigarettes smoked per day, alcohol drinks per day, year at first entry and baseline (at first observation) years worked. For h..c.patic clinical chemistry tests, serum triglycerides was also considered a covariate (Olsen et al 2~la). RESULTS Provided in Table 1 ar~ cross-sectional analyses of the study subjects who participated in each of the three years (1994/95, 1997 and 2000) stratified by Iocatioa. As reported previously in the complete cross-sectional analyses of these programs (Olsen et al 1998: 1999; 2001), Antwerp employees in this longitudinal investigation were younger, had lower BMIs and drank more alcoholic beverages than Decatur employees. They also had consistently lower triglyceride and alkaline phosphatase levels and hi~et HDL and total bilirubin levels. Decatur employees, on average, had serum PFOS levels 3M MN02482174 1799.0012 3M Company Page 13 of 63 that were higher by approximately 0..5 ppm in each cross-sectional analysis. Similar findings were observed for PFOA except with the 1997 data where the two populations had comparable mean PFOA levels. Provided in the following two tables are the cross-sectional analyses for the three subcohorts by location. Among Antwerp employees (Table 2), each of the three subcohorts had lower mean serum PFOS levels in 2000 than at their year of entry whereas there were no consistent changes across subcohorts with PFOA. Among the three Decatur subcohorts (Table 3) mean PFOS values declined over time but PFOA levels tended m increase. Provided in tables 4 through 30 are the mixed model coefficient estimates, standard errors,p:values and 9:5% confidence intervals from testing potential determinants of lipid and hepatic clinical chemistry change. The natural log was used for all dependent variables. Tables 4 through 6 contain the analyses for cholesterol. There was no change in cholesterol associated with PFOS (Table 4). Overall PFOA was positively associated with cholesterol as the main effect coefficient was significantly positive but its interaction with time (years variable) was negative (Table 5). Provided in Tables 5A through 5D arc separate analyses for Antwerp for all subjects (’Table 5A) and by each subcohort. The PFOA and cholesterol association appeared to primarily reside with the 21 Antwerp employees in subcohor~ A (Table 5B). This finding can also b~ observed in Table 2 as the subcohon’s mean PFOA levels went from 1.32 ppm. to 2.37 ppm and then declined to 2.06 ppm at the same time their cholesterol values rose from 208 m~dL to 226 m~dL to 229 mg/dL. Them were no associations between cholesterol and PFOA observed among 3M MN02482175 1799.0013 3M Company Pag~ 14 of 63 the Decatur employee population (Table 5E) nor were then significant associations between TOF and Antwerp or Decatur but when the two sites were combined there was a significant positive association between TOF and cholesterol (Tables 6, 6A and 68). There were no significant associations between PFOS, PFOA or TOF with I-IDL (Tables 7 through 9). BM], alcoholic drin~ per day and cigarettes smoked per day were the most si~ificant associations with I-IDL Tri#ycerides were not significantly associated with PFOS over time when both Antwerp and Decatur populations were examined together (Table 10). However, among the combined Antwerp and Decatur populations, PFOA was positively associated with triglycerides (Table 11) as seen with the significant positive coefficient for the main effect of PFOA. and the nonsignificant positive main coefficient of years and the negative coefficient for their interaction (PFOA x years). The significant main effect of PFOA was the consequence of the Antwerp population (Table 11A) and primarily