SEARCH

SEARCH BY CITATION

Keywords:

  • birth weight;
  • early life;
  • epidemiology;
  • leukemia

Abstract

  1. Top of page
  2. Abstract
  3. Material and methods
  4. Results
  5. Discussion
  6. References

A growing body of evidence suggests that childhood leukemia may be initiated in utero when lymphoid and myeloid cells are not fully differentiated and are particularly susceptible to malignant transformation. A fixed effects meta-analysis examining the association between birth weight and childhood leukemia was conducted including 32 studies and 16,501 cases of all types of leukemia (OL), 10,974 cases of acute lymphoblastic leukemia (ALL), and 1,832 cases of acute myeloid leukemia (AML). The odd ratios (OR) for the association of high birth weight with OL, ALL and AML were 1.35 (95% CI: 1.24, 1.48), 1.23 (95% CI: 1.15, 1.32), and 1.40 (95% CI: 1.11, 1.76), respectively, compared with normal birth weight. Low birth weight was not associated with overall and ALL leukemia, but with AML (OR = 1.50; 95% CI: 1.05, 2.13). Per 1000 g increase in birth weight, the OR for OL was 1.18 (95% CI: 1.13, 1.23) and ALL 1.18 (95% CI: 1.12, 1.23). The combined available evidence from observational studies suggests that high birth weight is associated with an increased risk of overall leukemia and ALL. For AML the risk may be elevated at both high and low extremes of birth weight, suggesting a U-shaped association. © 2008 Wiley-Liss, Inc.

Leukemia is the most common malignancy affecting children and accounts for 26% of all childhood cancers in the United States.1, 2 There is a growing body of evidence indicating that childhood leukemia is initiated in utero. Acute lymphoblastic leukemia (ALL) may acquire the first of 2 key mutations in utero due to the particular susceptibility of lymphoid cells in the liver and bone marrow that results from their predifferentiated state.3, 4 As the fetus grows, B-cell precursors rapidly proliferate and are exposed to high levels of circulating growth factors during embryogenesis. Simultaneously, the lymphocytes are maturing and are capable of responding to a broad range of antigens, which may permanently alter a cell to become more susceptible to malignant transformation.3, 4 Likewise, stem cells that give rise to the myeloid cell line may also be susceptible to circulating growth factors and hormones.5 Growth factors and hormones act to increase stem cell pool size, which increases the total number of replicating cells at risk for conversion into tumor cells. This may increase the risk of leukemia. One factor related to stem cell pool size is birth weight, which has been postulated as a risk factor for childhood leukemia.6

A meta-analysis conducted in 2002 by Hjalgrim et al. including 18 studies suggested an increased risk of overall leukemia and 1 leukemia subtype, ALL, for children with birth weight <4 kg versus ≥4 kg. However, no significant association was found for a subset of childhood leukemia, acute myeloid leukemia (AML). We conducted an updated meta-analysis, including all available clinical evidence to date, to assess the influence of high, low and normal birth weight categories and per kilogram of birthweight on the likelihood of developing childhood leukemia and 2 common leukemia subtypes (ALL and AML). Other leukemia types were not considered separately because they are rare in children.7

Material and methods

  1. Top of page
  2. Abstract
  3. Material and methods
  4. Results
  5. Discussion
  6. References

Relevant articles were identified via Medline and EMBASE search engines; published articles, letters, abstracts and review articles were considered. The search included articles published before September 1, 2008. Relevant publications were identified through Medline with search strings “Leukemia”[MeSH] and “Birth Weight”[MeSH]; “Risk Factors”[MeSH] and “Leukemia”[MeSH] and “Infant, Newborn”[MeSH]. EMBASE was searched with the string (‘birth’/exp OR ‘birth’) and (‘weight’/exp OR ‘weight’) and (‘leukemia’/exp OR ‘leukemia’). Additional articles were identified via cross-referencing using a validated “snow-balling” technique.8

To be eligible for inclusion, case-control or cohort studies had to report odds ratios, rate ratios, or hazard ratios with confidence intervals for associations of leukemia with high, low or normal birth weight strata. If effect estimates were unavailable, unadjusted effect estimates were calculated from the data provided when possible. A comprehensive review of the literature returned 32 studies for meta-analysis.

