PFOA Exposure and CAD, Hypertension, and High Cholesterol
PFOA Exposure and CAD, Hypertension, and High Cholesterol
Characteristics of study participants overall and by cohort are shown in Table 1. In the combined cohort, 54% were female, 97% were of white race, 71% were overweight or obese, 52% reported ever smoking, and 14% reported ever having a type 2 diabetes diagnosis. The median follow-up duration was 32.6 years starting at the age in 1952 or age 20 years, and 4.4 years starting at the time of the C8 Health Project. The median PFOA serum concentration measured in 2005–2006 was 26.1 ng/mL. Median estimated serum PFOA concentrations (see Supplemental Material, Figure S2 http://ehp.niehs.nih.gov/wp-content/uploads/122/12/ehp.1307943.s001.508.pdf) were highest in the late 1990s and early 2000s (Winquist et al. 2013). Table 2 reports the number of people reporting the outcomes of interest, the number with reported current medication or medical records validation, and the number included in the analyses. The mean age at first diagnosis was 47 years for hypertension among those who reported prescription medication use, 50 years for hypercholesterolemia among those who reported prescription medication use, and 56 years for coronary artery disease among those for whom the condition was validated.
Hypertension. In the primary retrospective analysis, combining sexes and ages, there was no clear evidence of an association between cumulative PFOA exposure and hypertension [hazard ratios (HRs) and 95% confidence intervals (CIs) by increasing quintile: 1.00 (reference), 1.10 (95% CI: 1.02, 1.19), 1.10 (95% CI: 1.02, 1.18), 1.05 (95% CI: 0.97, 1.12), 0.98 (95% CI: 0.91, 1.06)] (Figure 1). Results of analysis using the yearly exposure metric were similar (data not shown). In age- and sex-stratified analyses (Figure 1), there was some evidence of increasing hypertension hazard with increasing cumulative or yearly exposure among females in the 20- to 39-year age group and males in the 40- to 59-year age group, but trends in hazard across exposure quintiles were not monotonic, and tests for trend using the logged exposure metrics were not statistically significant (at α = 0.05). Prospective analyses starting at the age in 2005–2006, and conditioning on absence of hypertension to that point, also showed no clear evidence of a positive association between cumulative (see Supplemental Material, Figure S3 http://ehp.niehs.nih.gov/wp-content/uploads/122/12/ehp.1307943.s001.508.pdf) or yearly PFOA exposure and hypertension, and showed HRs for some quintiles relative to quintile 1 that were < 1.
(Enlarge Image)
Figure 1.
HRs and 95% CIs for hypertension in the primary retrospective analysis for the combined cohorts, cumulative exposure. Quintile (Q) cut points (μg/mL per year) were < 0.111, 0.111 to < 0.191, 0.191 to < 0.471, 0.471 to < 2.763, ≥ 2.763. The analysis included 11,798 cases of self-reported hypertension with medication. Models were stratified by single-year birth year and were either stratified by sex or controlled for sex and the interaction between sex and age. Models also controlled for years of schooling (not time-varying; < 12 years, high school diploma/GED, some college, or ≥ bachelor's degree), race (white vs. nonwhite or missing), smoking (time-varying; current, former, none), smoking duration (time-varying), smoking pack-years (time-varying linear term created by multiplying the self-reported number of packs smoked per day by the smoking duration to that point), regular alcohol consumption (time-varying; current, former, none), BMI (at time of first study survey; underweight, normal, overweight, obese), and self-reported type 2 diabetes (time-varying according to reported age at diagnosis).
Retrospective analyses restricted to community cohort members had results similar to those of the primary analysis (see Supplemental Material, Figure S4 http://ehp.niehs.nih.gov/wp-content/uploads/122/12/ehp.1307943.s001.508.pdf); trends across PFOA exposure quintiles were not statistically significant (at α = 0.05). Results of retrospective analysis restricted to nonsmokers and analysis starting at the "qualifying age" were similar to those of the primary analysis (not shown). In retrospective analyses varying the year in which the analysis was ended, an apparent trend (although not completely monotonic) of increasing hypertension hazard with increasing cumulative PFOA exposure quintile was seen for sexes and ages combined in analyses ending in 1987; this trend became progressively less clear with longer follow-up after the time of peak exposure (see Supplemental Material, Figure S5 http://ehp.niehs.nih.gov/wp-content/uploads/122/12/ehp.1307943.s001.508.pdf).
