Serum concentrations of per- and polyfluoroalkyl substances and risk of renal cell carcinoma in the Multiethnic Cohort Study

Abstract

Per- and polyfluoroalkyl substances (PFAS) are environmentally persistent organic pollutants detectable in the serum of most U.S. adults. We previously reported a positive association between serum perfluorooctanoate (PFOA) concentrations and risk of renal cell carcinoma (RCC) within the Prostate, Lung, Colorectal and Ovarian (PLCO) Cancer Screening Trial, comprising predominantly White individuals enrolled in 1993–2001. To extend our investigations to a larger and more racially and ethnically diverse population, we conducted a nested case-control study of serum PFAS concentrations and RCC within the Multiethnic Cohort Study. We measured pre-diagnostic serum concentrations of nine PFAS among 428 rCc cases and 428 individually matched controls. We estimated odds ratios (ORs) and 95% confidence intervals (CIs) for risk of RCC in relation to each PFAS using conditional logistic regression, adjusting for RCC risk factors and other PFAS. PFOA was not associated with RCC risk overall [doubling in serum concentration, OR_continuous=0.89 (95%CI=0.67,1.18)]. However, we observed suggestive positive associations among White participants [2.12 (0.87,5.18)] and among participants who had blood drawn before 2002 [1.49 (0.77,2.87)]. Furthermore, higher perfluorononanoate (PFNA) concentration was associated with increased risk of RCC overall [fourth vs. first quartile, OR=1.84 (0.97,3.50), P_trend=0.04; OR_continuous=1.29 (0.97,1.71)], with the strongest association observed among African American participants [OR_continuous=3.69 (1.33,10.25)], followed by Native Hawaiian [2.24 (0.70,7.19)] and White [1.98 (0.92,4.25)] participants. Most other PFAS were not associated with RCC. While PFOA was not associated with RCC risk overall in this racially and ethnically diverse population, the positive associations observed among White participants and those with sera collected before 2002 are consistent with previous PLCO findings. Our study also provided new evidence of a positive association between PFNA and RCC risk that was strongest in African American participants. These findings highlight the need for additional epidemiologic research investigating PFAS exposures and RCC in large racially and ethnically diverse populations.

1. Introduction

Per- and polyfluoroalkyl substances (PFAS) are a group of synthetic chemicals used since the 1940s in a broad range of industrial and consumer applications, such as nonstick cookware, food packaging, water-resistant fabric, cosmetics, and firefighting foams (1). Despite decreased production of certain PFAS in the United States since the early 2000s, PFAS are highly persistent in the environment and remain ubiquitous contaminants of soil and water around the globe (2). Human exposure to PFAS occurs mainly via consumption of contaminated water and food and, to a lesser extent, inhalation of indoor air and dust and dermal contact (3). Once absorbed, some PFAS can accumulate in the human body over long periods of time, with serum elimination half-lives ranging from 2–8 years (4–6). According to nationally representative data from the National Health and Nutrition Examination Survey (NHANES) between 1999 and 2008, more than 95% of the general U.S. population of 12 years of age or older had detectable concentrations of four major PFAS in their blood, including perfluorooctanoate (PFOA), perfluorooctane sulfonate (PFOS), perfluorohexane sulfonate (PFHxS), and perfluorononanoate (PFNA) (7, 8). Notably, serum concentrations of several PFAS have been found to differ by race and ethnicity. For example, elevated concentrations of PFOS and PFNA, especially in middle-aged or older populations, have been observed among non-Hispanic Black compared to non-Hispanic White persons in NHANES and other studies (7–10). Data are more limited for other racial and ethnic groups; however, recent (since 2011) NHANES reports (11) and a study in midlife women (9) also reported higher serum PFNA concentrations among Asian American compared to non-Hispanic White individuals.

Given their widespread exposure and high environmental persistence, PFAS have emerged as a major public health concern worldwide, with a growing body of literature linking specific PFAS to adverse health outcomes, including cancer (12–14). In 2014, the International Agency for Research on Cancer (IARC) classified PFOA—the only PFAS evaluated by IARC to date—as “possibly carcinogenic to humans” (Group 2B) based in part on limited epidemiologic evidence of an association with kidney cancer (15, 16). This conclusion was mainly drawn from studies that reported higher kidney cancer incidence or mortality among PFOA-exposed workers at a fluoropolymer production plant in West Virginia (17), and those residing in the surrounding communities who were highly exposed to PFOA through contaminated drinking water (18, 19). More recently, our group conducted the first prospective study to investigate associations between measured pre-diagnostic serum concentrations of PFAS, including PFOA and seven other PFAS, and risk of renal cell carcinoma (RCC; the most common form of kidney cancer) in a cohort with PFAS serum concentrations comparable to those observed in the general U.S. population (10). In that nested case- control study of 324 RCC cases and 324 matched controls (89% non-Hispanic White) within the Prostate, Lung, Colorectal and Ovarian (PLCO) Cancer Screening Trial, we observed a statistically significant association between higher serum PFOA concentrations and increased RCC risk, including after adjustment for several other PFAS. We also observed positive associations with RCC risk for PFOS and PFHxS, although those associations were attenuated in mutually adjusted analyses. Importantly our findings from the PLCO cohort provided evidence in support of PFOA as a renal carcinogen and demonstrated for the first time that general population exposures may contribute to RCC development. The incidence of RCC in the U.S. differs by race and ethnicity, with higher rates of disease in African American and American Indian / Alaskan Native populations than among non-Hispanic White Americans, while Latino and Asian and Pacific Islander populations experience lower rates (20). In the current investigation, we sought to replicate our findings from PLCO in a larger and more racially and ethnically diverse population by conducting a nested case-control study to evaluate RCC risk in relation to pre-diagnostic serum concentrations of PFOA, PFOS, PFHxS, and six other PFAS in the Multiethnic Cohort Study (MEC).

2. Methods

2.1. Study population

Details of the MEC study design and data collection procedures have been described previously (21). Briefly, the MEC is a large racially, ethnically, and socioeconomically diverse prospective cohort of more than 215,000 men and women aged 45–75 years at baseline (1993–1996) from five different racial and ethnic groups (African Americans, 16% of participants; Japanese Americans, 26%; Latinos, 22%; Native Hawaiians, 6%; and Whites, 23%) living in Hawaii and California (primarily Los Angeles County). Each participant completed a self-administered mailed questionnaire that inquired about demographic factors, anthropometric measures, personal behaviors, medical conditions including history of hypertension, diet, and family history of cancer. Assignment of study participants’ race and ethnicity was based on their questionnaire responses (21). Incident cancers were identified through record linkages to the Surveillance, Epidemiology, and End Results Program’s Hawaii Tumor Registry and the California State Cancer Registry.

