Body Size and Composition and the Risk of Lymphohematopoietic Malignancies

Abstract

Whether certain aspects of body size and composition are associated with risk of lymphohematopoietic malignancies is uncertain ( 1 – 13 ) . Only one study has reported results for measures other than height, weight, or body mass index (BMI; kg/m ² ) ( 5 , 6 ) , and apart from two studies ( 3 , 4 ) , measurement of obesity was based on self-reported measures ( 1 , 2 , 5 – 7 , 10 , 13 ) or on hospital discharge diagnoses of obesity ( 8 , 9 , 12 ) . We examined whether direct measurements of body size were associated with the incidence of lymphohematopoietic malignancies using a prospective cohort study of adults.

The Melbourne Collaborative Cohort Study recruited 41 528 people between 1990 and 1994 ( 14 – 16 ) to investigate the role of nutritional and other lifestyle factors in cancer. The study was approved by The Cancer Council Victoria Human Research Ethics Committee, and written informed consent was obtained from all subjects. Subjects diagnosed with lymphohematopoietic malignancies before baseline (n = 105), with incomplete baseline measurements (n = 214), or who had either died, left Victoria, or had a diagnosis of lymphohematopoietic malignancies in the first 2 years of follow-up (n = 300) were excluded, leaving 40 909 people available for analysis.

Height, weight, and waist and hips circumferences were measured using standard procedures ( 17 ) , and BMI and waist-to-hips ratios were computed. We used bioelectrical impedance analysis using a BIA-101A RJL system analyzer (RJL systems, Detroit, MI) to estimate fat-free mass ( 18 ) . Fat mass (weight − fat-free mass) and percent fat (fat mass divided by weight) were subsequently calculated. Information on country of birth, alcohol, smoking, physical activity, and highest level of education was obtained by structured interview. Vital status and place of residence were obtained from electoral rolls, electronic phone books, and death records until 31 December 2003. At this time, 721 people had left Victoria (1.7%) and 2681 (6.5%) had died.

Case patients were identified from notifications of diagnoses of malignant neoplasms of lymphatic and hematopoietic tissue (International Classification of Diseases 10th revision morphology codes 959–998) to the population-based Victorian Cancer Registry. The main outcomes were myeloid leukemia and lymphoproliferative malignancies (and subgroups such as non-Hodgkin lymphoma [NHL], multiple myeloma, and lymphocytic leukemia).

Cox proportional hazards regression models, with age as the time axis ( 19 ) , were used to estimate hazard ratios (HRs) and 95% confidence intervals (CIs). Tests based on Schoenfeld residuals and graphical methods using Kaplan–Meier curves ( 20 ) showed no evidence that proportional hazard assumptions were violated for any of the anthropometric measures. Because the hazard ratios from analyses excluding the first 2 years of follow-up were higher for myeloid leukemia than those without this exclusion, results presented are based on follow-up that began 2 years after baseline and ended at diagnosis of lymphohematopoietic malignancies or cancer of unknown primary site, death, the date last known to be in Victoria, or 31 December 2003, whichever came first.

Analyses were adjusted for country of birth (Australia and United Kingdom or Greece and Italy), sex, education, and smoking status. Adjustments for physical activity and alcohol consumption did not appreciably change the hazard ratios, so these variables were not included in final analyses. In a sensitivity analysis, all people diagnosed with any cancer (apart from the cancer of interest) were censored at diagnosis. This did not materially change the hazard ratios (data not shown).

Statistical analyses were performed using Stata/SE 8.2 (Stata Corporation, College Station, TX). All P values are two-sided, and P <.05 was considered as statistically significant.

