SEARCH

SEARCH BY CITATION

Keywords:

  • body mass index;
  • case–control studies;
  • prospective cohort studies;
  • meta-analysis;
  • non-Hodgkin's lymphoma;
  • obesity;
  • review

Abstract

  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. References

Obesity is associated with altered immune and inflammatory responses and it may therefore influence the risk of non-Hodgkin's lymphoma. However, epidemiologic findings on obesity in relation to non-Hodgkin's lymphoma have been inconsistent. We conducted a meta-analysis to summarize the epidemiologic evidence on the association between excess body weight and risk of non-Hodgkin's lymphoma. Relevant studies were identified by searching MEDLINE (1966 to February 2007) and the reference lists of retrieved publications. We included cohort and case–control studies that reported relative risk (RR) estimates with 95% confidence intervals (CIs) for the association of body mass index (BMI) with non-Hodgkin's lymphoma incidence or mortality. A random-effects model was used to combine results from individual studies. Sixteen studies (10 cohorts and 6 case–control studies), with 21,720 cases, met the inclusion criteria. Compared to individuals of normal weight (BMI < 25.0 kg/m2), the summary RRs of non-Hodgkin's lymphoma were 1.07 (95% CI, 1.01–1.14) for overweight individuals (BMI between 25 and 30 kg/m2) and 1.20 (95% CI, 1.07–1.34) for those who were obese (BMI ≥≥≥≥ 30.0 kg/m2). Meta-analysis stratified by histologic subtypes showed that obesity was associated with a statistically significant increased risk of diffuse large B-cell lymphoma (RR, 1.40; 95% CI, 1.18–1.66; n = 6 studies) but not of follicular lymphoma (RR, 1.10; 95% CI, 0.82–1.47; n = 6 studies) or small lymphocytic lymphoma/chronic lymphocytic leukemia (RR, 0.95; 95% CI, 0.76–1.20; n = 3 studies). These findings indicate that excess body weight is associated with an increased risk of non-Hodgkin's lymphoma, especially of diffuse large B-cell lymphoma. © 2007 Wiley-Liss, Inc.

Non-Hodgkin's lymphoma is a heterogeneous group of malignant diseases originating from lymphocytes.1 Apart from a small fraction of the cases caused by severe immunosuppression, autoimmunity and HIV infection, the causes of the majority of non-Hodgkin's lymphoma are largely unknown.1 Obesity is related to altered immune function and a chronic inflammatory response,2 and several studies have linked certain autoimmune and chronic inflammatory conditions to increased risk of non-Hodgkin's lymphoma.3 Obesity may also cause changes in the metabolism of endogenoushormones, which could distort the normal balance between cell proliferation, differentiation and apoptosis.4 Thus, obesity may be a risk factor for non-Hodgkin's lymphoma. However, epidemiologic findings on the relation between excess body weight and risk of non-Hodgkin's lymphoma have been inconsistent.

The aim of our study was to conduct a systematic review and meta-analysis of cohort and case–control studies to examine the association between overweight and obesity, as measured by body mass index (BMI), and risk of non-Hodgkin's lymphoma. We also evaluated whether the relation between obesity and non-Hodgkin's lymphoma varied by histologic subtype.

Methods

  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. References

Study selection

We searched the MEDLINE databases for studies on BMI and non-Hodgkin's lymphoma that were published from 1966 to February 2007, using the search terms body mass index, BMI, or obesity combined with lymphoma. The search was restricted to studies on human participants. No language restrictions were imposed. We also performed a manual search of references cited by relevant articles.

Studies were eligible for inclusion if they met the following criteria: (i) cohort or case–control study in which non-Hodgkin's lymphoma incidence or mortality was an outcome; (ii) the exposure of interest was BMI (body weight in kilograms divided by the square of height in meters); and (iii) relative risks (RRs) or odds ratios with their 95% confidence intervals (CIs) (or data to calculate them) were reported. Cohorts of hospitalized patients with a discharge diagnosis of obesity were not included. When there were multiple published reports from the same study population, we included the one with the largest sample size.

