Abstract
Background
Obesity is a chronic inflammatory condition strongly associated with the risk of numerous cancers. We examined the association between circulating high-sensitivity C-reactive protein (hsCRP), a biomarker of inflammation and strong correlate of obesity, and the risk of three understudied obesity-related cancers in postmenopausal women: ovarian cancer, kidney cancer, and multiple myeloma.
Methods
Participants were 24,205 postmenopausal women who had measurements of baseline serum hsCRP (mg/L) in the Women’s Health Initiative (WHI) CVD Biomarkers Cohort, a collection of four sub-studies within the WHI. Incident cancers were identified over 17.9 years of follow-up (n = 153 ovarian, n = 110 kidney, n = 137 multiple myeloma). hsCRP was categorized into study-specific quartiles. Adjusted Cox regression models were used to estimate hazard ratios (HR) and 95% confidence intervals (CI) for associations of baseline hsCRP with the risk of these cancers.
Results
There was no clear association between baseline hsCRP concentration and the risk of ovarian cancer (quartile 4 vs. 1: HR 0.87, 95% CI 0.56–1.37), kidney cancer (HR 0.95, 95% CI 0.56–1.61), or multiple myeloma (HR 0.82, 95% CI 0.52–1.29). HRs for 1 mg/L increases in hsCRP also approximated the null value for each cancer.
Conclusions
The results of this study suggest that elevated CRP is not a major risk factor for these obesity-related cancers (ovarian or kidney cancers, or multiple myeloma) among postmenopausal women. Given the importance of elucidating the mechanisms underlying the association of obesity with cancer risk, further analysis with expanded biomarkers and in larger or pooled prospective cohorts is warranted.
Keywords: C-reactive protein, Obesity, Women’s health, Renal cell carcinoma, Multiple myeloma, Ovarian cancer
Introduction
Inflammation is hypothesized to play an important role in cancer development [1–4]. Indeed, a variety of inflammatory conditions or exposures are associated with increased risks of several cancers [3]. One important biomarker of systemic inflammation is C-reactive protein (CRP), an acute-phase reactant protein that is produced in the liver and found in the blood in response to inflammatory signaling [4]. CRP has been associated with systemic inflammation in a number of inflammatory conditions, as well as in cardiovascular disease and diabetes [4]. In addition, CRP is strongly associated with obesity, which itself is a chronic inflammatory state, in both sexes and in different populations [4, 5].
Prospective biomarker studies have reported positive associations of circulating CRP with risks of several common cancers [6, 7]. In a nested case-control study in Greece, Trichopoulos et al. [8], reported positive associations between plasma CRP and risks of cancers in the liver, lung, skin, bladder, and kidney. A recent meta-analysis of prospective studies reported a strong positive association between CRP and ovarian cancer risk [9].
The Women’s Health Initiative (WHI) is a prospective study that was designed to examine common causes of morbidity and mortality among postmenopausal women, including cancer [10]. Previous reports from the WHI have identified positive associations between circulating CRP and risks of breast [11], colorectal [12], and endometrial cancer [13]. Notably, obesity is a common risk factor for each of these diseases [14]. Herein, we examined the association of baseline high-sensitivity CRP (hsCRP) with the risk of three less-common obesity-related cancers [14], ovarian, kidney, and multiple myeloma, which have not been previously studied in the WHI and for which the scientific evidence is limited.
Methods
Women’s Health Initiative
The WHI consists of a clinical trial (CT) (Trial registration: https://clinicaltrials.gov/ identifier, NCT00000611) and an observational study (OS). Detailed methods of the study are given elsewhere [10, 15, 16]. Briefly, 161,808 women, aged 50–79 years, were recruited at 40 US clinical centers between 1 September 1993 and 31 December 1998. The WHI CT included three overlapping components: two placebo-controlled hormone therapy trials [estrogen-alone (n = 10,739) and estrogen plus progestin (n = 16,608)]; a dietary modification compared to usual diet trial (n = 48,836); and a calcium/vitamin D supplementation placebo-controlled trial (n = 36,282) [17–19]. Participants in the OS were 93,676 women who were screened for participation in the CT but were ineligible or unwilling to participate, or who were directly recruited [20]. After the original WHI study ended in 2005, two WHI extension studies ending in 2010 and 2015, respectively, were carried out to collect additional follow-up data regarding new morbidity as well as mortality. Women provided written informed consent for participation in both the original and extension studies. Human Subjects Review Committees at all participating institutions approved the WHI study protocol.
Data collection
WHI participants attended baseline-screening visits, during which they completed self-administered questionnaires that elicited detailed information on demographics, medical, and reproductive history, family history of cancer, physical activity, and other risk factors. Height (cm) and weight (kg) were measured by clinic staff, and used to determine body mass index (BMI; kg/m2).
