Childhood body mass index at diagnosis and its association with non-Hodgkin lymphoma in the United States: a children’s oncology group data analysis project

Abstract

Childhood obesity has historically increased with time. Associations are noted between obesity and cancer in adults; but little investigation has been done in children. We conducted a matched case control study to assess the association between body mass index (BMI) and Non-Hodgkin Lymphoma (NHL) in children. Case information is from Children’s Oncology Group and controls from National Health and Nutrition Examination Survey. There was no association between obesity and NHL overall, but patients less than 10 years, showed a positive association if outside of the normal BMI range, both obesity and underweight. Additionally, there was an association between NHL and underweight overall, in males, and in non-Hispanic/Latino children. Obesity showed an inverse association with NHL in females and children older than 10. We were not able to assess causality. Disease causing weight loss prior to diagnosis likely skewed baseline weights lower. Additional studies with BMI prior to diagnosis are needed.

Introduction:

Non Hodgkin Lymphoma (NHL) is a malignant condition that originates in the lymphoid tissue and is derived from B cell progenitors, T cell progenitors, mature B cells, or mature T cells [1]. It is the 4^th most common pediatric malignancy [2] and incidence of NHL is increasing over time[3]. There are many different subtypes of NHL including Burkitt lymphoma, diffuse large B cell lymphoma, T/B lymphoblastic lymphoma, and anaplastic large cell lymphoma [2]. Typically in adult patients, NHL is low grade and slow growing but in pediatric patients we often see more aggressive and high grade forms of the disease [1].

Many studies have used body mass index (BMI) to define obesity and to assess the association between body size and different types of cancers, particularly in adult patients. Epidemiological studies have found strong associations between obesity and 13 adult malignancies [4], with obesity now estimated to account for approximately 20% of adult cancer incidence [5]. For NHL specifically, multiple previous studies in adult populations have observed a positive association between increased BMI and malignancy [6–9]. There are inherent differences between adult and pediatric NHL with adult NHL being both indolent or aggressive and pediatric NHL consisting mostly of aggressive disease. However, it remains unclear if there is a similar association between pediatric obesity and NHL.

The goal of this research is to investigate the association between pediatric obesity and NHL incidence. The prevalence of pediatric obesity has historically increased with time. However recently, there has been a decline in obesity in the youngest age groups (less than 5 years), stabilization in those ages 6–11, but continued increases in obesity in adolescents and young adults.[10] incidence continues to increase over time. Currently, about 20% of children age 2–19 are considered obese, and prevalence increases with age [11]. Given this, we examined the risk of NHL in pediatric patients based on their BMI category at diagnosis.

Methods:

Data Sources:

Cases of Non-Hodgkin Lymphoma were obtained from the Children’s Oncology Group (COG). Cases were enrolled in COG studies: AALL1731, ANHL12P1, 9404, 9517, 9917, ANHL0131, ANHL1131, AALL1732, AALL0434, AALL0932, 9315, and 9317. See supplemental table 1 regarding the study titles listed above. We obtained data regarding cancer diagnosis, BMI (calculated using weight and height, collected for dosing purposes), age at diagnosis, race, and ethnicity, which were collected as part of the study protocol and obtained. NHL stage and metastatic disease status were collected for the statistical analysis as well. Data on controls was obtained from the NHANES which was chosen given its national representation as well as ability to match based on race/ethnicity.

Study Population:

Eligible study cases were those diagnosed with NHL between 2 years old and 19 years old, resided in the United States and, had no missing BMI values that were calculated using Centers for Diseases Control and Prevention (CDC) Growth Charts for children and adolescents. Height and weight for cases was taken at diagnosis for chemotherapy dosing purposes. Due to the lack of meaningful interpretation for BMI in infants under two years of age, children less than 2 years of age were not included in this analysis [12]. Diagnosis years of cases ranged from 2003 to 2021. Controls were healthy participants interviewed by NHANES between 1999 to 2020 during their biennial survey.

Eligible cases of NHL (n=706) were identified from COG. From NHANES, 54,740 controls were identified. Descriptive characteristics of matched cases and controls can be seen in Table 1. COG cases were matched to the NHANES controls using a ratio of 1:4. Matching was done based on age at diagnosis or age at interview time for controls (within 1 year), sex, race/ethnicity, and year of NHL diagnosis or NHANES sampling (within 5 years).

Table 1.

