Plasma Adiponectin and Soluble Leptin Receptor and Risk of Colorectal Cancer: A Prospective Study

Abstract

Adipokines are adipocyte-secreted hormones that may mediate the etiologic link between obesity and colorectal cancer (CRC); however, the evidence from large prospective studies is limited. We prospectively evaluated the association of plasma adiponectin and soluble leptin receptor (sOB-R) with CRC risk within the Nurses’ Health Study (1990–2008) and the Health Professionals Follow-up Study (1994–2008) among 616 incident CRC cases and 1,205 controls selected using risk-set sampling and matched on age and date of blood draw. In unconditional logistic regression with adjustment for matching factors and multiple risk factors, plasma adiponectin was significantly associated with reduced risk of CRC among men, but not among women. Compared to men in the lowest quartile of adiponectin, men in the highest quartile had a relative risk for CRC of 0.55 (95% confidence interval: 0.35, 0.86; P_trend = 0.02). The corresponding relative risk in women was 0.96 (95% confidence interval: 0.67, 1.39; P_trend = 0.74). Plasma sOB-R was not associated with overall CRC risk in either men or women. A significant heterogeneity was noted in the association between sOB-R and CRC by subsite in women (P-heterogeneity = 0.004); sOB-R was significantly associated with increased risk of rectal cancer but not colon cancer. These findings support a role for adiponectin in colorectal carcinogenesis in men. Further studies are warranted to confirm these associations and elucidate potential underlying mechanisms.

INTRODUCTION

Obesity, particularly central adiposity, is an acknowledged risk factor for colorectal cancer (CRC) (1). However, the etiologic mechanisms underlying this link have not been fully elucidated. Adipose tissue is an active endocrine organ that produces a range of hormones, collectively termed adipokines. Accumulating evidence suggests that some adipokines, namely adiponectin and leptin, might mediate the association between adiposity and CRC (2).

Adiponectin, a 30-kDa protein hormone predominantly secreted by white adipose tissues (3), circulates in humans as a trimer, hexamer and high-molecular weight (HMW) form. Circulating adiponectin is negatively correlated with obesity (4). The mechanisms underlying the reduced levels of circulating adiponectin among obese individuals may involve the abnormal hormonal milieu, enhanced oxidative stress and the pro-inflammatory state that prevail in obesity (5). Adiponectin also has significant anti-inflammatory and insulin-sensitizing effects (6). Both inflammation and insulin resistance have been suggested as potential mechanisms which underlie the association between obesity and CRC (7). In addition, adiponectin has direct anti-carcinogenic effects including inhibition of cell growth and induction of apoptosis (8). Although a few epidemiologic studies have examined the association between circulating adiponectin and CRC, evidence remains inconclusive (9).

Another adipokine, leptin, is a pivotal regulator of energy balance with demonstrated effects on neuroendocrine and immune function, and possibly carcinogenesis (10). Although leptin appears to have both mitogenic and anti-apoptotic properties in colon cancer cell lines (11), it does not promote in vivo growth of colon tumors (12). These results suggest that other regulating factors might modulate the activity of leptin. Soluble leptin receptor (sOB-R, also known as LEPR), a principal circulating binding protein in humans, has been shown to regulate the bioavailability of free leptin (13). In humans, plasma sOB-R has been inversely associated with obesity, insulin resistance and diabetes (14), each of which has been implicated in the etiology of CRC (15). The only prospective study examining sOB-R and CRC reported a significantly inverse association between circulating sOB-R and CRC risk, while no association was seen for leptin (16).

To extend these findings, we performed a nested case-control study within two prospective cohort studies, the Nurses’ Health Study (NHS) and the Health Professionals Follow-up Study (HPFS), and investigated the association of plasma adiponectin and sOB-R with risk of CRC. An earlier examination of plasma adiponectin in the HPFS observed a significant inverse association with CRC; however that analysis was limited by the number of cases (n = 179), short follow-up (n = 8 years), and lack of women. In the present study, we offer results that encompass both men and women over 13–18 years of follow-up and 616 documented CRC cases.

MATERIALS AND METHODS

Study Population

We drew participants from 2 prospective cohort studies: the NHS (started in 1976, n = 121,700 women aged 30–55 years) and the HPFS (initiated in 1986, n = 51,529 men aged 40–75 years). Detailed descriptions of the cohorts are provided elsewhere (17, 18). Briefly, in both cohorts, follow-up questionnaires were administered biennially to collect and update medical, lifestyle, and other health-related information; validated food frequency questionnaires (FFQs) were completed every 4 years to update dietary information. The follow-up rates exceeded 90% for both cohorts. We requested written permission to acquire medical records and pathology reports from participants who reported CRC. A study physician, blinded to exposure information, reviewed records to extract information on histological type, anatomic location, and stage of the cancer.

Between 1989 and 1990, 32,826 women from the NHS and between 1993 and 1995, 18,225 men from the HPFS returned a blood specimen on ice packs by overnight courier. Our procedures for blood collection, handling and storage have been previously summarized (19, 20). Among participants who provided plasma samples, we confirmed 360 CRC cases in the NHS after blood collection through October 1, 2008, and 287 incident CRC cases in the HPFS after blood draw through January 1, 2008. Using risk set sampling, we randomly selected up to 2 controls for each case matched on age (within 2 years) and month/year of blood donation from eligible participants who were alive and free of cancer (except for non-melanoma skin cancer) at the time of diagnosis of the CRC case. The study protocol was approved by the institutional review board of the Brigham and Women’s Hospital.