Hypercholesterolemia. The primary retrospective analysis, combining ages and sexes, showed an elevated hazard of hypercholesterolemia for exposure quintiles 2–5 relative to quintile 1 (Figure 2). For cumulative exposure, HRs increased in quintile 2 and remained elevated with little further increase across subsequent quintiles (HRs and 95% CIs by increasing quintile): 1.00 (reference), 1.24 (95% CI: 1.15, 1.33), 1.17 (95% CI: 1.09, 1.26), 1.19 (95% CI: 1.11, 1.27), 1.19 (95% CI: 1.11, 1.28) (test for trend using log cumulative exposure p = 0.005). Analyses using the yearly exposure metric also showed an increasing hazard of hypercholesterolemia with increasing quintiles (HRs and 95% CIs by increasing quintile): 1.00 (reference), 1.07 (95% CI: 1.01, 1.15), 1.11 (95% CI: 1.04, 1.19), 1.05 (95% CI: 0.99, 1.13), 1.20 (95% CI: 1.12, 1.28) (test for trend using log yearly exposure p < 0.001). In age and sex-specific analyses, the increased hazard for hypercholesterolemia was most pronounced among men 40–59 years of age (HRs and 95% CIs by increasing quintile): cumulative: 1.00 (reference), 1.38 (95% CI: 1.21, 1.56), 1.32 (95% CI: 1.17, 1.50), 1.31 (95% CI: 1.16, 1.48), 1.44 (95% CI: 1.28, 1.62) (test for trend using log cumulative exposure p < 0.001); yearly: 1.00 (reference), 1.16 (95% CI: 1.03, 1.30), 1.19 (95% CI: 1.05, 1.34), 1.16 (95% CI: 1.03, 1.31), 1.38 (95% CI: 1.23, 1.54) (test for trend using log yearly exposure p < 0.001). In analyses combining ages, tests for interaction between log exposure and sex were statistically significant (cumulative p = 0.048, yearly p = 0.004). Tests for interaction with age were also statistically significant (cumulative exposure, sexes combined, p = 0.015, males p = < 0.001; yearly exposure, males p = 0.032). Because of these interactions, analyses stratified by sex and age are preferred to the overall analyses. Prospective analyses starting at the age in 2005–2006 and conditioning on absence of hypercholesterolemia to that point showed no evidence of a positive association between hypercholesterolemia and higher PFOA exposure levels for the cumulative (see Supplemental Material, Figure S3 http://ehp.niehs.nih.gov/wp-content/uploads/122/12/ehp.1307943.s001.508.pdf) or yearly exposure metrics with Bayesian calibration (some observed HRs relative to quintile 1 were < 1), in analyses combining ages and sexes, or in analyses by age and sex (data not shown).
(Enlarge Image)
Figure 2.
HRs and 95% CIs for hypercholesterolemia in the primary retrospective analysis for the combined cohorts, cumulative exposure. Quintile (Q) cut points (μg/mL per year) were < 0.142, 0.142 to < 0.234, 0.234 to < 0.630, 0.630 to < 3.579, ≥ 3.579. The analysis included 9,653 cases of self-reported hypercholesterolemia with medication. Models were stratified by single-year birth year and were either stratified by sex or controlled for sex and the interaction between sex and age. Models also controlled for years of schooling (not time-varying; < 12 years, high school diploma/GED, some college, or ≥ bachelor's degree), race (white vs. nonwhite or missing), smoking (time-varying; current, former, none), smoking duration (time-varying), smoking pack-years (time-varying linear term created by multiplying the self-reported number of packs smoked per day by the smoking duration to that point), regular alcohol consumption (time-varying; current, former, none), BMI (at time of first study survey; underweight, normal, overweight, obese), and self-reported type 2 diabetes (time-varying according to reported age at diagnosis).