Descriptive Demographic and Clinical Characteristics of Non-Hodgkin Lymphoma Matched Cases and Controls

	Non-Hodgkin lymphoma
Characteristics	COG cases (n=706)	NHANES controls (n=2824)
Sex
Male	486 (68.8)	1944 (68.8)
Female	220 (31.2)	880 (31.2)
Race
Hispanic	110 (15.6)	440 (15.6)
Non-Hispanic White	413 (58.5)	1652 (58.5)
Non-Hispanic Black	100 (14.2)	400 (14.2)
Other or Unknown	51 (7.2)	204 (7.2)
Non-Hispanic Asian	32 (4.5)	128 (4.5)
Ethnicity
Hispanic or Latino	110 (15.6)	440 (15.6)
Not Hispanic or Latino	545 (77.2)	2180 (77.2)
Unknown	51 (7.2)	204 (7.2)
Age in Months
Months	132.1 (56.7)	126.0 (58.5)
Age Categories in Years
Child (2–<10)	300 (42.5)	1223 (43.3)
Adolescent (10–<20)	406 (57.5)	1601 (56.7)
SES
Low
High
Stage at Diagnosis
Local	104 (17.2)	--
Regional	162 (26.8)	--
Distant	205 (33.9)	--
Unknown	134 (22.1)	--
Not applicable/Not answered	101	--
BMI Groups
Underweight	45 (6.4)	125 (4.4)
Normal Weight	435 (61.6)	1784 (63.2)
Overweight	121 (17.1)	396 (14.0)
Obese	105 (14.9)	105 (14.9)

Abbreviations: COG (Children’s Oncology Group); NHANES (National Health and Nutrition Examination Survey); BMI (body mass index); SD (standard deviation); SES (Socioeconomic status).

Note: Data displayed as n (%) for categorical variables and n (SD) for continuous variables; Other or Unknown – Multi-Racial, White with unknown ethnicity, Black with unknown ethnicity, Asian with unknown ethnicity, Other/unknown race with unknown ethnicity, Native Populations or Pacific Islander with unknown ethnicity; BMI Groups – underweight (<5^th percentile), normal weight (5^th–85^th percentile), overweight (85^th–95^th percentile), obese (≥95^th percentile); percentages may not add up to 100% due to rounding.

Main and confounding variables:

Based on CDC growth charts for children and adolescents, BMI was calculated for cases and controls using the SAS Macro codes provided by CDC according to weight and height values of study participants. Compared to percentile curves of the BMI distribution in the CDC growth charts, the study cases and controls were classified into normal (5th-84th percentile), underweight (<5th percentile), overweight (85th-95th percentile) or obese (>95th percentile) according to their sex and age. The normal BMI group (5^th–84^th percentile) was the referent. Sex, ethnicity, race, and age at diagnosis were selected as the main covariates in in examining whether BMI is associated with NHL development.

Outcomes:

Association between NHL and BMI was the main outcome measure. We also assessed each of the interactions between BMI and sex, ethnicity, age at diagnosis, and stage at diagnosis.

Statistical analysis:

Descriptive frequencies and percentages were used to illustrate distributions of sex, race, ethnicity, age at diagnosis, BMI groups, and stage at diagnosis in the study cases and controls. Chi-square/Fisher’s test was used to assess and compare statistical differences of prognostic characteristics as well as sex by BMI groups. BMI was defined as obese (>95^th percentile), overweight (85^th–95^th percentile), normal weight (5^th–84^th percentile), and underweight (<5^th percentile), with normal as the referent.

Conditional logistic regression was first used to access association of BMI and NHL adjusting for sex, race, ethnicity, and age between matched cases and controls. The association of BMI with NHL was further assessed using conditional logistic regression between matched cases and controls stratified by sex, ethnicity, child’s diagnosis age (<10 years old vs >10–19 years old), and stage at diagnosis adjusting for any of the covariates (sex, race, ethnicity, age) that were not used for stratification.

The interactions between sex, ethnicity, child’s diagnosis age (<10 years old vs >10–19 years old), stage at diagnosis and BMI as a categorical variable were tested using likelihood ratio tests (p<0.05 considered statistically significant) for all multivariable adjusted conditional logistic regression models. Odds ratios and 95% confidence intervals were reported for all models. P-values less than 0.05 were considered statistically significant. SAS version 9.4 (SAS Institute Inc. Cary, NC.) was used to conduct statistical analysis.