Biospecimen collection in MEC first began in 1994 with cancer case-control studies, where samples were collected from cases after diagnosis and from a representative set of controls. Subsequently, from 2001 to 2006, the prospective MEC biorepository was created by asking surviving cohort members to provide blood and urine specimens (22). The controls from the earlier studies and participants with sera in the prospective MEC biorepository were eligible for inclusion in the current study. Blood samples were processed within 4 hours of collection, and aliquoted into 0.5-mL cryotubes that were stored in the vapor phase of liquid nitrogen (−186°C). The majority of participants contributing to the biorepository provided fasting blood samples (~90% for >10 hours). The study was conducted in accordance with the Declaration of Helsinki; participants provided written informed consent for biospecimen collection, and the study protocol was approved by Institutional Review Boards at the University of Southern California and the University of Hawaii. De-identified samples and data were provided to National Cancer Institute (NCI) investigators and designated laboratories under a material transfer agreement between the NCI and the University of Hawaii and the University of Southern California. Analyses conducted by NCI and the designated laboratories did not constitute engagement in human subjects research given the de-identified nature of the transferred data and specimens.

2.2. Case and control selection

We included all incident cases of RCC with an available pre-diagnostic serum sample that were diagnosed through 2018. Incident RCC cases were defined as the International Classification of Diseases for Oncology, Third Revision (ICD-O-3) code C64.9, with the exclusion of transitional cell carcinomas (ICD-O-3 morphology codes 8120 and 8130). Controls were individually matched to cases with a 1:1 ratio on sex, race and ethnicity, study center (Hawaii, California), age at serum collection (±1 year), date of serum collection (±1 year), time of serum collection (±3 hours), and fasting status (0-<6, 6-<8, 8-<10, ≥10 hours). Matching criteria were gradually relaxed for a small subset (7%) of the case-control pairs, with nearly all sets matched within 3 years for both age at serum collection and date of serum collection. Eligible controls were alive and free of RCC as of the age of diagnosis of their matched case. For each case and control, we selected one parent aliquot of serum. In total, 428 RCC cases and 428 controls (n=856) were included for analysis.

2.3. PFAS and creatinine measurements

We measured serum concentrations of 11 PFAS: perfluorooctane sulfonamide (FOSA), 2-N-methyl-perfluorooctane sulfonamido acetate (MeFOSAA), 2-N-ethyl- perfluorooctane sulfonamido acetate (EtFOSAA), PFHxS, linear PFOS (n-PFOS), sum of perfluoromethylheptane sulfonic acid isomers (Sm-PFOS), linear PFOA (n-PFOA), sum of branched PFOA isomers (Sb-PFOA), PFNA, perfluorodecanoate (PFDA), and perfluoroundecanoate (PFUnDA). The limit of detection (LOD) was 0.1 μg/L for all PFAS; concentrations below the LOD were assigned a value of the $\text{LOD}/\sqrt{2}$ (0.071 μg/L). We calculated the concentrations of PFOA and PFOS by summing the concentrations of their respective isomers [i.e., PFOA (n-PFOA and Sb-PFOA) and PFOS (n-PFOS and Sm-PFOS)]. Serum PFAS concentrations were quantified at the Centers for Disease Control and Prevention (CDC, Atlanta, GA) using online solid-phase extraction coupled to high-performance liquid chromatography- isotope dilution-tandem mass spectrometry, as described previously (23). Samples from each matched case/control set were placed next to each other within the same batch. To evaluate measurement reproducibility, we included replicate quality control (QC) samples among batches (n=96; ~11% of the total number of test samples) from pooled serum from subjects recruited in Hawaii, selected to have similar distributions of race and ethnicity, age, and sex as the cases and controls. The QC samples were coded such that laboratory personnel were blinded to the replicates. All PFAS evaluated in this study had within-batch coefficients of variation ≤20% except PFUnDA (21%).

To address concerns that kidney function may affect serum PFAS concentrations and could potentially confound or modify the associations with RCC (24–26), we also measured serum creatinine and calculated estimated glomerular filtration rate (eGFR) using the 2021 CKD-EPI creatinine equation (27).

2.4. Statistical analysis

As preliminary analyses, we calculated Spearman rank correlations between PFAS concentrations among controls and examined the distributions of participant characteristics by case-control status using chi-squared tests. We also estimated adjusted geometric least-squares mean concentrations of each PFAS across categories of participant characteristics [sex, race and ethnicity, study center, calendar year of blood draw, age at blood draw, fasting status, smoking status at baseline, body mass index (BMI) at or close to time of blood collection, history of hypertension at or close to blood collection, eGFR, and (among female participants) years since menopause] among controls (n=428) using multivariable linear regression.

For our main analyses, we calculated odds ratios (ORs) and 95% confidence intervals (CIs) for the risk of RCC in relation to serum concentrations of each PFAS using conditional logistic regression among the matched case-control pairs. PFAS were modeled both as continuous variables (log2-transformed) and across categories defined using quartiles of concentrations among controls as cut-points. For PFAS with ≥20% non-detectable concentrations (EtFOSAA, PFUnDA), we assigned subjects with concentrations below the LOD as the referent group and created categories using control tertiles as cut-points. For FOSA, because 76% of results were non-detectable, we created a binary variable (non-detectable, detectable) for the analysis. Linear trends across categories of each PFAS (except FOSA) were assessed using a Wald test by modelling within-category median concentrations as a continuous variable. In addition to conditioning on the matched factors, all models were adjusted for established RCC risk factors, including BMI (<18.5, 18.5-<25, 25-<30, ≥30 kg/m², missing), eGFR (<60, 60- <90, ≥90 mL/min/1.73 m², missing), smoking status (never, former, current, missing), and history of hypertension (yes, no); missing values were assigned as separate categories. To assess independent associations of each PFAS with RCC risk, we performed analyses additionally adjusting for serum concentrations of other PFAS, including those with a priori evidence of associations in the PLCO investigation (log2- transformed concentrations of PFOA, PFOS, and PFHxS) (10) and those found to be associated with RCC risk in the current investigation [PFNA (log2-transformed) and FOSA (non-detectable, detectable)].