The WHI CVD biomarkers cohort
Participants who had measurements of baseline CRP, fasting serum glucose and insulin, and other clinical parameters that were made in four sub-studies within WHI were assembled into the CVD Biomarkers Cohort (n = 24,558) to explore relationships between these biomarkers and CVD outcomes. Some sub-studies entailed selecting a random sample; others selected participants based on specific age and race/ethnicity criteria (Supplemental Fig. 1). As such, participants of the CVD Biomarkers Cohort were less likely to be white (47 vs. 82%), had fewer years of education (23.3 vs. 28.3% post-graduate education), were more likely to smoke (9.5 vs. 7.0%), and had a higher prevalence of diabetes (9.7 vs. 6.0%) than participants in the parent WHI OS + CT cohort. They were also less likely to use oral contraceptives (36.6 vs. 41.3%) or menopausal hormone therapy (39.4 vs. 56.0%). For the current study, n = 353 women were excluded due to missing hsCRP data, leaving n = 24,205 available for analysis. For analyses of ovarian cancer, we additionally excluded n = 4,430 women who reported bilateral oophorectomy and n = 49 who reported a prior diagnosis of ovarian cancer, leaving n = 19,726 for analysis. Data regarding prior history of kidney cancer or multiple myeloma at baseline were unavailable in the WHI.
Follow-up for cancer and censoring
Incident invasive cancer cases were self-reported annually in the OS; in the CT, they were reported semi-annually until 2005, and annually thereafter. Cases were confirmed by medical record review by physician adjudicators. After a median follow-up of 17.9 years, n = 153 invasive ovarian, n = 110 invasive kidney, and n = 137 multiple myelomas were identified within this sub-sample. Participants were right-censored from the analysis at the earliest of the following events: end of original follow-up for participants who were not enrolled in the WHI Extension Study, withdrawal from the study, death, loss of contact, or 31 December 2015, the last date of the WHI Extension study data adjudication.
Biospecimen collection and laboratory measures
Fasting blood samples were collected from WHI participants at their baseline-screening visit. Women were asked not to take hypoglycemic agents or non-steroidal anti-inflammatory drugs ≤ 48 h prior to the visit, except those taken regularly; and to abstain from smoking (< 1 h) or vigorous activity (≤ 12 h) prior to the visit. Blood was drawn with the participant in a seated position and processed locally according to standardized protocols for venipuncture [15]. Biospecimens were centrifuged, separated by layers, labeled, and stored at −70 °C until shipped monthly on dry ice to the WHI central repository (McKesson BioServices, Rockville, MD), where samples are held for long-term storage at −80 °C [15]. All laboratory measurements were made at the Advanced Research and Diagnostic Laboratory at the University of Minnesota. Blinded quality control samples of pooled blood from healthy women were run with each batch of study samples (5%; n = 1,075). Coefficients of variation (SD/x−) for all reported biomarkers were ≤ 3%. hsCRP (mg/L) was measured in serum by immunoassay on a Roche Modular P Chemistry analyzer using a latex particle-enhanced immunoturbidimetric assay kit (Roche Diagnostics; Indianapolis, IN) and read on the Roche Modular P Chemistry analyzer.
In addition to hsCRP, serum cholesterol and glucose were also measured. Serum high-density lipoprotein (HDL; mg/dL) cholesterol was measured using the HDLC plus 3rd generation direct method (Roche Diagnostics Corporation) on the Roche Modular P Chemistry analyzer. Serum triglycerides (mg/dL) were measured using the Triglyceride GB reagent (Roche Diagnostics Corporation) on the Roche Modular P Chemistry analyzer. Low-density lipoprotein (LDL; mg/dL) cholesterol was calculated in serum specimens with a triglyceride value < 400 mg/dL using the formula of Friedewald et al. [21]. For the vast majority of women (n = 22,314), fasting serum glucose was measured using the Gluco-quant Glucose/hexokinase reagent (Roche Diagnostics) on the Roche Modular P Chemistry analyzer. In the remaining women, glucose was determined by the hexokinase method on the Hitachi 747 (Boehringer Mannheim Diagnostics; Indianapolis, IN) [22].