Results:

Included in this analysis were 706 cases of NHL and 2,824 matched controls. Our cohort was 68.8% male, reflecting the typically higher rate of NHL among males relative to females [2]. We did not observe a statistically significant increased association with NHL for those that were overweight or obese. Those patients that were underweight showed a 1.84-fold increased association with NHL (95% CI: 1.24, 2.74). This can be seen in Table 2. There was also an association between BMI in the underweight category and NHL among males (OR 1.76, 95% CI:1.11, 2.79). Females alone showed a decreased association between obese BMI and NHL (OR 0.57, 95% CI: 0.35, 0.94). Younger patients (<10 years old) with either underweight (OR 2.43, 95% CI: 1.37, 4.30) or overweight (OR 1.51, 95% CI: 1.00, 2.28), showed an increased association with NHL. In those greater than 10 years of age, obesity had a decreased association with NHL with OR 0.70 (95% CI: 0.50, 0.96). Being overweight, obese, or underweight did not impact the association of NHL in Hispanic and Latino groups but in the non-Hispanic and Latino cohort there was a 1.97 times increased association if underweight (95% CI: 1.24, 3.13). Being underweight was also associated with both local disease and distant disease, however there was a stronger effect estimate for being underweight and having distant disease with an OR of 2.69 (95% CI: 1.33, 5.44) for distant disease compared to 2.19 (95% CI: 1.14, 4.24) for local disease. These results can be viewed in Table 3.

Table 2.

Conditional Logistic Regression Odds Ratios and 95% Confidence Intervals for Non-Hodgkin Lymphoma Risks by Body Mass Index in Matched Cases and Controls

	BMI Category
	Underweight	Normal weight	Overweight	Obese	p-trend
Non-Hodgkin lymphoma risk
N (Cases/Controls)	143 (45/98)	2203 (435/1768)	570 (121/449)	614 (105/509)
Multivariable-adjusted OR (95% CI)*	1.84 (1.24, 2.74)	1.00	1.06 (0.83, 1.35)	0.85 (0.66, 1.09)	0.032

Abbreviations: BMI (body mass index).

^*Adjusted for sex (male, female), Race (categorical – Asian, Black or African American, White, Native Populations or Pacific Islander, White), ethnicity (Hispanic or Latino, Not Hispanic or Latino, unknown), age in months (continuous), socioeconomic status (low, high).

^#Interactions between BMI and age in months were tested.

Table 3.

Conditional Logistic Regression Odds Ratios and 95% Confidence Intervals for Non-Hodgkin Lymphoma Risks by Body Mass Index Stratified by Demographics in Matched Cases and Controls

	BMI Category
	Underweight	Normal weight	Overweight	Obese	p-trend
Sex
Male
N (Cases/Controls)	107 (33/74)	1507 (291/1216)	389 (82/307)	427 (80/347)
Multivariable-adjusted OR (95% CI)a	1.76 (1.11, 2.79)	1.00	1.05 (0.77, 1.42)	0.99 (0.74, 1.33)	0.35
Female
N (Cases/Controls)	36 (12/24)	696 (144/552)	181(39/142)	187 (25/162)
Multivariable-adjusted OR (95% CI)a	2.12 (0.98, 4.60)	1.00	1.08 (0.71, 1.65)	0.57 (0.35, 0.94)	0.013
Ethnicity
Hispanic or Latino
N (Cases/Controls)	19 (6/13)	297 (58/239)	99 (24/75)	135 (22/113)
Multivariable-adjusted OR (95% CI)b	1.54 (0.54, 4.41)	1.00	1.23 (0.68, 2.23)	0.89 (0.50, 1.58)	0.58
Not Hispanic or Latino
N (Cases/Controls)	105 (33/72)	1757 (347/1410)	424 (87/337)	439 (78/361)
Multivariable-adjusted OR (95% CI)b	1.97 (1.24, 3.13)	1.00	1.00 (0.75, 1.34)	0.87 (0.65, 1.17)	0.06
Child’s age
< 10 vears old
N (Cases/Controls)	73 (26/47)	994 (177/817)	223 (54/169)	221 (43/178)
Multivariable-adjusted OR (95% CI)c	2.43 (1.37, 4.30)	1.00	1.51 (1.00, 2.28)	1.07 (0.70, 1.61)	0.65
10 + years old
N (Cases/Controls)	70 (19/51)	1209 (258/951)	347 (67/280)	393 (62/331)
Multivariable-adjusted OR (95% CI)c	1.38 (0.78, 2.45)	1.00	0.88 (0.64, 1.21)	0.70 (0.50, 0.96)	0.010
Stage at Diagnosis
Local/regional
N (Cases/Controls)	52 (17/35)	826 (164/662)	233 (45/188)	219 (40/179)
Multivariable-adjusted OR (95% CI)d	2.19 (1.14, 4.24)	1.00	1.02 (0.68, 1.53)	0.92 (0.62, 1.38)	0.25
Distant
N (Cases/Controls)	42 (17/25)	650 (121/529)	148 (36/112)	185 (31/154)
Multivariable-adjusted OR (95% CI)d	2.69 (1.33, 5.44)	1.00	1.26 (0.80, 1.99)	0.93 (0.59, 1.47)	0.25

Abbreviations: BMI (body mass index).

