Parental, In Utero, and Early-life Exposure to Benzene and the Risk of Childhood Leukemia: A Meta-analysis

Frolayne M. Carlos-Wallace; Luoping Zhang; Martyn T. Smith; Gabriella Rader; Craig Steinmaus

Am J Epidemiol. 2016;183(1):1-14.

Abstract and Introduction

Abstract

Benzene is an established cause of adult leukemia, but whether it is associated with childhood leukemia remains unclear. We conducted a meta-analysis in which we reviewed the epidemiologic literature on this topic and explored causal inference, bias, and heterogeneity. The exposure metrics that we evaluated included occupational and household use of benzenes and solvents, traffic density, and traffic-related air pollution. For studies of occupational and household product exposure published from 1987 to 2014, the summary relative risk for childhood leukemia was 1.96 (95% confidence interval (CI): 1.53, 2.52; n = 20). In these studies, the summary relative risk was higher for acute myeloid leukemia (summary relative risk (sRR) = 2.34, 95% CI: 1.72, 3.18; n = 6) than for acute lymphoblastic leukemia (sRR = 1.57; 95% CI: 1.21, 2.05; n = 14). The summary relative risk was higher for maternal versus paternal exposure, in studies that assessed benzene versus all solvents, and in studies of gestational exposure. In studies of traffic density or traffic-related air pollution published from 1999 to 2014, the summary relative risk was 1.48 (95% CI: 1.10, 1.99; n = 12); it was higher for acute myeloid leukemia (sRR = 2.07; 95% CI: 1.34, 3.20) than for acute lymphoblastic leukemia (sRR = 1.49; 95% CI: 1.07, 2.08) and in studies that involved detailed models of traffic pollution (sRR = 1.70; 95% CI: 1.16, 2.49). Overall, we identified evidence of associations between childhood leukemia and several different potential metrics of benzene exposure.

Introduction

Millions of people in the United States and worldwide are exposed to benzene, either in occupational settings or environmentally.^[1,2] The International Agency for Research on Cancer has classified benzene as a human carcinogen based primarily on research that linked occupationally exposed workers to elevated risks of acute myeloid leukemia (AML).^[3] Children and fetuses might also be exposed to benzene either through air pollution, exposure sources near residences (such as gas stations or automotive repair facilities), maternal workplace exposure during pregnancy, or home use of products that contain benzene. Despite the sufficient evidence linking benzene to leukemia in adults, it is less clear whether there is an association in exposed children.

Childhood cancer is one of the leading causes of death by disease in children,^[4] and acute leukemia accounts for 30% of childhood malignancies.^[5] The most common form is acute lymphoblastic leukemia (ALL), with AML being less common but more fatal.^[6] Risk factors include genetic conditions such as Li-Fraumeni syndrome or Down syndrome, radiation therapy, chemotherapy, and possibly tobacco smoke.^[5,7] To date however, the causes of most childhood leukemias are unknown, and no single chemical agent has been conclusively linked to this cancer.

Although the links between benzene and childhood leukemia have been evaluated in several studies,^[8–12] the results have been mixed. The reasons for this are not well defined but could be attributed to differences in study design, exposure levels, leukemia subtypes, and the methods used to assess benzene exposure. Another possible reason is that exposures in children might be much lower than occupational exposures in adults. Studies of low exposures are likely to have relative risks closer to 1.0, and inadequate sample sizes, bias, or confounding can be especially important when increases in relative risks are small.^[13]

We examined current epidemiologic studies of parental or childhood exposure to benzene and childhood leukemia. Our goals were to perform a systematic literature review, evaluate each study for the basic tenets of epidemiologic validity, and summarize this literature using meta-analysis. Another goal was to evaluate sources of heterogeneity, including heterogeneity based on study design, confounding, leukemia subtype, and methods and timing of exposure assessment.