A preliminary analysis was conducted using all 32 studies evaluating the associations between leukemia and 2 leukemia subtypes (ALL and AML) with both high and low birth weight. Because these analyses displayed significant heterogeneity, its source was explored. One study was identified that contributed significantly to interstudy variability.9 Our study differed substantially from other studies because it included individuals up to the age of 30 rather than restricting cases to childhood leukemia. Restricting the population to children is important because the risk factor profile for leukemia may differ for adults and children, and intrauterine factors such as birth weight may be more relevant for the risk of childhood leukemia than for adult-onset leukemia. Moreover, disease outcomes can vary greatly by age of diagnosis.10 A sensitivity analysis excluding the Roman et al. study yielded little alteration of effect estimates, but greatly improved heterogeneity. Thus, the Roman et al. study was dropped from the final analyses, leaving 31 studies for inclusion.

The main analyses examine the association of high birth weight with the risk for leukemia. However, in a subset of the studies which provided information on low birth weight, additional analyses were conducted on the association of low birth weight with leukemia risk. After analyses using high and low birth weight compared to normal birth weight were performed, additional analyses were performed for each kilogram increase in birth weight. Where per kilogram effect estimates and confidence errors were not provided, effect estimates were calculated from the raw data where there were 3 or more birth weight categories.

Most studies reported leukemia incidence as the outcome, but 3 studies reported leukemia mortality.11–13 A sensitivity analysis excluding these studies yielded little alteration of effect estimates or heterogeneity. An additional sensitivity analysis excluding the 4 included cohort studies yielded little observed change in heterogeneity or effect estimate. When the same study population was used in 2 articles, the earlier study was excluded from the analysis. In 1 study that provided mean birth weights with standard deviations,14 cases and controls above and below 4 kg were estimated assuming a normal birth weight distribution.15 Using methods described by Hjalgrim et al., odds ratios for an association between high birth weight and leukemia were calculated using case-referent data in 2 studies by calculating odds ratios as observed divided by expected cases.11, 15, 16 Normal birth weight was usually defined as 2.5–4 kg or 3–3.5 kg. In all but 2 studies, low birth weight was defined as <2.5 kg. In studies published by Cnattingius et al. and Hjalgrim et al., low birth weight was reported as <1.5 kg. Only effect estimates for ALL and AML were reported due to the rarity of other types of leukemia in children.

Overall leukemia

Of the 31 studies identified for analyses, 14 reported overall leukemia effect estimates for the association between high birth weight and leukemia risk. Effect estimates and confidence intervals were extracted for analysis. ALL estimates were used to estimate overall leukemia estimates where overall data were unavailable for the low birth weight analysis, because ALL rates are similar to overall leukemia rates and ALL represents the large majority of leukemias in children.7, 15 Eleven studies were included to calculate overall leukemia effect estimates for low birth weight; ALL estimates from 6 of these were used as a proxy for overall leukemia effect estimates. Twenty-one studies were used to calculate risk of overall leukemia per kilogram of birth weight.

Acute lymphoblastic leukemia

ALL was measured most frequently because it is by far the most common type of childhood leukemia, representing 80% of total leukemia incidence.7 Effect estimates and confidence intervals were extracted. Of the 31 included studies, 12 studies provided effect estimates for low birth weight categories compared to normal birth weight and 23 reported ALL effect estimates for high birth weight. Sixteen studies were used to calculate risk of ALL per kilogram of birth weight.

Acute myeloid leukemia

Effect estimates and confidence intervals were extracted from 15 studies that reported AML effect estimates for low or high birth weight. However, due to evidence of a positive association between low birth weight and AML but not ALL, studies using binary birth weight exposure comparing high birth weight to normal and low birth weight, which effectively combined the low birth weight category with the intermediate birth weight category, were excluded from the AML high birth weight analysis. After exclusion, 7 studies were used in the analysis of AML effect estimates. Additionally, an AML per kilogram increase in birth weight analysis was not performed because risk for AML was determined to be nonlinear.