We identified 13 cohort studies4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16 and 9 case–control studies17, 18, 19, 20, 21, 22, 23, 24, 25 with data that were potentially eligible for inclusion in the meta-analysis. Two cohort studies were excluded because the outcome (lymphoma) included non-Hodgkin's lymphoma and Hodgkin's disease combined8 or no results for non-Hodgkin's lymphoma were reported.4 Three studies13, 17, 18 were excluded because their results were reported in subsequent articles with larger sample size.14, 19, 23 For 1 Canadian study with 2 overlapping reports,20, 21 we included the report with the largest sample size20 in the primary analyses and the other one21 in subgroup analysis by non-Hodgkin's lymphoma subtypes (results by histologic subtype were not provided in the report with the larger sample size). A total of 16 independent studies5, 6, 7, 9, 10, 11, 12, 14, 15, 16, 19, 20, 22, 23, 24, 25 were included in the present meta-analysis.

Data extraction

For each study, the following data were extracted: first author's last name; publication year; country in which the study was performed; study design; source of controls (in case–control studies); sample size; participant characteristics (sex and age); method of assessment of body weight and height (self-reported or measured); type of outcome (incidence or mortality); non-Hodgkin's lymphoma classification; variables adjusted for in the statistical analysis; and the RRs with their 95% CIs. RRs in each subgroup, according to sex, non-Hodgkin's lymphoma subtype and category of BMI, were extracted. From each study, we extracted the RRs that were controlled for the greatest number of potential confounders.

Statistical analysis

RR was used as the measure of the relation between BMI and non-Hodgkin's lymphoma. Because the absolute risk of non-Hodgkin's lymphoma is low, the odds ratios in case–control studies approximate the RR26; we report all results as RRs for simplicity. The RRs and corresponding standard errors (derived from the CIs) from individual studies were logarithmically transformed to stabilize variance and normalize the distributions. To examine associations between overweight (BMI between 25 and 30 kg/m2) and obesity (BMI ≥ 30 kg/m2 in all but 2 studies9, 16; Table I) and non-Hodgkin's lymphoma, we pooled the log RRs from each study for the category representing overweight or obesity versus the reference category (BMI < 25 kg/m2). For 4 studies7, 15, 22, 25 that reported RRs for more than 1 category of BMI that fell into the category for obesity in our meta-analysis, we pooled these RRs with inverse variance weight and used the pooled estimate in the meta-analysis. Study-specific RRs were pooled using the DerSimonian and Laird random-effects model.27 Thus, each summary RR was a weighted average of the study-specific RRs, where the weight for each study is the inverse of the sum of the within-study variance for that study, and the between-study variance.

Table I. Characteristics of Cohort and Case–Control Studies of Body Mass Index and Non-Hodgkin's Lymphoma
StudyCountrySex and ageCases/controls or cohort sizeNon-Hodgkin's lymphoma definitionBMI (kg/m2) categoriesAdjustments
BMI referenceBMI overweightBMI obesity
  • BMI, body mass index (the weight in kilograms divided by the square of height in metres); ICD, International Classification of Diseases; WHO, World Health Organization.

  • 1

    Based on fatal cases of non-Hodgkin's lymphoma.

  • 2

    Relative risks for BMI categories above 30 kg/m2 were pooled with inverse variance weight and the pooled estimate was used in the meta-analysis of obesity.

  • 3

    BMI categories for men and women, respectively.

  • 4

    Odds ratios for women were further adjusted for age at menarche, number of live births, age at first pregnancy and menopausal status.

