Height and Body Mass Index and Risk of Lymphohematopoietic Malignancies in Two Million Norwegian Men and Women

Reports on the association between obesity and lymphohematopoietic malignancies (LHMs) have been inconsistent. The present study aimed at exploring this association for specific disease lymphohematopoietic entities in a large Norwegian cohort. Height and weight were measured in two million Norwegian men and women aged 20– 74 years during 1963–2001. During follow-up, 24,500 cases of LHMs were observed. Relative risks of disease were estimated by Cox proportional hazards regression. The risk of LHMs overall increased moderately by increasing body mass index and height in both sexes. The relative risk of LHMs per five-unit increase in body mass index was 1.11 (95% confidence interval (CI): 1.08, 1.14) in men and 1.08 (95% CI: 1.05, 1.11) in women. For each 10-cm increase in height, the relative risk was 1.19 (95% CI: 1.16, 1.22) in men and 1.16 (95% CI: 1.12, 1.20) in women. Separate analyses for different lymphohematopoietic malignancies did not reveal any group’s being particularly strongly associated with body mass index. A modest increase in the risk of LHMs combined was observed with increasing height. The moderate associations between height and body mass index and LHMs found in the present study indicate that the observed increase in overweight/obesity plays only a minor role in explaining the increase in the incidence of LHMs.

Several types of cancer have been associated with obesity (1).The most consistent associations have been found for cancers of the colon, breast (in postmenopausal women), endometrium, kidney, esophagus, and gastric cardia (2).
The incidence rate of non-Hodgkin's lymphoma, comprising a major proportion of lymphohematopoietic malignancies (LHMs), has increased during the last three decades in Norway (3), as in many other parts of the world (4).Obesity has increased worldwide in the same period, and the association between obesity and risk of non-Hodgkin's lymphoma has been investigated in several studies (5)(6)(7)(8)(9)(10)(11)(12)(13)(14).Inclusions of other specific LHMs in these studies vary.Only MacInnis et al. (10) include a wide range of entities, and results have been inconsistent.
In an American cohort study of 40,000 women aged 55-69 years at baseline, no association between obesity and risk of non-Hodgkin's lymphoma was observed (6).A large casecontrol study (3,700 cases and 3,200 controls) did not find any association between body mass index and non-Hodgkin's lymphoma or Hodgkin's lymphoma (8).Other studies have shown a positive association between obesity and risk of LHMs, both case-control studies (7,9,11,13) and cohort studies (5,10,12,14).
The aim of the present study was to explore the associations between height and body mass index and the risk of lymphohematopoietic malignancies in a Norwegian cohort of more than two million men and women with a long followup time using a population-based cancer registry.The large study size allowed analysis of individual lymphohematopoietic diseases.

MATERIALS AND METHODS
The study population and the methods used in the present study have been described previously (15).