In retrospective analyses restricted to community cohort members, there was again evidence of a higher hazard of hypercholesterolemia with higher PFOA exposure in analyses combining sexes and ages (see Supplemental Material, Figure S6 http://ehp.niehs.nih.gov/wp-content/uploads/122/12/ehp.1307943.s001.508.pdf). Retrospective analyses restricted to nonsmokers (not shown) showed overall patterns of association similar to the primary analysis. Retrospective analyses starting at the "qualifying age" (not shown) also showed similar patterns of association, with significant tests for trend with log cumulative and yearly exposures in analyses combining ages and sexes, and for males 40–59 years of age. If analyses were ended in 1987, a strong trend of progressively increasing hazard of hypercholesterolemia with increasing PFOA exposure was seen in analyses combining ages and sexes, for both the cumulative (log-linear trend test p < 0.001) and yearly metrics (log-linear trend test p < 0.001) (see Supplemental Material, Figure S7 http://ehp.niehs.nih.gov/wp-content/uploads/122/12/ehp.1307943.s001.508.pdf for yearly metric results). In analyses ending in progressively later years, the pattern of increasing hypercholesterolemia hazard with increasing PFOA exposure became progressively weaker for both exposure metrics.
Coronary Artery Disease. In the primary retrospective analysis, there was no clear association between PFOA exposure and coronary artery disease (Figure 3) (combining sexes and ages, HRs and 95% CIs by increasing quintile): cumulative 1.00 (reference), 1.26 (95% CI: 1.10, 1.45), 1.17 (95% CI: 1.02, 1.35), 0.99 (95% CI: 0.86, 1.14), 1.07 (95% CI: 0.93, 1.23); yearly 1.00 (reference), 1.02 (95% CI: 0.90, 1.16), 1.06 (95% CI: 0.93, 1.21), 0.95 (95% CI: 0.84, 1.09), 0.97 (95% CI: 0.85, 1.11). Among males 20–39 years of age, there appeared to be higher hazards of coronary artery disease in quintiles 2–5 relative to quintile 1, but CIs were wide for this age group, and there was no significant trend (log linear test for trend, cumulative exposure p = 0.64). Prospective analyses also showed no evidence of increasing hazards for coronary artery disease with increasing PFOA exposure, and showed HRs for quintiles 2–5 relative to quintile 1 that were often < 1 (see Supplemental Material, Figure S3 http://ehp.niehs.nih.gov/wp-content/uploads/122/12/ehp.1307943.s001.508.pdf).
(Enlarge Image)
Figure 3.
HRs and 95% CIs for coronary artery disease in the primary retrospective analysis for the combined cohorts, cumulative exposure. Quintile (Q) cut points (μg/mL per year) were < 0.147, 0.14 to < 0.248, 0.248 to < 0.717, 0.717 to < 5.058, ≥ 5.058. The analysis included 2,468 cases of validated coronary artery disease. Models were stratified by single-year birth year and were either stratified by sex or controlled for sex and the interaction between sex and age. Models also controlled for years of schooling (not time-varying; < 12 years, high school diploma/GED, some college, or ≥ bachelor's degree), race (white vs. nonwhite or missing), smoking (time-varying; current, former, none), smoking duration (time-varying), smoking pack-years (time-varying linear term created by multiplying the self-reported number of packs smoked per day by the smoking duration to that point), regular alcohol consumption (time-varying; current, former, none), BMI (at time of first study survey; underweight, normal, overweight, obese), and self-reported type 2 diabetes (time-varying according to reported age at diagnosis).
In retrospective analyses restricted to community cohort members, patterns of associations were similar to those seen for the combined cohorts, with no clear associations (see Supplemental Material, Figure S8 http://ehp.niehs.nih.gov/wp-content/uploads/122/12/ehp.1307943.s001.508.pdf). Analyses restricted to nonsmokers and analyses starting at the "qualifying age" (not shown) were also similar. Analyses varying the year in which the analysis ended did not show clear monotonic trends across exposure quintiles, but had a suggestion of an association with coronary artery disease for cumulative and yearly PFOA exposure (for the sexes combined and among men) when the analysis was ended in years closer to peak PFOA exposures (see Supplemental Material, Figure S9 http://ehp.niehs.nih.gov/wp-content/uploads/122/12/ehp.1307943.s001.508.pdf, for yearly exposure results among men) (for analyses ending in 1987, log linear tests for trend: ages and sexes combined, cumulative p = 0.07, yearly p = 0.03; men, ages combined cumulative p = 0.11, yearly p = 0.061). Primary retrospective analysis for myocardial infarction also showed patterns of associations that were similar to those seen for coronary artery disease overall (data not shown).