s~bcohort A (table 1 IB) and, to a lesser extent subcohort B (Table 11C), but not subcohort C (Table 11D). Therefore, the association appeared to be related to the Antwerp workers who wet, enrolled in this longitudinal cohort beginning in 1995, but not 1997. There was not a significant association between PFOA and triglycerides a~.ong Decatur workers (Table 11E). Among the Antwerp subcohort A, their mean triglycerid~ levels rose from 85 m#dL to 115 m~dL to 123 m~ddL at the same time their PFOA levels increased from 1.32 ppm to 2.3"~ ppm and then declined to 2.06 ppm. Although the main effect for TOF was significantly positive, the interaction term with time (years) was not significant (Table 12). Again. this association was more consistent for Antwerp employees (Table ’ 12A) than Decatur employees (Table 12B). 3M MN02482176 1799.0014 3M Company Among the hepatic clinical chemistry tests that ~vere adjusted for the various charting demographic factors and u’igl~,cerid¢ levels, there were no si~ificant associations between PFOS, PFOA and TOF with changes in alkaline phosphatase (Tables 13 - 15), GGT (Tables 16-18), AST (Tables 19-21), ALT (Tables 22-24), total bilirubin (Tables 25 - 27) or direct bilirubin (Tables 28-30). Observations a0parent in Tables 2 and 3 can also be seen in these mixed model analyses. For example, the two raost significant predictors of alkaline phosphatase were rime (years) and location (as seen with the lower values among Antwerp employees). For ALT, entry period was also significant as it reflected the higher reference range values for ALT that were used in 1994/95 than in subsequent years. DISCUSSION These analyses were the first longitudinal assessment of the fluor0chemical medical surveillance program at 3M’s Antwerp and Decatur manufacturing sites. Overall. we observed no associations that were consistent with the toxicological evidence that PFOS produces a hypolipid~mi¢ effect at threshold dosages in rats and primates (3M Company 2000; Haughom and Spydevold 1992: Ikeda et al 1987; Pastoor et al 1987; Seacat et al 2001a: 2001b: Sohlenius et al 1993). Our results did suggest a positive association between temporal changes in cholesterol and triglycerides and PFOA; however this is also inconsistent with the toxicolo~¢al evidence that PFOA may result in a hypolipidemJc effect in rats (Haughom and Spydevold 1992; Pastoor et al 1987) but produced no effect on blood lipids in primates (Butenhoffet al 2001). 3M MN02482177 1799.0015 ~M Co~[~ny P~ge 16 Even though we were able to ~rform a lon~tu~n~ ~sessment, ~ere we~ seve~l li~tations to our an~yses. We we~ li~ted to 175 em~oyees of which only 41 (24. ~rcent) p~cipated in all thee su~eillance ye~. Although a veater absolute number of Decatur employees (but not ~nt-wis¢) have p~icipat¢d during each ye~, for this lon~din~ ~ssment th¢~ we~ more An~e~ (57 ~rcent) ~an Decatur (43 ~ent) employees. Antwe~ employees have had lower se~m PFOS level by approximately 0.5 ppm (Olsen ~t ~ 1998; Olsen et ~ 1999a; 1999b; 2~la; 2~ib; 2~1c). ~e~ we~ insufficient num~ of ferule employees for my mganin~l lon~mdinM ~ysis. Given ~e v~ability inhe~nt in ~e ~ic~ me~ ~sen et ~ 2~1) ~d ~ ~ff¢~nt labo~tofies ~ se~m PFOS ~d PFOA l~vgls may have systematic e~or inco~o~ted in each m~u~ment ~at we we~ unable to ~sess ~ blo~ s~ples we~ m~yzed only at ~e time of ~e su~eillmce p~. ~is sys~mafic e~or may have m~ked ~s~iafions with lipid or hepatic clinical che~s~es, akhough ~e range of PFOS ~d PFOA me~md w~ ~lafively consistent ~mu~out ~e study time ~dod. B~ause 3M h~ announced a ph~e~ut of the p~duction of pe~uo~tanyl ¢h~mis~-related mater,s, we doubt ~at ~e~ will ~ many more subj~ in ~e ~m~ ~at cm ~ inclu~d in ~s Ion~m~n~ ~sessment. ~so, ~ fin~n~ f~m ~is ~ssmen~ would suggest that ~mm PFOS levels have eider ~m~ned constmt or ~lined slimily over time ~ong these 175 employees. On ~e o~er h~d. ~m PFOA levels appemd to trend upw~, on average, by epp~ximately 0.5 to 1.0 ppm for ~ese ~mp[oye~s. Ano~er li~tafion is ~e f~t that ~e se~m PFOS md PFOA levels me~u~d in these employees we~ lower ~ those that cause effec~ in la~to~ ~imals. 