“Case-Control Study: Birth Weight and Risk of Childhood Acute Lymphoblastic Leukemia in Arizona, Illinois, and Kentucky”

Abstract

Objective:

To confirm the previously-reported increased risk of leukemia among macrosomic children (those whose birth weight was greater than 4000 grams).

Methods:

Birth certificates of Arizona, Illinois, and Kentucky children diagnosed with acute lymphoblastic leukemia (ALL) before age five were matched with birth certificates from leukemia-free children of the same sex, race, and ethnicity who were born in the same county on or about the same day. Odds ratios for ALL among children of low (<2.5 kg) or high (>4 kg) birth weight were calculated by conditional logistic regression.

Results:

Children with high birth weight had an elevated risk of ALL in the first five years of life (OR=1.28; 95% CI: [1.01-1.61]). The excess risk was confined to non-Hispanic whites (OR=1.77; 95% CI: [1.27-2.48]), both boys (OR=1.57; 95% CI: [1.01-2.45]) and girls (OR=2.10; 95% CI: [1.26-3.52].

Conclusion:

This study confirms the association between high birth weight and ALL previously reported by other studies in children of European ancestry. The literature on maternal risk factors for both macrosomia and ALL is reviewed, with maternal overnutrition emerging as a plausible risk factor for both outcomes.

Introduction:

Approximately 2,400 US children and adolescents are diagnosed with acute lymphoblastic leukemia (ALL) each year; the etiology remains elusive, but one well-documented risk factor is the direct association of ALL incidence and high birthweight. ¹ The odds ratio for childhood ALL was estimated at 1.23 (95% CI: [1.15, 1.32]) for birthweight greater than 4000 grams versus “normal” birthweight (various definitions) in a pooled analysis of data from twenty-two studies. ^2-24 Eight of the aforementioned studies were in the United States, where racial and ethnic diversity make the analysis more complex; results have varied, perhaps because of different matching variables. ^17-24 An early case-control study in the greater Denver, Colorado metropolitan area (frequency-matched on age and sex), did not detect any effect of birthweight on risk of ALL, nor did a secondary analysis of a case-control study (frequency-matched on year of birth, county of birth, maternal age, sex, and ethnicity) of leukemia in parts of three states where Department of Energy facilties were located. ^17,25 However, strong effects were detected by separate case-control studies in the states of Minnesota (matched on date of birth and county), New York (matched only on year of birth), and Washington (matched only on year of birth), and also in the nine-state “E15” study (matched on age and race but not sex). ^18-21

Birthweight had no effect on risk of ALL in two studies in California, one statewide (matched only on sex and date of birth) and one (matched on sex, race, ethnicity, date of birth, and county of birth for some matched sets) confined to the northern part of the state. ^22,23 However, the earliest case-control study in California (matched on sex, race, county, and month of birth) found an increased risk of ALL among macrosomic girls (OR=1.73), but not boys (OR=1.11). ²⁶ In a statewide study of thousands of cases, the ALL odds ratio among California children born during 1986-2007 increased from 1.51 (95% CI: [1.15, 1.98]) for birthweights of 2500-3500 grams, and 1.50 (95% CI: [1.13, 1.98]) for birthweights of 3500-4500 grams, to 1.70 (95% CI: [1.08, 2.68]) for birthweights greater than 4500 grams, compared with those weighing less than 2500 grams. ²⁷ A statewide case-control study in Texas, matched only on year of birth, found a strong effect of birthweight. ²⁴ In a subsequent statewide case-control study in Texas, risk of ALL was elevated significantly among Hispanic children who weighed more than 4000 grams at birth (OR=1.79; 95% CI: [1.19, 2.42]), and non-significantly among non-Hispanic black (OR=1.59; 95% CI: [0.36, 6.93]) and non-Hispanic white (OR=1.25; 95% CI: [0.90, 1.74]) macrosomic children. ²⁸ A hospital-based case-control study in Texas likewise found significantly elevated risk of ALL among macrosomic children (OR=1.93; 95% CI: [1.16, 3.19]), with the strongest association among Hispanic boys (OR=3.18; 95% CI: [1.16, 8.73]). ²⁹

In view of the variation in the findings of statewide studies using different methods of control selection and matching, we elected to assess the association of birth weight and childhood ALL by analyzing birth certificate data from cases and precisely-matched controls in Arizona, Illinois, and Kentucky; our findings are reported here, along with a focused review of pertinent etiologic literature on childhood ALL.

