Overweight and obesity and incidence of leukemia: A meta-analysis of cohort studies

Abstract

We conducted a meta-analysis to summarize the available evidence from cohort studies on the association between excess body weight and incidence of leukemia. Studies were identified by searching the MEDLINE and EMBASE databases (1966–July 2007) and by examining the references of retrieved articles. A random-effects model was used to combine the results from individual studies. We identified 9 cohort studies with data on body mass index (BMI) or obesity in relation to incidence of leukemia. Compared with nonoverweight individuals (BMI < 25 kg/m²), the summary relative risks (RRs) of leukemia were 1.14 [95% confidence interval (CI), 1.03–1.25] for overweight individuals (BMI 25–30 kg/m²) and 1.39 (95% CI, 1.25–1.54) for obese (BMI ≥ 30 kg/m²) individuals. On a continuous scale, a 5 kg/m² increase in BMI was associated with a 13% increased risk of leukemia (RR, 1.13; 95% CI, 1.07–1.19). In a meta-analysis of 4 studies reporting results on subtypes of leukemia, the summary RRs associated with obesity were 1.25 (95% CI, 1.11–1.41) for chronic lymphocytic leukemia, 1.65 (95% CI, 1.16–2.35) for acute lymphocytic leukemia, 1.52 (95% CI, 1.19–1.95) for acute myeloid leukemia and 1.26 (95% CI, 1.09–1.46) for chronic myeloid leukemia. This meta-analysis indicates that excess body weight is associated with an increased risk of developing leukemia. © 2007 Wiley-Liss, Inc.

Excess body weight has been associated with increased risk of a wide range of malignancies, including cancer of the endometrium, breast (in postmenopausal women), colon, kidney, esophagus, pancreas, liver and gallbladder.1-4 Emerging epidemiologic evidence also suggests that excess body weight may be a risk factor for hematologic cancers, including non-Hodgkin's lymphoma,5 multiple myeloma6 and possibly leukemia. The aim of this study was to summarize the available evidence from cohort studies on the relations of overweight and obesity, as measured by body mass index (BMI), with the incidence of leukemia.

Material and methods

Study selection

We searched the MEDLINE and EMBASE databases for relevant studies that were published from 1966 to July 2007, using the search terms body mass index, BMI, overweight or obesity combined with leukemia. We also reviewed the reference lists of retrieved articles to search for additional studies. No language restrictions were imposed.

Studies were included if: (i) they presented data from cohort studies; (ii) the exposure of interest was BMI (body weight in kilograms divided by the square of height in meters) or obesity; (iii) the outcome was incidence of leukemia; and (iv) reported relative risks (RRs) with 95% confidence intervals (CIs). When there were multiple published reports from the same study population, only the most recent publication was included in the meta-analysis. A total of 871 articles were identified through the literature search. Of these, 14 publications were potentially relevant for inclusion in the meta-analysis. Five studies were excluded because of case-control design,7 duplicate publications from the same study population8 or the outcome was mortality from leukemia.9-11 Thus, 9 cohort studies12-20 with data on BMI or obesity in relation to incidence of leukemia were included.

Data extraction

For each study, the following information was extracted: publication data, country in which the study was performed, sex, sample size, International Classification of Disease codes for leukemia, assessment of body weight and height (measured versus self-reported), covariates controlled for in the analysis and RRs with corresponding 95% CIs for each category of BMI. We extracted from each study the RRs that were adjusted for the greatest number of potential confounders.

Statistical analysis

To examine associations of overweight and obesity with the risk of leukemia, we combined the log-RRs from each study for the category representing overweight (BMI between 25 and 30 kg/m²) or obesity (BMI ≥ 30 kg/m² or a discharge diagnosis of obesity) versus the reference category (BMI < 25 kg/m² or nonobese). When results were reported separately for subtypes of leukemia,17, 19, 20 the subtype-specific RRs were pooled, and the pooled estimate was used in the meta-analysis of total leukemia. For the dose-response meta-analysis (per 5 kg/m² increment in BMI), we used the method proposed by Greenland and coworkers21, 22 to compute study-specific slopes from the correlated log-RRs across BMI categories. Study-specific RRs were pooled using the DerSimonian and Laird random-effects model,23 which incorporates both within-study and between-study variability. Details of the method used to compute study-specific slopes have been described in a published meta-analysis of BMI and pancreatic cancer.2

Statistical heterogeneity among studies (i.e., variation between studies in the association being observed) was assessed with the Q and I² statistics.24 For the Q statistic, p < 0.1 was considered statistically significant. Publication bias was evaluated with the Egger's regression test.25 All statistical analyses were performed using Stata, version 9.0 (StataCorp, College Station, TX).