Cohort studies
 Zhang et al.5United StatesWomen aged 34–60 years199/88,410ICD-8 code 202<21.025.0–29.9≥30.0Age, smoking, area of residence, height, energy
 Cerhan et al.6United StatesWomen aged 55–69 years261/37,931Not specified<23.526.2–29.7≥29.8Age
 Calle et al.7United StatesMen and women aged 50–74 years2,384/900,0531ICD-9 codes 202.0–202.918.5–24.925.0–29.9≥30.02Age, race, marital status, education, smoking, physical activity, aspirin use, estrogen-replacement therapy (women), alcohol, dietary factors
 Oh et al.9KoreaMen aged ≥20 years190/781,283Not specified18.5–22.925.0–26.9≥27.0Age, area of residence, smoking, exercise, alcohol
 MacInnis et al.10AustraliaMen and women aged 27–75 years170/40,909ICD-10 morphology codes 959, 967–971<25.025.0–29.9≥30.0Age, country of birth, education
 Rapp et al.11AustriaMen and women aged 19–94 years148/145,931ICD-9 codes 200, 20218.5–24.925.0–29.9≥30.0Age, occupational group, smoking
 Lukanova et al.12SwedenMen and women aged 29–61 years89/68,786Not specified18.5–24.925.0–29.9≥30.0Age, calendar year, smoking
 Samanic et al.14SwedenMen aged 18–67 years1,077/362,552ICD-7 codes 200, 20218.5–24.925.0–29.9≥30.0Age, smoking
 Engeland et al.15NorwayMen and women aged 20–74 years8,512/1,999,978Not specified18.5–24.925.0–29.9≥30.02Age, birth cohort
 Chiu et al.16United StatesMen and women aged 15–90 years129/35,4201ICD-8 codes 200, 202; ICD-9 codes 200, 202.0–202.2, 202.8–202.9<24.1; <21.0326.3–28.6; 23.3–26.23≥28.6; ≥26.23Age
Case–control studies
 Skibola et al.19United StatesMen and women aged 21–74 years1,301/2,400Not specified20.0–25.025.0–29.9≥30.0Age, sex
 Pan et al.20CanadaMen and women aged 20–76 years1,668/5,039ICD-9 codes 200, 202<25.025.0–29.9≥30.0Age, sex, province of residence, education, smoking, physical activity, alcohol, dietary factors4
 Chang et al.22Sweden and DenmarkMen and women aged 18–74 years3,045/3,158WHO Classification18.5–24.925.0–29.9≥30.02Age, sex, country
 Bosetti et al.23ItalyMen and women aged 14–85 years633/1,769Not specified18.5–24.925.0–29.9≥30.0Age, sex, study center, area of residence, education, smoking
 Willett et al.24EnglandMen and women aged 18–64 years699/914WHO Classification18.5–24.925.0–29.9≥30.0Age, sex, area of residence
 Cerhan et al.25United StatesMen and women aged 21–74 years1,219/977Not specified15.0–24.925.0–29.9≥30.02Age, sex, race, study center

For the dose–response meta-analysis of BMI, we used the method proposed by Greenland and Longnecker28, 29 to compute study-specific slopes (linear trends) from the correlated log RRs across categories of BMI. We used an increase of 5 kg/m2 in BMI, which corresponds to 15.7 kg for a man of an average height (1.77 m) and 13.5 kg for a woman (1.64 m). Summary RR estimates were obtained from a random-effects model27 applied to the study-specific dose–response slopes. We checked for nonlinearity of the dose–response relationship between BMI and non-Hodgkin's lymphoma by estimating polynomial models. This was done by using the “pool-first” method described by Greenland and Longnecker.29 We found that the best-fitting model was a linear model.

Statistical heterogeneity among studies was evaluated with the Q and I2 statistics.30 For the Q statistic, statistical significance was set at p < 0.1. We conducted subgroup analyses by study design, sex, geographic region (North America vs. Europe), assessment of body weight and height (self-reported vs. measured), outcome (incidence vs. mortality in cohort studies) and histologic subtype. We used funnel plots (i.e., plots of study results against precision) to assess publication bias, and tested symmetry of the funnel plot as suggested by Egger et al.31 All statistical analyses were performed with Stata software, version 9.0 (StataCorp, College Station, TX).

Results

  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. References

The 10 cohort studies5, 6, 7, 9, 10, 11, 12, 14, 15, 16 (involving 13,159 cases and 4,461,253 study participants) and 6 case–control studies19, 20, 22, 23, 24, 25 (8,561 cases and 14,254 controls) that were included in the meta-analysis were published between 1999 and 2006 (Table I). Of these, 6 were from the United States, 7 from Europe and 1 each from Canada, Korea and Australia. Body weight and height were measured by researchers in 7 studies9, 10, 11, 12, 14, 15, 16 and self-reported in 9 studies.5, 6, 7, 19, 20, 22, 23, 24, 25 The outcome was mortality from non-Hodgkin's lymphoma in 2 studies7, 16 and incidence of non-Hodgkin's lymphoma in the remaining studies. Among the case–control studies, 5 used population-based controls19, 20, 22, 24, 25 and 1 used hospital-based controls.23