Study population
In a number of health surveys during 1963-2001, height and weight were measured in 2,001,719 persons (963,709 men and 1,038,010 women) aged 20-74 years, after excluding 215 measurements where height was below 120 cm or weight was below 20 kg.The height and weight were measured in a standardized way by a trained staff mainly from the same institution, the Norwegian Mass Screening Services (now included in the Norwegian Institute of Public Health).Height was measured without shoes, and weight was measured with the subjects wearing light clothing.Height and weight measurements using nonstandard methods (persons wearing shoes, disabled persons, and so on) were excluded.
Most of the health surveys conducted in Norway in the period 1963-1975 were included in a nationwide screening program aimed at detecting tuberculosis (16).This screening was compulsory for all inhabitants aged 15 years or above, and the attendance was about 85 percent.
In 1963-1964 and in 1972-2001, height and weight were measured in other health surveys (17,18), the main objective of which was to assess risk factors for coronary heart diseases.The attendance in the mid 1970s was 85-90 percent but decreased to about 75 percent in the mid-1990s (18).
Deaths, emigrations, and incident cases of LHMs in these cohorts were identified by linkage to the Death Registry at Statistics Norway (19) and to the Cancer Registry of Norway (3).The two registries are population based, covering the entire Norwegian population.A unique 11-digit identification number assigned to all individuals living in Norway after 1960 simplified the linkages.
Since 1953, cancer registration has been mandatory in Norway, and reporting of incident cancer cases to the Cancer Registry of Norway has been nearly complete (3).
Persons with a diagnosis of LHM prior to the height and weight measurements were excluded (1,108 persons).In the analyses, the persons were followed up from the date of measurement until the date of diagnosis, emigration, age 100 years, death, or until December 31, 2002, whichever occurred first.Altogether, 2,000,611 persons (963,071 men and 1,037,540 women) were eligible for the study.A very small number of these (84 men and 103 women) were lost to follow-up.
The cases were categorized as follows: 1) lymphoproliferative malignancies, that is, malignant lymphoma (non-Hodgkin's lymphoma including Waldenstro ¨m's macroglobulinemia, Hodgkin's lymphoma, and malignant lymphoma not otherwise specified), lymphatic leukemia (acute lymphatic leukemia and chronic lymphatic leukemia), plasma cell neoplasms (plasma cell myeloma and plasmocytoma), and lymphoproliferative malignancies not otherwise specified; 2) myeloproliferative malignancies, that is, acute myeloid leukemia, chronic myeloid leukemia, and other and unspecified myeloproliferative malignancies including chronic myelomonocytic leukemia, monocytic leukemia not otherwise specified, myeloid leukemia not otherwise specified, and myeloproliferative disease not otherwise specified; and 3) lymphohematopoietic malignancies not otherwise specified including leukemia not otherwise specified.Myelodysplastic syndrome was not included.
From the tuberculosis screening program during 1963-1975, only data on height and weight measurements were available in addition to time of measurement, age, and sex.Most health surveys carried out during 1972-2001 included information about smoking habits (self-administered questionnaire) as well.The study subjects were divided into never, former, and current smokers.Information on smoking habits was available for 646,422 persons (322,364 men and 324,058 women) after exclusions.The same exclusion criteria and follow-up procedure were used for this subsample as in the main analysis.
Tests for trend in the risk of disease by body mass index and height were performed by including body mass index and height, respectively, as continuous variables.
To explore whether body mass index had a different impact on the incidence rates of the diseases for different age groups, analyses stratified on both age at measurement and attained age were performed.These analyses did not add any information.Further, analyses were performed separately for different disease entities.
The proportionality assumption in the Cox model was assessed by inspecting log-minus-log plots and the results from stratified analyses.All analyses were performed using the SPSS statistical program package (22).The results were presented as relative risks of disease with 95 percent confidence intervals.
The hazard function of malignancy by body mass index in the Cox model was estimated by use of penalized spline functions in S-PLUS software (23), with 4 df.Body mass index was not categorized in these analyses, but cutpoints were defined in the estimation of splines by the limits between the categories used in the main analyses.The relative risk of point a compared with point b on the x-axis can be calculated by using figure 1 to find the corresponding y values, y a and y b , and then the relative risk by e y a ÿy b .