Matching and adjusting factors considered potentially relevant for these studies were date of birth, gender, mother's ethnicity, child's ethnicity, birthplace, age at diagnosis, gestational age, maternal education, maternal age, socioeconomic status, tobacco use, parity and calendar period. However, no studies were excluded on the basis of these factors. Although most studies controlled for gender and date of birth, other matching and adjusting factors varied.

Statistical analysis

The meta-analyses were performed with 3 endpoints: overall leukemia, ALL and AML. A sensitivity analysis excluded studies that used binary birth weight categories for the high birth weight effect estimates, most commonly < 4 kg versus ≥4 kg. The sensitivity analysis suggested little effect on observed heterogeneity or the summary effect estimates. The meta command was used in STATA version 10 for each category using the log of the effect estimates and accompanying standard error. Results were exponentiated and displayed in a standard forest plot. Heterogeneity was also assessed using a Q-test and Galbraith plots. Though both random and fixed effects estimates were calculated, a fixed effects model was chosen for the final analysis. Publication bias was examined with Egger's and Begg's tests and the accompanying Egger's publication bias plots and Begg's funnel plots using the metabias command in STATA.

Results

  1. Top of page
  2. Abstract
  3. Material and methods
  4. Results
  5. Discussion
  6. References

Description of studies

The estimates represent more than 16,501 cases of all types of leukemia, 10,974 cases of ALL, and 1,832 cases of AML accrued between 1947 and 2005. The studies were published between 1962 and 2008. All were case-control studies, with the exception of 3 cohorts.

Of the 31 studies included studies, 12 reported effect estimates for comparisons of low birth weight to normal birth weight. Using these studies, an analysis was conducted to determine whether low birth weight is associated with leukemia. Eleven studies were used to calculate overall leukemia effect estimates, 11 for ALL and 9 for AML effect estimates.

Table I provides information on study design and duration, numbers of cases and controls, sources of birth weight information, age ranges of children, birth weight strata, effect estimates with confidence intervals, and matching and adjusting factors taken into account in each study's analysis (Table I). In 22 studies birth weight information was obtained from birth records and in 11 studies this information was obtained interviews and/or questionnaires administered to parents and patients.