Methods:

This study was approved by the Institutional Review Boards at the University of Louisville, and at the state health departments in Kentucky, Arizona, and Illinois before data collection began. Cases were children diagnosed with ALL before the fifth birthday, as ascertained by their respective statewide population-based cancer registries. Link Plus software available from the Centers for Disease Control (CDC) was used to link cancer registry and birth certificate files, and birth certificates were obtained for all subjects who were born in the same state in which they were diagnosed. Kentucky Cancer Registry staff provided data on 90 cases diagnosed during 1995-2002 and successfully linked with Kentucky birth certificates, including 85 non-Hispanic whites (45 boys and 40 girls), 2 non-Hispanic black boys, 1 non-Hispanic black girl, 1 non-Hispanic boy of another race, and 1 Hispanic boy. These 90 cases were matched (4:1) with 360 controls of the same sex, race, and ethnicity, who were born in the same county on or about the same day. One investigator visited the Arizona Cancer Registry to abstract data for 262 cases who were diagnosed in Arizona during 1990-2002 and successfully linked with Arizona birth certificates, including 148 non-Hispanic whites (72 boys and 76 boys), 5 non-Hispanic blacks (four boys and one girl), 20 non-Hispanic children of some other race (13 boys and seven girls), and 89 Hispanics (58 boys and 31 girls). These 262 cases were matched (5:1) with 1307 controls of the same sex, race, and ethnicity, who were born in the same county on or about the same day. Illinois Cancer Registry staff provided data for 401 children diagnosed during 1989-2001 and successfully linked with Illinois birth certificates. There were 100 non-Hispanic whites (56 boys and 44 girls), 100 non-Hispanic blacks (58 boys and 42 girls), 101 non-Hispanic children of some other race (59 boys and 42 girls), and 100 Hispanics (53 boys and 47 girls). These 401 cases were matched (4:1) to 1592 controls of the same sex, race, and ethnicity, who were born in the same county on or about the same day. Birth weights were categorized as low (less than 2500 grams), normal (2500-4000 grams), or high (greater than 4000 grams); odds ratios (OR), and 95% confidence intervals (CI) for ALL were calculated by conditional logistic regression using PROC LOGISTIC (SAS® 9.4, SAS Institute, Cary, NC).

Results:

Data from all three states were pooled in the primary analysis, since the available sample size from any single state was deemed insufficient for definitive estimation of odds ratios. Children weighing more than 4000 grams at birth had an elevated risk of ALL in the first five years of life (OR=1.28; 95% CI: [1.01, 1.61]), considering all races and both sexes in a combined analysis (data not shown). As shown in Table 1, this excess was statistically significant among non-Hispanic whites (OR=1.77; 95% CI: [1.27, 2.48]), both boys (OR=1.57; 95% CI: [1.01, 2.45]) and girls (OR=2.10; 95% CI: [1.26, 3.52]); a similar excess was also present, but not statistically significant, among Hispanic boys (OR=1.35; 95% CI: [0.80, 2.29]), but not among Hispanic girls (OR=0.41; 95% CI: [0.12, 1.38]). Conversely, birth weight less than 2500 grams was associated with a borderline-significant reduced risk of ALL among girls (OR=0.56; 95% CI: [0.32, 0.99]), but not among boys (OR=1.08; 95% CI: [0.68, 1.73]). Neither high nor low birth weight were associated with risk of ALL among non-Hispanic blacks, nor among non-Hispanic persons of other races (data not shown), regardless of gender.