1 2 3 4

Next Section

Abstract and Introduction
Methods
Results
Discussion
References
Appendix

Table 1. Studies Included in the Overall Analysis of Occupational and Household Exposure to Benzene and Childhood Leukemia, 1987–2014
First Author, Year (Reference No.)	RR	95% CI	Type of Leukemia	No. of Exposed Cases	Study Location	Study Years	Children's Ages, years	Exposure
First Author, Year (Reference No.)	RR	95% CI	Type of Leukemia	No. of Exposed Cases	Study Location	Study Years	Children's Ages, years	Parent	Period	Exposure Assessed
	Studies of Occupational Exposures
Abadi-Korek et al., 2006 (8)	2.11	1.10, 4.20	ALL	112^a	Israel	1984–2002	0–19	Either	Overall	Solvents, based on detailed histories of parental occupations and exposures, every job held, and specific exposures
Buckley et al., 1989 (44)	2.0	1.2, 3.8	AML	57	United States and Canada	1980–1984	0–18	Father	Overall	Solvents, based on lifetime occupational history; 52 agents in 9 categories
Castro-Jiménez and Orozco-Vargas 2011 (9)	11.65	2.98, 45.59	ALL	85^a	Colombia	2000–2005	0–15	Both	Unknown	Benzene, based on occupations, JEM, and expert evaluation; conception to diagnosis
Cocco et al., 1996 (45)	1.5	0.3, 8.0	ALL	3	Italy	1980–1989	Unknown	Father	Unknown	Solvents, based on occupations, JEM, and expert evaluation
Feingold et al., 1992 (37)	1.6	0.5, 5.8	ALL	9	United States	1976–1983	0–14	Father	Periconception	Benzene, based on occupations and expert evaluation
Feychting et al., 2001 (38)	1.23	0.39, 3.85	Leukemia	3	Sweden	Children born 1976, 1977, 1981, 1982	0–14	Father	Preconception	Benzene, based on father's occupation in the census 1–2 years before the child's birth, JEM, and expert evaluation
Infante-Rivard et al., 1991 (46)	0.62	0.20, 1.91	ALL	6	Spain	1983–1985	0–15	Mother	Gestation	Solvents, based on occupation, specific substances during pregnancy, and expert evaluation
Infante-Rivard et al., 2005 (39)	1.47	0.25, 8.85	ALL	790^a	Canada	1980–2000	0–9, 0–14	Mother	Any or ever before birth	Benzene, based on occupations and specific exposures from 2 years before pregnancy to birth and on expert evaluation (highest exposure category)
Lowengart et al., 1987 (49)	1.00	0.33, 3.06	Leukemia	123^a	United States	1980–1984	0–10	Father	Overall	Petroleum chemicals, based on occupations and specific agents from 1 year before conception to 1 year before diagnosis
McKinney et al., 1991 (31)	4.00	0.30, 118	Leukemia and NHL	2	United Kingdom	1974–1988	0–14	Mother	Preconception	Benzene, based on occupations and specific questions about benzene
McKinney et al., 2008 (47)	2.0	0.7, 5.8	AML	256^a	United Kingdom	1991–1996	0–14	Mother	Gestation	Solvents, based on occupations and expert evaluation
Miligi et al., 2013 (43)	3.1	0.8, 12.4	Leukemia	6	Italy	1998–2001	0–10	Mother	Overall	Benzene, based on occupations, specific chemical exposures, and expert evaluation
Perez-Saldivar et al., 2008 (50)	2.81	0.48, 16.43	Leukemia (acute)	9	Mexico	1999–2000	0–15	Father	Preconception	Work as a mechanic, based on occupational history before conception
Reid et al., 2011 (40)	0.93	0.42, 2.09	ALL	9	Australia	2003–2006	0–14	Mother	Before birth	Benzene, based on occupations and expert evaluations
Shu et al., 1988 (41)	4.0	1.8, 9.3	AML	11	China	1974–1986	0–15	Mother	Gestation	Benzene, based on occupations and specific exposures
Shu et al., 1999 (42)	0.7	0.3, 1.6	ALL	11	United States	1989–1993	0–15	Mother	Any	Benzene, based on occupations and specific exposures
Smulevich et al., 1999 (48)	3.1	1.5, 6.3	Leukemia	20	Russia	1986–1988	0–14	Mother	Any	Solvents, based on occupations and job categories
	Studies of Household Product Use
Kishi et al., 1993 (51)	2.73	1.21, 6.31	ALL	147^a	Japan	1980–1990	0–15	Mother	Gestation	Benzene, based on home exposures
Scélo et al., 2009 (10)	2.23	1.01, 4.93	AML	100^a	United States	1995–2005	0–14	Family	Overall	Solvents, based on personal interviews on household use, from gestation to age 3 years (higher frequency category)
Slater et al., 2011 (11)	2.33	1.30, 4.18	AML	29	United States and Canada	1996–2006	0–1	Mother	Any	Petroleum products, based on personal interview about household exposures 1 month before and during pregnancy

Abbreviations: ALL, acute lymphoblastic leukemia; AML, acute myeloid leukemia; CI, confidence interval; JEM, job exposure matrix; NHL, non-Hodgkin lymphoma; RR, relative risk estimate.
^aTotal number of cases, including unexposed cases.