Results
Cohort Characteristics
Characteristics of study participants overall and by cohort are shown in Table 1. In the combined cohort, 54% were female, 97% were of white race, 71% were overweight or obese, 52% reported ever smoking, and 14% reported ever having a type 2 diabetes diagnosis. The median follow-up duration was 32.6 years starting at the age in 1952 or age 20 years, and 4.4 years starting at the time of the C8 Health Project. The median PFOA serum concentration measured in 2005–2006 was 26.1 ng/mL. Median estimated serum PFOA concentrations (see Supplemental Material, Figure S2 http://ehp.niehs.nih.gov/wp-content/uploads/122/12/ehp.1307943.s001.508.pdf) were highest in the late 1990s and early 2000s (Winquist et al. 2013). Table 2 reports the number of people reporting the outcomes of interest, the number with reported current medication or medical records validation, and the number included in the analyses. The mean age at first diagnosis was 47 years for hypertension among those who reported prescription medication use, 50 years for hypercholesterolemia among those who reported prescription medication use, and 56 years for coronary artery disease among those for whom the condition was validated.
Survival Analysis Results
Hypertension. In the primary retrospective analysis, combining sexes and ages, there was no clear evidence of an association between cumulative PFOA exposure and hypertension [hazard ratios (HRs) and 95% confidence intervals (CIs) by increasing quintile: 1.00 (reference), 1.10 (95% CI: 1.02, 1.19), 1.10 (95% CI: 1.02, 1.18), 1.05 (95% CI: 0.97, 1.12), 0.98 (95% CI: 0.91, 1.06)] (Figure 1). Results of analysis using the yearly exposure metric were similar (data not shown). In age- and sex-stratified analyses (Figure 1), there was some evidence of increasing hypertension hazard with increasing cumulative or yearly exposure among females in the 20- to 39-year age group and males in the 40- to 59-year age group, but trends in hazard across exposure quintiles were not monotonic, and tests for trend using the logged exposure metrics were not statistically significant (at α = 0.05). Prospective analyses starting at the age in 2005–2006, and conditioning on absence of hypertension to that point, also showed no clear evidence of a positive association between cumulative (see Supplemental Material, Figure S3 http://ehp.niehs.nih.gov/wp-content/uploads/122/12/ehp.1307943.s001.508.pdf) or yearly PFOA exposure and hypertension, and showed HRs for some quintiles relative to quintile 1 that were < 1.
(Enlarge Image)
Figure 1.
HRs and 95% CIs for hypertension in the primary retrospective analysis for the combined cohorts, cumulative exposure. Quintile (Q) cut points (μg/mL per year) were < 0.111, 0.111 to < 0.191, 0.191 to < 0.471, 0.471 to < 2.763, ≥ 2.763. The analysis included 11,798 cases of self-reported hypertension with medication. Models were stratified by single-year birth year and were either stratified by sex or controlled for sex and the interaction between sex and age. Models also controlled for years of schooling (not time-varying; < 12 years, high school diploma/GED, some college, or ≥ bachelor's degree), race (white vs. nonwhite or missing), smoking (time-varying; current, former, none), smoking duration (time-varying), smoking pack-years (time-varying linear term created by multiplying the self-reported number of packs smoked per day by the smoking duration to that point), regular alcohol consumption (time-varying; current, former, none), BMI (at time of first study survey; underweight, normal, overweight, obese), and self-reported type 2 diabetes (time-varying according to reported age at diagnosis).
Retrospective analyses restricted to community cohort members had results similar to those of the primary analysis (see Supplemental Material, Figure S4 http://ehp.niehs.nih.gov/wp-content/uploads/122/12/ehp.1307943.s001.508.pdf); trends across PFOA exposure quintiles were not statistically significant (at α = 0.05). Results of retrospective analysis restricted to nonsmokers and analysis starting at the "qualifying age" were similar to those of the primary analysis (not shown). In retrospective analyses varying the year in which the analysis was ended, an apparent trend (although not completely monotonic) of increasing hypertension hazard with increasing cumulative PFOA exposure quintile was seen for sexes and ages combined in analyses ending in 1987; this trend became progressively less clear with longer follow-up after the time of peak exposure (see Supplemental Material, Figure S5 http://ehp.niehs.nih.gov/wp-content/uploads/122/12/ehp.1307943.s001.508.pdf).