Using a lower threshold (3800 grams) for high birthweight in girls did not result in additional statistically-significant associations (data not shown). A separate secondary analysis of the Arizona data (not shown) detected statistically-significant associations between high birth weight and risk of ALL among children of all races, both sexes (OR=1.73; 95% CI: [1.18, 2.53]), boys of all races (OR=1.90; 95% CI: [1.20, 3.01]), non-Hispanic children of both sexes (OR=2.14; 95% CI: [1.31, 3.49]), and non-Hispanic boys (OR=2.20; 95% CI: [1.15, 4.21]), along with borderline elevated risk among macrosomic Hispanic children, both boys (OR=1.96; 95% CI: [0.97, 3.95]) and girls (OR=2.09; 95% CI: [0.97, 4.49]). A similar secondary analysis of the Illinois data (not shown) detected a statistically-significant reduced risk of ALL among black girls of low birth weight (OR=0.33; 95% CI: [0.13, 0.83]).

Discussion:

Our study provides additional confirmation of the association of high birth weight with the risk of childhood ALL, reported previously in the meta-analysis cited in the Introduction. ² Several international studies that were not included in the 2009 meta-analysis have also supported an increased risk of ALL associated with high birthweight and will be reviewed below; other studies recently conducted in the United States were already reviewed in the Introduction. The risk of ALL increased with increasing birthweight (for every 500-gram increase, OR=1.08; 95% CI: [1.05, 1.12]) in a case-control study of children aged 1-9 years in England and Wales. ³⁰ More recent literature was reviewed in a commentary by Ross, which accompanied the latter article. ³¹ Every increase of one standard deviation in the “proportion of optimal birthweight” was associated with a 1.18-fold increased risk of pre-B-cell ALL (95% CI: [1.03, 1.35]) in a case-control study in Australia. ³²A subsequent analysis of pooled data from studies in the US and the UK as well as Australia found an overall odds ratio of 1.16 (95% CI: [1.09, 1.24]) per increase of one standard deviation increase in proportion of optimal birthweight. When data from twelve studies in the same consortium were pooled, risk was likewise increased among infants deemed “large for gestational age” (OR=1.21; 95% CI: [1.11, 1.32]). ³³

The association of macrosomia with risk of childhood ALL will be just another interesting scientific fact, except insofar as it can be applied to promote health and prevent disease among pregnant women and their offspring. It is presumed that every expectant mother wants her offspring to be born at a healthy weight, yet not so big as to increase the risk of cancer – but what are the prenatal determinants of birthweight, what are the prenatal determinants of childhood cancer, and to what extent do they coincide? In particular, which components of the maternal diet, if any, may increase the risk of macrosomia or leukemia in the offspring, and what sort of prenatal diet might reduce the risks of macrosomia, leukemia, and other adverse health outcomes? A previous report on the inverse correlation of childhood anemia and leukemia included a wide-ranging review of the literature on descriptive and analytic epidemiology of childhood ALL. ³⁴ A more focused review of the literature on maternal prenatal exposures, including diet, and risk of macrosomia and childhood ALL, will follow.

The risk factors for macrosomia, and its associated risk of childhood ALL, must be rooted in nutritional, metabolic, and hormonal, differences in the intrauterine environment, which must necessarily depend on maternal prenatal health and nutritional status. The “Pedersen hypothesis” holds that maternal hyperglycemia, and the concomitant reactive fetal hyperinsulinemia, lead to rapid intrauterine growth or frank macrosomia among the offspring of diabetic (and hyperglycemic) mothers. ³⁵ Insulin-like growth factor I (IGF-1) stimulates B-lymphocyte proliferation, and an etiologic role for IGF-1 in childhood ALL has been postulated. ^36,37 Furthermore, a California study found an increased risk of ALL among the offspring of mothers with pre-pregnancy diabetes (OR=1.37; 95% CI: [1.11, 1.69]); a higher proportion of cases (12.1%) than controls (10.3%) had high birthweight, but the odds ratio not was reported. ³⁸ Previous studies have found a non-significant increased risk of ALL associated with maternal diabetes in New York State (OR=1.37; 95% CI: [0. 91, 2.18]) and Washington State (OR=1.4; 95% CI: [0.8, 1.7]), but not in Minnesota (OR=1.00; 95% CI: [0.48, 2.08]). ^19,20,39 Gestational diabetes in particular was not associated with ALL among children under 20 years of age in Washington State (OR=2.3; 95% CI: [0.5, 11.0]), nor among children under 15 years of age in Greece (no odds ratio was reported), nor among children under five years of age in Western Australia (OR=0.46; 95% CI: [0.07, 3.32]). ^20,40,41