3M MN02482178 1799.0016 3M Company Page 17 ~f63 In summary, a longitudinal analysis over a six year time period of 175 Antwerp and Decatur male employees did not show significant changes, consistent with toxicological data, of lipid or hepatic clinical chemistry values associated with either PFOS or PFOA. The PFOS and PFOA serum levels measured in these employees were lower than those that cause effects in laboratory animals. 3M MN02482179 1799.0017 3M Company Page lg of 63 REFERENCES 3M Company (2000). SIDS Initial assessment report: Perfluorooctane sulfonic acid and its salts. St. Paul (~N):3M Company, (unpublished report), Minneapolis (MN):University of Minnesota, (unpublished report). AndersonDJ, MuIvana DE (1997a). Analytical Report for the Determination of Perfluorooctanoate and Perfluorooetanesulfonate in Human Serum by LC/MS. Ithaca NY:Advanced Bioanalytieal Services Inc; August 1997. Anderson DI, Mulvana DE (1997b). Analytical Report for the Determination of Perfluorooctanoate and Perfluorooetanesulfonate in Human Serum by LC/MS. Ithaca N~’:Advanced Bioanalytical Services Inc; September 1997. Butenhoff.lL, Costa G, Elcombe C, Farrar D, Hansen K, Iwai H, lung R, Kennedy G, Lieder P. Olsen GW, Thomford P. Toxicity of ammonium perfluorooetanoate (APFO) in cynomolgus monkeys after 26 weeks of oral dosing. (unpublished report) Hansen KJ, Clem¢ia I.A, Ellefson ME, Johnson JI-IO (2001). Compound-specific, quantitative characterization oF organic fluorochermeals in biological matrices. Environ Sci Technol 35:766-770. Haughom B, Spydevold 0 (1992). The mechanism underlying the hypolipemic effect of perfluoroetanoic acid (PFOA), perfluoroctanesulphonie acid (PFOSA) and elofibric acid. Biochemica et Biophysica Acta 1128:65-72. lkeda T. Fukuda K, Moil I, Enomoto M, Komai T, Suga T (t987). Induction of cytochrome P-450 and peroxisome proliferation in rat liver by perfluorinated octanesulfonic acid. In: Peri.rosraes in Biology. and Medicine. (I-~ Fahmi and H Sies, eds) New York:Springer Verlag, pp 304-308. Johnson J’D, Wolter J’I’, Cola.izy OF,, Rethwill PA, Nelson RM. Quantification of Perfluorooctanoate and Perfluorooetanesulfonate in Human Serum Using Ion-Pair Extraction and High Performance Liquid Chromatography-Thermospray Mass Stt~ctrometry with Automated Sample Preparation. St. Paul, I~-N:3M Env Lab, 1996. Litteii RC. Pendergast J, Natarajan R (2000). Modoling eovariance structure in the analysis of’repeated measures data. Stat Meal 19:1793-1819. Lit~ell RC. Milliken G~, Stroup WW, Wolfinger RD. SAS System for Mixed Models. Cary NC:SAS Institute Inc., 199(5, 633 pp. Olsen GW. Burris J’M, Mandel J’l-I, Zobol LR (1998). An epidemiologic investigation of clinical chemistries, hematology and hormones in relation to serum levels of perfluorooctane sulfonate in male fluorochemical production employees. St. Paul:3M Company (unpublished report). 