Hypercholesterolemia. The primary retrospective analysis, combining ages and sexes, showed an elevated hazard of hypercholesterolemia for exposure quintiles 2–5 relative to quintile 1 (Figure 2). For cumulative exposure, HRs increased in quintile 2 and remained elevated with little further increase across subsequent quintiles (HRs and 95% CIs by increasing quintile): 1.00 (reference), 1.24 (95% CI: 1.15, 1.33), 1.17 (95% CI: 1.09, 1.26), 1.19 (95% CI: 1.11, 1.27), 1.19 (95% CI: 1.11, 1.28) (test for trend using log cumulative exposure p = 0.005). Analyses using the yearly exposure metric also showed an increasing hazard of hypercholesterolemia with increasing quintiles (HRs and 95% CIs by increasing quintile): 1.00 (reference), 1.07 (95% CI: 1.01, 1.15), 1.11 (95% CI: 1.04, 1.19), 1.05 (95% CI: 0.99, 1.13), 1.20 (95% CI: 1.12, 1.28) (test for trend using log yearly exposure p < 0.001). In age and sex-specific analyses, the increased hazard for hypercholesterolemia was most pronounced among men 40–59 years of age (HRs and 95% CIs by increasing quintile): cumulative: 1.00 (reference), 1.38 (95% CI: 1.21, 1.56), 1.32 (95% CI: 1.17, 1.50), 1.31 (95% CI: 1.16, 1.48), 1.44 (95% CI: 1.28, 1.62) (test for trend using log cumulative exposure p < 0.001); yearly: 1.00 (reference), 1.16 (95% CI: 1.03, 1.30), 1.19 (95% CI: 1.05, 1.34), 1.16 (95% CI: 1.03, 1.31), 1.38 (95% CI: 1.23, 1.54) (test for trend using log yearly exposure p < 0.001). In analyses combining ages, tests for interaction between log exposure and sex were statistically significant (cumulative p = 0.048, yearly p = 0.004). Tests for interaction with age were also statistically significant (cumulative exposure, sexes combined, p = 0.015, males p = < 0.001; yearly exposure, males p = 0.032). Because of these interactions, analyses stratified by sex and age are preferred to the overall analyses. Prospective analyses starting at the age in 2005–2006 and conditioning on absence of hypercholesterolemia to that point showed no evidence of a positive association between hypercholesterolemia and higher PFOA exposure levels for the cumulative (see Supplemental Material, Figure S3 http://ehp.niehs.nih.gov/wp-content/uploads/122/12/ehp.1307943.s001.508.pdf) or yearly exposure metrics with Bayesian calibration (some observed HRs relative to quintile 1 were < 1), in analyses combining ages and sexes, or in analyses by age and sex (data not shown).
(Enlarge Image)
Figure 2.
HRs and 95% CIs for hypercholesterolemia in the primary retrospective analysis for the combined cohorts, cumulative exposure. Quintile (Q) cut points (μg/mL per year) were < 0.142, 0.142 to < 0.234, 0.234 to < 0.630, 0.630 to < 3.579, ≥ 3.579. The analysis included 9,653 cases of self-reported hypercholesterolemia with medication. Models were stratified by single-year birth year and were either stratified by sex or controlled for sex and the interaction between sex and age. Models also controlled for years of schooling (not time-varying; < 12 years, high school diploma/GED, some college, or ≥ bachelor's degree), race (white vs. nonwhite or missing), smoking (time-varying; current, former, none), smoking duration (time-varying), smoking pack-years (time-varying linear term created by multiplying the self-reported number of packs smoked per day by the smoking duration to that point), regular alcohol consumption (time-varying; current, former, none), BMI (at time of first study survey; underweight, normal, overweight, obese), and self-reported type 2 diabetes (time-varying according to reported age at diagnosis).