Laboratory Assays

Although this study was an extension to the previous study (19), we re-measured plasma adiponectin for all participants using enzyme-linked immunosorbentq assays (ELISA) from ALPCO Diagnostics (Salem, NH) (21). For sOB-R, we used the ELISA from R&D Systems (Minneapolis, MN), as previously described (14). Samples from case patients and their matched control participants were analyzed in the same batch. Quality controls samples were randomly interspersed among the case-control samples. Personnel blinded to quality control and case-control status conducted all assays. The inter-batch coefficients of variation from quality control samples were 8.6% for adiponectin and 11.5% for sOB-R. In our previous study, we measured total adiponectin for a subset of participants using another ELISA (Linco Research, St. Charles, MO) (19), which had a correlation of 0.79 with the measurements using the present assay. Biomarkers were assayed in a single run in the HPFS, and measured in two runs in the NHS. To account for possible laboratory variation over time, we used the run-specific cutoff points for association analysis in the NHS.

Other biomarkers previously measured and described in detail elsewhere were used in the analysis of this study, including C-peptide, insulin-like growth factor (IGF)-1, IGF binding protein (IGFBP)-3, high-sensitivity C-reactive protein (CRP), interleukin (IL)-6, the soluble tumor necrosis factor receptor 2 (sTNFR-2), and 25-hydroxyvitamin D (25(OH)D) (22, 23).

Assessment of Dietary and Lifestyle Factors

As in previous analyses (22), we used information collected from biennial questionnaires on major lifestyle factors for CRC, such as body weight, physical activity, smoking, family history of CRC, endoscopic screening, multivitamin use, and aspirin and nonsteroidal anti-inflammatory drug (NSAID) use. Body mass index (BMI) defined as weight in kilograms divided by the square of height in meters (kg/m²) was calculated to assess overall adiposity. In optional questions completed by 69% of the NHS women in 1986 and 65% of the HPFS men in 1987, we instructed participants to measure their waist at the umbilicus and their hips at the largest circumference between the waist and thighs while standing and without measuring over bulky clothing. We have previously shown that these self-reported measurements compared to technician measurements are reasonably accurate (24).

Using a previously validated assessment (25, 26), physical activity was calculated by summing the products of time spent at each recreational or leisure-time activity with the average metabolic equivalent (MET) for that activity. Dietary information was obtained from the validated FFQs administered in 1980, 1984, 1986 and 1990 in the NHS, and from the FFQs in 1986, 1990 and 1994 in the HPFS (27). To represent the overall dietary pattern, we calculated a summary score based on individual food intake for each participant according to the Dietary Approaches to Stop Hypertension (DASH) diet, which features high intakes of fruit, vegetables, legumes, and nuts; moderate amounts of low-fat dairy products; and low amounts of animal protein and sweets (28). Adherence to the DASH diet has been associated with reduced risk of CRC in the two cohorts (29).

Statistical Analysis

We employed the extreme Studentized deviate Many-Outlier procedure to identify statistical outliers in biomarker measurements (30). After excluding outliers and participants whose plasma failed in laboratory assays, we included 346 cases and 686 controls for the adiponectin analysis; 340 cases and 371 controls for the sOB-R analysis in the NHS, since we initially matched only one control for each case for sOB-R measurements. In the HPFS, a total of 270 cases and 519 controls were included for analyses of both adiponectin and sOB-R.

We compared means (standard deviation) and medians (interquartile ranges) of continuous variables for case and control participants using paired t test and Wilcoxon signed rank test, respectively. We used conditional logistic regression to compare categorical variables. We calculated the age- and sex-adjusted Spearman partial correlation coefficients to assess the relationships of plasma adiponectin and sOB-R with lifestyle factors and other biomarkers among control participants.

We categorized the plasma markers into quartiles within each cohort on the basis of the distribution in the controls, and estimated relative risks (RRs) and 95% confidence intervals (95% CIs) for CRC using logistic regression. Tests for trend were conducted using the median value for each quartile as a continuous variable in the regression models. We obtained similar results using conditional logistic regression models or unconditional logistic regression models with adjustment for matching factors; we thus present the results from unconditional logistic regression because for subgroup analyses unconditional regression allows us to utilize all of the controls and has enhanced power.

We then conducted multivariable analyses in men and women separately with adjustment for potential confounders, including family history of CRC, endoscopic screening, history of polyp, multivitamin use, smoking, alcohol consumption, physical activity, regular aspirin/NSAID use, plasma 25-hydroxyvitamin D, and DASH score. In women, we additionally adjusted for menopausal status and current postmenopausal hormone use. In sensitivity analyses, we also controlled for individual food or nutrient intake, including red meat, and energy-adjusted intake of folate, calcium and total fiber, instead of DASH score. Because the results using DASH score or individual items were essentially the same, we used DASH score in our multivariate models to maximize statistical power. To better approximate long-term lifestyle and nutritional status, we used cumulative averages through the time of blood collection in our analyses. Missing information was carried forward from available information from prior questionnaires.

We performed stratified analyses to evaluate whether observed associations varied by lifestyle factors or other markers. To test for multiplicative interaction, we included cross-product terms for stratification factors and biomarkers to our models. We also examined possible heterogeneity in the relationship between biomarkers and CRC according to cancer subsite using a polytomous logistic regression model. To calculate P-heterogeneity between case groups, we performed a likelihood ratio test comparing the model in which the association with biomarkers was allowed to vary between the case groups to a model in which all the associations were held constant.