Table 2. Studies Included in the Overall Analysis of Traffic Density, Traffic-Related Air Pollution Models, or Gas Station Residential Proximity and Childhood Leukemia, 1999–2014
First Author, Year (Reference No.)	RR	95% CI	Type of Leukemia	No. of Exposed Cases	Study Location	Years	Child Age, Years	Exposure
First Author, Year (Reference No.)	RR	95% CI	Type of Leukemia	No. of Exposed Cases	Study Location	Years	Child Age, Years	Period	Exposure Assessed or Model Used
	Studies of Traffic Density or Traffic-Related Air Pollution Models
Amigou et al., 2011 (52)	3.1	0.8, 11.6	AML	3	France	2003–2004	0–14	At diagnosis	Composite (density, proximity, and nitrogen dioxide)
Badaloni et al., 2013 (53)	1.30	0.79, 2.12	Leukemia	74	Italy	1998–2001	0–10	Lifetime (never moved)	Land-use regression models (nitrogen dioxide)
Crosignani et al., 2004 (54)	4.31	1.47, 12.62	Leukemia	7	Italy	1978–1997	0–14	At diagnosis	California LINE Source Dispersion Model, version 4
Harrison et al., 1999 (55)	1.16	0.74, 1.72	Leukemia	24	United Kingdom	1990–1994	0–15	At diagnosis	Traffic density
Heck et al., 2014 (60)	1.94	0.89, 4.19	AML	66	United States	1990–2007	0–5	Birth	Air monitoring data for benzene (per IQR)
Langholz et al., 2002 (61)	1.4	0.7, 3.0	Leukemia	46	United States	1978–1984	0–10	Residence of longest duration	Traffic density
Pearson et al., 2000 (62)	8.28	2.09, 32.8	Leukemia	8	United States	1976–1983	0–14	At diagnosis	Traffic density (distance-weighted)
Raaschou-Nielsen et al., 2001 (56)	0.4	0.1, 1.6	Leukemia	870^a	Denmark	1968–1991	0–14	Childhood (all residences)	Operational Street Pollution Model (benzene)
Steffen et al., 2004 (12)	1.1	0.8, 1.6	Leukemia	155	France	1995–1999	0–14	Childhood (all residences)	Traffic density (heavy traffic roads)
Vinceti et al., 2012 (57)	1.92	0.64, 5.78	AML	19	Italy	1998–2009	0–14	At diagnosis	California LINE Source Dispersion Model, version 4 (per 1 μg/m³ benzene)
Visser et al., 2004 (58)	4.9	1.6, 11.4	ALL	5	Netherlands	1989–1997	0–14	At diagnosis	Traffic density (traffic intensity score)
Whitworth et al., 2008 (63)	2.02	1.03, 3.96	AML	31	United States	1995–2004	0–19	At diagnosis	Assessment System for Pollution Exposure Nationwide
	Studies of Residential Proximity to Gas Stations
Brosselin et al., 2009 (59)	1.1	0.5, 2.5	AML	7	France	2003–2004	0–18	Any	Residential proximity to garages or gas stations based on personal interviews; all residences from pregnancy to diagnosis
Harrison et al., 1999 (55)	1.48	0.65, 2.93	Leukemia	8	United Kingdom	1990–1994	0–15	At diagnosis	Living within 100 m of a gas station
Steffen et al., 2004 (12)	7.7	1.7, 34.3	AML	4	France	1995–1999	0–14	Childhood	Proximity to repair garages or gas stations, based on interviews about residential history from conception to diagnosis

Abbreviations: ALL, acute lymphoblastic leukemia; AML, acute myeloid leukemia; CI, confidence interval; IQR, interquartile range; RR, relative risk.
^aTotal number of cases, including unexposed cases.

Table 3. Results of the Meta-Analyses of Studies of Childhood Leukemia and Parental and Early Childhood Exposure to Benzene in Occupational and Household Products, 1987–2014
Group and Subgroup	No. of Studies	Fixed Effects		Shore-Adjusted^a	Random Effects^a		Heterogeneity
Group and Subgroup	No. of Studies	sRR	95% CI	95% CI	sRR	95% CI	χ²	P Value
Outcome
All studies combined	20	1.96	1.60, 2.41	1.53, 2.52	1.94	1.48, 2.53	28.42	0.08
AML only	6	2.34	1.72, 3.18	NA	NA	NA	2.08	0.84
ALL only	14	1.57	1.30, 1.90	1.21, 2.05	1.63	1.21, 2.21	24.73	0.03
Parent
Mother	13	1.96	1.52, 2.55	1.39, 2.78	1.94	1.34, 2.80	21.55	0.04
Father	14	1.23	1.08, 1.39	1.07, 1.41	1.27	1.08, 1.50	16.23	0.24
Exposure metric
Maternal occupational exposure
Benzene	7	1.71	1.14, 2.58	0.91, 3.24	1.94	0.97, 3.91	14.51	0.02
Solvents	7	1.17	0.94, 1.46	0.84, 1.63	1.25	0.87, 1.80	13.66	0.03
Paternal occupational exposure
Benzene	6	1.18	1.00, 1.41	0.94, 1.50	1.27	0.89, 1.80	9.43	0.09
Solvents	9	1.18	1.00, 1.39	NA	NA	NA	6.65	0.57
Household product use^b	6	1.67	1.24, 2.26	1.01, 2.78	1.88	1.10, 3.21	14.16	0.01
Adjustments^c
Smoking	8	1.94	1.43, 2.64	1.30, 2.90	1.94	1.27, 2.97	11.97	0.10
Socioeconomic status	13	1.94	1.50, 2.51	1.39, 2.71	1.95	1.37, 2.71	20.44	0.06
Time of exposure
Preconception
Maternal	6	1.32	0.92, 1.89	0.84, 2.08	1.38	0.83, 2.27	7.98	0.16
Paternal	8	1.31	1.11, 1.54	NA	NA	NA	6.64	0.47
Periconception
Maternal	3	1.29	0.97, 1.71	NA	NA	NA	1.26	0.53
Paternal	4	1.13	0.94, 1.37	NA	NA	NA	1.53	0.67
Gestation (maternal exposure)	9	2.06	1.51, 2.81	1.41, 3.02	1.93	1.29, 2.88	12.12	0.15
Postnatal	6	1.51	0.99, 2.30	NA	NA	NA	3.95	0.56