Results

The 9 cohort studies12-20 that were included in the meta-analysis were published between 1994 and 2007 (Table I). Five were conducted in Europe, 2 in the United States, 1 in Australia and 1 in Korea. Weight and height were directly measured in 5 studies13, 17-20 and self-reported in 1 study15; 3 studies were based on a discharge diagnosis of obesity.12, 14, 16

Table I. Characteristics of Cohort Studies Included in the Meta-Analysis of BMI and Leukemia Incidence

Study	Country	Study participants	Cases	ICD codes for leukemia	BMI (kg/m^2) categories			Adjustments
					BMI reference	BMI overweight	BMI obesity
Møller et al., 199412	Denmark	43,965 men and women	51	Not specified	Nonobese	–	Obese11 Discharge diagnosis of obesity.	Age
Tulinius et al., 199713	Iceland	11,366 men	33	ICD-7 code 204	–22 Results were only presented per unit increase in BMI (kg/m^2).	–22 Results were only presented per unit increase in BMI (kg/m^2).	–22 Results were only presented per unit increase in BMI (kg/m^2).	Age
Wolk et al., 200114	Sweden	28,129 men and women	39	ICD-7 code 204	Nonobese	–	Obese11 Discharge diagnosis of obesity.	Age, calendar year
Ross et al., 200415	United States	37,627 women	194	Not specified	18.5–24.9	25.0–29.9	≥30.0	Age, physical activity
Samanic et al., 200416	United States	4,500,700 male veterans	9790	ICD-8/9 codes 204–208	Nonobese	–	Obese11 Discharge diagnosis of obesity.	Age, calendar year
MacInnis et al., 200517	Australia	40,909 men and women	106	ICD-10 morphology codes 9821, 9823, 9861, 9863, 9866–9868, 9874, 9891, 9895, 9945	<25.0	25.0–29.9	≥30.0	Age, country of birth, education
Oh et al., 200518	Korea	781,283 men	200	Not specified	18.5–22.9	25.0–29.933 Relative risks for BMI categories 25.0–26.9 kg/m^2 and 27.0–29.9 kg/m^2 were pooled, and the pooled estimate was used in the meta-analysis.	≥30.0	Age, area of residence, family history of cancer, smoking, exercise, alcohol
Engeland et al., 200719	Norway	1,999,978 men and women	6564	Not specified	18.5–24.9	25.0–29.9	≥30.044 Relative risks for BMI categories above 30 kg/m^2 were pooled, and the pooled estimate was used in the meta-analysis of obesity.	Age, birth cohort
Fernberg et al., 200720	Sweden	336,381 men	372	Not specified	18.6–24.9	25.0–29.9	≥30.0	Age, snuff use, smoking

• ICD = International Classification of Diseases.
• 1 Discharge diagnosis of obesity.
• 2 Results were only presented per unit increase in BMI (kg/m²).
• 3 Relative risks for BMI categories 25.0–26.9 kg/m² and 27.0–29.9 kg/m² were pooled, and the pooled estimate was used in the meta-analysis.
• 4 Relative risks for BMI categories above 30 kg/m² were pooled, and the pooled estimate was used in the meta-analysis of obesity.