Statistical analyses
hsCRP data were log-transformed for normality. All serum measures, including hsCRP, HDLC, LDLC, triglycerides, and glucose were categorized into study-specific quartiles in order to account for differences in the biomarkers’ distributions in the sub-studies. For hsCRP, the lower bound of the upper quartile was 4.70–7.38 mg/L, depending on the WHI sub-study (Table 1). Overall quartiles were also considered, to address the possibility that absolute hsCRP concentration is important. Frequencies and percentages of participants’ baseline characteristics were compared by quartiles of baseline hsCRP using t tests and χ2 tests as appropriate. Age-adjusted ratios for associations between participants’ baseline characteristics and hsCRP were calculated using generalized linear regression models. Estimates are expressed as eβ which represents the ratio of the adjusted geometric mean for hsCRP in a given category of a variable relative to the referent category for that variable. Adjusted Cox proportional hazards (CPH) models were used to estimate hazard ratios (HR) and 95% confidence intervals (CI) for associations between baseline serum hsCRP and risk of ovarian cancer, kidney cancer, or multiple myeloma. Age-adjusted CPH models and multivariable-adjusted results are presented. A final multivariable model was determined by including characteristics that were associated with hsCRP and cancer status in bivariate analyses. Models were reduced using backward selection with p < 0.05 as the cutoff for retention, with the exception of age, which was retained in all models. Final multivariable models for each cancer are given in footnotes of Table 2. Cox regression models estimating associations for hsCRP as a continuous variable for overall (rather than study-specific) hsCRP quartiles are additionally adjusted for WHI sub-study. Schoenfeld residuals were used to verify that the proportionality assumption of CPH models was not violated [23]. A number of sensitivity analyses were performed: As CRP can be affected by acute infections, an analysis restricted to women with hsCRP < 10 mg/L was performed. We have previously observed artificially increased CRP among users of menopausal hormone therapy due to hepatic first-pass metabolism [11]. In a second sensitivity analysis, regression models were restricted to non-users of estrogen-alone therapy or use of non-steroidal anti-inflammatory drugs (NSAIDs). Third, we excluded women with a baseline history of any cancer from the analysis.
Table 1.
Distribution of hsCRP in the WHI CVD biomarkers cohort and its component studies
| Study | High-sensitivity C-reactive protein (mg/L) | ||||
|---|---|---|---|---|---|
| n | Quartile 1 | Median | Quartile 3 | IQR | |
| WHI CVD Biomarkers Cohort | 24,205 | 1.30 | 2.86 | 6.03 | 4.73 |
| CVD Biomarkers for SNP Health Association Resource (SHARe) | 11,963 | 1.57 | 3.50 | 7.38 | 5.81 |
| CVD Biomarkers in Non-minority Hormone Therapy Participants | 10,161 | 1.12 | 2.35 | 4.77 | 3.65 |
| CVD Biomarkers in WHI Native Americans | 588 | 1.63 | 3.74 | 6.25 | 4.62 |
| Long Life Study-Phase III Biomarkers and GWAS | 1,493 | 1.16 | 2.27 | 4.70 | 3.54 |
IQR interquartile range
Table 2.
Associations between participants’ baseline characteristics and serum hsCRP in the Women’s Health Initiative CVD biomarkers cohort, n = 24,205
| Characteristics | Age-adjusted ratio (95% CI)a | Study-specific quartiles of serum high-sensitivity C-reactive protein (mg/L), n (%) |
p Valueb | |||
|---|---|---|---|---|---|---|
| Quartile 1, n = 6,053 | Quartile 2, n = 6,057 | Quartile 3, n = 6,043 | Quartile 4, n = 6,052 | |||
| Age, years [mean (SD)] | 0.98 (0.98–0.99) | 64.28 (7.41) | 64.31 (7.40) | 64.06 (7.20) | 63.25 (7.07) | < 0.0001 |
| Race | ||||||