Abbreviations: ALL, acute lymphoblastic leukemia; AML, acute myeloid leukemia; CI, confidence interval; NA, not applicable; sRR, summary relative risk.
^aThe Shore-adjusted and random-effects models were used if the χ² statistic was greater than the degrees of freedom (the number of studies minus 1).
^bIncludes benzene, aromatic hydrocarbon, solvent, and petroleum product use.
^cIncludes adjusted results and results from studies in which the authors stated that the adjustments had little or no impact on the results.

Table 4. Results of the Meta-Analyses of Studies of Childhood Leukemia and Traffic Density, Traffic-Related Air Pollution Models, or Gas Station Residential Proximity, 1999–2014
Group and Subgroup	No. of Studies	Fixed Effects		Shore-Adjusted^a	Random Effects^a		Heterogeneity
Group and Subgroup	No. of Studies	sRR	95% CI	95% CI	sRR	95% CI	χ²	P Value
Outcome
All studies combined	12	1.48	1.23, 1.78	1.10, 1.99	1.80	1.29, 2.52	27.49	<0.01
AML only	4	2.07	1.34, 3.20	NA	NA	NA	0.40	0.94
ALL only	7	1.49	1.20, 1.84	1.07, 2.08	1.79	1.21, 2.65	14.96	0.02
Exposure metric
Traffic density	11	0.97	0.88, 1.07	0.82, 1.15	1.25	0.96, 1.62	29.92	<0.01
Traffic density (excluding Heck et al. (85))^b	10	1.17	0.99, 1.40	0.89, 1.55	1.38	0.96, 1.62	23.26	0.01
Traffic pollution models (e.g., CALINE4)^c	7	1.70	1.25, 2.30	1.16, 2.49	1.77	1.17, 2.68	9.38	0.15
Adjustments^d
Socioeconomic status	9	1.49	1.19, 1.86	1.07, 2.08	1.77	1.20, 2.62	18.05	0.02
Area
Europe	8	1.35	1.09, 1.68	0.95, 1.93	1.65	1.09, 2.52	19.51	0.01
United States	4	2.02	1.35, 3.00	1.20, 3.38	2.13	1.25, 3.64	5.02	0.17
Time of exposure
Birth or gestation only	4	0.92	0.75, 1.14	0.73, 1.17	0.94	0.73, 1.22	3.93	0.27
Postnatal (any)	13	1.36	1.15, 1.60	1.05, 1.76	1.59	1.19, 2.13	29.19	<0.01
At diagnosis only	9	1.46	1.19, 1.80	1.03, 2.08	1.94	1.30, 2.91	22.91	<0.01
Residential proximity to gas stations	3	1.59	0.95, 2.67	0.70, 3.62	1.88	0.78, 4.55	5.08	0.08

Abbreviations: ALL, acute lymphoblastic leukemia; AML, acute myeloid leukemia; CALINE4, California LINE Source Dispersion Model, version 4; CI, confidence interval; NA, not applicable; sRR, summary relative risk.
^aThe Shore-adjusted and random-effects models were used if the χ² statistic was greater than the degrees of freedom (the number of studies minus 1).
^bHeck et al. (85) contributed 69% of the total weight in the traffic-density analysis.
^cIncludes studies with more detailed modeling of traffic pollution or ambient air levels of benzene.
^dIncludes adjusted results and results from studies in which the authors stated that the adjustments had little or no impact on results. Few of these studies were adjusted for tobacco smoking.