We used SAS version 9.2 (SAS Institute, Inc, Cary, NC) for all analyses with the exception of the polytomous logistic regression model, for which we used STATA version 11.0 (StataCorp, College Station, TX). All statistical tests were two-sided and P < 0.05 was considered statistically significant.

RESULTS

In both cohorts, CRC cases had a significantly higher waist circumference, and were less likely to use aspirin and consume folate than controls (Table 1). In women, compared to CRC patients, control participants tended to use postmenopausal hormone, had higher calcium intake and DASH score. In men, CRC patients were more likely to have family history of CRC and to be obese than controls. Plasma concentrations of adiponectin and sOB-R significantly differed between cases and controls in men (P = 0.001), but not in women (P = 0.18 and 0.82, respectively). With respect to other plasma biomarkers, as compared with controls, CRC cases had significantly lower 25(OH)D levels.

Table 1.

Baseline Characteristics of Study Participants in the Nurses’ Health Study (1990) and the Health Professionals Follow-up Study (1994)

Baseline characteristics	Women			Men
	Cases (n = 346)	Controls (n = 686)	P value	Cases (n = 270)	Controls (n = 519)	P value
Mean age at blood draw (SD), year	59.0 (6.7)	59.0 (6.7)	0.57	65.8(8.3)	65.7(8.3)	0.28
Colorectal cancer in a parent or sibling, %	13.9	12.0	0.43	19.6	13.9	0.03
History of previous endoscopy, %	12.1	15.2	0.20	56. 7	67.1	0.004
History of polyp, %	7.51	4.96	0.11	14.1	14.8	0.75
Postmenopausal, %	85.8	86.9	0.51	-	-	-
Current use of hormones, %a	38.1	46.3	0.03	-	-	-
Current multivitamin use, %	34.8	38.9	0.22	47.4	52.0	0.26
Regular aspirin use (≥2 tablets/week), %b	38.4	46.1	0.02	41.9	48.6	0.05
Regular NSAID use (≥2 tablets/week), %	17.2	18.3	0.73	11.5	12.1	0.87
Current smoker, %	14.5	12.3	0.31	5.00	4.87	0.95
Mean body mass index (SD), kg/m2	26.0 (4.93)	25.5 (4.72)	0.13	26.2 (3.05)	25.4 (2.71)	<0.001
Mean waist circumference (SD), inch	31.9 (4.71)	31.3 (4.35)	0.05	38.6 (3.51)	37.5 (3.32)	<0.001
Mean waist-to-hip ratio (SD)	0.79 (0.09)	0.78 (0.08)	0.24	0.96 (0.05)	0.94 (0.05)	<0.001
Mean physical activity (SD), MET-hours/week	16.6 (19.0)	16.9 (21.3)	0.78	31.9 (27.0)	31.0 (25.1)	0.65
Mean daily intakes (SD)
Alcohol, g	5.67 (9.63)	5.48 (9.44)	0.75	12.3 (14.9)	12.0 (14.8)	0.71
Folate, μg	416 (204)	453 (239)	0.01	494 (210)	521 (228)	0.09
Calcium, mg	995 (553)	1075 (565)	0.04	920 (387)	926 (342)	0.67
Total fiber, g	18.6 (5.80)	18.9 (5.72)	0.37	22.0 (6.33)	22.6 (6.47)	0.18
Red meat as main dish, servings	0.30 (0.18)	0.29 (0.17)	0.48	0.27 (0.22)	0.26 (0.18)	0.52
Mean DASH score (SD)	23.7 (4.14)	24.4 (4.37)	0.02	24.2 (4.62)	24.7 (4.55)	0.10
Median adiponectin (IQR), μg/mL	8.04 (5.31–11.08)	8.19 (5.85–10.64)	0.18	4.99 (3.30–6.89)	5.32 (3.71–7.70)	0.001
Median sOB-R (IQR), ng/mLc	32.6 (26.8–39.0)	32.0 (27.2–39.7)	0.82	25.1 (20.8–29.3)	26.3 (21.9–31.5)	0.001
Median CRP (IQR), mg/Lc	1.52 (0.65–3.28)	1.67 (0.71–3.61)	0.004	1.34 (0.67–2.62)	1.13 (0.60–2.21)	0.96
Median IL-6 (IQR), pg/mLc	1.16 (0.81–1.90)	1.15 (0.78–1.79)	0.27	1.60 (0.99–2.65)	1.40 (0.94–2.26)	0.54
Median sTNFR-2 (IQR), ng/mLc	2.65 (2.24–3.14)	2.58 (2.17–3.08)	0.80	2.73 (2.34–3.22)	2.73 (2.35–3.32)	0.18
Median C-Peptide (IQR), ng/mlc	1.93 (1.34–2.84)	1.82 (1.27–2.77)	0.98	2.32 (1.60–3.49)	2.09 (1.40–3.25)	0.36
Median 25(OH)D (IQR), ng/mLc	24.0 (17.5–30.0)	26.1 (19.6–32.1)	<0.001	27.5 (22.4–33.4)	28.8 (23.1–34.2)	0.09
Median IGF-1/IGFBP-3 ratio (IQR)c	0.15 (0.11–0.18)	0.14 (0.11–0.18)	0.08	0.13 (0.07–0.16)	0.12 (0.07–1.16)	0.19

Abbreviations: 25(OH)D, 25-hydroxyvitamin D; CRP, C-reactive protein; DASH, Dietary Approaches to Stop Hypertension; NSAID, non-steroidal anti-inflammatory drug; IGF-1, insulin-like growth factor-1; IGFBP-3, insulin-like growth factor binding protein-3; IL-6, interleukin-6; IQR, inter-quartile range; MET, metabolic equivalent; SD, standard deviation; sOB-R, soluble leptin receptor; sTNFR-2, soluble tumor necrosis factor receptor 2.