| White | 1.00 referent | 2,866 (47.35) | 2,826 (46.66) | 2,838 (46.96) | 2,786 (46.03) | < 0.0001 |
| Black | 1.42 (1.37–1.47) | 2,122 (35.06) | 2,026 (33.45) | 2,103 (34.80) | 2,376 (39.26) | |
| Hispanic/Latina | 1.26 (1.21–1.31) | 918 (15.17) | 1,058 (17.47) | 954 (15.79) | 744 (12.29) | |
| Native American | 1.36 (1.24–1.49) | 147 (2.43) | 147 (2.43) | 148 (2.45) | 146 (2.41) | |
| WHI studies | ||||||
| Clinical trials | 1.00 Referent | 4,437 (73.30) | 4,569 (75.43) | 4,569 (75.61) | 4,544 (75.08) | 0.01 |
| Observational study | 1.13 (1.10–1.17) | 1,616 (26.70) | 1,488 (24.57) | 1,474 (24.39) | 1,508 (24.92) | |
| Body Mass Index, kg/m2 | ||||||
| < 25.0 | 1.00 Referent | 2,772 (46.10) | 1,443 (23.99) | 863 (14.38) | 543 (9.06) | < 0.0001 |
| 25.0–29.9 | 1.76 (1.70–1.82) | 2,204 (36.65) | 2,514 (41.79) | 2,198 (36.62) | 1,538 (25.67) | |
| 30.0–34.9 | 2.66 (2.56–2.76) | 782 (13.01) | 1,434 (23.84) | 1,789 (29.81) | 1,826 (30.47) | |
| ≥ 35.0 | 4.12 (3.96–4.29) | 255 (4.24) | 625 (10.39) | 1,152 (19.19) | 2,085 (34.80) | |
| Age at menarche, years | ||||||
| ≤ 10 | 1.00 Referent | 350 (5.81) | 386 (6.40) | 448 (7.45) | 544 (9.05) | < 0.0001 |
| 11–12 | 0.87 (0.82–0.92) | 2,387 (39.64) | 2,486 (41.24) | 2,423 (40.27) | 2,564 (42.63) | |
| 13–14 | 0.81 (0.77–0.86) | 2,575 (42.77) | 2,497 (41.42) | 2,468 (41.02) | 2,258 (37.55) | |
| ≥ 15 | 0.85 (0.80–0.91) | 709 (11.78) | 659 (10.93) | 678 (11.27) | 648 (10.77) | |
| Parity | ||||||
| Nulliparous/nulligravida | 1.00 Referent | 738 (12.30) | 655 (10.90) | 621 (10.36) | 611 (10.18) | < 0.0001 |
| 1 | 1.06 (1.00–1.12) | 665 (11.08) | 586 (9.75) | 579 (9.55) | 625 (10.42) | |
| 2–4 | 1.04 (0.99–1.08) | 3,556 (59.26) | 3,562 (59.27) | 3,545 (59.15) | 3,473 (57.88) | |
| ≥ 5 | 1.16 (1.10–1.22) | 1,042 (17.36) | 1,207 (20.08) | 1,248 (20.82) | 1,291 (21.52) | |
| Age at first birth, years | ||||||
| Nulliparous/nulligravida | 1.00 Referent | 738 (14.11) | 655 (12.39) | 621 (11.99) | 611 (11.67) | < 0.0001 |
| < 20 | 1.23 (1.17–1.30) | 887 (16.96) | 1,117 (21.13) | 1,119 (21.60) | 1,307 (24.96) | |
| 20–29 | 1.01 (0.96–1.06) | 3,154 (60.29) | 3,099 (58.63) | 3,023 (58.35) | 2,944 (56.22) | |
| ≥ 30 | 0.98 (0.92–1.05) | < 0.0001 | ||||
| Age at menopause, years | ||||||
| ≤ 45 | 1.00 Referent | 1,588 (28.90) | 1,777 (32.51) | 1,838 (33.86) | 2,116 (38.90) | |
| 46–50 | 0.82 (0.80–0.85) | 1,863 (33.91) | 1,787 (32.69) | 1,740 (32.05) | 1,629 (29.94) | |
| ≥ 51 | 0.82 (0.79–0.85) | 2,043 (37.19) | 1,902 (34.80) | 1,851 (34.09) | 1,695 (31.16) | |
| Duration of estrogen therapy, years | ||||||
| Non-user | 1.00 Referent | 4,517 (74.64) | 4,352 (71.85) | 4,192 (69.37) | 4,026 (66.52) | < 0.0001 |
| < 5 | 1.10 (1.05–1.14) | 833 (13.76) | 822 (13.57) | 827 (13.69) | 912 (15.07) | |
| ≥ 5 | 1.32 (1.27–1.37) | 702 (11.60) | 883 (14.58) | 1,024 (16.95) | 1,114 (18.41) | |
| Duration estrogen + progestin therapy, years | ||||||
| Non-user | 1.00 Referent | 5,224 (86.32) | 5,278 (87.14) | 5,295 (87.62) | 5,447 (90.00) | < 0.0001 |
| < 5 | 0.79 (0.75–0.83) | 495 (8.18) | 419 (6.92) | 428 (7.08) | 306 (5.06) | |
| ≥ 5 | 0.94 (0.88–1.00) | 333 (5.50) | 360 (5.94) | 320 (5.30) | 299 (4.94) | |
| Physical activity, MET-h/week | < 0.0001 | |||||
| Inactive | 1.00 Referent | 1,103 (18.24) | 1,380 (22.79) | 1,484 (24.57) | 1,855 (30.69) | |
| 0.1–5.24 | 0.95 (0.91–0.98) | 1,264 (20.90) | 1,465 (24.20) | 1,536 (25.43) | 1,724 (28.52) | |
| 5.25–14.49 | 0.79 (0.76–0.82) | 1,709 (28.26) | 1,557 (25.72) | 1,570 (26.00) | 1,415 (23.41) | |
| ≥ 14.50 | 0.65 (0.62–0.67) | 1,972 (32.61) | 1,652 (27.29) | 1,449 (23.99) | 1,050 (17.37) | |
| Cigarette smoking, pack-years | < 0.0001 | |||||
| Non-smoker | 1.00 Referent | 3,237 (55.27) | 3,234 (54.99) | 3,081 (52.70) | 2,906 (49.44) | |
| 0.1–6.82 | 1.03 (0.99–1.07) | 940 (16.05) | 930 (15.81) | 913 (15.62) | 881 (14.99) | |