^aAdjusted for race (categorical – Asian, Black or African American, White, Native Populations or Pacific Islander, White), ethnicity (Hispanic or Latino, Not Hispanic or Latino, unknown), age in months (continuous), socioeconomic status (low, high).

^bAdjusted for sex (male, female), Race (categorical – Asian, Black or African American, White, Native Populations or Pacific Islander, White), age in months (continuous), socioeconomic status (low, high).

^cAdjusted for sex (male, female), Race (categorical – Asian, Black or African American, White, Native Populations or Pacific Islander, White), age in months (continuous), ethnicity (Hispanic or Latino, Not Hispanic or Latino, unknown), socioeconomic status (low, high).

^dAdjusted for sex (male, female), Race (categorical – Asian, Black or African American, White, Native Populations or Pacific Islander, White), ethnicity (Hispanic or Latino, Not Hispanic or Latino, unknown), age in months (continuous), socioeconomic status (low, high).

^#Interactions between BMI and sex, ethnicity, and child’s age (categorized into less than or more than 10 years old) were tested. Further, among cohorts stratified by stage at diagnosis, interactions between BMI and age in months (continuous) were tested.

Discussion:

Our study examined BMI at diagnosis and its association with pediatric NHL. There is substantial data in adult populations regarding obesity and its association with overall malignancy risk. Obesity has been associated with NHL in adults [6,7] but less information is available on how BMI is associated with pediatric NHL. In our analysis, we did not find an association between obesity and NHL. We did see that there may be an association between those with underweight BMI and NHL.

There are already established risk factors for NHL in the literature. Black children and young adults typically experience higher rates of NHL than white children, whereas all other racial and ethnic groups experience lower overall rates of lymphoma in general when compared to white children [13]. Males have increased incidence rates of most childhood cancers, including NHL and each of the other lymphoma types [14]. There are also birth characteristics which have shown increased association with NHL risk. Older maternal age, being large for gestational age, preterm labor, and premature rupture of membranes have each been associated with increased NHL risk [15–17]. We observed a positive association of NHL with obesity in young children who were less than 10 years of age at diagnosis. This could be due to alterations in the immune system or increased inflammation associated with early adiposity that could create a conducive environment for aberrant cell proliferation [18,19]. There could also be epigenetic changes related to early nutritional status which could lead to development of malignancies such as NHL [20]. However, given our study design, we are not able to address causality and further studies are needed to address this.

Additionally, BMI at diagnosis may be shifted lower due to the disease process itself, including weight loss prior to definitive diagnosis which may help to explain lack of association between NHL and obesity overall. Given that pediatric NHL tends to be more aggressive than adult NHL, with increased likelihood of systemic symptoms, our documented weights may not reflect a patient’s baseline weight status prior to initiation of the disease process. We also saw a stronger association with being underweight and having distant or more severe disease. This is likely because those with distant disease are more likely to have systemic symptoms including significant weight loss prior to diagnosis leading to being underweight at that time [21].

In adults there is evidence that being underweight at time of diagnosis or having “cancer cachexia” with weight loss caused from the malignancy, particularly in lymphomas, has a negative effect on progression free and overall survival [22]. In another study, patients who were underweight at diagnosis were seen to have overall poor survival outcomes compared to those with a normal BMI. Those who were overweight or obese had better survival outcomes [21]. This was again seen in a cohort study where those with elevated BMI were found to have better overall survival [23]. These are all studies in adult patients. There are, however, other studies that show an association between obesity in patients with NHL having worse overall survival [24] and in a meta-analysis looking at pediatric outcomes in leukemia, being obese was associated with worse overall outcomes [25]. Within the current literature, it remains unclear whether BMI influences overall survival outcomes in adult and in pediatric patients with NHL. Our current analysis did not examine survival of NHL as it relates to BMI at time of diagnosis.

Our study has potential limitations including that it is a retrospective case control study with BMI data taken only at diagnosis. This likely does not account for recent weight changes due to the systemic symptoms associated with NHL prior to diagnosis, the severity of which could be different in each patient depending on their individual disease severity. Despite these limitations, our study is one of the first to look at the association between BMI and NHL in pediatric patients. Further investigation should be done to help enhance our understanding of the various factors contributing to the development of NHL. Being underweight at diagnosis could be a predictor of outcome given its likely association with more aggressive disease and should be evaluated further. Additional studies to assess the association between NHL and BMI as well as to look at outcome differences based on pediatric BMI at diagnosis are warranted.