^aPercentage is among postmenopausal women.

^bA standard tablet contains 325-mg aspirin.

^cIn the Nurses’ Health Study, 340 cases and 371 controls were available for sOB-R analysis, and some paricipants had missing values on the measurements of other biomarkers (3 women for IL-6, 3 women for sTNFR-2, 15 women for C-peptide, 22 women for 25(OH)D, and 21 women for IGF-1/IGFBP-3 ratio). In the Health Professionals Follow-up Study, there were 1, 3 and 1 men without measurements of CRP, C-peptide, and IGF-1/IGFBP-3 ratio, respectively.

As shown in Table 2, plasma levels of total adiponectin and sOB-R had a significantly positive correlation (r = 0.29 and 0.37 in women and men, respectively; P < 0.001). Both markers demonstrated significantly inverse correlations with BMI, waist circumference, waist-to-hip ratio, inflammatory markers, and C-peptide. In contrast, physical activity, DASH score and plasma 25(OH)D appeared to be positively correlated with adiponectin and sOB-R levels. Alcohol consumption was positively correlated with plasma adiponectin levels, whereas pack-years of smoking and IGF-1/IGFBP-3 ratio displayed inverse correlations with sOB-R. These correlations did not appear to differ by gender.

Table 2.

Age-Adjusted Spearman Partial Correlation Coefficients of Plasma Adiponectin and Soluble Leptin Receptor with Lifestyle Factors and Biomarkers among Control Participants in the Nurses’ Health Study (1990) and the Health Professionals Follow-up Study (1994)^a

Variable	Adiponectin, μg/mL		Soluble Leptin Receptor, ng/mL
	Women	Men	Women	Men
Soluble Leptin Receptor, ng/mL	0.29b	0.37b	-	-
Body mass index, kg/m2	−0.31b	−0.24b	−0.37b	−0.40b
Waist circumference, inch	−0.32b	−0.21b	−0.33b	−0.38b
Waist-to-hip ratio	−0.32b	−0.26b	−0.34b	−0.26b
Physical activity, MET-hours/week	0.05d	0.06d	0.09d	0.15b
Pack-years of smoking	−0.02d	−0.01d	−0.15c	−0.15b
Alcohol consumption, g/day	0.11c	0.08	−0.03d	<0.001d
DASH score	0.03d	0.09d	0.09d	0.20b
CRP, mg/L	−0.25b	−0.17b	−0.16b	−0.18b
IL-6, pg/mL	−0.22b	−0.11c	−0.23b	−0.21b
sTNFR-2, ng/mL	−0.14b	−0.11c	−0.19b	−0.11c
25(OH)D, ng/mL	0.09c	0.08d	0.03d	0.14b
C-Peptide, ng/mL	−0.38b	−0.30b	−0.44b	−0.30b
IGF-1/IGFBP-3 ratio	0.01d	0.02d	−0.25b	−0.08d

Abbreviations: 25(OH)D, 25-hydroxyvitamin D; CRP, C-reactive protein; DASH, Dietary Approaches to Stop Hypertension; MET, metabolic equivalent = (caloric need/kilogram body weight per hour activity)/(caloric need/kilogram body weight per hour at rest); IGF-1, insulin-like growth factor-1; IGFBP-3, insulin-like growth factor binding protein-3; IL-6, interleukin-6; sTNFR-2, soluble tumor necrosis factor receptor 2;.

^aCorrelation analysis for biomarkers was restricted to the participants with measurement information available, as described in the footnote of Table 1.

^bP<0.001

^cP< 0.05

^dP≥0.05

Table 3 shows the associations of plasma adiponectin and sOB-R with risk of CRC in both cohorts. Adiponectin was not associated with CRC risk in women, but was significantly associated with reduced risk of CRC in men. After adjusting for matching factors and multiple risk factors for CRC, men in the highest quartile (Q4) of adiponectin had a 45% lower risk of CRC than those in the lowest quartile (Q1) (95% CI: 0.35, 0.86, P_trend = 0.02). Further adjustment for BMI did not essentially alter the results. However, adding waist circumference instead of BMI to the multivariate model attenuated the inverse association between adiponectin and CRC (for Q4 versus Q1: multivariable RR = 0.78, 95% CI: 0.46, 1.32, P_trend = 0.39). For sOB-R, no association was observed with CRC risk in either women or men after adjustment for major risk factors of CRC including BMI (for Q4 versus Q1: RR = 1.23, 95% CI: 0.77, 1.97, P_trend = 0.53 in women; RR = 0.77, 95% CI: 0.48, 1.24, P_trend = 0.33 in men).

Table 3.

