Obesity and colon cancer: Does leptin provide a link?
Obesity, a risk factor for colorectal cancer, is associated with elevated serum levels of leptin, the adipocyte-derived hormone, and insulin. Experimental and epidemiologic studies have indicated a role for insulin in the pathogenesis of colon cancer, and recent experimental studies have suggested a similar role for leptin. In a case-control study nested in the Janus Biobank, Norway, we measured serum levels of leptin and C-peptide (a marker of pancreatic insulin secretion) in cryopreserved prediagnostic sera from men (median age, 45 years) who were diagnosed with cancer of the colon (n = 235) or rectum (n = 143) after blood collection (median time, 17 years), and among 378 controls matched for age and date of blood collection. Conditional logistic regression analyses showed an approximately 3-fold increase in colon cancer risk with increasing concentrations of leptin up to an odds ratio (OR) of 2.72 (95% CI = 1.44–5.12) for top vs. bottom quartile (ptrend = 0.008). The corresponding OR for C-peptide was 1.81 (95% CI = 0.67–4.86; ptrend = 0.19). The risk estimates remained unchanged after mutual adjustment. No association of hormone levels with rectal cancer risk was found. Reproducibility of hormone measurements assessed by intraclass coefficients (ICCs) for paired samples taken 1 year apart was high for leptin (ICC = 0.82) but lower for C-peptide (ICC = 0.30). Our results suggest that leptin is a risk factor for colon cancer, and that leptin may provide a link between obesity and colon cancer. Leptin may be directly involved in colon tumorigenesis or it may serve as a sensitive and robust marker of an obesity-induced adverse endocrine environment. Only weak support for an association of insulin with colon cancer was found. © 2003 Wiley-Liss, Inc.
Obesity, a consequence of a positive energy balance caused by high intake of an energy-dense diet and a sedentary lifestyle, is a risk factor for colorectal cancer.1, 2 The association has been stronger and more consistent for men than for women as demonstrated in a recent metaanalysis,1 although the association may be of similar magnitude in premenopausal women as for men.3, 4 Obesity-induced changes in hormonal metabolism may be a link to cancer risk.1 Circulating levels of insulin are increased in obesity, and insulin has been postulated to be such a link.5 The anabolic signals of insulin can promote tumor development by inhibiting apoptosis and by stimulating cell proliferation.6 In contrast, data from epidemiologic studies are limited and only 2 prospective studies to date have shown an association between insulin exposure and colon cancer risk,7, 8 while another prospective study found no significant association.9
The adipose tissue is an important endocrine organ, in which several hormones are produced, among them leptin.10 Circulating leptin levels are closely related to the percentage and amount of adipose tissue (correlations of body mass index, BMI, with circulating levels ranging from 0.5 to 0.8) and leptin is also related to other factors in the metabolic syndrome, including insulin resistance and serum insulin levels. Leptin conveys information to the brain about the size of energy stores and leptin levels are increased by overfeeding. Energy restriction, a well-established protective factor against cancer,11 decreases leptin levels.11, 12 Leptin is also involved in regulation of blood pressure, angiogenesis and wound healing.12, 13 In recent experimental studies, leptin stimulated growth of colon cancer cells.14, 15, 16 In view of these data, we hypothesized that leptin is a link between obesity and colon cancer. We describe the results from a nested case-control study on the association of prediagnostic serum levels of leptin and C-peptide with cancers of the colon and the rectum.
MATERIAL AND METHODS
The Janus Project in Norway was started in 1973 and contains blood samples from about 600,000 subjects. The samples have been collected from men who participated in county health examinations, mostly for cardiovascular diseases, and from blood donors. The participants in the health examinations were recruited from several counties in Norway. The blood donors were from the Red Cross Blood Donor Center in Oslo. The blood collection took place during office hours, participants were not required to fast and fasting times were not recorded. Serum samples were stored at −25°C.17, 18
Incident cases of colorectal cancer and deaths were identified through linkage with the Norwegian cancer and all-cause mortality registries. Among 1,105 incident male cases of colorectal cancer, 400 men with no previous malignancy who were diagnosed with colorectal cancer more than 3 months after recruitment were randomly selected. No serum sample was available for 17 of these cases. One case with polyps and 4 cases with tumors of the appendix were excluded from analysis.