We also conducted analyses stratified by matching factors [race and ethnicity (African American, Japanese American, Latino, Native Hawaiian, White), sex (male, female), study center (Hawaii, California), calendar year of blood draw (<2002, ≥2002; before and after the phaseout or decreased production of certain PFAS and subsequent declines in concentrations in the general U.S. population (7)), age at blood draw (<65, >65 years), and fasting status (<10, ≥10 hours)], smoking status (never, ever), eGFR (<60, 60-<90, ≥90 mL/min/1.73 m²), history of hypertension (no, yes), and BMI (<25, 25- <30, ≥30 kg/m²) and evaluated statistical significance of multiplicative interactions using the Wald test for cross-product model terms. For analyses stratified by non-matching factors (smoking status, eGFR, history of hypertension, and BMI), we broke the matched case-control sets and fit unconditional logistic regression models adjusting for matching factors. We also investigated associations stratified by median years of follow- up (<7, ≥7 years) and restricted to clear cell RCC subtype. As sensitivity analyses, we restricted to RCC cases diagnosed >2 years after the time of blood draw (376 cases and 376 controls) and participants without any history of cancer prior to blood collection (353 cases and 351 controls). We also repeated analyses restricting to participants with no missing data on BMI, smoking, and/or eGFR (404 cases and 404 controls). Lastly, we conducted analyses restricting to post-menopausal women with >10 years since menopause at blood collection (127 cases and 133 controls) as menstruation has been related to serum PFAS concentrations.(28) All statistical analyses were performed with SAS software version 9.4 (SAS Institute, Cary, NC) and RStudio Version 1.3.1093 (http://www.rstudio.com/). All statistical tests were two-sided.

3. Results

Selected characteristics of the RCC cases and controls are presented in Table 1. Participants were racially and ethnically diverse (African American 17%, Japanese American 25%, Latino 27%, Native Hawaiian 12%, White 19%), with a median age at blood draw of 67 years. The majority of participants were male (63%), provided serum in 2002 or later (79%), and had fasted for ≥10 hours prior to blood collection (85%). Nearly all female participants (99%) were post-menopausal at blood collection and 87% of them became menopausal at least 10 years before blood draw. Cases were more likely to be overweight or obese and have a history of hypertension compared to controls (P=0.03 and 0.002, respectively). Nearly 30% of study participants had impaired kidney function (eGFR <60 mL/min/1.73 m²), although no difference between cases and controls was observed.

Table 1.

Selected characteristics [n (%)] of renal cell carcinoma cases and controls in the Multiethnic Cohort Study

Characteristic	Controls (n = 428)	Cases (n = 428)	P a
Sex
Male	268 (62.6)	268 (62.6)	—
Female	160 (37.4)	160 (37.4)
Race and ethnicity
African American	72 (16.8)	72 (16.8)	—
Japanese American	107 (25.0)	107 (25.0)
Latino	116 (27.1)	116 (27.1)
Native Hawaiian	50 (11.7)	50 (11.7)
White	80 (18.7)	80 (18.7)
Otherb	3 (0.7)	3 (0.7)
Study center
Hawaii	211 (49.3)	211 (49.3)	—
California	217 (50.7)	217 (50.7)
Age at blood draw, years
<60	82 (19.2)	82 (19.2)	—
60 to <65	92 (21.5)	93 (21.7)
65 to <70	91 (21.3)	89 (20.8)
70 to <75	86 (20.1)	88 (20.6)
≥75	77 (18.0)	76 (17.8)
Calendar year of blood draw
1994–2001	91 (21.3)	91 (21.3)	—
2002	50 (11.7)	79 (18.5)
2003	122 (28.5)	112 (26.2)
2004	106 (24.8)	87 (20.3)
2005–2006	59 (13.8)	59 (13.8)
Fasting status
<10 hoursc	63 (14.7)	63 (14.7)	—
≥10 hours	365 (85.3)	365 (85.3)
History of hypertensiond
No	229 (53.5)	183 (42.8)	0.002
Yes	199 (46.5)	245 (57.2)
Body mass indexd, kg/m2
<18.5	9 (2.1)	2 (0.5)	0.03
18.5 to <25	139 (32.5)	113 (26.4)
25 to <30	165 (38.6)	194 (45.3)
≥30	114 (26.6)	117 (27.3)
Unknown	1 (0.2)	2 (0.5)
Cigarette smoking statuse
Never	174 (40.7)	188 (43.9)	0.71
Former	181 (42.3)	175 (40.9)
Current	65 (15.2)	56 (13.1)
Unknown	8 (1.9)	9 (2.1)
eGFR, mL/min/1.73 m2
<60	116 (27.1)	124 (29.0)	0.85
60-<90	257 (60.0)	244 (57.0)
≥90	52 (12.1)	57 (13.3)
Unknown	3 (0.7)	3 (0.7)

Abbreviations: eGFR, estimated glomerular filtration rate

^aP-value for difference between cases and controls, calculated using chi-squared test for history of hypertension and smoking status and Fisher’s exact test for body mass index and eGFR. Not reported for matching factors (sex, race and ethnicity, study center, age at blood draw, calendar year of blood draw, and fasting status).

^bIncludes Chinese, Filipino, Korean and other ethnic groups.

^cIncludes fasting <10 hours or unknown.

^dBased on information reported at or close to time of blood draw.

^eReported at baseline.

Bold indicates statistical significance at P<0.05.

PFHxS, PFOS, PFOA, and PFNA were detected in ≥97% of participants samples. MeFOSAA, EtFOSAA, PFDA, and PFUnDA had non-detectable concentrations ranging from 15–30%, and the majority of participants had non-detectable FOSA concentrations (76%). The nine PFAS were moderately correlated with one another in pair-wise comparisons among controls (Supplementary Table S1); some of the strongest correlations were observed between PFNA and PFDA (p=0.84) and PFUnDA (p=0.64), PFOA-PFOS (p=0.61), PFOA-PFNA (p=0.57), and PFOS-PFHxS (p=0.55).

Among controls, we observed that the adjusted geometric least-squares mean concentrations of PFOS, PFNA, PFDA, PFUnDA, and MeFOSAA significantly differed by race and ethnicity (Supplementary Table S2). Compared to White participants, African American and Native Hawaiian participants had higher PFOS, PFNA, PFDA, and MeFOSAA concentrations and Japanese American participants had higher PFNA, PFDA, and PFUnDA concentrations. PFAS concentrations among Latino participants were generally lower than in the other racial and ethnic groups. Concentrations of PFDA and PFUnDA were higher among participants recruited from Hawaii compared to those from California (P=0.01 and 0.0001, respectively). We observed significant declining patterns of PFOS, MeFOSAA, and EtFOSAA by calendar year of blood draw, while concentrations of PFNA and PFDA increased slightly by calendar year (Supplementary Table S2). We observed higher PFAS concentrations among men in general and particularly for PFHxS (P=0.003), PFOS (P=0.02), and MeFOSAA (P=0.03). In addition, concentrations of most PFAS except PFHxS and EtFOSAA decreased marginally by increasing age at blood draw. PFAS concentrations were similar across categories of eGFR, smoking, BMI, and hypertension.