Table I. Included Study Characteristics
Study (first author, year of publication, country)Study designFollow-up periodTotal leukemia casesALL casesAML cas'esControlsSource of birth weight informationBirth weight categories (g)Hazard (OR, IR, HR), with 95% CIAge RangeMatching factorsAdjusting factors
Koifman 2008, Brazil17Case-Control1999–200520114853440Birth RegistryOverall:ORInfantsNoneGender, income, maternal age, hormone intake, pesticide exposure during pregnancy
<25001.52 (0.80, 2.89)
2500-29991
3000-39991.63 (1.08, 2.47)
>39991.59 (0.79, 3.17)
ALL:OR
<25001.50 (0.72, 3.11)
2500-29991
3000-39991.68 (1.03, 2.76)
>39992.28 (1.08, 4.75)
AML:OR
<25001.59 (0.46, 5.16)
2500-29991
3000-39991.48 (0.64, 3.28)
>3999-
Dorak 2007, England18Case-control1968–1992 225 1163Hospital recordsALL:OR<15NoneNone
<25001.29 (0.64, 2.60)
2500-29991
3000-34991.0 (0.66, 1.52)
3500-39991.03 (0.66, 1.6)
>40001.2 (0.67, 2.15)
Shuz 2007, Germany19Case-control1992–1994 6501052057Self-administered questionnaire by parentsALL:OR<15Gender, year of birth, communityGender, age of diagnosis, degree of urbanization, socioeconomic status
<25001.23 (0.71, 2.10)
2500-40001.00
>40001.41 (1.08, 1.84)
AML:OR
<25001.58 (0.55, 4.59)
2500-40001.00
>40001.56 (0.88, 2.79)
Spector 2007, USA20Case-control1996–2002240  255Telephone interviews with mothersOverall:OR<1Year of birthGender, race, maternal education
≤32031
3204-35151.29 (0.74, 2.22)
3516-38541.36 (0.79, 2.37)
≥38551.49 (0.86, 2.59)
McLaughlin 2006, USA21Case-control1985–2001 122269667Hospital recordsALL:RR<10Year of birthBirth year, gender, race and ethnicity, maternal age
<25000.73 (0.50, 1.06)
2500-29990.77 (0.61, 0.97)
3000-34991
3500-39991.07 (0.91, 1.26)
4000-44991.10 (0.86, 1.39)
≥45001.10 (0.67, 1.73)
AML:RR
<25001.94 (0.88, 4.00)
2500-29991.77 (1.04, 2.96)
3000-34991
3500-39991.59 (1.03, 2.49)
4000-44991.90 (1.04, 3.36)
≥45003.89 (1.63, 8.26)
Podvin 2006, USA22Case- control1981–2003595376854980Birth RecordsOverall:OR<20Year of birthMaternal age
<25001.3 (0.9, 1.9)
2500-39991
≥40001.4 (1.1, 1.8)
ALL:OR
<25001.2 (0.8, 1.8)
2500-39991
≥40001.6 (1.2, 2.0)
AML:OR
<25002.2 (1.1, 4.4)
2500-39991
≥40001.2 (0.7, 2.1)
Ma 2005, USA23Case- control1995–2003 3135360In-person interview of mothers and birth registriesALL:OR<15Date of birth, gender, mother's race, Hispanic statusALL adjusted for household income, maternal education; AML unadjusted
<25001
2500-39991.03 (0.53, 1.97)
≥40001.04 (0.52, 2.10)
AML:OR
<25000.35 (0.04, 1.25)
2500-39991
≥40000.82 (0.33, 2.01)
Roman 2005, UK24Case- control1992–1996119610131634753Hospital recordsOverall:OR<15Gender, month and year of birth, region of residence at diagnosisGender, age, study region
<40001
≥40001.2 (1.0, 1.4)
ALL:OR
<40001
≥40001.1 (0.9, 1.4)
AML:OR
<40001
≥40001.6 (1.1, 2.5)
Hjalgrim 2004, Denmark, Sweden, Norway, Iceland25Case- control1984–1999 18172801372Birth RegistriesALL:OR<15Gender, birth month and yearNone
<15000.78 (0.38, 1.60)
1500-19990.71 (0.37, 1.37)
2000-24990.85 (0.61, 1.20)
2500-29990.79 (0.66, 0.95)
3000-34990.83 (0.73, 0.94)
3500-39991
4000-44991.05 (0.9, 1.23)
≥45001.19 (0.09, 1.58)
AML:OR
<15007.29 (1.48, 36.05)
1500-19991.84 (0.33, 10.15)
2000-24990.88 (0.34, 2.24)
2500-29990.87 (0.53, 1.41)
3000-34990.87 (0.63, 1.19)
3500-39991
4000-44991.12 (0.76, 1.66)
≥45000.95 (0.45, 2.04)
Jourdan-Da Silva 2004, France26Case-control1995–1998 40865567Self-administered questionnaire for mothersALL:OR<16NoneAge, gender, regional distribution
<25001.0 (0.6, 1.9)
2500-29991.1 (0.7, 1.6)
3000-34991
3500-39990.8 (0.6, 1.1)
≥40001.4 (0.8, 2.3)
AML:OR
<25000.3 (0.003, 2.1)
2500-29990.9 (0.4, 1.0)
3000-34991
3500-39990.6 (0.3, 1.2)
≥40000.8 (0.2, 2.8)
Lee 2004, Singapore27Cohort1992–199864N/A CohortBirth RegistryOverall:IRR<6NoneGender, gestational age, birth order, maternal age
2500-35001
>35001.6 (1.1, 2.3)
ALL:IRR
2500-35001
>35001.3 (0.6, 2.5)
Paltiel 2004, Israel28Cohort1964–1976654124CohortHospital recordsOverall:HR<16NoneAML and Overall unadjusted. ALL adjusted for gender, sibling > 3.5 kg, sibship size, maternal origin, SES
<30001
3000-34992.21 (1.0, 5.1)
3500-39993.09 (1.3, 7.2)