Maternal intake of carbohydrates with a low glycemic index was associated with reduced risk of fetal macrosomia. ^42,43 Shifting the focus to specific food items, maternal milk consumption has been associated with increased risk of fetal macrosomia. ⁴⁴ Maternal diet has also been studied with respect to its effect on subsequent risk of childhood ALL. For example, the risk of childhood ALL was inversely associated with maternal consumption of fruits (OR 0.71, 95% CI: [0.49, 1.04]), vegetables (OR 0.53, 95% CI: [0.33, 0.85]), and protein sources (OR 0.40, 95% CI: [0.18, 0.90]), including beans (OR 0.83, 95% CI: [0.70, 0.99]) in a case-control study of Northern California children. ⁴⁵ The risk of childhood ALL in a case-control study in Greece was increased by maternal consumption of (potentially diabetogenic) sugars and syrups (for each quintile increase, OR=1.32; 95% CI: [1.05, 1.67]), and was also increased by consumption of (saturated fatty-acid-rich) meat and meat products (for each quintile increase, OR=1.25; 95% CI: [1.00, 1.57]), but was decreased by consumption of (polyunsaturated fatty-acid-rich) fish and seafood (for each quintile increase, OR=0.72; 95% CI: [0.59, 0.89]). ⁴⁶ Childhood ALL risk was not associated with maternal consumption of ham / bacon / sausage (OR=1.5; 95% CI: [0.7, 3.0]), hamburgers (OR=1.2; 95% CI: [0.5, 2.7]), lunch meats (OR=1.0; 95% CI: [0.5, 2.0]), or charcoal-broiled meats (OR=1.0; 95% CI: [0.5, 1.9]), in a case-control study in Southern California; however, there was a “suggestive but not statistically significant” increasing trend (p=0.1) as monthly maternal consumption of hot dogs increased from zero (referent group) to 1-3.9 servings (OR=0.9), to 4-11.9 servings (OR=1.8) to twelve or more servings (OR=2.4; 95% CI: [0.7, 8.1]). ⁴⁷ A case-control study in Denver, Colorado found no association of childhood ALL with maternal consumption of hot dogs (OR=0.9; 95% CI: [0.4, 1.8]), hamburgers (OR=1.2; 95% CI: [0.5, 2.7]), lunch meats (OR=1.0; 95% CI: [0.5, 2.0]), or charcoal-broiled foods (OR=1.0; 95% CI: [0.5, 1.9]), nor with maternal consumption of ham / bacon / sausage (OR=1.5; 95% CI: [0.7, 3.0]), except among the small subgroup of mothers who did not take prenatal vitamins (OR=2.9; 95% CI: [1.1, 7.9]); the authors speculated that nitrosamines might form in-vivo among women who consumed ham, bacon, or sausage, except insofar as prenatal antioxidant vitamins could inhibit in-vivo nitrosation. ⁴⁸