In this population, PFOA was not associated with RCC risk overall, when serum concentrations were modelled categorically [fourth vs. first quartile, OR_Q4=1.04 (95% CI=0.60,1.81)] or continuously ([doubling in concentration, OR_continuous=0.89 (0.67,1.18), Table 2]. However, we observed a suggestive positive association among White participants [OR_continuous=2.12 (0.87,5.18), Table 3; ORs ranging from 2.1 to 3.6 for quartiles 2–4 vs. quartile 1, Supplementary Table S3] and among participants who provided their serum before 2002 [OR_continuous=1.49 (0.77,2.87)], although differences by race and ethnicity (P_interaction=0.65) and calendar year (P_interaction=0.98) were not statistically significant. We also observed some evidence of heterogeneity across categories of eGFR (P_interaction=0.05), with a suggestive inverse association among those with impaired kidney function ([eGFR<60, OR_continuous=0.59 (0.34,1.01)] and non-significant positive associations among those with normal kidney function [e.g., for eGFR≥90, OR_continuous=1.25 (0.39,4.04); Table 3].

Table 2.

Odds ratios (ORs) and 95% confidence intervals (CIs) evaluating PFAS serum concentrations and risk of renal cell carcinoma in the Multiethnic Cohort Study

PFAS (μg/L)	N control	N case	OR (95% CI)a	OR (95% CI)b
PFOA c
≤3.27	106	107	1	1
>3.27–4.47	102	99	1.09 (0.73,1.64)	1.26 (0.80,1.97)
>4.47–6.22	113	122	1.05 (0.70,1.56)	1.26 (0.78,2.05)
>6.22	107	100	0.88 (0.58,1.35)	1.04 (0.60,1.81)
P-trend			0.46	0.75
Continuousd	428	428	0.87 (0.73,1.05)	0.89 (0.67,1.18)
PFOS c
<16.65	107	118	1	1
16.65-<25.05	107	105	0.90 (0.60,1.35)	1.05 (0.66,1.66)
25.05-<36.40	106	100	0.89 (0.57,1.37)	0.99 (0.58,1.68)
≥36.40	108	105	0.83 (0.53,1.30)	0.93 (0.51,1.72)
P-trend			0.44	0.72
Continuousd	428	428	0.92 (0.79,1.07)	0.95 (0.74,1.23)
PFHxS
<1.6	105	121	1	1
1.6-<2.4	103	116	0.94 (0.64,1.38)	0.97 (0.64,1.49)
2.4-<3.55	113	88	0.65 (0.44,0.98)	0.67 (0.43,1.06)
≥3.55	107	103	0.79 (0.52,1.21)	0.84 (0.50,1.41)
P-trend			0.27	0.54
Continuousd	428	428	0.84 (0.74,0.96)	0.82 (0.69,0.98)
PFNA
≤0.5	121	119	1	1
>0.5–0.8	120	113	0.88 (0.59,1.32)	1.13 (0.70,1.82)
>0.8–1.1	85	78	0.90 (0.57,1.42)	1.24 (0.71,2.16)
>1.1	102	118	1.29 (0.78,2.12)	1.84 (0.97,3.50)
P-trend			0.24	0.04
Continuousd	428	428	1.03 (0.85,1.24)	1.29 (0.97,1.71)
PFDA
≤0.1	121	128	1	1
>0.1–0.2	128	101	0.77 (0.52,1.14)	0.91 (0.57,1.46)
>0.2–0.4	106	119	1.15 (0.74,1.79)	1.37 (0.77,2.42)
>0.4	73	80	1.26 (0.75,2.14)	1.42 (0.70,2.89)
P-trend			0.24	0.26
Continuousd	428	428	1.05 (0.88,1.24)	1.11 (0.85,1.43)
PFUnDA
<LOD (non-detectable)	111	130	1	1
0.1–0.2	139	118	0.75 (0.50,1.13)	0.75 (0.48,1.17)
>0.2–0.4	77	66	0.75 (0.43,1.31)	0.79 (0.42,1.48)
>0.4	101	114	1.05 (0.61,1.82)	0.88 (0.45,1.72)
P-trend			0.34	0.80
Continuousd	428	428	0.97 (0.84,1.12)	0.90 (0.75,1.08)
MeFOSAA
≤0.3	117	118	1	1
>0.3–0.6	114	103	0.93 (0.63,1.36)	0.94 (0.64,1.40)
>0.6–1.1	96	104	1.07 (0.71,1.63)	1.13 (0.73,1.74)
>1.1	101	103	1.16 (0.74,1.81)	1.22 (0.76,1.97)
P-trend			0.42	0.34
Continuousd	428	428	1.00 (0.88,1.13)	1.00 (0.87,1.14)
EtFOSAA
<LOD (non-detectable)	100	90	1	1
0.1-<0.3	102	112	1.16 (0.76,1.76)	1.15 (0.75,1.78)
0.3-<0.7	112	104	1.07 (0.66,1.72)	1.12 (0.67,1.86)
≥0.7	114	122	1.35 (0.76,2.37)	1.35 (0.73,2.49)
P-trend			0.34	0.40
Continuousd	428	428	1.03 (0.92,1.16)	1.03 (0.90,1.18)
FOSA e
Non-detectable	331	319	1	1
Detectable	95	107	1.58 (0.95,2.60)	1.78 (1.05,3.04)

Abbreviations: EtFOSAA, 2-N-ethyl-perfluorooctane sulfonamido acetate; FOSA, perfluorooctane sulfonamide; MeFOSAA, 2-N-methyl-perfluorooctane sulfonamido acetate; PFAS, per- and polyfluoroalkyl substances; PFDA, perfluorodecanoate; PFHxS, perfluorohexane sulfonate; PFNA, perfluorononanoate; PFOA, perfluorooctanoate; PFOS, perfluorooctane sulfonate; PFUnDA, perfluoroundecanoate.

^aAnalyses were conducted using conditional logistic regression models of matched case-control sets adjusting for smoking status (never, former, current, missing), body mass index (<18.5, 18.5-<25, 25-<30, ≥30 kg/m², missing), history of hypertension (yes, no), and eGFR (<60, 60-<90, ≥90 mL/min/1.73 m², missing).