| 6.83–21.38 | 1.06 (1.02–1.10) | 917 (15.66) | 923 (15.69) | 921 (15.75) | 937 (15.94) | |
| ≥ 21.39 | 1.18 (1.14–1.23) | 763 (13.03) | 794 (13.50) | 931 (15.93) | 1,154 (19.63) | |
| Alcohol consumption, servings/week | < 0.0001 | |||||
| Non-drinker | 1.00 Referent | 2,804 (46.42) | 3,057 (50.60) | 3,210 (53.37) | 3,545 (58.89) | |
| 0.1–0.71 | 0.87 (0.83–0.90) | 977 (16.17) | 984 (16.29) | 968 (16.09) | 919 (15.27) | |
| 0.72–3.15 | 0.79 (0.76–0.82) | 1,012 (16.75) | 908 (15.03) | 906 (15.06) | 791 (13.14) | |
| ≥ 3.16 | 0.70 (0.67–0.72) | 1,248 (20.66) | 1,093 (18.09) | 931 (15.48) | 765 (12.71) | |
| Prevalent diabetes | < 0.0001 | |||||
| No | 1.00 Referent | 5,752 (95.11) | 5,598 (92.48) | 5,390 (89.28) | 5,099 (84.38) | |
| Yes | 1.77 (1.69–1.85) | 296 (4.89) | 455 (7.52) | 647 (10.72) | 944 (15.62) | |
| Serum glucose, mg/dL2 | < 0.0001 | |||||
| Quartile 1 | 1.00 Referent | 2,286 (37.77) | 1,778 (29.35) | 1,525 (25.24) | 1,234 (20.39) | |
| Quartile 2 | 1.11 (1.07–1.15) | 1,593 (26.32) | 1,491 (24.62) | 1,327 (21.96) | 1,082 (17.88) | |
| Quartile 3 | 1.33 (1.28–1.38) | 1,428 (23.59) | 1,599 (26.40) | 1,584 (26.21) | 1,496 (24.72) | |
| Quartile 4 | 2.00 (1.93–2.08) | 746 (12.32) | 1,189 (19.63) | 1,607 (26.59) | 2,240 (37.01) | |
| Serum triglycerides, mg/dL2 | < 0.0001 | |||||
| Quartile 1 | 1.00 Referent | 2,416 (39.91) | 1,510 (24.93) | 1,145 (18.95) | 1,044 (17.25) | |
| Quartile 2 | 1.43 (1.38–1.49) | 1,594 (26.33) | 1,535 (25.34) | 1,480 (24.49) | 1,479 (24.44) | |
| Quartile 3 | 1.67 (1.61–1.73) | 1,187 (19.61) | 1,552 (25.62) | 1,630 (26.97) | 1,624 (26.83) | |
| Quartile 4 | 1.94 (1.87–2.02) | 856 (14.14) | 1,460 (24.10) | 1,788 (29.59) | 1,905 (31.48) | |
| Serum HDL cholesterol, mg/dL2 | < 0.0001 | |||||
| Quartile 1 | 1.00 Referent | 870 (14.38) | 1,404 (23.18) | 1,752 (28.99) | 2,185 (36.10) | |
| Quartile 2 | 0.77 (0.74–0.80) | 1,386 (22.90) | 1,595 (26.33) | 1,648 (27.27) | 1,668 (27.56) | |
| Quartile 3 | 0.66 (0.63–0.68) | 1,578 (26.07) | 1,555 (25.67) | 1,483 (24.54) | 1,216 (20.09) | |
| Quartile 4 | 0.50 (0.49–0.53) | 2,218 (36.65) | 1,503 (24.81) | 1,160 (19.20) | 983 (16.24) | |
| Serum LDL cholesterol, mg/dL2 | < 0.0001 | |||||
| Quartile 1 | 1.00 Referent | 1,716 (28.47) | 1,442 (24.14) | 1,416 (23.79) | 1,609 (27.04) | |
| Quartile 2 | 1.01 (0.98–1.05) | 1,547 (25.66) | 1,444 (24.18) | 1,494 (25.11) | 1,433 (24.08) | |
| Quartile 3 | 1.03 (0.99–1.07) | 1,457 (24.17) | 1,538 (25.75) | 1,449 (24.35) | 1,443 (24.25) | |
| Quartile 4 | 1.10 (1.06–1.14) | 1,308 (21.70) | 1,549 (25.93) | 1,592 (26.75) | 1,465 (24.62) | |
Derived from a generalized linear regression model of ln(hsCRP) as the outcome and adjusted for baseline age. Estimates are expressed as eβ, which represent the ratio of adjusted geometric means for hsCRP in a given characteristic category relative to the referent category
p Values derived from t tests (age) or χ2 tests
Results
Baseline participant characteristics were assessed across study-specific quartiles of serum hsCRP (Table 1). Consistent with the literature, BMI was strongly positively associated with hsCRP, with the proportions of class 1 and class 2 obesity increasing across increasing hsCRP quartiles. Relative to normal-weight women, class 1 and class 2 obese women had 2.66 (95% CI 2.56–2.76)- and 4.12 (3.96–4.29)-fold higher hsCRP. Associations between other baseline characteristics and hsCRP were weaker. Among them, increased parity, duration of estrogen therapy, cigarette smoking, prevalent diabetes, and increased categories of serum glucose and triglycerides were associated with increased hsCRP; whereas later age at menopause and higher levels of physical activity, alcohol consumption, or serum HDL cholesterol were each associated with lower hsCRP.