In retrospective analyses restricted to community cohort members, there was again evidence of a higher hazard of hypercholesterolemia with higher PFOA exposure in analyses combining sexes and ages (see Supplemental Material, Figure S6 http://ehp.niehs.nih.gov/wp-content/uploads/122/12/ehp.1307943.s001.508.pdf). Retrospective analyses restricted to nonsmokers (not shown) showed overall patterns of association similar to the primary analysis. Retrospective analyses starting at the "qualifying age" (not shown) also showed similar patterns of association, with significant tests for trend with log cumulative and yearly exposures in analyses combining ages and sexes, and for males 40–59 years of age. If analyses were ended in 1987, a strong trend of progressively increasing hazard of hypercholesterolemia with increasing PFOA exposure was seen in analyses combining ages and sexes, for both the cumulative (log-linear trend test p < 0.001) and yearly metrics (log-linear trend test p < 0.001) (see Supplemental Material, Figure S7 http://ehp.niehs.nih.gov/wp-content/uploads/122/12/ehp.1307943.s001.508.pdf for yearly metric results). In analyses ending in progressively later years, the pattern of increasing hypercholesterolemia hazard with increasing PFOA exposure became progressively weaker for both exposure metrics.
Coronary Artery Disease. In the primary retrospective analysis, there was no clear association between PFOA exposure and coronary artery disease (Figure 3) (combining sexes and ages, HRs and 95% CIs by increasing quintile): cumulative 1.00 (reference), 1.26 (95% CI: 1.10, 1.45), 1.17 (95% CI: 1.02, 1.35), 0.99 (95% CI: 0.86, 1.14), 1.07 (95% CI: 0.93, 1.23); yearly 1.00 (reference), 1.02 (95% CI: 0.90, 1.16), 1.06 (95% CI: 0.93, 1.21), 0.95 (95% CI: 0.84, 1.09), 0.97 (95% CI: 0.85, 1.11). Among males 20–39 years of age, there appeared to be higher hazards of coronary artery disease in quintiles 2–5 relative to quintile 1, but CIs were wide for this age group, and there was no significant trend (log linear test for trend, cumulative exposure p = 0.64). Prospective analyses also showed no evidence of increasing hazards for coronary artery disease with increasing PFOA exposure, and showed HRs for quintiles 2–5 relative to quintile 1 that were often < 1 (see Supplemental Material, Figure S3 http://ehp.niehs.nih.gov/wp-content/uploads/122/12/ehp.1307943.s001.508.pdf).
(Enlarge Image)
Figure 3.
HRs and 95% CIs for coronary artery disease in the primary retrospective analysis for the combined cohorts, cumulative exposure. Quintile (Q) cut points (μg/mL per year) were < 0.147, 0.14 to < 0.248, 0.248 to < 0.717, 0.717 to < 5.058, ≥ 5.058. The analysis included 2,468 cases of validated coronary artery disease. Models were stratified by single-year birth year and were either stratified by sex or controlled for sex and the interaction between sex and age. Models also controlled for years of schooling (not time-varying; < 12 years, high school diploma/GED, some college, or ≥ bachelor's degree), race (white vs. nonwhite or missing), smoking (time-varying; current, former, none), smoking duration (time-varying), smoking pack-years (time-varying linear term created by multiplying the self-reported number of packs smoked per day by the smoking duration to that point), regular alcohol consumption (time-varying; current, former, none), BMI (at time of first study survey; underweight, normal, overweight, obese), and self-reported type 2 diabetes (time-varying according to reported age at diagnosis).
In retrospective analyses restricted to community cohort members, patterns of associations were similar to those seen for the combined cohorts, with no clear associations (see Supplemental Material, Figure S8 http://ehp.niehs.nih.gov/wp-content/uploads/122/12/ehp.1307943.s001.508.pdf). Analyses restricted to nonsmokers and analyses starting at the "qualifying age" (not shown) were also similar. Analyses varying the year in which the analysis ended did not show clear monotonic trends across exposure quintiles, but had a suggestion of an association with coronary artery disease for cumulative and yearly PFOA exposure (for the sexes combined and among men) when the analysis was ended in years closer to peak PFOA exposures (see Supplemental Material, Figure S9 http://ehp.niehs.nih.gov/wp-content/uploads/122/12/ehp.1307943.s001.508.pdf, for yearly exposure results among men) (for analyses ending in 1987, log linear tests for trend: ages and sexes combined, cumulative p = 0.07, yearly p = 0.03; men, ages combined cumulative p = 0.11, yearly p = 0.061). Primary retrospective analysis for myocardial infarction also showed patterns of associations that were similar to those seen for coronary artery disease overall (data not shown).