^bAdditionally adjusted for serum concentrations of PFOA (log2-transformed), PFOS (log2-transformed), PFHxS (log2-transformed), PFNA (log2-transformed), and FOSA (non-detectable, detectable, missing).

^cPFOA and PFOS here represent the sum of their corresponding linear and branch isomers.

^dContinuous odds ratios (95% confidence intervals) for renal cell carcinoma in relation to a 1-unit increase in serum PFAS concentrations on the log base 2 scale, corresponding to a doubling in analyte concentrations.

^eFOSA analysis was conducted using 426 cases and 426 controls because of missing FOSA measurements [1 case and 1 control; we additionally excluded their respective matched control (n=1) and case (n=1) for conditional logistic regression].

Bold indicates statistical significance at P<0.05.

Table 3.

Odds ratios (ORs)^a and 95% confidence intervals (CIs) evaluating PFAS serum concentrations and risk of renal cell carcinoma stratified by selected characteristics in the Multiethnic Cohort Study

Characteristic	N controls	N cases	PFOAb	PFOSb	PFHxS	PFNA	FOSA
All	428	428	0.89 (0.67,1.18)	0.95 (0.74,1.23)	0.82 (0.69,0.98)	1.29 (0.97,1.71)	1.78 (1.05,3.04)
Race and ethnicity
African American	72	72	1.01 (0.51,1.98)	0.40 (0.20,0.79)	0.70 (0.38,1.31)	3.69 (1.33,10.25)	1.05 (0.26,4.19)
Japanese	107	107	1.00 (0.47,2.13)	1.54 (0.91,2.62)	0.69 (0.47,1.01)	0.82 (0.41,1.61)	2.28 (0.56,9.29)
Latino	116	116	1.04 (0.53,2.03)	1.30 (0.68,2.49)	0.64 (0.42,0.99)	1.03 (0.56,1.89)	2.05 (0.63,6.66)
Native Hawaiian	50	50	0.57 (0.21,1.55)	0.59 (0.20,1.73)	1.05 (0.59,1.88)	2.24 (0.70,7.19)	6.69 (0.86,51.9)
White	80	80	2.12 (0.87,5.18)	0.36 (0.13,0.95)	0.83 (0.56,1.23)	1.98 (0.92,4.25)	1.35 (0.38,4.83)
P-interactionc			0.65	0.18	0.84	0.96	0.81
Sex
Male	268	268	1.05 (0.72,1.53)	0.94 (0.67,1.34)	0.80 (0.64,1.02)	1.22 (0.84,1.76)	1.04 (0.56,1.94)
Female	160	160	0.67 (0.39,1.12)	0.95 (0.60,1.49)	0.80 (0.58,1.09)	1.59 (0.96,2.63)	11.72 (2.84,48.38)
P-interactionc			0.65	0.89	0.97	0.81	0.005
Study center
Hawaii	211	211	0.96 (0.61,1.49)	0.92 (0.63,1.36)	0.88 (0.70,1.11)	1.31 (0.88,1.95)	1.65 (0.79,3.47)
California	217	217	0.88 (0.60,1.31)	0.94 (0.65,1.35)	0.75 (0.56,0.99)	1.31 (0.86,2.00)	1.97 (0.89,4.34)
P-interactionc			0.29	0.41	0.17	0.55	0.78
Calendar year of blood draw d
<2002	90	90	1.49 (0.77,2.87)	0.77 (0.40,1.48)	0.67 (0.42,1.06)	0.89 (0.49,1.62)	1.06 (0.36,3.08)
2002+	336	336	0.80 (0.56,1.13)	0.96 (0.73,1.28)	0.85 (0.70,1.03)	1.43 (1.01,2.01)	2.48 (1.19,5.16)
P-interactionc			0.98	0.52	0.42	0.25	0.17
Age at blood draw d
<65	170	170	0.84 (0.51,1.41)	0.97 (0.59,1.57)	0.69 (0.51,0.94)	1.22 (0.72,2.07)	2.84 (1.01,8.04)
≥65	249	249	0.92 (0.63,1.33)	0.94 (0.68,1.29)	0.95 (0.76,1.19)	1.30 (0.91,1.84)	1.51 (0.77,2.96)
P-interactionc			0.32	0.30	0.03	0.28	0.88
Years of follow-up
<7 years	206	206	0.86 (0.57,1.31)	0.74 (0.52,1.05)	0.88 (0.69,1.13)	1.58 (1.04,2.41)	2.73 (1.19,6.30)
≥7 years	222	222	0.91 (0.60,1.37)	1.28 (0.84,1.94)	0.73 (0.56,0.96)	1.11 (0.75,1.64)	1.26 (0.61,2.61)
P-interactionc			0.66	0.23	0.85	0.86	0.41
eGFR e
<60	116	124	0.59 (0.34,1.01)	0.97 (0.58,1.64)	0.93 (0.68,1.27)	1.53 (0.90,2.59)	1.73 (0.84,3.57)
60-<90	257	244	1.11 (0.77,1.60)	0.97 (0.68,1.38)	0.82 (0.65,1.03)	1.43 (1.03,1.97)	1.19 (0.72,1.96)
≥90	52	57	1.25 (0.39,4.04)	0.51 (0.20,1.35)	0.56 (0.26,1.20)	1.53 (0.53,4.42)	2.02 (0.50,8.26)
P-interactionc			0.05	0.13	0.10	0.54	0.88

Abbreviations: eGFR, estimated glomerular filtration rate, FOSA, perfluorooctane sulfonamide; PFAS, per- and polyfluoroalkyl substances; PFHxS, perfluorohexane sulfonate; PFNA, perfluorononanoate; PFOA, perfluorooctanoate; PFOS, perfluorooctane sulfonate.

^aUnless otherwise specified, analyses were conducted using conditional logistic regression models of matched case-control sets adjusting for smoking status (never, former, current, missing), body mass index (<18.5, 18.5-<25, 25-<30, ≥30 kg/m², missing), history of hypertension (yes, no), eGFR (<60, 60-<90, ≥90 mL/min/1.73 m², missing), PFOA (log2- transformed), PFOS (log2-transformed), PFHxS (log2-transformed), PFNA (log2-transformed), FoSa (non-detectable, detectable, missing). Continuous odds ratios (95% confidence intervals) for renal cell carcinoma in relation to a 1-unit increase in serum PFAS concentrations on the log base 2 scale, corresponding to a doubling in analyte concentrations. For FOSA analysis, ORs for detectable FOSA vs. non-detectable FOSA are presented.

^bPFOA and PFOS here represent the sum of their corresponding linear and branch isomers.