Associations between baseline serum hsCRP concentrations and risk of ovarian cancer, kidney cancer, and multiple myeloma are given in Table 3. We observed no evidence of an association with any of the three cancers. Contrasting the highest to the lowest study-specific quartile of hsCRP, HRs for ovarian cancer, kidney cancer, and multiple myeloma were 0.87 (95% CI 0.56–1.37), 0.95 (95% CI 0.56–1.61), and 0.82 (95% CI 0.52–1.29), respectively. Continuous HRs representing 1 mg/L increases in hsCRP also approximated the null value for each of the three cancers. Further adjustment for BMI had minimal effect on point estimates (Table 3). In sensitivity analyses, restriction of analyses to women with hsCRP < 10 mg/L, or non-users of estrogen therapy or NSAIDs, or women with no history of cancer at baseline also did not alter point estimates (data not shown). Associations remained null when overall—rather than study-specific—hsCRP quartiles were considered. Specifically, adjusted HRs for contrasts of hsCRP quartiles four to one were 0.97 (95% CI 0.63–1.50), 0.88 (95% CI 0.50–1.55), and 0.84 (95% CI 0.53–1.33), for ovarian cancer, kidney cancer, and multiple myeloma, respectively.
Table 3.
Associations of serum hsCRP with risk of ovarian cancer, kidney cancer, and multiple myeloma in the Women’s Health Initiative CVD biomarkers cohort
| Ovary |
Kidney |
Multiple myeloma |
||||
|---|---|---|---|---|---|---|
| n Cases | HR (95% CI) | n Cases | HR (95% CI) | n Cases | HR (95% CI) | |
| Age-adjusted models | ||||||
| Study-specific quartiles of hsCRP (mg/L) | ||||||
| Quartile 1 | 45 | 1.00 Referent | 26 | 1.00 Referent | 44 | 1.00 Referent |
| Quartile 2 | 33 | 0.77 (0.49–1.20) | 20 | 0.79 (0.44–1.41) | 35 | 0.81 (0.52–1.27) |
| Quartile 3 | 39 | 0.96 (0.62–1.46) | 30 | 1.20 (0.71–2.03) | 24 | 0.57 (0.35–0.94) |
| Quartile 4 | 36 | 0.97 (0.62–1.48) | 34 | 1.42 (0.85–2.37) | 34 | 0.87 (0.55–1.35) |
| p Trend = 0.96 | p Trend = 0.08 | p Trend = 0.27 | ||||
| Continuousa | 1.02 (0.88–1.18) | 1.17 (0.98–1.40) | 0.99 (0.85–1.16) | |||
| Multivariable-adjusted models | ||||||
| Study-specific quartiles hsCRP (mg/L) | ||||||
| Quartile 1 | 45 | 1.00 Referentb | 26 | 1.00 Referentc | 44 | 1.00 Referentd |
| Quartile 2 | 33 | 0.72 (0.46–1.14) | 20 | 0.65 (0.36–1.16) | 35 | 0.79 (0.50–1.23) |
| Quartile 3 | 39 | 0.87 (0.56–1.35) | 30 | 0.88 (0.52–1.51) | 24 | 0.54 (0.33–0.90) |
| Quartile 4 | 36 | 0.87 (0.56–1.37) | 34 | 0.95 (0.56–1.61) | 34 | 0.82 (0.52–1.29) |
| p Trend = 0.73 | p Trend = 0.80 | p Trend = 0.19 | ||||
| Continuousa | 0.99 (0.85–1.15) | 1.02 (0.85–1.23) | 0.98 (0.83–1.15) | |||
| Multivariable-adjusted models including body mass index | ||||||
| Study-specific quartiles of hsCRP (mg/L) | ||||||
| Quartile 1 | 45 | 1.00 Referente | 26 | 1.00 Referente | 44 | 1.00 Referente |
| Quartile 2 | 33 | 0.70 (0.44–1.10) | 20 | 0.63 (0.35–1.14) | 35 | 0.73 (0.47–1.15) |
| Quartile 3 | 39 | 0.80 (0.51–1.26) | 30 | 0.85 (0.49–1.47) | 24 | 0.48 (0.29–0.80) |
| Quartile 4 | 36 | 0.79 (0.49–1.29) | 34 | 0.89 (0.51–1.57) | 34 | 0.67 (0.41–1.10) |
| p Trend = 0.46 | p Trend = 0.97 | p Trend = 0.05 | ||||
| Continuousa | 0.96 (0.81–1.14) | 1.00 (0.82–1.22) | 0.93 (0.78–1.11) | |||
Continuous measures additionally adjusted for WHI sub-study
Additionally adjusted for WHI OS/CT participation and serum triglycerides
Additionally adjusted for physical activity, history of diabetes, and serum triglycerides
Additionally adjusted for serum triglycerides
Additionally adjusted for body mass index
Discussion
In this prospective investigation among postmenopausal women, we found no evidence to support a positive association of serum hsCRP with the risk of three obesity-related cancers. For kidney cancer and multiple myeloma, in particular, our results add to a little-researched area.