RESULTS
The 2,000,424 persons (962,987 men and 1,037,437 women) included in this study were on average followed for 23 years (maximum: 40 years), totaling 45.6 million person-years (table 1).The mean age at start of follow-up was 44 years.During follow-up, 24,526 cases of LHMs  were diagnosed.The mean age at diagnosis was 68 years.The patients were diagnosed on average 18 years after the height and weight measurements.The proportions of obese persons were 6 percent in men and 13 percent in women.We observed a modest increase in the risk of LHMs with both increasing body mass index and height (table 2), when adjusting for age at measurement and birth cohort.On average, each five-unit increase in body mass index increased the risk by 11 percent in men (95 percent confidence interval (CI): 8, 14) and 8 percent in women (95 percent CI: 5, 11).For each 10-cm increase in height, the risk increased by 19 percent (95 percent CI: 16, 22) in men and 16 percent (95 percent CI: 12, 20) in women.
Separate analyses revealed that the risk of both lymphoproliferative and myeloproliferative malignancies increased with increasing body mass index and height (table 2).Among individual lymphoproliferative malignancies, the most convincing association between body mass index and disease was seen in plasma cell neoplasms (table 3), while the association between height and disease was most convincing in non-Hodgkin's lymphoma, Hodgkin's lymphoma, and chronic lymphatic leukemia.Among myeloproliferative malignancies, an increasing risk of acute myeloid leukemia was seen with increasing height (table 4).There was a positive association between body mass index and chronic myeloid leukemia in men.
The associations between body mass index or height and the risk of LHMs were quite similar when analyzing persons aged 20-49 years and 50-74 years at measurement separately (data not shown).
The hazard functions of the risk of LHMs by body mass index and height were estimated using spline functions (figures 1 and 2).The risk increased steadily with body mass index and height for LHMs in both sexes.The downward trend in relative risk in men with a body mass index above 35 is connected with a large degree of uncertainty (only 46 cases of LHMs were observed in men with a body mass index of 35).
To exclude the possibility that, at the time of the body mass index measurement, weight was influenced by presence of an undiagnosed malignancy, we analyzed the data omitting the first 5 years of follow-up.The associations between body mass index or height and the overall risk of LHMs were marginally changed.On average, each five-unit increase in body mass index increased the risk of LHMs by 13 percent (95 percent CI: 10, 17) in men and 10 percent (95 percent CI: 7, 12) in women.For each 10-cm increase in height, the risk increased by 19 percent (95 percent CI: 16, 23) in men and 15 percent (95 percent CI: 11, 19) in women.
The association between smoking and the risk of the individual disease entities was explored in the subsample with known smoking habits.However, an effect of smoking on the risk of the different disease categories was small, if at all, existing in this subsample.Hence, we did not adjust for smoking in the evaluation of height and body mass index and the risk of LHMs.

DISCUSSION
In the present study of more than two million persons, we found a moderate and consistent positive association between body mass index and the risk of lymphohematopoietic malignancies in both sexes.Height was positively associated with the risk of LHMs as well.
The study is, by far, the largest study on this topic, including almost 25,000 cases of LHMs.The large number of cases allowed separate analyses on individual LHM entities.Further, the study was based on height and weight measurements obtained on average 18 years prior to diagnosis.The study subjects were recruited from population-based studies with high attendance.
A weakness of the present study is the inability to evaluate possible confounders as educational level, alcohol, and dietary factors.Information on smoking habits was available for a subcohort only.However, since there was little or no effect of smoking on the risk of LHMs in this subsample, smoking was not included in the analysis on the effect of body mass index and height on LHMs.In a relatively recent American study, Cerhan et al. ( 6) adjusted for a range of possible confounders, but the associations between body mass index and height and risk of non-Hodgkin's lymphoma were not influenced by the adjustment.
Body mass index is a rough measure of adiposity.However, there is a very good correlation between body mass index and the percentage of body fat in large populations (1).Further, measurements were taken at one point in time, so that changes in body mass index over time could not be assessed.
The associations observed in this study are small in magnitude.However, the findings are relatively consistent across sexes and across different disease entities.If the observed associations are true, relatively large studies are necessary to obtain statistical significance of the associations.
The results from cohort studies on the association between body mass index and the risk of LHMs are inconsistent.While Rapp et al. (12) found an increasing risk of non-Hodgkin's lymphoma with increasing body mass index in women but not in men, Cerhan et al. (6) did not find any such association.Rapp et al. followed 145,000 Australians for an average of 9.9 years, while Cerhan et al. included 40,000 American women.The largest cohort study so far was conducted by Calle et al. (5).They explored the associations between body mass index and mortality from several cancers in more than 900,000 Americans.They observed a positive association between body mass index and risk of death due to non-Hodgkin's lymphoma in both sexes.In the present study, a modest increase in the risk of non-Hodgkin's lymphoma was observed in men.
In a recent cohort study, MacInnis et al. (10) observed a more than five times higher risk of myeloid leukemia in overweight and obese persons compared with persons having a body mass index of less than 25.The hazard ratios were somewhat higher for chronic myeloid leukemia than for acute myeloid leukemia.The risk of lymphoproliferative malignancies showed little relation with body size.In the present study, we observed an overall modest increase in risk of both lymphoproliferative and myeloproliferative malignancies with increasing body mass index as well as with increasing height.As with MacInnis et al. (10), we observed higher relative risks for chronic myeloid leukemia than for acute myeloid leukemia with increasing body mass index.
Several case-control studies have been performed to explore the association between body mass index and risk of LHMs.Case-control studies typically rely on self-reported measurements.Cases are asked about their height and weight before they were diagnosed.Hence, the results may be seriously flawed by recall bias (1).It is unknown whether such recall bias would be differential or nondifferential between cases and controls.In the largest case-control study (8) that included more than 3,000 patients, no association was observed between body mass index and the risk of non-Hodgkin's lymphoma or Hodgkin's lymphoma.In another case-control study, Cerhan et al. (7) divided non-Hodgkin's lymphoma into diffuse and follicular disease entities.They found a positive association between body mass index and risk of diffuse non-Hodgkin's lymphoma, but not between body mass index and follicular non-Hodgkin's lymphoma or non-Hodgkin's lymphoma overall.
LHMs are a complex group of diseases.Studies have used different classifications of LHMs.Most studies have explored the associations between body mass index and risk of non-Hodgkin's lymphoma, and the results have been inconsistent.The biologic mechanisms for a possible association include insulin-like growth factors, levels of leptin, and nutritional alterations (10).
The association between height and the risk of LHMs has received little attention.Studies have shown both positive associations (7) and no association (7,8,10) between adult height and risk of non-Hodgkin's lymphoma.Chang et al. (8) did not find any association between height and the risk of either non-Hodgkin's lymphoma or Hodgkin's lymphoma.However, they found a higher risk of chronic lymphatic leukemia in the tallest persons than in the shortest persons.The present study also observed an increasing risk of chronic lymphatic leukemia with increasing height.Further, a modest increase in the risk of LHMs combined was observed with increasing height, being most pronounced for non-Hodgkin's lymphoma, Hodgkin's lymphoma, chronic lymphatic leukemia, and acute myeloid leukemia.The asso-ciations with height might indicate that nutritional factors in early life can influence the risk of these disease entities.
In the present study, we observed a moderate positive association between body mass index and the risk of lymphohematopoietic malignancies.Separate analyses for different disease entities did not reveal any single disease more strongly associated with body mass index than the others.We observed a modest increase in the risk of lymphohematopoietic malignancies with increasing height.The association was most pronounced for non-Hodgkin's lymphoma, Hodgkin's lymphoma, chronic lymphatic leukemia, and acute myeloid leukemia.The moderate associations between height and body mass index and LHMs found in the present study indicate that the observed increase in overweight/obesity plays only a minor role in explaining the increased incidence of LHMs.