^cP-value for multiplicative interaction evaluated using the Wald test for a cross-product term between PFAS and stratifying factor.

^dFor analysis of calendar year of blood draw and age at blood draw, unmatched sets were excluded from the analysis (n=4 and n=18, respectively).

^eAnalyses were conducted using unconditional logistic regression models adjusting for matching factors [sex, race and ethnicity (African American, Japanese, Latino, Native Hawaiian, White, other), study center (Hawaii, California), calendar year of blood draw (1994–2001, 2002, 2003, 2004, 2005–2006), age at blood draw (<60, 60-<65, 65-<70, 70-<75, ≥75), fasting (≥10 hours, <10 hours or missing)], smoking status (never, former, current, missing), body mass index (<18.5, 18.5-<25, 25-<30, ≥30 kg/m², missing), history of hypertension (yes, no), eGFR (<60, 60-<90, ≥90 mL/min/1.73 m², missing), PFOA (log2-transformed), PFOS (log2-transformed), PFHxS (log2-transformed), PFNA (log2-transformed), and FOSA (non- detectable, detectable, missing).

Bold indicates statistical significance at P<0.05.

Furthermore, higher PFNA concentrations were associated with increased risk of RCC overall, with a statistically significant exposure-response trend [OR_Q4=1.84 (0.97,3.50), P_trend=0.04; OR_continuous=1.29 (0.97,1.71), Table 2], adjusting for RCC risk factors and other PFAS. The strongest association with PFNA was observed among African American participants [OR_continuous=3.69 (1.33,10.25), Table 3], and we also observed elevated but non-statistically significant risk estimates among Native Hawaiian [OR_continuous=2.24 (0.70,7.19)] and White [OR_continuous=1.98 (0.92,4.25)] participants. Positive associations with PFNA were also most apparent among participants with blood drawn in 2002 or later [1.43 (1.01,2.01); P_interaction=0.25, Table 3], those with blood samples collected <7 years prior to diagnosis [1.58 (1.04,2.41); P_interaction=0.86], and those who fasted for ≥10 hours [1.58 (1.15,2.17); P_interaction=0.005, Supplementary Table S4]. Patterns of associations with PFNA overall and by race and ethnicity were similar after further adjustment for additional PFAS that were correlated with PFNA concentrations (PFDA and PFUnDA; Supplementary Table S5).

We also observed a positive association for FOSA (detectable vs. non-detectable) overall [OR=1.78 (1.05,3.04), Table 2] after adjustment for other PFAS. This association was only evident among women [11.72 (2.84,48.38) vs. 1.04 (0.56,1.94) among men; P_interaction=0.005, Table 3]. An inverse association with RCC risk was found for continuous concentrations of PFHxS overall [OR_continuous=0.82 (0.69,0.98), Table 2] and particularly among Latino individuals [0.64 (0.42,0.99); P_interaction=0.84, Table 3] and individuals aged <65 years at blood draw [0.69 (0.51,0.94); P_interaction=0.03]. Other PFAS were not associated with RCC overall (Table 2), although for PFOS we observed a suggestive positive association in Japanese American participants [OR_continuous=1.54 (0.91,2.62), Table 3] and statistically significant inverse associations among African American [0.40 (0.20,0.79)] and White [0.36 (0.13,0.95); P_interaction=0.18] participants.

We observed no statistically significant differences in associations by other stratifying factors, including hypertension, smoking status, and BMI (Supplementary Table S4). In addition, we observed similar results in sensitivity analyses restricting to RCC cases diagnosed >2 years after serum collection, participants without any history of cancer prior to blood collection, participants with no missing data on BMI, smoking, and/or eGFR, and post-menopausal women with >10 years since menopause at blood collection.

4. Discussion

In this nested case-control study of 428 cases and 428 matched controls from a racially and ethnically diverse population-based cohort from Hawaii and California, we found no association between pre-diagnostic serum PFOA concentrations and RCC risk overall. However, we observed suggestive associations between higher PFOA concentrations and increased RCC risk among White participants and among individuals with pre-diagnostic serum samples collected prior to year 2002; such observations are notable because our previous study within the PLCO cohort consisted of a predominantly non-Hispanic White population enrolled in 1993–2001 (10). In the current study, we also observed a suggestive positive association between serum PFNA concentrations and RCC risk overall, particularly in analyses adjusted for other PFAS. Race and ethnicity-specific analyses for PFNA further revealed a statistically significant positive association with RCC risk among African American participants and positive but non-statistically significant associations among Native Hawaiian and White participants. Furthermore, participants with detectable serum concentrations of FOSA, a precursor of PFOS (29, 30), had a statistically significantly elevated risk of RCC compared to those with non-detectable FOSA concentrations. In stratified analyses, this association was only apparent in women. Finally, we found an inverse association between continuous serum concentrations of PFHxS and RCC risk overall and no associations overall for PFOS or other PFAS.

Serum concentrations of PFAS among controls in our study, overall and by demographic factors such as sex and age, are generally within the ranges observed in NHANES during a similar time period (7, 8, 11), suggesting that exposures among MEC participants reflect exposures experienced by the general U.S. population. Consistent with other studies (7–11), we observed racial and ethnic differences in concentrations of several PFAS, even after accounting for study site and other factors potentially related to PFAS concentrations. Notably, African American and Native Hawaiian participants had the highest concentrations of PFOS, while African American, Japanese American, and Native Hawaiian participants had higher concentrations of PFNA relative to Latino and White participants. Similar to our findings in this older population, elevated concentrations of PFOS and PFNA have also been observed among non-Hispanic Black individuals aged ≥55 years compared to non-Hispanic White individuals and Mexican Americans in NHANES (7), and among Black (vs. White) control participants (ages ≥55 years) in PLCO (10). Likewise, PFNA concentrations were elevated among Japanese American (vs. White) women in a multiethnic study of middle-aged women across the U.S. (9) and among Asian Americans (all ethnicities combined) in NHANES (11). Little is known about PFAS exposure among Native Hawaiian and Pacific Islander (NHPI) populations; the higher concentrations observed among Native Hawaiian participants in our study warrant further investigation in larger samples of this population. Racial and ethnic differences in PFAS concentrations likely reflect differences in sources of exposure, including drinking water supply, dietary consumption (e.g., packaged foods, fish/shellfish), and consumer product use (9, 31–33). Future research investigating race and ethnicity-specific determinants of PFAS concentrations is needed to better understand these disparities and identify potentially modifiable factors for reducing exposure.