3M MN02482180 1799.0018 3M Company Page 19 of 63 Olsen GW, Burris JM, Mandel JH, Zobel LR (1999a). Serum perfluorcoctane sulfonate and hepatic and lipid clinical chemistry tests in fluorochemical production employees. J’OEM 41 (9):799-806. Olsen GW, Logan PW, Simpson CA, Hansen KJ, Bun’is J’M, Burlew MM, Schumpert JC, Mandel J’FI (1999b). Fluorochemical exposure assessment of Decatur chemical and film plant employees. St. Paul:3M Company (unpublished report). Olsen GW, Bun’is JM, Burlew MM, Mandel J’H (2000). Plasma cholecystokinn and hepatic enzymes, cholesterol and lipoproteins in ammonium perfluorooctanoate production workers. Drug Chem Toxicol 23(4):603-620. Olsen OW, Buflew MM, Bards JM, Mandel J’H (2001a). A cross-sectional analysis of serura p~rfluorooctansulfonate (PFOS) and perlfuorooctanoate (PFOA) in relation to clinical chemistry, thyroid hormone, hematology and urinalysis results from male and female employee participants of the 2000 Antwerp and Decatur fluorochemical medical surveillance program. St: Paul (MN):3M Company (unpublished report). Olsen GW, Schmiclder MN. Tier~ns JM, Logan PW, Burris JM, Burlew MM, Lundberg JK, Mandel JH (2001b). Descriptive summary of serum fluorochemicai levels among employee participants of the year 2000 Antwerp fluorochemical medical surveillance program. St. Paul:3M Company (unpublished report). Olsen GW, Logan PW, Simpson CA, Burns JM, Burlew lVh-~, Lundberg JK, Mandel JI-I "(’2-001c). Descriptive summary of serum fluorochemical levels among employee participants of the year 2000 Decatur fluorochemical medical surveillance program. St Paul:3M Company (unpublished report). Pastoor TP. L~e KP, Peril ~L~., Gillies Pl (1987). Biochemical and morpholo~cal studies of ammonium perfluorooctanoate-induced hepatomegaly and peroxisome proliferation. Exp Mol Pathol 47:98-109. Seacat AM. Thomford PL Hansen KJ, Cleraen LA, Case MT, Butenhoff YL (2001a). Sub-chronic dietary toxicity, of potassium peffluorooctanesulfonic acid in rats. Toxicol Sci (submitted 200ia). Seacat AM. Th0ford PJ’, Hansen KJ’, Olsen GW, Case MT. ButenhoffJL. Subchronic toxicity studies on pe~uorooctanesulfonate potassium salt in cynomolgus monkeys. Toxicol Sci (submitted. 2001b). Sohlenius AK, Eriksson AM, Hogstrom C, Kimland M, DcPierre JW (1993). Perfluorooctane sulfonic acid is a potent inducer of peroxisornal fatty acid B-oxidation and other activities known to be affected by pcroxisome proliferators in mouse liver. Pharmacol Toxicoi 72:90-93. 3M MN02482181 1799.0019 l ~lllO~ll)lli~ O)ara~terislk-s and (:linical Ol~inislries of Ant~rp ~lld D~calur Male Who I~ailicipat~ in Two or More Medical Surveilla.c~ ~x;ll,i,ations I]ctween 1994195 191J4! 1905 2111XI I g97 . .~(~a_!_u,r.!N =J!)~. . A.~!W~.!rD.(_I~.-~_.55.) ..... .l_)¢~.:~_!~ir (N = .’i_:’i) ^nlwer ~ (N=IOO) I’)ecatur (N = 75) .. Mean SI) Me;In SD Mean SI) Mea. SI) 1.7g 1.42 1.2fi I.R5 I.M I. 16 1.07 1.67~’ 1.39 1.9~~ 1.0fl 1.54 1.61 1.41 1.17 1.43 1.21 1.83" !.53 3.(~’ 1.77 I.gX 1.4X 2.17 1.76 1.74 1.24 2.3-I~’ 1.72 Mean SI ) Mean I)F{).~ I.g7 I.gt~ 2.62= I)F(’)A 1.08 1.53 TOF 1.4~t, 1.77 Age 3(, {,.(, 43’~ 32 6.8 4.1’i 7.3 .18 7,8 47~ 7.0 I]MI 2~.~ 2.4 ~8.0’~ 232 2.4 29. I’= 4..3 24.7 2.1~ 2’).0’l 4. I I. I 0.3 0.9’l I.(1 O. I O. I I. I’l 1.0 (I. I 0.2 S.7 211.4~ 7.0 5.2 I( ,0’1 6. I 9. I ft. I 20.Y 8.5 4 7 I0~’ 5 7 8 Cl~oleslerol 217 43 219 I 1DL ~4’~ 13 43 "rriglycerides I I I 7) AIk Phos 72 ( ~( ;’1" AIc(,hol Ilaselh]e Years I. I’~ I I J) 7.1 Worked (’igaretles I 5 10 5 5 II 202 45 214 35 220 41 213 411 13 49" Ii 43 I0 53’~ 12 44 I0 2~~ 122 lOg 53 181’~ 112 131 gO 17g~" 115 I~ 103~ 27 68 15 g7’l 20 5g 14 74" 2I 37 2~ 47 24 23 I I 33" 27 25 19 2() 18 AS’I" 2~ I I 31= ]I 26 6 26 8 24 7 25 7 AIJI" 44 I7 49 30 12 32 15 23 I 1 32a 14 ’r.lal Ifiliruhi. (I.Q 0.4 03 0.2 0.8I’ 0.4 0.6 0.2 1.0’) (I.3 ().~ 0.2 I)ired lliliruhi. (I.2 0.4 0.2 0.(H 0. l)’ 0.(!7 (}. 1 0.04 0. I~ 11.05 0, I 0.(}6 a. p < .fl.S h. p < .01 r. i) < 3M_M N02482182 1799.0020 Tabh." 2. (’nl~.~-.";cctional A,alysb; o1 Mean lind Standard I.)c.vhaion olScrunl I’FOS, I’I"OA, 1"( )F. I)¢nlographk: (’h;traclerislJc.