FIGURE 1 .
FIGURE 1.The logarithm of the relative risk of lymphohematopoietic malignancies, with 95% confidence intervals, by body mass index from penalized spline functions with 4 df, adjusted for birth cohort and age at measurement, Norway, 1963-2001.The study population was a cohort of persons aged 20-74 years at inclusion.A, graph for men; B, graph for women.

FIGURE 2 .
FIGURE 2. The logarithm of the relative risk of lymphohematopoietic malignancies, with 95% confidence intervals, by height from penalized spline functions with 4 df, adjusted for birth cohort and age at measurement, Norway, 1963-2001.The study population was a cohort of persons aged 20-74 years at inclusion.A, graph for men; B, graph for women.

TABLE 1 .
Number of observed cases of lymphohematopoietic malignancies, personyears, and overall incidence rates* in a cohort aged 20-74 years at inclusion,Norway,  1963Norway,   -2001 * Number of cases of lymphohematopoietic malignancies per 100,000 person-years.

TABLE 3 .
Relative risk of lymphoproliferative malignancies, with 95% confidence intervals, obtained in Cox regression analysis,Norway, 1963Norway,  -2001*   * * Age at measurement and birth cohort were included in the model in addition to either body mass index or height.The study population was a cohort of persons aged 20-74 years at inclusion.yBody mass index and height, respectively, were included as continuous variables.

TABLE 4 .
Relative risk of myeloproliferative malignancies, with 95% confidence intervals, obtained in Cox regression analysis, Norway, 1963-2001* Age at measurement and birth cohort were included in the model in addition to either body mass index or height.The study population was a cohort of persons aged 20-74 years at inclusion.yBody mass index and height, respectively, were included as continuous variables.