The previously observed association between PFOA and RCC risk in the PLCO cohort (10) was not replicated overall in our current investigation within MEC. Differences between these two cohorts, particularly in the racial and ethnic distributions and time periods of blood collection, may have contributed to this discrepancy. While most participants in PLCO were non-Hispanic White (89%; vs. 6% Black and 5% other) (10), the MEC targeted recruitment from five main racial and ethnic groups (African American, Japanese American, Latino, Native Hawaiian, and White) (21), with Latino and Japanese Americans together comprising ~50% of all participants in this study. Notably, when analyses were restricted to White participants (19%), we observed a greater than two-fold increase in RCC risk associated with a doubling in serum PFOA concentration; despite a lack of statistical significance, the magnitude of this association is similar, and even slightly stronger, compared to that in PLCO (OR=1.7) (10). In contrast, we found no evidence of an association among each of the other racial and ethnic groups. Racial and ethnic variations in exposure sources and pathways could have played a role in the RCC association, with higher serum PFOA concentrations noted among non-Hispanic White compared to other groups in several studies (8, 9, 31); however, our data did not reveal such a difference. To our knowledge, this was the first study to evaluate the PFOA-RCC association by race and ethnicity and among non-White individuals. Additional research in both White and other specific racial and ethnic groups may help clarify whether this association is most relevant among White individuals and whether the results are generalizable to other populations. Furthermore, nearly 80% of participants in this study had blood drawn in 2002 or later (i.e., after the phaseout of certain PFAS), which likely explains the generally lower PFAS concentrations than those observed in PLCO. In fact, although suggestive, our finding for PFOA and RCC risk among those with sera collected before 2002 is consistent with the association observed in PLCO (10). In addition to the generally higher serum PFOA concentrations that were observed in the general U.S. population during this time period (7), it is possible that the measured PFOA concentrations among these participants in our study could also reflect more stable estimates of longer-term PFOA exposure relative to those with later serum collections when PFOA concentrations were declining in the U.S. population overall. Future studies that can evaluate associations between PFOA and RCC in populations with a wider range of serum concentrations and with exposures assessed during different secular time periods may help to clarify these relationships. Another difference between the MEC and PLCO study populations is the higher prevalence of eGFR<60 in MEC compared with PLCO (28% vs. 7%), which may be attributable in part to the older age at blood draw in MEC than PLCO (≥65 years: 59% vs. 36%, respectively). We note that there was some evidence of heterogeneity in our findings for PFOA by eGFR category, with a suggestive inverse association among participants with eGFR<60 and non-significant positive associations in those with higher eGFR.

Our investigations within the PLCO (10) and MEC cohorts are, to our knowledge, the only studies to date to assess directly measured serum PFOA (and other PFAS) concentrations in relation to kidney cancer risk in the general population. Other epidemiologic studies, including those cited in lARC’s previous evaluation of PFOA, have been conducted in occupational settings or in highly exposed communities (14, 16, 34). In a study of workers from a PFOA-producing chemical plant in West Virginia, an excess in deaths from kidney cancer was observed among those with high model-estimated cumulative serum PFOA concentrations (17), whereas another occupational cohort study in Minnesota found no evidence of elevated kidney cancer mortality or incidence among workers highly exposed to ammonium perfluorooctanoate (ammonium salt of PFOA), estimated using air monitoring data (35). Two studies that included partially overlapping populations were conducted among residents of Mid-Ohio Valley communities exposed to PFOA-contaminated drinking water; both studies reported increased kidney cancer risk associated with higher estimated PFOA serum concentrations, which were predicted using historical PFOA emission data, residential history, and drinking water source and consumption patterns (18, 19). Taken together, several epidemiologic studies of populations that have experienced relatively high exposures to PFOA support an association between PFOA and RCC risk, and the evidence from studies measuring PFAS concentrations comparable to concentrations observed in the general population is limited to PLCO and the current study and warrants additional research. Additionally, as all previous studies were conducted in populations consisting largely of non-Hispanic White individuals, associations in non-White populations need to be further investigated.

Potential mechanisms through which PFOA may contribute to RCC development include induction of oxidative stress, alterations in inflammatory pathways, suppression of immune function, and modulation of receptor-mediated activity (16, 36–38). In particular, evidence from human cells in vitro and animal experiments suggests that PFOA can promote activation of peroxisome proliferator-activated receptor alpha (16, 36), a nuclear receptor protein involved in lipid metabolism and oxidative stress pathways and has been implicated in nephrotoxicity and renal carcinogenesis (39–41). Moreover, PFOA exposure has been shown to induce renal tubular epithelial hypertrophy or hyperplasia and increased kidney weights in rodent models (36).

Our study provided novel evidence of a positive association between PFNA and RCC risk, which had the largest effect estimates among African American participants, followed by Native Hawaiian and White participants. This study is the first to present results of PFNA-RCC relationships stratified by race and ethnicity, with statistically significant or suggestive associations observed in both White and non-White individuals. Our previous investigation in PLCO found a suggestive positive association for PFNA (doubling in serum concentration, OR=1.19, 95% CI=0.91, 1.55), although the observed association became null after adjusting for other PFAS (10). Compared to results in PLCO, we observed a stronger association with PFNA in White participants (doubling in serum concentration, OR=1.98, 95% CI=0.92, 4.25), even after adjusting for other PFAS. No other previous studies have examined PFNA-RCC relationships using pre-diagnostic serum samples in the general population. In analyses stratified by calendar year of blood draw, we only observed associations among participants with blood drawn in 2002 or later; this is notable because of the observed increase in PFNA concentrations in the general U.S. population during this time period that was also apparent in our study population (7, 33). The association with RCC was also more apparent among cases diagnosed within 7 years of follow-up; whether this reflects PFNA exposure during an etiologically relevant time period for RCC development is unclear, and as such confirmation of these findings for PFNA is needed. PFNA is a suspected genotoxic carcinogen that induces DNA damage. In vitro studies reported that exposure to PFNA induced high levels of 8-hydroxy-2-deoxyguanosine, a biomarker of oxidative DNA damage, in human lymphoblastoid cells and increased DNA strand breaks in human hepatoma cells (42–44). Some experimental studies reported toxicologic effects of PFNA on kidney health. An animal study of rats demonstrated that prenatal exposure to PFNA resulted in lower renal nephron endowment and elevated blood pressure in offspring (45). Another animal study of male rats observed altered blood plasma levels of diacylglycerols, phosphatidylcholines, and cholesterol derivatives (46).