Several prospective studies have examined associations between circulating CRP and ovarian cancer risk. In a recent meta-analysis of seven prospective studies including over 2,000 ovarian cancer cases, Zeng et al. [9], reported that women with the highest third of circulating hsCRP levels—approximately > 9.75 mg/L—had a 91% increased ovarian cancer risk relative to women with the lowest third (meta-RR 1.91, 95% CI 1.51–2.40). Our results are not wholly inconsistent with these findings. The cutpoints for the highest hsCRP quartiles in the WHI CVD biomarkers cohort sub-studies (i.e., 4.66–7.37 mg/L) were much lower than for the upper tertiles among studies included in the metaanalysis. Indeed, our finding of no association is consistent with respect to hsCRP distribution and results for the middle third of the biomarker in the meta-analysis (≈ 1–10 mg/L; meta-RR 1.13, 95% CI 0.96–1.33) [9]. Taken together with prior reports, it may be that only higher levels of CRP are associated with ovarian cancer risk.
Studies examining associations of CRP with risks of kidney cancer [8] or multiple myeloma [24] are lacking. In a small Greek nested case-control study which focused on total cancer risk, a 1 SD (3.2 mg/L) increase in plasma CRP was associated with 48% increased kidney cancer risk (OR 1.48, 95% CI 1.11–1.96). The finding was based upon n = 10 kidney cancer cases. Our finding of no association, based upon n = 110 kidney cancer cases in postmenopausal women, contrasts strongly with this result. Our observation of no relation between hsCRP and multiple myeloma risk agrees with the only prior study on the topic. In 2012, Birmann et al. [24], reported no association between circulating CRP and multiple myeloma risk (n cases = 493; Quartile 4 vs. 1, OR 1.0, 95% CI 0.7–1.4), using data pooled as part of the Multiple Myeloma Cohort Consortium and which included 197 cases and 394 controls from the WHI (n = 54 cases and n = 91 non-cases overlapped with this study). When restricted to non-overlapping participants, findings were unchanged (continuous HR 1.03, 95% CI 0.84–1.27). Our findings suggest that alternative mechanisms beside obesity’s effect on chronic inflammation may be important in the etiology of these two diseases among postmenopausal women.
This study has a number of strengths, including its prospective design and the high reliability of the biomarker assays. There are also limitations. Chief among them is the potential for measurement error, as hsCRP was measured at different dates in the participating WHI sub-studies. We attempted to address this by examining sub-study-specific quartiles of exposure [25]. Similarly, CRP was measured at a single time point, which could also contribute to measurement error. Despite consistency with prior results, such error could nevertheless explain the null findings reported herein. Also, individuals regularly taking anti-inflammatory medications may have favored null associations. We are further limited in statistical power to detect modest associations, due to the relatively small number of incident cases of the three rare cancers examined within the cohort. Another limitation of our study is that the CVD Biomarkers Cohort sub-sample is not a random sample of the WHI. Women included in the sub-sample differed from the total WHI study population. However, while our results are not applicable to all of WHI, analyses conducted within the sub-sample should be valid.
The results of this study suggest that CRP is not a major factor in the etiologies of multiple myeloma, ovarian or kidney cancer among postmenopausal women. However, given the importance of elucidating the mechanisms underlying the association of obesity with cancer risk, further investigation evaluated in larger samples of women and interrogating additional inflammatory biomarkers as well as alternative pathways is warranted.
Supplementary Material
Acknowledgments
This work is supported by the National Heart, Lung, and Blood Institute, National Institutes of Health, and U.S. Department of Health and Human Services grants HHSN2682011000046C, HHSN268201100001C, HHSN268201100002C, HHSN268201100003C, and HHSN268201100004C. The funding source had no role in the collection of data, the analysis, interpretation, or writing of this report, or in the decision to submit the article for publication.
Footnotes
Electronic supplementary material The online version of this article (https://doi.org/10.1007/s10552-018-1061-9) contains supplementary material, which is available to authorized users.
Conflict of interest The authors have no competing interests to declare.