Relative Risk of Colorectal Cancer According to Plasma Adiponectin and Soluble Leptin Receptor in the Nurses’ Health Study (1990–2008) and Health Professionals Follow-up Study (1994–2008)

Qa	Medianb	No. of Cases	No. of Controls	Model 1c	Model 2d	Model 3e
				RR (95% CI)	RR (95% CI)	RR (95% CI)
Adiponectin, μg/mL
Women
Q1	4.56, 4.01	103	172	1.00 (Referent)	1.00 (Referent)	1.00 (Referent)
Q2	7.05, 7.28	77	172	0.75 (0.52, 1.08)	0.80 (0.55, 1.17)	0.82 (0.56, 1.20)
Q3	9.50, 9.07	74	171	0.72 (0.50, 1.05)	0.71 (0.49, 1.04)	0.74 (0.51, 1.10)
Q4	12.7, 12.5	92	171	0.90 (0.63, 1.29)	0.96 (0.67, 1.39)	1.01 (0.69, 1.49)
Ptrendf				0.59	0.74	0.99
Men
Q1	3.00	88	129	1.00 (Referent)	1.00 (Referent)	1.00 (Referent)
Q2	4.50	61	130	0.68 (0.46, 1.03)	0.69 (0.46, 1.05)	0.71 (0.47, 1.08)
Q3	6.18	75	130	0.83 (0.56, 1.24)	0.88 (0.59, 1.33)	0.92 (0.61, 1.39)
Q4	9.95	46	130	0.51 (0.33, 0.79)	0.55 (0.35, 0.86)	0.61 (0.38, 0.96)
Ptrendf				0.007	0.02	0.07
Soluble Leptin Receptor, ng/mL
Women
Q1	23.7, 25.7	84	94	1.00 (Referent)	1.00 (Referent)	1.00 (Referent)
Q2	29.0, 30.7	72	92	0.88 (0.57, 1.35)	0.89 (0.58, 1.39)	0.93 (0.60, 1.45)
Q3	34.6, 37.1	102	93	1.24 (0.82, 1.87)	1.32 (0.86, 2.03)	1.40 (0.91, 2.18)
Q4	44.8, 46.0	82	92	0.99 (0.65, 1.51)	1.11 (0.71, 1.75)	1.23 (0.77, 1.97)
Ptrendf				0.79	0.81	0.53
Men
Q1	19.4	83	129	1.00 (Referent)	1.00 (Referent)	1.00 (Referent)
Q2	24.2	70	130	0.83 (0.56, 1.25)	0.80 (0.53, 1.21)	0.85 (0.56, 1.29)
Q3	28.2	66	130	0.79 (0.52, 1.18)	0.80 (0.53, 1.23)	0.89 (0.58, 1.38)
Q4	36.2	51	130	0.61 (0.40, 0.93)	0.66 (0.42, 1.04)	0.77 (0.48, 1.24)
Ptrendf				0.02	0.09	0.33

Abbreviation: CI, confidence interval; Q, quartile; RR, relative risk.

^aQuartiles and their medians of plasma markers were based on the distribution among the control participants. For the Nurses’ Health Study in women, the run-specific cutoff points were used.

^bFor the Nurses’ Health Study in women, the median values of each quartile separately for the two runs (1990–2004, 2006–2008) are given.

^cAdjusted for matching factors (age at blood draw and date of blood draw).

^dAdjusted for matching factors (age at blood draw and date of blood draw), fasting status of blood collection, colorectal cancer in parent or sibling, prior lower gastrointestinal endoscopy, history of polyp, regular use of multivitamins, pack-years of smoking (never smoking, <10, 10–24, 25–49, ≥50 pack-years), alcohol consumption (0, 0–5, 5–15, >15g/day), physical activity (tertile, metabolic equivalent hours per week), regular aspirin/NSAID use (≥2 tablets/week), plasma 25-hydroxyvitamin D (tertile, ng/mL), and Dietary Approaches to Stop Hypertension (DASH) score (quartile). In NHS, menopausal status and current postmenopausal hormone use were additionally adjusted.

^eAdditionally adjusted for body mass index (kg/m²).

^fTests for trend were conducted using the median values for each quartile of analyte.

We also examined whether adjustment for other biomarkers may influence the associations of adiponectin and sOB-R with CRC risk in both cohorts. Among participants with available biomarker measurements, the associations for adiponectin and sOB-R remained essentially unchanged when plasma C-peptide, IGF-1/IGFBP3 ratio, CRP, IL-6 and sTNFR-2 were included in the multivariable model (Model 3 in Table 3) individually or in combination. For example, in men the multivariable RRs of CRC comparing extreme quartiles of adiponectin were 0.64 (95% CI: 0.40, 1.03, P_trend = 0.13) after adjusting for C-peptide, and 0.63 (95% CI: 0.40, 1.00, P_trend = 0.11) after adjusting for IGF-1/IGFBP3 ratio.

We also investigated the joint associations for adiponectin and sOB-R, and did not observe any significant interaction between the two markers in either women or men (P_interaction = 0.14 and 0.80, respectively) (Supplementary Table 1). Compared to individuals in the lowest tertiles of both markers, those in the highest tertile of adiponectin and lowest tertile of sOB-R showed a substantially decreased risk of CRC. However, no further decrease was seen with increasing levels of sOB-R among those in the highest tertile of adiponectin.

We further performed analyses by cancer site (Table 4). For adiponectin, no association was found with risk of either colon or rectal cancer in women (P_{heterogeneity} = 0.36). In contrast, in men the highest quartile of plasma adiponectin was significantly associated with reduced risk of colon cancer and non-significantly with rectal cancer. For sOB-R, we observed a significant difference in its association with cancer risk by subsite in women (P_{heterogeneity} = 0.004). A significantly increased risk of for Q4 compared to Q1 was found for rectal cancer but not for colon cancer. In contrast, sOB-R had no significant association with either colon or rectal cancer among men (P_{heterogeneity} = 0.72).

Table 4.