>40003.96 (1.4, 10.9)
ALL:HR
<30001
3000-34994.04 (1.2, 13.5)
3500-39993.44 (1.0, 12.2)
>40004.06 (0.9, 18.2)
AML:IRR
<30001.40 (0.38, 5.21)
3000-34991
3500-39993.62 (2.51, 20.82)
>40004.85 (2.60, 36.09)
Murray 2002, Ireland29Cohort1975–1997 34227CohortBirth RegistriesALL:RR<14NoneNone
<35001
≥35001.34 (1.01, 1.79)
Okcu 2002, USA30Case-control199510483 245Birth CertificatesOverall:OR<5Year of birthYear of birth, gender, gestational age, maternal age, tobacco use, parity, and race/ethnicity
2500-39991
≥40001.7 (0.9, 3.0)
ALL:OR
2500-39991
≥40001.5 (0.9, 2.5)
Reynolds 2002, USA31Case-control1988–1997 190 58Birth CertificatesALL:OR<5Date of birth, genderGestational age
2500-39991
≥40001.09 (0.9, 1.31)
Shu 2002, USA32Case-control1989–1993 1839 1985InterviewALL:OR<15Age at diagnosis, gestational age, raceSocioeconomic status, maternal age, race
≤30001
3001-35001.9 (0.9, 1.3)
3501-40001.1 (0.9, 1.4)
>40001.4 (1.1, 1.8)
Mckinney 1999, Scotland33Case-control1991–1994 124 236Hospital RecordsOverall:OR<15Gender, maternal age, gestational ageNone
<25001.84 (0.67, 5.04)
2500-34991
≥35001.12 (0.73, 1.73)
ALL:OR
<25001.52 (0.52, 4.43)
2500-34991
≥35001.17 (0.74, 1.85)
Shu 1999, USA32Case-control1989–1993  456538Hospital recordsAML:OR1-14NoneAge at diagnosis, gestational age, race
<40001
≥40000.9 (0.62, 1.30)
Suminoe 1999, Japan16Case-control1985–1994 496177Case- referentBirth registryALL:OR<19NoneGender
<40001
≥40001.13 (0.67, 1.93)
AML:OR
<40001
≥40002.34 (1.23, 4.42)
Petridou 1997, Greece53Case-control1993–1994153  300Interview administered questionnaireOverall:OR<15Gender, age at diagnosis, gestational ageNone
<25001
≥40003.8 (1.1, 12.9)
Roman 1997, England9Case-control1962–199214311315286Hospital recordsOverall:OR<30Gender, maternal age, birthplaceNone
<25001.1 (0.4, 2.9)
2500-35001
>35001.2 (0.8, 1.8)
ALL:OR
<25001.3 (0.5, 3.7)
2500-35001
>35000.9 (0.6, 1.5)
AML:OR
<25000.0 (0.0, 11.7)
2500-35001
>35006.2 (1.3, 29.8)
Westergaard 1997, Denmark35Cohort1968–1992 40565CohortBirth registryALL:RR<15NoneGender, gestational age, maternal age, birth order, calendar period
<25100.74 (0.43, 1.26)
2510-30090.94 (0.68, 1.28)
3010-35091
3510-40091.22 (0.96, 1.55)
4010-45091.59 (1.17, 2.17)
≥45101.50 (0.79, 2.17)
AML:RR
<2510-
2510-30090.51 (0.22, 1.20)
3010-35091
3510-40091.12 (0.62, 2.02)
4010-45091.66 (0.83, 3.31)
≥4510-
Yeazel 1997, USA, Canada, Australia36Case-control1982–1989 1455232816Questionnaire for mothersALL:OR<4Area of originMaternal age, birth order, gestational age, gender
≤27971
2798-32911.2 (0.9, 1.6)
3292-35471.3 (1.0, 1.8)
3548-38591.1 (0.8, 1.6)
>38591.8 (1.2, 2.5)
Cnattingius 1995, Sweden37Case-control1973–1989 610 3061Birth RegistryALL:OR<17Gender, Maternal ageGestational age
<15001.8 (0.5, 6.8)
1500-19990.7 (0.2, 2.3)
2000-24991.0 (0.5, 1.8)
2500-29991.1 (0.8, 1.5)
3000-34991
3500-39991.2 (0.97, 1.5)
4000-44991.1 (0.8, 1.4)
≥45001.7 (1.1, 2.7)
Savitz 1994, USA38Case-control1976–1983 68 208In-person InterviewALL:OR<15Gender, gestational ageNone
<25000.8 (0.3, 2.0)
2500-40001
>40000.7 (0.2, 2.3)
Shu 1994, China32Case-control1986–1991166  166InterviewAML:OR<15Gender, age at diagnosis, gestational ageNone
<40001
≥40000.9 (0.62, 1.3)
Shu 1988, China39Case-control1974–1986309  618In-person InterviewOverall:OR<15Gender, age at diagnosisBirth order, birthplace
≤30001
3001-34991.1 (0.7, 1.5)
≥35001.7 (1.2, 2.6)
ALL:OR
≤30001
3001-34991.0 (0.7, 1.6)
≥35001.5 (0.9, 2.3)
Robison 1987, USA40Case-control1969–? 219 1744Birth registryALL:ORChildrenAge at birth, birthplaceNone
<40001
≥40000.83 (0.54, 1.28)
Daling 1984, USA11Case-control1974–198237  Case-referentBirth RecordsOverall:OR<2NoneGender, calendar period
<40001
≥40002.73 (1.35, 5.50)
Shaw 1984, USA14Case-control1975–1980255  510Birth CertificatesOverall:OR<16Gender, maternal age, birthplaceNone
<40001
≥40001.34 (0.87, 2.04)
Fasal 1971, USA12Case-control1959–1965800  810Birth registryOverall:OR1-9Gender, maternal age, birthplace, birth order, raceNone
<40001
≥40001.46 (1.11, 1.93)
MacMahon 1962, USA13Case-control1947–19581323  1301Birth certificatesOverall:OR<12Maternal age, birthplaceNone
<40001
≥40001.18 (0.95, 1.46)