References
- 1.Coussens LM, Werb Z (2002) Inflammation and cancer. Nature 420:860–867 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Hanahan D, Weinberg RA (2011) Hallmarks of cancer: the next generation. Cell 144:646–674 [DOI] [PubMed] [Google Scholar]
- 3.Chaturvedi AK, Moore SC, Hildesheim A (2013) Invited commentary: circulating inflammation markers and cancer risk-implications for epidemiologic studies. Am J Epidemiol 177:14–19 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Brenner DR, Scherer D, Muir K et al. (2014) A review of the application of inflammatory biomarkers in epidemiologic cancer research. Cancer Epidemiol Biomark Prev 23:1729–1751 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Choi J, Joseph L, Pilote L (2013) Obesity and C-reactive protein in various populations: a systematic review and meta-analysis. Obes Rev 14:232–244 [DOI] [PubMed] [Google Scholar]
- 6.Heikkila K, Harris R, Lowe G et al. (2009) Associations of circulating C-reactive protein and interleukin-6 with cancer risk: findings from two prospective cohorts and a meta-analysis. Cancer Causes Control 20:15–26 [DOI] [PubMed] [Google Scholar]
- 7.Bittoni MA, Focht BC, Clinton SK, Buckworth J, Harris RE (2015) Prospective evaluation of C-reactive protein, smoking and lung cancer death in the Third National Health and Nutrition Examination Survey. Int J Oncol 47:1537–1544 [DOI] [PubMed] [Google Scholar]
- 8.Trichopoulos D, Psaltopoulou T, Orfanos P, Trichopoulou A, Boffetta P (2006) Plasma C-reactive protein and risk of cancer: a prospective study from Greece. Cancer Epidemiol Biomark Prev 15:381–384 [DOI] [PubMed] [Google Scholar]
- 9.Zeng F, Wei H, Yeoh E et al. (2016) Inflammatory markers of CRP, IL6, TNFalpha, and soluble TNFR2 and the risk of ovarian cancer: a meta-analysis of prospective studies. Cancer Epidemiol Biomark Prev 25:1231–1239 [DOI] [PubMed] [Google Scholar]
- 10.The Women’s Health Initiative Study Group (1998) Design of the Women’s Health Initiative clinical trial and observational study. Control Clin Trials 19:61–109 [DOI] [PubMed] [Google Scholar]
- 11.Gunter MJ, Wang T, Cushman M et al. (2015) Circulating adipokines and inflammatory markers and postmenopausal breast cancer risk. J Natl Cancer Inst. 10.1093/jnci/djv169 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Toriola AT, Cheng TY, Neuhouser ML et al. (2013) Biomarkers of inflammation are associated with colorectal cancer risk in women but are not suitable as early detection markers. Int J Cancer 132:2648–2658 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Wang T, Rohan TE, Gunter MJ et al. (2011) A prospective study of inflammation markers and endometrial cancer risk in postmenopausal hormone nonusers. Cancer Epidemiol Biomark Prev 20:971–977 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Lauby-Secretan B, Scoccianti C, Loomis D, Grosse Y, Bianchini F, Straif K (2016) Body fatness and cancer-viewpoint of the IARC working group. N Engl J Med 375:794–798 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Anderson GL, Manson J, Wallace R et al. (2003) Implementation of the Women’s Health Initiative study design. Ann Epidemiol 13:S5–S17 [DOI] [PubMed] [Google Scholar]
- 16.Hays J, Hunt JR, Hubbell FA et al. (2003) The Women’s Health Initiative recruitment methods and results. Ann Epidemiol 13:S18–S77 [DOI] [PubMed] [Google Scholar]
- 17.Stefanick ML, Cochrane BB, Hsia J, Barad DH, Liu JH, Johnson SR (2003) The Women’s Health Initiative postmenopausal hormone trials: overview and baseline characteristics of participants. Ann Epidemiol 13:S78–S86 [DOI] [PubMed] [Google Scholar]
- 18.Ritenbaugh C, Patterson RE, Chlebowski RT et al. (2003) The Women’s Health Initiative Dietary Modification trial: overview and baseline characteristics of participants. Ann Epidemiol 13:S87–S97 [DOI] [PubMed] [Google Scholar]
- 19.Jackson RD, LaCroix AZ, Cauley JA, McGowan J (2003) The Women’s Health Initiative calcium-vitamin D trial: overview and baseline characteristics of participants. Ann Epidemiol 13:S98–S106 [DOI] [PubMed] [Google Scholar]
- 20.Langer RD, White E, Lewis CE, Kotchen JM, Hendrix SL, Trevisan M (2003) The Women’s Health Initiative Observational Study: baseline characteristics of participants and reliability of baseline measures. Ann Epidemiol 13:S107–S121 [DOI] [PubMed] [Google Scholar]
- 21.Friedewald WT, Levy RI, Fredrickson DS (1972) Estimation of the concentration of low-density lipoprotein cholesterol in plasma, without use of the preparative ultracentrifuge. Clin Chem 18:499–502 [PubMed] [Google Scholar]
- 22.Peterson JI, Young DS (1968) Evaluation of the hexokinase-glucose-6-phosphate dehydrogenase method of determination of glucose in urine. Anal Biochem 23:301–316 [DOI] [PubMed] [Google Scholar]
- 23.Schoenfeld D (1982) Partial residuals for the proportional hazards regression model. Biometrika 69:239–241 [Google Scholar]
- 24.Birmann BM, Neuhouser ML, Rosner B et al. (2012) Prediagnosis biomarkers of insulin-like growth factor-1, insulin, and interleukin-6 dysregulation and multiple myeloma risk in the Multiple Myeloma Cohort Consortium. Blood 120:4929–4937 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Key TJ, Appleby PN, Allen NE, Reeves GK (2010) Pooling biomarker data from different studies of disease risk, with a focus on endogenous hormones. Cancer Epidemiol Biomark Prev 19:960–965 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.