We observed a positive association with RCC for FOSA, which was not measured in the earlier PLCO study. However, this association should be interpreted cautiously given the large proportion of non-detectable concentrations (76%) and the heterogeneity in effect size in sex-stratified analyses. Notably, serum concentrations of FOSA, a precursor to PFOS, are no longer measured in NHANES because of the high proportion of non-detectable results since the 2003–2004 NHANES cycle (11), suggesting that FOSA is no longer detected in the majority of the general U.S. population.

We observed an unexpected inverse association between serum PFHxS concentrations and RCC risk, overall and with similar patterns across most racial and ethnic groups. Additionally, while PFOS was not associated with RCC risk overall in our study, statistically significant inverse associations were observed among African American and White participants. Secular trends of declining serum concentrations of PFHxS and PFOS have been observed in the general U.S. population since the early 2000s (7), and concentrations of these PFAS were lower among participants in this study compared with those in PLCO and other studies in highly exposed communities.(12, 47) There is limited and inconsistent epidemiologic evidence for the associations of PFHxS and PFOS with kidney cancer risk. In the PLCO cohort, positive associations with RCC risk were observed for both PFHxS and PFOS (doubling in serum concentration, ORs=1.27 and 1.39, respectively); however, they were attenuated and lost statistical significance in analyses adjusted for other PFAS (10). More recently, in a large Swedish cohort study of individuals who lived in Ronneby, a municipality whose drinking water supplies were historically contaminated with PFAS (predominantly PFOS and PFHxS) from firefighting foams used at a nearby military airfield, a moderately increased risk of kidney cancer was observed among those who resided in contaminated (vs. uncontaminated) areas (48). In another study of three Australian communities exposed to PFOS and PFHxS through firefighting foams, elevated incidence of kidney cancer relative to an external comparison population was observed in one but not the other two communities (49). In light of these findings and mechanistic studies supporting potential roles of PFOS and PFHxS in kidney health and carcinogenesis (e.g., PFOS-induced endothelial permeability through actin filament remodeling)(36, 37, 50), our findings may be attributable to differences in the magnitude of exposure to these PFAS during etiologically relevant time periods in our study compared with prior studies, or they could also be due to chance.

A strength of this study is that we were able to evaluate the relationship between PFAS and RCC risk in a cohort with greater racial and ethnic diversity than in previous studies, and our investigation included populations that have been understudied with respect to the health effects of PFAS exposures; there is limited epidemiologic evidence related to PFAS and cancer risk in non-White individuals, particularly NHPI and Asian Americans. In addition, NHANES biomonitoring data do not provide PFAS concentrations for NHPI, although information among Asian Americans has been added since the 2011 survey (11). Other study strengths include our use of pre-diagnostic serum PFAS concentrations, a direct assessment of exposure to specific PFAS, and analyses of individual PFAS, with and without adjustment for other PFAS. We also measured serum creatinine to adjust for kidney function (eGFR) in the analyses. This study also had some limitations. Our analyses stratifying by racial and ethnic groups had limited statistical power. We did not correct for multiple comparisons given a priori evidence of associations for some PFAS in previous studies; however, findings due to chance cannot be ruled out, and the associations observed in certain subgroups, particularly those defined by race and ethnicity, require confirmation in future studies. Another limitation is the reliance on PFAS concentrations measured from a single point in time for exposure assessment. In an investigation of intra-individual variability in PFAS concentrations within PLCO (51) using serial samples (baseline, one and five years after baseline), concentrations of most PFAS were consistent (intraclass correlation coefficient >0.7), suggesting that single-serum measurements are informative surrogates of exposure status across several years. Since nearly all female participants in our study were post-menopausal at blood draw, our findings may not be generalizable to pre-menopausal women whose serum PFAS concentrations are more likely to be affected by menstrual and reproductive factors.(28) Lastly, although our analyses adjusted for eGFR as a marker of kidney function, we did not have data on urine or serum albumin levels and were not able to take albuminuria into account. Albuminuria has been associated with lower serum PFAS concentrations, as well as an increased risk of some cancers, possibly including kidney cancer.(52–56) However, it has been reported that the prevalence of albuminuria is higher among individuals with eGFR<60 compared to those with eGFR ≥60 (33% and 8%, respectively), and most participants in our study had eGFR≥60.(57) As such, the impact of the lack of adjustment for albuminuria was likely minimal, in particular for those with no evidence of diminished kidney function. We might expect any potential underestimation of true associations between PFAS and RCC to be most apparent among those with low eGFR, as was observed in our stratified analyses of PFOA.

5. Conclusions

In this large nested-case control study of serum PFAS and RCC risk in a racially and ethnically diverse population, PFOA was not associated with RCC risk overall, although we found suggestive positive associations among White participants and those who had blood drawn before 2002, which are consistent with our previous findings in PLCO. We also provided new evidence of a positive association between PFNA and RCC risk, which was strongest in African American participants followed by Native Hawaiian and White participants. Our findings demonstrate the importance for epidemiologic studies to be more inclusive of non-White participants in order to better understand differences in PFAS exposures and their relationships with cancer risk in various racial and ethnic groups.

Highlights.

• We evaluated serum PFAS and RCC risk in a racially and ethnically diverse cohort.

• Concentrations of several PFAS differed by race and ethnicity (e.g., higher PFOS and PFNA in African American vs. White participants)

• PFOA was suggestively associated with increased RCC risk among White participants.

• We found a strong positive PFNA-RCC association among African American participants.

Abbreviations

EtFOSAA

2-N-ethyl-perfluorooctane sulfonamido acetate

MeFOSAA

2-N-methyl-perfluorooctane sulfonamido acetate

BMI

Body mass index

CIs

Confidence intervals

eGFR

Estimated glomerular filtration rate

IARC

International Agency for Research on Cancer

ICD-O-3

International Classification of Diseases for Oncology Third Revision

LOD

Limit of detection

n-PFOS

Linear perfluorooctane sulfonate

n-PFOA

Linear perfluorooctanoate

MEC

Multiethnic Cohort Study

NCI

National Cancer Institute

NHANES

National Health and Nutrition Examination Survey

NHPI

Native Hawaiians and Pacific Islanders

ORs

Odds ratios

PFAS

Per- and polyfluoroalkyl substances

PFDA

Perfluorodecanoate

PFHxS

Perfluorohexane sulfonate

PFNA

Perfluorononanoate

FOSA

Perfluorooctane sulfonamide

PFOS

Perfluorooctane sulfonate

PFOA

Perfluorooctanoate

PFUnDA

Perfluoroundecanoate

PLCO

Prostate, Lung, Colorectal and Ovarian

QC

Quality control

Sb-PFOA

Sum of branched perfluorooctanoate isomers

Sm-PFOS

Sum of perfluoromethylheptane sulfonic acid isomers