Relative risk estimates of leukemia for overweight and obese individuals compared with those of normal weight for individual studies and all studies combined are shown in Figure 1. Compared with normal weight, the summary RRs of leukemia were 1.14 (95% CI, 1.03–1.25) for overweight and 1.39 (95% CI, 1.25–1.54) for obesity. There was evidence of heterogeneity among study-specific results for obesity but not overweight (Fig. 1). The positive relation between obesity and risk of leukemia remained after excluding the 3 cohort studies based on a discharge diagnosis of obesity12, 14, 16 (RR, 1.23; 95% CI, 1.12–1.34). Obesity was associated with a statistically significant increased risk of leukemia both among men (RR 1.46; 95% CI, 1.36–1.56; n = 6 studies) and women (RR 1.19; 95% CI, 1.09–1.31; n = 4 studies). There was no indication of publication bias (Egger's test: p = 0.61 for overweight; p = 0.47 for obesity). On a continuous scale, a 5 kg/m² increment in BMI was associated with a statistically significant 13% increased risk of leukemia (RR, 1.13; 95% CI, 1.07–1.19; n = 6 studies), with no heterogeneity among studies (p = 0.35).

Figure 1

Associations of overweight and obesity with incidence of leukemia.

Meta-analysis of 4 studies that provided results by leukemia subtypes15, 16, 19, 20 found that obesity was associated with a statistically significant increased risk of all subtypes, including chronic lymphocytic leukemia, acute lymphocytic leukemia, acute myeloid leukemia and chronic myeloid leukemia (Fig. 2). There was significant heterogeneity only among study-specific results for acute myeloid leukemia (Fig. 2).

Figure 2

Association of obesity with incidence of leukemia subtypes. Tests for heterogeneity: chronic lymphocytic leukemia, p = 0.21, I² = 31.2%; acute lymphocytic leukemia, p = 0.37, I² = 6.8%; acute myeloid leukemia, p < 0.001, I² = 79.3%; chronic myeloid leukemia, p = 0.51, I² = 0%.

Discussion

This meta-analysis of cohort studies provides evidence that excess body weight may be a risk factor for leukemia. Summary results indicate that overweight and obese individuals have a 14 and 39%, respectively, greater risk of leukemia compared to nonoverweight individuals. Obesity was directly associated with the risk of leukemia in both men and women and with all subtypes of leukemia.

The major strength of this meta-analysis is the inclusion of prospective cohort studies, which eliminated the possibility that the observed association between excess body weight and risk of leukemia was due to recall and selection biases. This meta-analysis also has some potential limitations. First, the possibility that the results were affected by confounding from other risk factors cannot be ruled out. Only 2 studies adjusted for smoking.18, 20 However, any confounding from smoking would probably tend to attenuate the positive association between body weight and risk of leukemia, and the increase in risk observed in this meta-analysis may have been a conservative estimate. Another limitation is that only 4 studies reported results by subtypes of leukemia, leading to relatively low statistical power in analysis relating obesity to leukemia subtypes. Finally, in a meta-analysis of published studies, publication bias could be a problem because small studies with null results tend not to be published. Although we found no evidence of such bias, the test for publication bias has limited power when the number of studies is small. The presence of possible publication bias could have led to an overestimation of the true relationship between excess body weight and risk of leukemia.

Obesity has also been associated with an increased risk of mortality from leukemia. In a large prospective cohort study of more than 900,000 US adults,9 men with a BMI of 35 kg/m² or more had a statistically significant 70% higher risk of death from leukemia; no association was observed among women. Results from another cohort study of 35,420 US adults showed a positive dose-response relationship between BMI and mortality from leukemia in both men and women.11

The biologic mechanism underlying the observed relations of overweight and obesity with increased risk of leukemia is unclear. A metabolic consequence of obesity is insulin resistance followed by an increased pancreatic insulin secretion.26 Insulin may promote tumorigenesis directly through insulin receptors in (pre)neoplastic target cells, or indirectly by increasing bioavailable insulin-like growth factor-I (IGF-I) levels.26 Almost all normal and neoplastic hematopoietic cells express the IGF receptors.27 IGF-I is involved in hematopoeisis and is mitogenic for cell lines of myeloid and lymphoid leukemias.27 The increased risk of leukemia in obese individuals may also be due to impaired immune function and chronic inflammation associated with obesity.28

In summary, findings from this meta-analysis of prospective studies provide evidence that excess body weight may increase the risk of developing leukemia. Whether the relation varies by leukemia subtypes warrants further investigation.