Meta-analysis of all studies found that compared to individuals with normal weight (BMI < 25 kg/m2), those who were overweight had a 7% greater risk of non-Hodgkin's lymphoma (Fig. 1) and those who were obese had a 20% greater risk (Fig. 2). There was statistically significant heterogeneity among the results of individual studies for overweight (p < 0.001, I2 = 57.3%) and obesity (p < 0.001, I2 = 72.1%). However, the summary estimates did not differ statistically significantly between cohort and case–control studies (p = 0.68 for overweight and p = 0.83 for obesity). We found no evidence of publication bias on the funnel plot or by Egger's test (p = 0.40 for overweight and p = 0.50 for obesity).

thumbnail image

Figure 1. Relative risks of non-Hodgkin's lymphoma associated with overweight. M, men; W, women. Relative risks are presented separately for men and women wherever this data were available. Overweight was defined as BMI between 25 and 30 kg/m2.

Download figure to PowerPoint

thumbnail image

Figure 2. Relative risks of non-Hodgkin's lymphoma associated with obesity. M, men; W, women. Relative risks are presented separately for men and women wherever this data were available. Obesity was defined as BMI ≥ 30 kg/m2 in all but 2 studies9, 16 (Table I).

Download figure to PowerPoint

Four studies7, 15, 22, 25 reported results for more than 1 category of BMI that fell into the category for obesity in our meta-analysis. When we combined the results from these studies, the RRs of non-Hodgkin's lymphoma were 1.14 (95% CI, 1.06–1.24) for BMI between 30 and 35 kg/m2 and 1.23 (95% CI, 1.08–1.39) for BMI ≥ 35 kg/m2, indicating that the risk of non-Hodgkin's lymphoma increases with increasing grade of obesity.

On a continuous scale, a 5 kg/m2 increase in BMI was associated with a 10% increased risk of non-Hodgkin's lymphoma (Table II). There was statistically significant heterogeneity among all studies and within strata defined by study design, sex and geographic region (Table II). The summary RR was higher for studies from North America than for European studies (Table II). Among cohort studies, the association between BMI (per 5 kg/m2 increase) and non-Hodgkin's lymphoma was nonsignificantly stronger (p = 0.14) for mortality (RR, 1.17; 95% CI, 1.02–1.33; n = 2 studies) than incidence (RR, 1.05; 95% CI, 1.00–1.10; n = 8 studies).

Table II. Summary Relative Risks of Non-Hodgkin's Lymphoma per 5 kg/m2 Increase in Body Mass Index
 Studies (n)Cases (n)RR (95% CI)Tests for heterogeneity
p-valueI2 (%)1
  • CI, confidence interval; RR, relative risk.

  • 1

    I2 is interpreted as the proportion of total variation contributed by between-study variation.

  • 2

    p-value for difference in the strength of the association between body mass index and non-Hodgkin's lymphoma between the strata.

All studies1621,7201.10 (1.05–1.16)<0.00172.9
Study design
 Cohort studies1013,1591.10 (1.03–1.17)<0.00169.4
 Case–control studies68,5611.12 (1.02–1.23)<0.00179.5
 p-value difference2  0.72  
Sex
 Men98,8011.17 (1.08–1.27)0.00464.2
 Women97,1571.12 (1.02–1.22)<0.00175.4
 p-value difference2  0.45  
Geographic region
 North America77,1611.15 (1.07–1.23)0.00167.8
 Europe714,1991.05 (0.99–1.12)0.0958.9
 p-value difference2  0.06  
Assessment of weight and height
 Self-reported911,4051.12 (1.05–1.20)<0.00174.4
 Measured710,3151.07 (0.99–1.15)0.0158.4
 p-value difference2  0.34  

One cohort study6 and 5 case–control studies19, 21, 22, 24, 25 reported results on diffuse large B-cell lymphoma and follicular lymphoma, and 3 studies6, 21, 22 also presented results on small lymphocytic lymphoma and B-cell chronic lymphocytic leukemia. Meta-analysis of these studies showed that obesity was associated with a statistically significant 40% increased risk of diffuse large B-cell lymphoma, without heterogeneity among studies, but was not associated with risk of follicular lymphoma or small lymphocytic lymphoma/B-cell chronic lymphocytic leukemia (Fig. 3). The difference in summary RR between subtypes was statistically significant (p = 0.03). Only 1 case–control study22 reported results on other subtypes.

thumbnail image

Figure 3. Relative risks of non-Hodgkin's lymphoma associated with obesity, by histologic subtypes. SLL, small lymphocytic lymphoma; CLL, B-cell chronic lymphocytic leukemia.