Summary estimates:

Tests for heterogeneity reported by Q-test were not statistically significant for the high birth weight analysis for overall leukemia, ALL or AML with Q13 = 18.3 (p = 0.15), Q22 = 33.6 (p = 0.06), Q8 = 13.5 (p = 0.1), respectively. Moreover, visual inspection of Galbraith plots, Egger's publication bias plots, and Begg's funnel plots suggested little evidence of publication bias.

The fixed effects model analysis yielded significant positive summary effect estimates for the association between high birth weight and leukemia, with p-values < 0.01 for overall leukemia, ALL, and AML. The summary effect estimates (ORs) for the association of high birth weight with overall leukemia, ALL and AML were 1.35 (95% CI 1.24, 1.48) (Fig. 1), 1.23 (95% CI 1.15, 1.32) (Fig. 2), and 1.40 (95% CI 1.11, 1.76) (Fig. 3), respectively.

thumbnail image

Figure 1. Studies on the risk of overall leukemia associated with high birth weight. OR, odds ratio; RR, relative risk; HR, hazard ratio; CI, confidence interval. The square indicates the OR, RR, or HR; the size of the square represents the statistical weight of each study; the bars represent 95% CIs.

Download figure to PowerPoint

thumbnail image

Figure 2. Studies on the risk of ALL associated with high birth weight. OR, odds ratio; RR, relative risk; HR, hazard ratio; CI, confidence interval. The square indicates the OR, RR, or HR; the size of the square represents the statistical weight of each study; the bars represent 95% CIs.

Download figure to PowerPoint

thumbnail image

Figure 3. Studies on the risk of AML associated with high birth weight. OR, odds ratio; RR, relative risk; HR, hazard ratio; CI, confidence interval. The square indicates the OR, RR, or HR; the size of the square represents the statistical weight of each study; the bars represent 95% CIs.