~ still (’li,fical { ’ht~lllisll’ies o1" Three Suhcro, lXS of Antwerp Male I]nlldO~ecs (A, I mid (") Who I)illlicipated ill Two or Mol’C Medical .~l.lrveill;lllCe 19o5 EX~lltlill;lliOlIS nelwel~n 1995 ;,rid 21}0tl 1997 _A_.(_.N.~. 21_) .... !_l (N = 45) ........ A,(.N_~.2_I)_. C (N = 34) Mean Mean Mean SD (? (N = 34) Mean SD 1.53. 1.211 1.31 11.87I’ 0.7 I 11.61 2.(16a 1.04 1. I 7 1.37 1.38 0.94 1.59 I1.91 2.41= 0.81 I.Sg 1.51 1.52 11.89 5 7 38= 6 42" 7 32~ 7 2.8 2.O 24.3 3.4 25.5" 2.7 23.9 2.4 1.7" t.3 1.2 I.I 0.8 11.6 6 8 4 8 5 8 9.9 4.3 6.3 5.3 4.9 1.32 2.2R (1.7 7 4.3 51 SD II (N=451 11.87 11.93 1.2 226’ A(N=21) 511 9 5.3 4.9 I 1.6 187 34 ~29 46 233=’ 38 196 31 dR II 56 12 52 II 52 13 115 69 1(14 41 123 65 154 95 105 55 67 19 69 12 55 17 (~1 14 58 12 25 I(1 22 I I " 23 13 3If 25 19 11 211 5 26 7 , 22 5 25 9 23 6 31 12 30 13 22 II 25 13 21 7 11.8 11.3 0.8 11.4 1.0 11,3 1,0 1}.3 11.9 0.3 O. I 0.1i5 11.2. ll.01l O.l 0.03 (I. I 0.06 O. I 0.()6 3M_MN02482183 1799.0021 Table 3. (’ross-,~ol:liOllal A,ialysi,~ of Mean ;llid SI;itRl:lrd I)eviation of Scrum I’1~O,";, Pf;OA, TOF, I )en~q~raphic Characlcrislics and Clinical Clmmistries of Three SubgroLil)S of Decatur Male E,nl)h)yees (A, B and Who I’articipaled in Two or Mol~ Medical ~urvoillance Examinalions Between 1994 and 1994 :, __ A_ (_N. ~_~(.1) ....... .B_(N = 2111 _ 2(100 1997 A .(N =.201 _ C (N = 35) A (N = 20) P, (N = 20) c (N = 35) Mean SI) Mean SD Mean Mean SD I’F( IS 2.07 I .(,7 3. I 3= 1.76 1.93 I.’/6 1.811 1.59 1.78 2.14 1.84 I.O3 1.51 0.98 PFOA 1.50 {I.~7 2.30= I. 14 1.41 I. 16 1.41 : 1.20 1.46 1.34 2.60~ 2.112 1.60 1,14 TOF 2.37 1.54 3.64~ !.79 2.22 1.78 ~.14 1.78 2.16 2.16 2.9R 1.76 2.(1~ 1.32 SD Mean SD Mean SD Mean SD Age 42 7 44 S 45 7 42 8 48 7 511 5 45 8 BMI 28.2 4.0 27.7 ~.3 28.3 4. I 29,6 4.5 28.9 4. I 28.7 3.3 29.3 4.5 I)rinks/day O. I (I.~ 0.4 0.8 (1.1 ILl ILl 0.2 O.11 ILl 11.2 0.3 (I. I 0.2 Cigarelle~day 6 13 14 15 ,I 3 8 I11 Baseline Years Wo~ed 19.6 K6 21.3 5.2 19.(; 8.6 21.3 211.2 IO.O Cholesterol 235" 34 21H 36 223 4I HDL 47~ 16 39 9 T~glycCride~ 180 8~ 229 AIk Plltls 98 29 (;GT 41 AST ALT 10 g.6 5 I I 2( .2 10.() 3I 208 37 45 I I 41 145 IB8 115 178 1117 24 85 22 25 52 22 35 32 15 29 6 52 33 47 11 19.6 15 5.2 228 39 199I’ 33 212 9 47 12 .’:18= 6 45 9 112 193 124 187 76 165 128 87 Ig 711 27 1~2 17 72 211 36 31 20 29 16 34 2.t 27 17 25 6 27 8 ~Z6 8 25 4 25 7 30 15 16 30 18 32 11 33 14 " 34 - Total Ililirubia 0.6 0.2 0.5 0.2 (I,fi 0.2 11,6 0.i 0.8 0.2 0.7 0.2~ 0.8 0.~ Direct Igillrubin (I.2= (~.115 IL2 0.~ 0. I 0,04 O. 1 11.05 O. 1 0.08 0. I 0.04 (I. 1 0.06 3M_M N02482184 1799.0022 ’l’abl~ 4. Mixed M~d¢l Co~fl’icicnl I.’.’Slhnal~s, ~landard Errors {SF.), P-Valu¢~ and 9.5% Confidence Limi s I’rOnl "l’csting Polcntial Determinants of Chol¢.qterol* Change Including PFOS and the [ntcr’-,ction with biumher of Years o1" Observation o[ Antwerp and Decatur Male Emp oyccs 95% Confidence Limits ....................... C__(.~l’ficieilt S!~. .l~-valtle. Lower Uj21~r <~.0001 4.618 5.118 ¯ 18 - 0.005 0.025 Inteacep! 4.868 PFOS O.(llO Years Obwerv:,tion 0.(1009 0.005 .84 - 0.008 1.010 PFOS x Years Obs - O.~H 0.1~)2 .83 - 0.0O4 0.003 Age O.(fi7 O.$)3 .0 ! 0.002 0.013 BMI 0.(~6 ().(~13 .07 - 0.0005 0.013 Dfink~day 0.014 0.012 .27 - 0.01 ! 0.038 - 0.~9 0.(~ I .41 - 0.003 0.001 ~ati~m* 0.(134 0.037 .36 - 0.039 O. 108 ~l~ Peri~)tl** 0.(164 0.028 .02 0.009 0.119 Baseline Years - 0.~ 0,003 .20 - 0.009 0.002 Cigaretle~day O. 12’7 ,. *natural log *Antw¢~ vs Decatur ** 1994195 vs 1997 3M_MN02482185 1799.0023 ’rahle 5. Mixed Mt~lel Coefficie,I Eslhnales, Sla. ndard Errors (SE}, P-Values and 95% Confidenc~ Li,nils from Testing I)olential D~terminanls of Cholesterol’ Change Including PFOA and the Inlet’action with Nnmber of Years of Ohsm’valion of Antwerp and Decatur Male Employees 95% Confidence Limila ............... .L.’,..m. I_".fic__i_e!