Relative Risk of Colorectal Cancer According to Plasma Adiponectin and Soluble Leptin Receptor by Cancer Subsite, in the Nurses’ Health Study (1990–2008) and Health Professionals Follow-up Study (1994–2008)

	Q1a	Q2		Q3		Q4		Ptrendc	Pheterogeneityd
	No. of Cases	No. of Cases	RR (95% CI)b	No. of Cases	RR (95% CI)b	No. of Cases	RR (95% CI)b
Adiponectin
Women									0.36
Colon cancer	86	56	0.72 (0.47, 1.09)	56	0.67 (0.44, 1.02)	75	0.95 (0.63, 1.43)	0.77
Rectal cancer	17	21	1.31 (0.65, 2.65)	18	1.18 (0.57, 2.49)	17	1.26 (0.59, 2.71)	0.59
Men									0.95
Colon cancer	68	48	0.72 (0.46, 1.14)	56	0.89 (0.57, 1.39)	35	0.60 (0.36, 0.98)	0.08
Rectal cancer	20	13	0.67 (0.32, 1.41)	19	1.03 (0.52, 2.03)	11	0.64 (0.29, 1.40)	0.41
Soluble Leptin Receptor
Women									0.004
Colon cancer	72	61	0.88 (0.55, 1.40)	76	1.14 (0.71, 1.81)	58	0.91 (0.55, 1.51)	0.54
Rectal cancer	12	11	1.08 (0.44, 2.66)	26	3.22 (1.44, 7.21)	24	3.45 (1.47, 8.07)	0.005
Men									0.72
Colon cancer	67	53	0.79 (0.50, 1.25)	50	0.84 (0.52, 1.34)	37	0.69 (0.42, 1.16)	0.19
Rectal cancer	16	17	1.07 (0.51, 2.23)	16	1.12 (0.53, 2.38)	14	1.10 (0.50, 2.41)	0.80

Abbreviation: CI, confidence interval; Q, quartile; RR, relative risk.

^aReference category (RR = 1).

^bAdjusted for the same variables as in model 3 in Table 3, but using polytomous logistic regression.

^cTests for trend were conducted using the median values for each quartile of analyte.

^dP_{heterogeneity} for associations with colon versus rectal cancer was estimated using likelihood ratio test by comparing the polytomous logistic regression model constraining common effects for all variables on both outcomes to the model allowing the effect for analyte to vary by outcome.

We subsequently conducted analyses according to selected subgroups (Table 5). The relationship between adiponectin and CRC significantly differed by CRP levels among women (P_interaction = 0.02), with a positive association in low CRP group and inverse association in high CRP group, although neither was statistically significant. In men, the inverse association between adiponectin and CRC appeared stronger among the subgroups that had lower physical activity levels, did not regularly use aspirin/NSAID, and had higher CRP or C-peptide levels (all P_trend < 0.05). However, formal tests for interaction did not attain significance for any of these factors. Similarly, sOB-R was significantly associated with reduced CRC risk among men who did not regularly use aspirin/NSAID or had higher sTNFR-2 levels (all P_trend < 0.05), although a test of interaction was not statistically significant.

Table 5.

Relative Risk of Colorectal Cancer Associated with Continuous Log-Transformed Concentrations of Plasma Adiponectin and Plasma Soluble Leptin Receptor, by Subgroups, in the Nurses’ Health Study (1990–2008) and the Health Professionals Follow-up Study (1994–2008)

	Adiponectin								Soluble Leptin Receptor
	Women				Men				Women		Men
	No. of cases	No. of controls	RR (95% CI)a	Ptrendb	No. of cases	No. of controls	RR (95% CI)a	Ptrendb	RR (95% CI)a	Ptrendb	RR (95% CI)a	Ptrendb
Body mass indexc,d
< 25	182	389	1.08 (0.69, 1.68)	0.74	102	254	0.74 (0.44, 1.24)	0.25	1.05 (0.45, 2.49)	0.91	0.57 (0.20, 1.64)	0.30
≥ 25	164	297	0.99 (0.64, 1.55)	0.97	168	265	0.77 (0.50, 1.20)	0.25	1.71 (0.59, 4.93)	0.32	0.93 (0.40, 2.15)	0.86
Pinteractione			0.60				0.69		0.89		0.38
Physical activityd,f
<15 in women, <25 in men	221	438	1.07 (0.73, 1.57)	0.72	132	257	0.62 (0.39, 0.97)	0.04	1.65 (0.75, 3.65)	0.21	0.44 (0.18, 1.10)	0.08
≥15 in women, ≥25 in men	125	248	0.82 (0.49, 1.39)	0.47	138	262	0.83 (0.51, 1.33)	0.43	0.81 (0.27, 2.43)	0.71	0.72 (0.29, 1.76)	0.47
Pinteractione			0.37				0.20		0.37		0.22
Aspirin/NSAID use
Non-regular user	187	321	1.00 (0.65, 1.54)	0.99	138	237	0.64 (0.41, 1.00)	0.05	1.41 (0.59, 3.39)	0.44	0.39 (0.16, 0.95)	0.04
Regular user	159	365	1.02 (0.66, 1.58)	0.93	132	282	0.73 (0.46, 1.16)	0.18	1.11 (0.43, 2.91)	0.82	0.84 (0.34, 2.08)	0.71
Pinteractione			0.95				0.90		0.78		0.43
Median CRP, mg/L
<1.67 in women, <1.13 in men	183	343	1.34 (0.85, 2.10)	0.20	117	259	0.94 (0.58, 1.55)	0.82	1.41 (0.59, 3.35)	0.44	0.66 (0.26, 1.67)	0.37
≥1.67 in women, ≥1.13 in men	163	343	0.74 (0.47, 1.15)	0.18	153	259	0.60 (0.38, 0.95)	0.03	0.99 (0.37, 2.66)	0.99	0.53 (0.22, 1.26)	0.15
Pinteractione			0.02				0.13		0.52		0.52
Median IL-6, pg/mL
<1.15 in women, <1.40 in men	169	342	0.90 (0.57, 1.44)	0.67	110	259	0.67 (0.40, 1.12)	0.13	1.07 (0.44, 2.60)	0.88	0.68 (0.25, 1.83)	0.44
≥1.15 in women, ≥1.40 in men	176	342	1.07 (0.70, 1.63)	0.75	160	260	0.70 (0.45, 1.07)	0.10	1.40 (0.53, 3.69)	0.50	0.58 (0.25, 1.37)	0.22
Pinteractione			0.61				0.83		0.80		0.81
Median sTNFR-2, ng/mL
<2.58 in women, <2.73 in men	152	343	1.25 (0.76, 2.05)	0.37	134	259	0.68 (0.42, 1.11)	0.12	0.97 (0.36, 2.63)	0.95	1.02 (0.40, 2.61)	0.97
≥2.58 in women, ≥2.73 in men	191	343	0.96 (0.64, 1.44)	0.84	136	260	0.72 (0.46, 1.12)	0.14	1.66 (0.70, 3.93)	0.25	0.31 (0.13, 0.75)	0.009
Pinteractione			0.44				0.99		0.82		0.17
Median C-peptide, ng/mL
<1.82 in women, <2.09 in men	153	337	1.07 (0.62, 1.83)	0.81	110	258	0.87 (0.52, 1.44)	0.59	0.86 (0.30, 2.48)	0.78	0.43 (0.15, 1.20)	0.11
≥1.82 in women, ≥2.09 in men	190	337	1.04 (0.70, 1.55)	0.85	160	258	0.59 (0.37, 0.94)	0.03	1.77 (0.74, 4.23)	0.20	0.59 (0.25, 1.39)	0.23
Pinteractione			0.78				0.15		0.20		0.87