Download figure to PowerPoint

The fixed effects estimates (ORs) for the association between low birth weight and leukemia yielded null associations for overall leukemia and ALL: 1.00 (95% CI 0.84, 1.20) (Fig. 4) and 0.97 (95% CI 0.81, 1.16) (Fig. 5), respectively. Neither overall nor ALL-specific effect estimates displayed significant heterogeneity, at Q9 = 9.72 (p = 0.37) and Q9 = 7.67 (p = 0.57), respectively. The fixed effects analysis for AML, on the other hand, indicated a positive association between low birth weight and AML; the summary effect estimate was 1.49 (95% CI 1.03, 2.15) (Fig. 6). However, a Q-test identified heterogeneity of borderline statistical significance, Q7 = 14.1 (p = 0.05).

thumbnail image

Figure 4. Studies on the risk of overall leukemia associated with low birth weight. OR, odds ratio; RR, relative risk; HR, hazard ratio; CI, confidence interval. The square indicates the OR, RR,or HR; the size of the square represents the statistical weight of each study; the bars represent 95% CIs.

Download figure to PowerPoint

thumbnail image

Figure 5. Studies on risk of ALL associated with low birth weight. OR, odds ratio; RR, relative risk; HR, hazard ratio; CI, confidence interval. The square indicates the OR, RR, or HR; the size of the square represents the statistical weight of each study; the bars represent 95% CIs.

Download figure to PowerPoint

thumbnail image

Figure 6. Studies on the risk of AML associated with low birth weight. OR, odds ratio; RR, relative risk; HR, hazard ratio; CI, confidence interval. The square indicates the OR, RR, or HR; the size of the square represents the statistical weight of each study; the bars represent 95% CIs.

Download figure to PowerPoint

The fixed effects estimates (ORs) for the risk of overall leukemia per kilogram increase in birth weight revealed possible heterogeneity where Q20 = 36.4 (p = 0.18). Fixed effects estimates for ALL alone revealed less heterogeneity where Q15 = 24.3 (p = 0.06). The fixed effects analysis also yielded significant positive associations for overall leukemia and ALL at 1.18 (95% CI 1.13, 1.23) (Fig. 7) and 1.18 (95% CI 1.12, 1.23) (Fig. 8).

thumbnail image

Figure 7. Studies on the risk of overall leukemia per kilogram birth weight. OR, odds ratio; RR, relative risk; HR, hazard ratio; CI, confidence interval. The square indicates the OR, RR, or HR; the size of the square represents the statistical weight of each study; the bars represent 95% CIs.

Download figure to PowerPoint

thumbnail image

Figure 8. Studies on the risk of acute lymphoblastic leukemia per kilogram birth weight. OR, odds ratio; RR, relative risk; HR, hazard ratio; CI, confidence interval. The square indicates the OR, RR, or HR; the size of the square represents the statistical weight of each study; the bars represent 95% CIs.

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Abstract
  3. Material and methods
  4. Results
  5. Discussion
  6. References

Although most individual studies included in this meta-analysis were too underpowered to yield significant results, the presented meta-analysis confirmed a significant positive association between high birth weight and overall leukemia risk. Moreover, the association between high birth weight and both ALL and AML subgroups of leukemia were also significantly positive. A previous meta-analysis conducted by Hjalgrim et al. found similar positive associations for high birth weight and leukemia to the present study's overall leukemia and ALL effect estimates, but was unable to report significant associations for AML. The present study includes nearly twice as many AML cases and is the first to report significant positive association between high birth weight and AML. Additionally, our meta-analysis is the first to examine the association between low birth weight and leukemia. We identified no association between low birth weight and ALL but a positive association with AML. This suggests that the risk of AML as a function of birth weight is U-shaped, being elevated in children of low or high birth weight whereas the ALL risk curve may not be. A dose-response relationship for every 1000 g increase in birth weight discovered positive associations for both ALL and overall leukemia.