One control was randomly selected from all participants alive and free of cancer at the time of diagnosis of the index case, within sets matching the case on age (± 1 year), date at blood sampling (± 2 months) and county. Thus, 378 colorectal cancer cases and 378 controls were available for analysis. Two samples that had been collected about 1 year apart (mean, 11.3 months; range, 3.5–12.9) from 80 subjects unrelated to the case-control sets were used to assess reproducibility of hormone measurements. The study was approved by the research ethics committees of Rikshospitalet, Oslo, Norway, and Umeå University Hospital, Umeå, Sweden.
Serum concentrations of hormones were determined by standard radioimmunoassays with reagents from Linco Res. (St. Louis, MO) for leptin and from DSL (Webster, TX) for C-peptide. Samples pertaining to matched study subjects were always analyzed together in the same batch, i.e., on the same day and with the same immunoassay kit, without knowledge of the case-control status. The mean intrabatch coefficient of variation, calculated from 4 quality control samples randomly inserted in each batch, was 4.8% for leptin and 6.8% for C-peptide. Duplicate measurements of leptin and C-peptide performed in 39 samples correlated almost perfectly (rS > 0.99 and rS = 0.98, respectively).
Spearman correlation coefficients were used to examine the cross-sectional relationship between serum levels of leptin and C-peptide. Intraclass correlations for repeated measurements were calculated from variance components by the mixed procedure from SAS (Cary, NC). Odds ratios (ORs) for disease were calculated by conditional logistic regression by quartile hormone levels for colon cancer and rectal cancer separately. Quartile cutoff points were determined on the variable distribution in cases and controls combined in subgroups of interest, i.e., according to anatomic site or duration of follow-up. Confidence intervals (95% CIs) were computed using the standard errors of the pertinent regression coefficients and assuming a normal probability distribution for the estimated coefficients. Likelihood ratio tests for linear trends in risk with increasing leptin concentrations were performed using scores 1, 2, 3 and 4 for the 4 quartile levels. Tests for heterogeneity of effect were based on comparison of statistical models with and without the interaction term between exposure and potential effect modifier. All analyses were performed using the PHREG procedure for proportional hazards regression from SAS. All statistical tests and corresponding p-values were 2-sided.
Baseline characteristics are shown in Table I. A majority of men (80%) were between 40 and 50 years at recruitment and only 2 cases were diagnosed less than 6 months after blood collection. There were no significant differences in serum levels of leptin or C-peptide between cases and controls. There was a moderately strong correlation between leptin and C-peptide (rS = 0.38). The median values (and interquartile range) in the sequential samples were for leptin 4.01 (3.05–5.58) and 3.85 ng/mL (2.79–5.35), respectively, and the corresponding values for C-peptide were 0.57 (0.38–1.25) and 1.21 ng/mL (0.71–2.46). Intraclass correlation (ICC) for the sequential samples was high for leptin (ICC = 0.83; 95% CI = 0.76–0.90) but lower for C-peptide (ICC = 0.30; 95% CI = 0.18–0.89).
Table I. Baseline Characteristics for Cases of Colon and Rectal Cancer and Matched Controls in the Janus Serum Bank
|Age, years||44.8 (41.4–47.8)||44.9 (41.3–47.7)||Matched|
|Follow-up time, years||17.1 (10.0–21.5)|| || |
|Storage time, years||26.6 (24.7–28.5)||26.6 (24.6–28.5)||0.19|
|Leptin, ng/mL||2.59 (1.76–4.3)||2.49 (1.61–3.82)||0.44|
|C-peptide, ng/mL||0.33 (0.19–0.76)||0.32 (0.17–0.80)||0.97|
We found a significant increase in risk of colon cancer with increasing levels of leptin (Table II). The risk was higher for left-sided compared to right-sided tumors (pinteraction cat = 0.015 with leptin in quartile categories and pinteraction cont = 0.15 with leptin as continuous variable). Risk was also stronger for cases with a shorter storage time than median (pinteraction cat = 0.09 and pinteraction cont = 0.018). The ORs for increasing levels of leptin among case control sets with less than median storage time (26.6 years) were 1.00 (ref), 1.45 (95% CI = 0.62–3.39), 2.01 (95% CI = 0.81–4.89) and 5.19 (95% CI = 1.81–14.84; ptrend = 0.002), whereas the corresponding ORs for case sets with storage time longer than 26.6 years were 1.00, 1.24 (95% CI = 0.55–2.76), 1.30 (95% CI = 0.56–3.01) and 1.72 (95% CI = 0.76–3.89; ptrend = 0.20). Increasing levels of C-peptide were also directly related to colon cancer risk, although this did not reach statistical significance (Table III). Risk estimates according to levels of leptin and C-peptide remained virtually unchanged after mutual adjustment. Concentrations of leptin and C-peptide were unrelated to risk of rectal cancer (Tables II and III).