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. References

This meta-analysis of published studies supports a statistically significant positive association between BMI and risk of non-Hodgkin's lymphoma. Overweight and obese individuals had a 7 and 20% greater risk of non-Hodgkin's lymphoma, respectively, than those of normal weight. Summary results based on a limited number of studies that reported results on non-Hodgkin's lymphoma subtypes showed that obesity was associated with an increased risk of diffuse large B-cell lymphoma but not of follicular lymphoma or small lymphocytic lymphoma/B-cell chronic lymphocytic leukemia.

As a meta-analysis of observational studies, our study has limitations that are based largely on the quality and availability of published studies. First, the possibility that the observed positive association between BMI and non-Hodgkin's lymphoma was due to bias (e.g., recall and selection bias) should be considered. Prospective cohort studies are less prone to bias than case–control studies because, in the prospective design, information on exposures is collected before the diagnosis of disease. However, in this meta-analysis, we did not detect significant differences between summary results from case–control and cohort studies. Second, individual studies may have failed to control for potential confounders. Failure to adjust for confounders may bias the results in either direction, toward exaggeration or underestimation of the true relationship. Thus, we cannot rule out uncontrolled confounding as a potential explanation for the observed association between excess body weight and risk of non-Hodgkin's lymphoma. A third limitation is that all cohort studies assessed body weight only once (at baseline), and changes in body weight during follow-up may have weakened the association between BMI and risk of non-Hodgkin's lymphoma. A fourth limitation of our analyses is that there was heterogeneity among results for total non-Hodgkin's lymphoma from individual studies in the overall analysis and within subgroups defined by study design, sex and geographic region. However, results for the specific subtype diffuse large B-cell lymphoma were homogenous. Hence, heterogeneity between studies may in part be related to different proportions of non-Hodgkin's lymphoma subtypes in different populations. The proportion of diffuse large B-cell lymphoma in the studies included this meta-analysis varied from about 2622 to 45%.24 Methodological differences between studies, including different assessment and classification of outcome, may also have contributed to heterogeneity between study results. Finally, in a meta-analysis of published studies, it is possible that an observed association is the result of publication bias, because small studies with null results tend not to be published. In this meta-analysis, we found no evidence for such bias.

The observed increase in non-Hodgkin's lymphoma risk associated with obesity may be related to changes in circulating levels of adipocytokines, including adiponectin, resistin and leptin. Besides their role in insulin resistance, these adipocyte-derived hormones are involved in immunity and inflammation.2 Several autoimmune and chronic inflammatory conditions have been associated with an increased risk of non-Hodgkin's lymphoma.32 Obesity is related to decreased adiponectin and resistin levels and increased leptin levels.2 Studies in vitro and in animal models indicate that adiponectin has anti-inflammatory properties and reduces cell proliferation, whereas leptin has proinflammatory properties and promotes the growth of certain cancer cells.2In vitro studies have shown that leptin stimulates the proliferation of normal hematopoietic cells33 as well as circulating monocytes producing proinflammatory cytokines, such as tumor-necrosis factor and interleukin-6.34 Genetic polymorphisms in leptin and leptin receptors have been found to influence the risk of non-Hodgkin's lymphoma.19, 24

Obesity also gives rise to insulin resistance and compensatory hyperinsulinemia.3 Insulin may mediate tumourigenic effects directly through insulin receptors in (pre)neoplastic target cells, or indirectly through alterations in endogenous hormone metabolism.3 For instance, elevated insulin levels lead to an increase in bioavailable insulin-like growth factor-I (IGF-I). In vitro studies have shown that both insulin and IGF-I act as growth factors that promote cell proliferation and inhibit apoptosis.35, 36, 37 Type 2 diabetes, which is associated with insulin resistance and increased pancreatic insulin secretion for long periods both before and after disease onset, has been associated with an elevated risk of non-Hodgkin's lymphoma.16, 38, 39, 40

In summary, this meta-analysis supports the hypothesis that overweight and obesity may be associated with an elevated risk of non-Hodgkin's lymphoma, particularly of diffuse large B-cell lymphoma. More research is necessary to evaluate whether the association between excess body weight and non-Hodgkin's lymphoma varies by histologic subtype and to clarify the underlying mechanisms involved.

References

  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. References