~ I...... SI~. p-v:due Lower Upper < .{X}O ! 4.561 5.063 .0~R~8 0.013 0.0.5 I InlorcepI ~1.812 O. 127 PFOA ().O32 0.0{F) Years Ohserv~ilion ILIX 5 O.(X)5 .24 - 0.004 0.014 - 0.005 0.~12 .005 - 0.009 - 0.002 Age O.008 0.003 .007 0.002 0.013 BM! O.(l~i7 0.003 .049 0.00004 0.013 Drinkshlay 0.014 0.012 .263 - 0.010 0.037 - 0.01) I 0.(X I .32 - 0.003 0.001 l..,Ocalion * O.(H ! (I.037 .27 - 0.032 O. 114 Entry Period** 0.068 0.027 .01 0.015 O. 122 Baseline Years Worked - 0.004 O.LR)3 .15 - 0.009 0.001 PFOA x Year~ Ohs CilPtrelles/d:ty : *natural log *Antwerp vs Decatur *’1994/95 vs 1997 3M_M N02482186 1799.0024 ’r-hie 5A. Mixed Model Coefficient F.stimates, Standard Errors {Sff-), P-Values and 95% Confidence Limits I’ronl Testing Polcnlial Determinants of Cholesterol~ Change Including PFOA and the lnteraclion with Number of Yca,~ of Observation of Antwerp Male Employees 93% Confidence Limits C,~fl~cicn! SF. p-value Lower lnterccpI 4.7 l0 0.149 < ’.~l 4.414 PFOA 0.029 0,012 .01 0.~ 0.053 Yearn Ohscrvation O.(g)5 0.(~6 .36 - 0.~ 0.017 - 0.(~13 0.{~)3 .20 - 0.~9 0.~2 Age 0.~ 0.~3 .~8 0.~3, 0.016 BM[ 0.~g 0.{~5 .14 - 0.~3 0.019 D~nk~y 0.022 0.013 .09 - 0.~ Ciga~tle~day O.{XX)7 OAX~2 .70 - 0.~3 0.~ ~lry Peri~al** 0.0~9 0:038 .(H 0.~ 0.1’53 B~elin¢ Years - 0.(~2 0.~ .67 - 0.Ol0 0.007 PFOA x Ycm~ Obs Upper Work~ ~naluml log **19~/95 vs 1997 3M_MN02482187 1799.0025 j Company Page 26 of 63 Table 5B. Mixed Model Coefficient Estimates, St=mdard Errors (SE), P-Values and 95% Confidence Limits fro,n Testing Polential Determinants of Cholesterol~ Change Including PFOA and the Interaction with Nt,mber of Years of Observation o1’ Anlwerp Subgrm,p A ( 1995, 1997 and 2000) Male Employees 95’$ Confidence Limits C~efficient SE p-value Lower Upper lniercepl 5.059 0.414 < .0001 4. ! 94 5.925 PFOA 0.044 ().t)2t) .03 0.004 0.084 Years Observation 0.037 0.015 .02 0.007 0.067 PFOA x Yenrs Obs - 0.013 0.005 .02 - 0.023 - 0.002 Age - 0.0007 0.012 .95 - 0.025 0.024 0.~4 0.012 .77 - 0.020 0.028 - 0.(116 0.019 .41 - 0.054 0.023 Cigarettes/day (1.003 (1.(XI5 .5 i - 0.007 0.0 ! 3 Baseline Yeat~ Worked 0.015 0.016 .36 - 0.018 0.047 Ddnks/,t~,y i ~natural log 3M_MN02482188 1799.0026 Table 5(:. Mixed Mud¢l Cocfl’icicnt I~timales, Slandard EiTors (S]~), P-Values and 95% Confidence Limils I’rom Tcsling Polenlial Dclerminanls of Choleslerol~’ Change Including PFOA and the lnleraclion wilh Number of Years of Observation of Antwerp Subgroup B (1995 and 2000) Male Employees 95% Conlidcnc~ Limits Lower Upper 4.34fi 5.330 .30 - O.0 i 6 0.052 0,008 .55 - 0.0 ! 1 0.021 - 0.002 0.004 .58 - 0,009 0.005 Ag© 0.(XH 0.{X)5 .40 - 0.006 0.015 BM! 0.014 O./X)9 .13 - 0.004 0.03 I Drinks/day 0.O I 7 0.022 .45 - 0.028 0.061 Ciga~elle~dqy 11.002 0.003 .44 - 0.004 0.009 Ba~eliue Years Worked - 0.11001 0,006 .99 - 0.011 0.01 I Interccpl PFOA Years Obscrvalion PFOA x Ye,’~rs Obs (2~fl~cienl SI.." 4.838 0.244 " ’ 0.018 0.017 (1.005 p-value < 0.001 ’natural log 3M_MN02482189 1799.0027 Table 5D. Mixed Model Coefficieni Hstimates, Standard ~rrors (SE), P-Values and 95% Confidence Limils from Tesling Potential Delerminaals .f Cholesterolu Change Including PFOA and Ih¢ Interaction with Number of Years of Observali.n of Anlwe~p Subgroup C (1997 and 2000) Male Employees 95% Confidence Limits C~x.’fficient SH p-wdue Lower Upper .0001 3.955 5.100 < lnlercepl 4.528 0.28 I PFOA 0.004 0.062 .95 - 0.125 0.132 - 0.010 0.026 .70 - 0.065 0.044 PPOA x Years Obs O.OI0 0.022 ¯ 65 - 0.035 0.055 Ag~ 0.016 0.006 .01 0.004 0.028 BMI 0.009 0.012 .43 - 0.015 