However, although the presented results are strong, 2 effect estimates reported borderline and strong evidence for heterogeneity. The strong evidence for heterogeneity in the calculation of overall leukemia risk per 1000 g increase in birth weight may be due the possibility that AML may exhibit a U-shaped risk for leukemia rather than a dose response relationship. Other variation between studies may be a result of differences in study design. Several of the included studies did not match or adjust for several potentially important confounders, including gestational age, maternal education, maternal age, socioeconomic status, mother's tobacco use, and mother's parity. Furthermore, only 3 cohort studies were included in this analysis. Additional cohort studies on the association between birth weight and cancer may shed light on the link between birth weight and leukemia. Although gender was examined as an effect modifier by 6 included studies,24, 25, 28, 31, 35, 41 only two24, 28 reported finding significant results. We were not able to address the question of effect modification by gender in this meta-analysis because it was not reported in the majority of studies included. Analyses stratified by the source of birth weight information revealed stronger associations for overall leukemia risk and high birth weight among participants for whom birth weight information was available from records, but no differences were found for other associations.

The studies for this meta-analysis were conducted almost exclusively in developed nations, with the exception of 2 studies examining data from China. Leukemia rates may be increasing in developed countries, but information from developing countries is not as readily available.42 The gross national product (and other economic indicators) is associated with high birth weight independent of ethnicity. Upon migration to the US, Chinese women who migrate when young typically grow taller and heavier and give birth to heavier babies.43, 44 This suggests that there may be link between a Westernized environment and higher birth weights, which may in turn lead to leukemia. Though Western environments may influence birth weights, a sensitivity analysis excluding the Chinese studies had little effect on heterogeneity or effect estimates (data not shown).

Recent studies suggest that a hallmark genetic event for acute leukemias, a chromosomal translocation, can be initiated in utero.45–50 Birth weight is, in part, a result of the intrauterine environment and has been associated with several growth factors, including insulin-like growth factor-I (IGF-1), sex-steroid hormones, and insulin-like growth factor II (IGF-1I).51, 52In utero, growth factors are associated with an increase in the total number of stem cells, which increases the total number of replicating cells at risk for conversion into tumor cells. This may increase the risk of leukemia.6 IGF-1, an important embryonic growth factor, can increase the stem cell pool in humans and animals. Several other hormones and growth factors associated with birth weight and stem cell size, including IGF binding protein-3, estriol, and testosterone, significantly increased the stem cell pool.51

Growth factors also may expand populations of preleukemic cells that already exhibit genetic abnormalities.5 IGF-1 stimulates growth through IGF-1 Receptor (IGF-1R), which in turn stimulates the Ras/Raf/MAP kinase signaling pathway.53, 54 Importantly, this receptor is expressed on all hematopoietically derived cells. In vitro, IGF-1 stimulates growth of lymphoid and myeloid cells.54–56 Additionally, IGF-1 may be antiapoptotic.5 Human and murine tumor cells and murine hematopoietic cells are protected from apoptosis-inducing agents when IGF-1R is overexpressed on cell surfaces.57–59

IGF-2 is a powerful intrauterine growth factor and over-expression of is the IGF-2 gene has been associated with AML60 and ALL.60, 61 IGF-2 is an imprinted gene, normally only expressed from the paternal allele.52 However, when loss of imprinting occurs, biallelic expression may lead to fetal overgrowth.52 Thus epigenetic changes may provide alternative mechanisms for the observed associations between high birth weight and leukemia.

The presented meta-analysis also suggests a positive association between low birth weight and AML. Fetal programming may directly or indirectly affect the growth hormone-IGF axis through hyperinsulinemia.62 In response to starvation, a permanent reprogramming occurs that causes GH or IGF-1 resistance. As a result of starvation, children born low birth weight have an exaggerated response to growth hormone releasing hormone (GHRH), which stimulates GH and may lead to increased IGF-1 bioactivity.61 Reprogramming leads to higher levels of circulating hormones in childhood, which may promote AML incidence.

Although our results suggest a positive association between low birth weight and AML, we found no association between low birth weight and ALL. This difference may be related to the fact that AML and ALL have distinct origins. ALL and AML involve malignant transformation of lymphocytes and myeloid cells, respectively. Different cell types may respond differently to the intrauterine stimuli.

In summary, this study confirms previously reported evidence of a positive association between high birth weight and overall leukemia risk, and specifically risk of ALL and a dose-response type relationship for birth weight with ALL and overall leukemia. Further, this study identifies a U-shaped association between birth weight and AML.

References

  1. Top of page
  2. Abstract
  3. Material and methods
  4. Results
  5. Discussion
  6. References