Table II. Odds Ratios and 95% Confidence Interval of Cancer of the Colon and of the Rectum for Quartiles of Leptin
|Colon|| || || || || |
| All|| || || || || |
| Quartile cut-pointsb||< 1.7||1.7 to < 2.6||2.6 to <4.1||≥ 4.1|| |
| Number of cases/controls||49/67||61/58||54/63||71/47|| |
| OR, crudec||1.00||1.65 (0.93–2.92)||1.25 (0.68–2.30)||2.72 (1.44–5.12)||0.008|
| OR, adjusted C-peptide||1.00||1.66 (0.94–2.93)||1.26 (0.69–2.32)||2.77 (1.47–5.23)||0.007|
| Right colon|| || || || || |
| Number of cases/controls||24/25||28/21||21/29||26/24|| |
| OR, crude||1.00||1.33 (0.61–2.93)||0.66 (0.26–1.65)||1.20 (0.44–3.32)||0.81|
| OR, adjusted C-peptide||1.00||1.33 (0.60–2.92)||0.65 (0.26–1.64)||1.18 (0.42–3.29)||0.78|
| Left colon|| || || || || |
| Number of cases/controls||25/38||28/35||31/32||42/21|| |
| OR, crude||1.00||1.43 (0.63–3.22)||1.83 (0.78–4.29)||3.73 (1.59–8.71)||0.002|
| OR, adjusted C-peptide||1.00||1.43 (0.63–3.25)||1.80 (0.77–4.24)||3.78 (1.61–8.86)||0.002|
|Rectum|| || || || || |
| Quartile cut-points||< 1.7||1.7 to < 2.4||2.4 to < 4.1||≥ 4.1|| |
| Number of cases/controls||36/35||38/34||34/37||35/37|| |
| OR, crude||1.00||1.08 (0.57–2.03)||0.89 (0.46–1.73)||0.91 (0.49–1.70)||0.68|
| OR, adjusted C-peptided||1.00||1.13 (0.60–2.15)||0.86 (0.44–1.68)||0.89 (0.48–1.66)||0.60|
Table III. Odds Ratios and 95% Confidence Interval of Cancer of the Colon and of the Rectum for Quartiles of C-Peptide
|Colon|| || || || || |
| All|| || || || || |
| Quartile cut-pointb||< 0.18||0.18 to < 0.31||0.31 to < 0.70||≥ 0.70|| |
| Number of cases/controls||52/65||63/54||60/58||60/58|| |
| OR, crude||1.00||1.68 (0.92–3.05)||1.53 (0.82–2.86)||1.81 (0.67–4.86)||0.19|
| OR, adjusted leptinc||1.00||1.69 (0.93–3.09)||1.53 (0.82–2.86)||1.74 (0.65–4.69)||0.21|
| Right colon|| || || || || |
| Number of cases/controls||24/25||25/25||24/25||26/24|| |
| OR, crude||1.00||1.04 (0.49–2.23)||1.12 (0.44–2.82)||3.35 (0.29–38.6)||0.65|
| OR, adjusted leptin||1.00||1.04 (0.49–2.23)||1.12 (0.44–2.82)||3.35 (0.29–38.6)||0.65|
| Left colon|| || || || || |
| Number of cases/controls||31/32||29/34||34/29||32/31|| |
| OR, crude||1.00||0.91 (0.41–2.02)||1.32 (0.57–3.07)||1.31 (0.41–4.17)||0.47|
| OR, adjusted leptin||1.00||0.91 (0.40–2.03)||1.27 (0.54–2.98)||1.23 (0.38–3.97)||0.55|
|Rectum|| || || || || |
| Quartile cut-points||< 0.18||0.18 to < 0.36||0.36 to < 1.06||≥ 1.06|| |
| Number of cases/controls||37/34||38/34||33/38||35/37|| |
| OR, crude||1.00||0.89 (0.41–1.95)||0.61 (0.24–1.55)||0.44 (0.10–1.99)||0.21|
| OR, adjusted leptin||1.00||0.90 (0.41–1.98)||0.63 (0.24–1.68)||0.46 (0.10–2.17)||0.27|
In this prospective study, we found a significant increase in risk of colon cancer with increasing serum levels of leptin and a nonsignificant increase in risk for C-peptide.