0.033 Drinks/d-y 0,033 0.032 ¯ 31 -0.033 0.100 Cigareltes/day - 0.0001 0.003 .97 - 0.007 0.006 Baseline Yea~s Work - 0.005 0.008 .49 - 0.021 0.011 Ycam Obscrv-tion ~natural log 3M_MN02482190 1799.0028 Table 5E. Mixed Model Coefficienl Estimates, Slandard Errors (SE), P-Values and 95% Confidence Limits from Testing Polcnlial Determinants of Cholesterol~ Change Including PFOA and the ]nteraclion wilh Number of Years of Observation of Decatur Male Employees 95% Confidence Limits Coefl~cien! Intercqa S ft, p-value Lower Upper 4.810 .5.630 5.220 0.206 ,~ .0001 0.016 0.016 .34 - 0.017 0.048 - 0.002 (I.(X)8 .77 - 0.017 0.01:3 - 0.003 0.003 .22 - 0.009 .0.002 Age 0.002 0.005 .70 - 0.008 0.012 BM! 0.003 0.004 .55 - 0.006 0.011 Drinks/day - 0.088 0.035 .01 - 0.158 - 0.018 Cigarettes/day - 0.002 (1.001 . [5 - 0.004 0.0007 Entry Period** 0.047 0.040 .24 - 0.032 O. 125 Baseline Years Worked - 0.002 ().1)04 .62 - 0.010 0.006 PFOA Years Obscrv,qion PFOA x Y~ms Obe ~natural log *’1994/9.5 vs 1997 3M_MN02482191 1799.0029 . Company P~g¢ 30 of 63 Table 6. Mixed Model Coefficie,! I~slimales, Standard Errors (SE), P-Values and 95% Confidence Limits from T~tiag Potential Detem)inanls of Cholesterol* Change Including TOF and the lnleraction with Number of Years of Oh~rvalion or Anlwerp and Decalur Male Employees 95% Confidence Limits Ct~l’ficien! SI~. p-value Lower < .l~01 4.577 5.079 0.006 0.035 .37 - 0.t~)5 0.014 .07 - 0.005 0.0003 Inlercepl 4.828 TOF 0.02 I Year~ ObservaliOll O.O(H O.IX)5 - 0,003 0.001 Age 0.007 0.003 .0 I 0.002 0.013 BMI 0.006 0.003 .05 - 0.0001 0.013 Drinks/day 0.012 0.0 i 2 .32 - 0.012 0.036 - 0.001 O.(,q) I .37 - 0.003 0.00! Location* 0.042 0.037 .26 - 0.032 0.115 Enlry Peri(~l** 0.063 0.027 .02 0.010 0. !7 Baseline Yeats Worked - 01}04 0.003 .17 - 0.009 0.002 TOF x Years Cigarettes/day 0.127 : anaturul log *Antwerp vs Decatur *’1994/95 vs 1997 3M_M N02482192 1799.0030 Table 6A. Mixed Modal Co~flicient I~timates. St, a,ldard Errors (SI~.), P-Values and 95% Confidence Limits fi’om Testing Potential Det,~rminants of Cholesterol~ Change Including TOP and the Interaction with Number of Years of Observation of Antwerp Male Employees 95% Confidence Limits Coefficienl SI~. p-value Lower Upper 4.424 .5.018 Intercept 4.’/21 0.150 < .O001 TOF 0.017 0.011 .12 - 0.004 0.038 Years Obscrvalion O.t104 O.lXI7 .55 - 0.009 0.017 TOF x Years Obs - 0.0008 0.002 .70 - 0.005 0.003 O.(X)9 0.003 .0 i 0.002 0.016 BMI 0.009 0.006 .12 - 0.002 0.020 Drinks/day 0.019 0.013 .14 - 0.006 0.045 Cigarettes/day 0.0008 0.002 .69 - 0.003 0.005 Entry Period** 0.070 0.038 .07 - 0.006 O. 145 Baseline Years Worked - O,O(k)g O.{X)4 .85 - 0.009 0.008 ’. ~naluml log **1~4/95 vs 1997 3M_MN02482193 1799.0031 Company Page 32 of 63 Table 6B. Mixed Model Cocfficien! l~limates, Slandard Errors (SE), P-Values and 95% Confidence Limits from ’l~sting Po!ential De.tcrminanls of Cholesterol~ Change Including TOF a~i the Interaction with Number of Ye=n~ of Observation of Deca!ur Male Employe~a 95% Confidence Limils ....................... _.(.’.~m I’__fii__ c e_!~! SE p-value Lower Upper 5.616 Intercept 5.214 0.202 < .000 I 4.gl I TOP 0.014 0.01 ! .19 - 0.007 Years Observalion - 0.002 0.008 .’/9 -0.018 0.013 ]’OF x Years Obs - 0.002 0.0(12 .22 °0.006 0.001 Age 0.002 0.005 .69 - 0.008 0.012 BMI 0.002 0.004 .55 -0.006 0,01 l Drinks/day - 0.090 0.035 .012 -0.160 - 0.020 Cigamues/day - 0.002 0.001 .16 ! - 0.004 Bl~eline Yem’s Worked - 0.002 0.004 .60 -0.010 : 0.006 Jnatural log *’!994/95 vs 1997 3M_M N02482194 1799.0032 .i CUliilllin y Page 33 of 63 Tal)lc 7. Mixed ldodcl Cocl’ficicn117.sli,ilales, Standard Errors (SE), P-Values and 95% Confidence Limits from Testing Potential Determinants of HDL# Change Including PFOS and the Interaction with Number of Yc;u,’s of Obscrvatitm of Anlwerp and Dccntur Male Employees 95% Confidence Limits lntercepl PFOS Year~ Olisel’Vlil iilli Age BM! Ddnk~day Cigarelte~day (?oefficienl SI~ 4.212 0.144 < .O001 " - 0.010 .24 0.