Abbreviations: BMI, body mass index; CI, confidence interval; CRP, C-reactive protein; IL-6, interleukin-6; MET, metabolic equivalent = (caloric need/kilogram body weight per hour activity)/(caloric need/kilogram body weight per hour at rest); NSAID, non-steroidal anti-inflammatory drug; RR, relative risk; sTNFR-2, soluble tumor necrosis factor receptor 2.

^aStratum-specific relative risks and 95% confidence intervals were estimated for the one-unit increase of log-transformed concentrations of adiponectin and soluble leptin receptor using logistic regression, with adjustment for the same covariates as in model 3 in Table 3. When stratified by regular aspirin/NSAID use, this variable was excluded from the multivariate model.

^bTest for trend was conducted using the continuous log-transformed concentrations of adiponectin and soluble leptin receptor.

^cWeight (kg)/height (m)².

^dWe further adjusted for the interacting variables in the continuous form to control for residual confounding in the stratified analysis.

^eMultiplicative interaction was evaluated between log-transformed total adiponectin and soluble leptin receptor concentrations and the stratified variables in a multivariable-adjusted logistic regression.

^fMetabolic equivalent (MET)-hours/week.

In sensitivity analyses, excluding CRC cases diagnosed within the first 2 years after blood draw yielded similar results. For example, the multivariable-adjusted RRs (as in model 3 of Table 3) of CRC comparing extreme quartiles of adiponectin were 1.07 (95% CI: 0.72, 1.59) in women and 0.59 (95% CI: 0.35, 0.98) in men, respectively. Further exclusion of CRC patients diagnosed within 4 years of blood collection did not essentially change the associations. The RRs of CRC comparing extreme quartiles of adiponectin were 1.02 (95% CI: 0.67, 1.54) in women and 0.57 (95% CI: 0.33, 1.00) in men, respectively. To assess the potential confounding effect of CRC screening, we restricted our analyses to participants without prior endoscopy, and the results remained unchanged (for Q4 versus Q1 of adiponectin: RR = 0.89, 95% CI: 0.60, 1.33 in women; RR = 0.43, 95% CI: 0.19, 0.94 in men). Excluding those with prior polyps also did not affect our associations (for Q4 versus Q1 of adiponectin: RR = 0.93, 95% CI: 0.63, 1.36 in women; RR = 0.56, 95% CI: 0.34, 0.92 in men). We also excluded participants with history of diabetes mellitus at baseline and the results did not materially change (data not shown).

DISCUSSION

In this prospective case-control study nested within two large cohorts, we found that high plasma adiponectin was associated with reduced risk of CRC among men, but not among women. Plasma sOB-R had no association with overall CRC risk in either men or women, but was positively associated with female rectal cancer.

Previously, we reported an inverse association between plasma adiponectin and male CRC risk in the HPFS after 8 years of follow-up (19). In the present study, we documented an additional 91 cases after 14 years of follow-up, confirming our prior results. Adiponectin has been hypothesized to protect against carcinogenesis by influencing insulin sensitivity and the inflammatory state (2), both of which have been implicated in the etiology of CRC (7). However, in this study, results for adiponectin were essentially unchanged after adjustment for C-peptide, a marker for insulin secretion, and CRP, IL-6, or sTNF-R2, biomarkers of inflammation. These results suggest alternative mechanisms for adiponectin beyond its insulin-sensitizing and anti-inflammatory actions by which it may influence CRC risk.