The participants' demographic characteristics are shown in Table 1 . Myeloid leukemia was positively associated with several aspects of body size ( Table 2 ). Incidence was approximately five times higher for overweight and obese participants than for those with BMI less than 25 kg/m ² (HR = 5.3, 95% CI = 1.9 to 15.2 and HR = 5.0, 95% CI = 1.6 to 15.5, respectively). The hazard ratio for the highest tertile was 2.9 (95% CI = 1.4 to 6.1) for fat-free mass and 2.6 (95% CI = 1.2 to 5.6) for fat mass. Waist circumference was also associated with increased risk (per 10-cm increase HR = 1.35, 95% CI = 1.06 to 1.72; P = .02), but the relationship was not monotonic. Although hazard ratios were higher for chronic myeloid leukemia (CML) (data not shown), numbers were too small to perform formal statistical tests of homogeneity between CML and acute myeloid leukemia (AML). All hazards ratios for lymphoproliferative malignancies (and subgroups) were close to unity ( Table 3 and Supplementary Tables 1 to 4 available at http://jncicancerspectrum.oxfordjourals.org/jnci/content/vol97/issue15 ).

We observed that increased risk of myeloid leukemia was associated with increased body girth but not stature. The incidence of myeloid leukemia for overweight and obese persons was five times greater than for those with a BMI less than 25 kg/m ² . People with an elevated non–fat component of weight or central adiposity were also at increased risk of myeloid leukemia. Conversely, the lymphoproliferative malignancies (including subgroups) showed little relationship with body size.

The study had several strengths and some limitations. Because few participants had left Victoria during the follow-up period, follow-up was 98.3% complete. Another strength of the study was that direct measures of body size were taken at baseline. Issues concerning the measurement of fat-free mass and fat mass have been addressed elsewhere ( 16 , 21 ) . In this study, measurement errors for fat-free mass and fat mass were generally small. To prevent a potential bias of altered body mass and composition that was due to undiagnosed disease, we excluded measurements made during the first 2 years of follow-up. However, we could not completely exclude patients with splenomegaly, which is commonly associated with myeloid leukemia, who could have consequently had a larger waist size prior to diagnosis. In addition, established etiologic exposures for lymphohematopoietic malignancies, such as radiation and benzene, were not recorded; therefore, we could not assess their possible impact. The major limitation of the study was the small number of case patients, leading to low statistical power for certain subgroup analyses.

Whether obesity is a risk factor for NHL remains uncertain. Some studies, including ours, have not shown any notable association ( 2 , 5 , 8 , 13 ) , whereas others have described a positive association ( 1 , 9 , 10 ) . There are reports ( 3 , 7 , 10 , 12 ) that obese persons are at an increased risk of multiple myeloma, although the associations have been generally weak, and we were unable to confirm this association.

Little has been reported on the association between leukemia and body size. Our finding that overweight and obese people are at higher risk than people of healthy weight of myeloid leukemia but not chronic lymphocytic leukemia (CLL) is concordant with the Iowa Women's Health Study ( 6 ) , which showed an association between BMI and AML, but not CLL. In our study, the association with body size appeared to be stronger for CML but, due to small numbers, this supposition could not be verified. Furthermore, male U.S. veterans with a history of hospitalization for obesity have been found to be at increased risk of multiple myeloma, CLL, and AML ( 12 ) .

Central adiposity increases the risk of several common epithelial cancers ( 21 ) , and it has been hypothesized to be related to the consequent chronic hyperinsulinemia and an associated increase in circulating insulin like growth factor 1 (IGF-1) that stimulates cell proliferation and inhibits apoptosis ( 22 ) . Growth factors, including IGF-1, have also been shown to be mitogenic in AML cell lines ( 23 ) . Central obesity also increases circulating levels of leptin ( 24 ) . Myeloid precursor cells in the bone marrow have both IGF and leptin receptors, and leptin is known to play a role in the modulation of the innate immune response, inflammation, and hematopoiesis ( 25 ) . Obesity may also influence immune function ( 26 , 27 ) , although little research comparing immune responses of lean and obese subjects has been reported ( 28 ) . Nutritional alterations, such as fasting and acute nutritional deprivation, occur frequently in obese persons and can both increase and decrease immunocompetence ( 29 ) .

In conclusion, we found that overall adiposity (including central) and nonadipose mass (or fat-free mass) were both associated with myeloid leukemia. However, they were not associated with any other lymphohematopoietic malignancies.

References