(1()2 .71 0.~ 0.013 - O.{g~ 0.~ I 3.930 4.495 - 0.026 0.012 o 0.005 0.00_.2 .22 - 0.002. 0.0! ! < o00OI - 0.025 0.010 0.037 0.090 .0005 - 0.007 - 0.002 .89 - 0.079 0.091 II.(~ - 0.017 Upper [)-viihie < Efllry Perilldi* 0.025 0.032 .44 - 0.038 O.Og8 Baseline Ycar~ Wo~ed - 0.~7 0.~3 .O3 - 0.013 - 0.0005 enatuml Io~ *Antweq) vs ** t9~95 vs 1997 3M_MN02482195 1799.0033 ¯ Company Page 34 of 63 Table 8, Mixed Model Co¢l’ficient Estimates, Standard Errors (SE), P-Values and 95% Confidence Limits from Testing Potential Determina,lts of" HDL’ Change Including PFOA and the Interaction with Number of Years of Ohservalilln of Aniwe,’p and Decatur Male Employees 95% Confidence Limits 12ocl’ficient__ Inlercepl _.qF. ...... ~p-value Lower Upl~r 3.929 4,.503 4.216 (). 1,15 ~,0001 " - (1.006 tl.(ll I ..’i6 - 0.027 0.0 11.011.5 I}.lltl$ {35 - 0.005 0.014 0.002 I1.1102 .40 - 0.006 0.002 Age 0.004 0.003 .28 - 0.003 0.010 BM1 - 0.017 0.(104 < .0001 - 0.024 - 0.010 0.tgl2 0.013 < .(RIO I 0.036 0.088 . - 0.1}(H 11.1101 .00(14 - 0.007 Localim~* O.i~g 0.043 .g5 - 0.076 0.093 Enlry Period** 0.02~ 0.032 .53 - 0.042 0.082 Baseline Years WorEed - 0.007 0.003 .04 - 0.013 - 0.0004 PITOA Years Ollsl~rvllliun Pl’q~li~. i ~Ciil-,,,l Ohg Drinks/day Cigarelies/day - ~ - ~natural lug *Antwerp ~sDe~alur *’1994/95 vs 1997 3M_M N02482196 1799.0034 Table 9. Mixed M~lel Coefficient Esli~nales, Slandard Errors (SE), P-Values and 95% Confidence Limils from Tesling Potential Determinants of HDL~ Change Including TOF and the Interaction with Number of Years of Ohservation of Anlwerp and Decatur Male Employees 95% Confidence Limits Coel’licienl Intercept __SL; p-value Lower Upper < .0001 3.931 4.502 .44 - 0.023 0.0 i 0 .47 - 0.007 0.014 - 0.004 0.002 0.003 0.010 4.217 0.14.S "- 0.006 0.(xi8 Years Ohservalion 0.004 0.(XI5 TOP x Yem’s Obs - 0.(R)(F) O.(N)I Age 0.004 O.(XL3 .2.5 BMI - 0.O 17 0.0(H < .000 I - 0.025 - 0.010 0.063 0.01 ~ < .0001 0.037 0.089 - 0.004 0.001 .0005 - 0,007 - 0.002 Localion* 0.007 0.043 .88 - 0.078 0.0’92 l~nlry Period** 0.021 0.032 .50 - 0.041 0.084 Baseline Years Worked - 0.(RI7 0.003 - 0.013 - 0.0004 TOP Drinks/day Cigarelles/day ~natural log *Antwerp vs Dc~:atur **1994/95 vs 1997 : - !. 3M_MN02482197 1799.0035 Page 36 of 63 Table I0. Mix~ Model Coefficicn! I~limales, Standard Errors (SI~), P-Values and 95% Confidence ~mils fi~m 3~sling Polcnlial Delcrminanls ol"F~glycc~d¢~ ~angc Including PFOS and the Interaction wiLh Numl~r of Years of Observation of AnI~=? and ~atur Male Employees 95% Confidence Limits Cocfficicn! lnlerccp! 2.730 0.335 PFOS O.1 25 0.( 2(! 0.(~ 0.O 13 Years Obsc~alion - Lower Upper 2.068 3.392 -0.015 0.065 .73 - 0.029 0.02 I .22 - 0.004 0.015 .~7 - 0.011 0.01g 0.048 0.083 - 0.094 0.035 0.005 0.017 - 0.143 0.247 - 0.055 0.234 - 0.010 0.019 < .0001 PFOS x Years Obs 0.~6 Age 0.(~3 0.~)8 BMI 0.(~ 0.~9 - 0.029 0,033 .37 Cigar~lle~day 0.011 0.~)3 .0002 ~ati=~n* 0.052 0.0~D .~0 FJ~I~ I~riod~* 0.089 0.073 Baseline Ymtrs Worked 0.(~15 0.~7 D~nk~day < ,50 ~natural log *Antwc~ vs Decatur ** 1994/95 vs 19~ 3M_MN02482198 1799.0036 I Lulill)a iy rage 37 of 63 ’Fahle I I, Mixed Model Coefficiem [Lslimales, Standard Errors (SE), P-Values and 95% Confidence Limits fn)m Te.~ling Polential Determinants ofT,’iglyceride~ Change Including PFOA and the Interaction with blumbcr 0f Years of Observation of Antwerp and Decatur Male Employees 95% Confidence Limits Coefficient t,;E p-value Lower Upper 2..539 0,332