We found a high reproducibility between samples collected around 1 year apart for the analysis of leptin but lower for C-peptide, in accordance with a previous study that also reported high reproducibility (r = 0.74) for leptin but not for insulin (r = 0.33) in samples taken up to 4 years apart.19 Some degradation may have occurred during the very extended storage time at −25°C in the Janus biorepository, a temperature too high for optimal preservation of peptides. To account for the effect of such a possible time-dependent degradation, we closely matched the controls to the cases for date of blood collection. We found that the risk associated with high levels of serum leptin was elevated in case-control sets with a storage time less than median (27 years), but not in sets with a longer storage time. The latter suggests the view that either some degradation might have occurred in the older samples that decreased our chances of finding an association to risk, or that the association between exposure and risk is stronger if exposure is measured closer to the time of diagnosis.
Our results only weakly support the hypothesis that long-term exposure to high levels of insulin is associated with increased risk of colon cancer. We used C-peptide as a marker of insulin exposure because serum C-peptide has a longer half-life than serum insulin, although both are highly dependent on fasting status.5, 7, 8, 20 Our ability to determine the effects of high chronic insulin exposure on cancer risk was limited by the nondefined fasting status of the study participants, the poor reproducibility of C-peptide measurements and a possible degradation of the C-peptide. Our results thus provide only weak support for the hypothesis that long-term exposure to high levels of insulin is associated with increased risk of colon cancer.
Serum concentrations of leptin are less affected by variations in fasting or nonfasting status than pancreatic insulin secretion and serum levels of insulin and C-peptide, but leptin does display a distinct circadian rhythm, with peak levels at night that are approximately 50% higher than nadir levels at noon.21 Serum leptin is strongly correlated to obesity, insulin resistance and serum insulin. One limitation of our study is that no obesity-related anthropometric measurement was available. However, the risk estimates for leptin were not affected by adjustment for BMI and plasma insulin in a smaller previous study relating plasma leptin to colon cancer risk, in which we found approximately the same increase in risk as in the present study.22 Furthermore, in the present study, relative risk estimates were unaffected by adjustment for C-peptide. Our finding that leptin concentrations were more strongly associated with cancers of the left colon than those of the right concurs with observations on the association of BMI to cancer risk by subsite.1, 3 Also, our study gave no support for a role for leptin or insulin in rectal cancer, in accordance with the view that cancers of the rectum are less strongly related to obesity than cancers of the colon.1
Our finding of an approximately 3-fold increase in risk of colon cancer for the highest levels of serum leptin may be a reflection of a direct role for leptin in colon tumorigenesis. The hypothesis that leptin is mechanistically related to the development of colon cancer is supported by several recent studies showing that leptin is a growth factor in several types of cells.13, 14, 15, 16 Importantly, leptin receptors are present in human colon cancer cell lines and also in human colon tumors, polyps and adjacent mucosa.14, 15In vitro, leptin stimulated proliferation and invasiveness of human colon cancer cells.14, 15, 16 In mice, administration of leptin was shown to increase colonic cell proliferation, and a high fat diet, increasing serum leptin, had an identical effect.14, 16 Leptin has also been demonstrated to stimulate angiogenesis,13 which is essential for tumor growth, invasion and metastasis.23
While a truly causal role of leptin in the development of colon tumors is plausible, we cannot rule out that leptin is simply an innocent bystander, being merely a correlate of other obesity-induced adverse alterations in metabolism that may be the true cause of tumor development. Despite the fact that we, in accordance with previous studies,10 observed about 3 times higher serum levels of leptin in women compared to men, we found no association between leptin levels and risk among women in a previous smaller study.22 Possibly, the stronger association seen among men despite their lower absolute levels of leptin is explained by the stronger association between leptin levels and abdominal obesity in men compared to women,10 suggesting that leptin is merely an innocent bystander.
In conclusion, our results suggest that leptin may be an important link between obesity and colon cancer risk. Whether leptin is directly involved in colon tumorigenesis, or whether leptin is merely a sensitive and robust marker of other obesity-induced hormonal aberrations, remains to be elucidated.
The Janus Serum Bank owned by the Norwegian Cancer Society provided the serum samples. Randi Elin Gislefoss is data coordinator of the Janus Serum Bank and Karin Hjertkvist, Department of Clinical Chemistry, Umeå University Hospital, performed the immunoassays.