One emerging micronutrient of interest is iron; an alternative explanation for the findings in the Denver study might be that animal products are rich in iron, while plant-based foods might interfere with iron absorption, and that iron might increase the risk of childhood ALL. ⁴⁸ A meta-analysis of clinical trials and cohort studies of iron supplementation and birth weight concluded that every 10-milligram increment in daily iron dose, up to 66 mg/day, was associated with a 15.1–gram (95% CI: [6.0, 24.2], trend p=0.005) increase in birth weight of the offspring, and that every 1-g/L increase in maternal prenatal hemoglobin was associated with a 14.0–gram (95% CI: [6.8, 21.8], trend p=0.002) increase in birth weight. ⁴⁹ High iron stores may explain why children with mutations (e.g., C282Y and H63D) in the hemochromatosis gene have increased risks of both macrosomia and leukemia. ⁵⁰ However, both overall use (OR=0.67; 95% CI: [0.47, 0.94]) and peripregnancy use (OR=0.72; 95% CI: [0.47, 1.09]) of iron supplements by the mothers were inversely associated with risk of childhood ALL in the Northern California Childhood Leukemia Study. ⁵¹ Risk of ALL was lower (OR=0.8; 95% CI: [0.7, 1.0]) in children born to mothers who took iron supplements during pregnancy in a multistate case-control study in the United States. ⁵² A non-significantly reduced risk of similar magnitude (OR=0.80; 95% CI: [0.51, 1.25]) was noted in a study in the United Kingdom. ⁹

Aside from diet, other maternal prenatal environmental exposures have also been implicated in the etiology of childhood ALL. Turning our attention back to the Northern California Childhood Leukemia Study, the risk of childhood ALL was approximately doubled in association with maternal history of pneumonia or influenza, both overall (OR=1.89; 95% CI: [1.24, 2.89]) and peripregnancy (OR=2.02; 95% CI: [1.28, 3.18]), and was more than quintupled by maternal history of sexually-transmitted diseases, both overall (OR=4.85; 95% CI: [1.24, 18.96]) and peripregnancy (OR=6.65; 95% CI: [1.37, 32.38]). ⁵¹ Maternal occupational exposure to “solvents, degreasers, or cleaning agents” almost doubled the risk of childhood ALL (before conception: OR=1.8; 95% CI: [1.3, 2.5]; during pregnancy: OR=1.6; 95% CI: [1.1, 2.3]), and the risk was also substantially increased by exposure to paints (before conception: OR=1.6; 95% CI: [1.2, 2.2]; during pregnancy: OR=1.7; 95% CI: [1.2, 2.3]) in a large nine-state case-control study conducted by the Children’s Cancer Group. ⁵³ Maternal occupational exposure to “paints and lacquers” was likewise implicated (before conception: OR=1.6; 95% CI: [1.1, 2.4]; during pregnancy: OR=2.0; 95% CI: [1.2, 3.3]) in a subsequent pooled analysis of three German case-control studies. ⁵⁴ The risk of childhood ALL was also approximately tripled by maternal occupational exposures to insecticides (OR=2.72; 95% CI: [1. 47, 5.04]) and herbicides (OR=3.62; 95% CI: [1.28, 10.3]). ⁵⁵ Since toxic exposures are more likely to be associated with low birthweight, however, it seems implausible that high birthweight is mediating the effect of toxic or otherwise noxious prenatal environmental exposures on the risk of ALL in the offspring. Overnutrition, rather than toxicity, seems more likely to increase the risk of macrosomia and the concomitant risk of childhood ALL.

Conclusion

Our data confirm the previously-reported association between macrosomia and childhood ALL in the under-five age group, but only in non-Hispanic whites. Our findings are strengthened by the use of birth certificate data from cases drawn from population-based cancer registries, meticulously matched with birth certificates of controls of the same sex, race, and ethnicity, born in the same county on or close to the same day. However, gestational age was not among the variables that we chose to abstract, and this precluded any assessment of weight for gestational age. A review of the literature on risk factors for macrosomia and childhood leukemia led us to question the role of iron nutriture in the etiology of childhood ALL. Future studies of the effect of maternal, fetal, and infant iron status on the risk of high birth weight and childhood ALL are warranted.