Adiponectin circulates in different forms in plasma. Recent evidence suggests that HMW adiponectin is the active form of this hormone with respect to insulin sensitivity and has a stronger relationship with lower risk of diabetes (21). Thus, HMW adiponectin has been hypothesized to be more closely related to CRC risk than total adiponectin. Surprisingly, however, the European Prospective Investigation into Cancer and Nutrition Study (EPIC) found that non-HMW adiponectin rather than HMW adiponectin was significantly associated with reduced risk of CRC (31). These results further suggest that adiponectin may exert its anticarcinogenic effect through mechanisms other than modulation of insulin sensitivity. Experimental studies have reported that adiponectin could directly inhibit CRC cell growth and proliferation (8, 32), and induce endothelial apoptosis (33). In parallel with these lines of evidence, genetic association studies have demonstrated that variants of adiponectin and its receptor genes are related to altered CRC risk (34, 35).

We did not find any significant association between total adiponectin and CRC incidence among women. In the EPIC study, the inverse association between total adiponectin and CRC risk was slightly stronger among men than among women (RR = 0.72 vs. 0.79), although non-HMW adiponectin was more strongly associated with reduced risk of CRC among women than among men (RR = 0.47 vs. 0.53) (31). A prospective study on colorectal adenoma, a well-established precursor lesion of CRC, also found that the inverse association of total adiponectin was restricted to men (36). Given the potential heterogeneity by gender, it is possible that failing to take into account sex-specific associations might partially explain the overall null relationship between adiponectin and CRC observed in two other prospective studies (37, 38).

Although the exact mechanism responsible for such heterogeneity by sex remains to be elucidated, the generally higher levels of adiponectin in women may contribute to the null association. Studies have found that the gender difference in adiponectin concentrations is independent of fat mass or distribution, and may result from the influence of sex-steroid hormones (39). Adiponectin levels are high in hypogonadal men and reduced by testosterone administration in both men and murine models (40). Gonadectomy increases total circulating adiponectin, and this change is inversely associated with circulating androgen, but not estradiol (41). On the other hand, obesity has been consistently associated with CRC risk, albeit with a weaker magnitude of association among women compared with men. Endogenous sex hormones have been differentially related to CRC risk in men and women (42). Thus, the heterogeneity in the association between adiposity and adiponectin and risk of CRC according to sex may also reflect a distinct influence of altered estrogen and testosterone concentrations related to adiposity (43, 44).

By subsite analysis, we found the association of adiponectin with colon cancer was stronger than with rectal cancer. This finding was expected as metabolic factors such as abdominal fatness, physical inactivity and hyperlipidemia were almost invariably associated with a larger increased risk for colon cancer than for rectal cancer (45). Consistent with the EPIC study, we also found the association between adiponectin and risk of CRC varied by CRP levels, such that the inverse relationship was stronger among individuals with high CRP levels than among those with low CRP concentrations. Since circulating CRP levels is an established marker of systematic inflammation and adiponectin possesses anti-inflammatory effect (2), it is possible that these results reflect a role for adiponectin in conferring a benefit against CRC development among people with chronic inflammatory states, but not in those with normal or low inflammation. Nevertheless, we did not find any interaction between adiponectin and other inflammatory markers than CRP. Thus, further studies are warranted to elucidate the potential influence of additional inflammatory cytokines on the anti-carcinogenic effect of adiponectin.

For sOB-R, we did not find any relationship with overall CRC development in either men or women; however, a positive association was observed with risk of rectal cancer in women. sOB-R, either the product of alternatively spliced mRNA species or proteolytic cleavage products of membrane-bound forms of leptin receptor, contains only extracellular domains that bind to circulating leptin and perhaps regulate the bioavailability of free leptin (46). In addition, some studies suggest that reduced amounts of sOB-R may reflect decreased expression of membrane-anchored leptin receptor and that sOB-R directly block leptin action, thus contributing to leptin resistance in obese individuals (47). Although leptin has been involved in adiposity-induced insulin resistance, its effect on colorectal carcinogensis is unclear and epidemiologic evidence is inconsistent (16, 48, 49). For sOB-R, only one prospective study investigated its relationship to CRC risk and observed a significantly inverse association (16). In contrast, studies on genetic variations in leptin receptors did not find any association with CRC risk (50, 51). Given the sparse data and inconsistent findings, more investigations are needed to discern the complex effects of sOB-R and leptin on CRC development. With regard to obesity and female rectal cancer, a recent meta-analysis found that abdominal adiposity assessed by waist circumference was significantly associated with increased risk of rectal cancer in women, with the magnitude even larger than in men (52). However, this result failed to replicate in either of the two subsequent large studies in the U.S. and Chinese women (53, 54). Thus, further studies on sex-specific rectal cancer risk associated with obesity are warranted.

Our study has several strengths, including prospective blood collection, measurement of multiple biomarkers related to CRC, high follow-up rate of participants, and detailed information on covariates which allowed us to adjust for potential confounding and evaluate possible effect modification. One limitation of the current study is the single measurement of plasma markers which may not represent their long-term levels. However, previous studies have shown that these markers are generally stable over time (55, 56). Other limitations include examination of only total adiponectin rather than specific forms of adiponectin, including HMW and non-HMW adiponectin, and sOB-R but not leptin.

In conclusion, in this prospective nested case-control study, plasma adiponectin was inversely associated with CRC risk in men but not women. This relationship was independent of BMI, inflammatory and other metabolic biomarkers. Plasma sOB-R did not appear to be related to the overall risk of CRC.