Pediatric leukemia and maternal occupational exposure to anticancer drugs: the Japan Environment and Children’s Study

Most pediatric cancers, including leukemias and brain tumors, occur in a sporadic form of unknown etiology. Prenatal genetic alterations are reportedly involved in the origin of leukemia cells developing in childhood as well as in young adults.^1,2 Beyond germline cancer predisposition, maternal exposure to environmental factors during pregnancy may influence carcinogenesis in fetuses. Several studies have described an association between leukemia in offspring and pregnant mothers’ exposure to hazardous factors, including pesticides,³ wood,⁴ hydrocarbons,⁵ solvents,^6,7 and air pollution.⁸ However, the effects of maternal occupational exposure on the development of leukemia in their offspring are controversial.^9,10 The discrepancies among the findings might be due to selection bias and recall/information bias from case-control study designs or a lack of clinical information about the patients. This being the case, no cohort study has yet pertinently succeeded in demonstrating an association between maternal exposure to anticancer drugs and leukemia in offspring.

Combination chemotherapy is the mainstay of treatment for pediatric leukemia and many solid tumors in childhood, despite recent advances in molecular target-based therapy, immunotherapy, radiotherapy, and hematopoietic cell transplantation. Cytotoxic agents effectively control proliferating primary cancer cells and have potential to induce the genetic alterations in noncancer cells, leading to second primary malignant neoplasm.^11,12 The administration of these agents further raises a substantial risk of carcinogenesis in the personnel handling these agents.¹³ Many studies have reported that occupational exposure of health care workers to these hazardous agents can affect their offspring and cause fetal loss or teratogenicity.^14-17 We have recently reported an association between maternal occupational exposure to ionizing radiation and infantile-onset neuroblastoma but not between exposure to medical agents and infant leukemia or brain tumors in a Japanese cohort.¹⁸ This result was derived from follow-up data in infants aged <12 months, and thus did not represent most pediatric patients with leukemia and brain tumors. Infant leukemia is distinct from common acute lymphoblastic leukemia (ALL).^19,20 Early-onset high-grade gliomas have a unique genetic background from adult gliomas.^21,22 Therefore, no study has yet demonstrated the missing link between the major pediatric-onset neoplasms and parental exposure to medical agents. Using a large-scale Japanese cohort, which released a new data set containing data for the first 3 years of life, we investigated the association between parental occupational exposure to medical agents and pediatric malignant neoplasm in children 3 years old or younger.

The Japan Environment and Children’s Study (JECS) is a nationwide multicenter prospective birth cohort study funded by the Ministry of Environment, Japan. The design of the active portion of the study has been published previously.^23,24 In brief, the original study included >100 000 fetuses from 15 regional centers in Japan between January 2011 and March 2014, and self-administered questionnaires were used to obtain information about the occupational exposure of parents. Detailed information regarding the mother and child was transcribed from medical records during the first trimester, at the time of delivery, and when the child was 1 month old. Postdelivery, data on the child’s health status, including any diseases, were gathered biannually via caregiver-completed questionnaires until the child reached 3 years old.

The JECS protocol was reviewed and approved by the Ministry of Environment’s Institutional Review Board for Epidemiological Studies (number 100910001) and the Ethics Committees of all participating institutions. Ethical approval for this study was an extension of the ethical approval for the JECS protocol. This study was conducted in accordance with the Declaration of Helsinki. Written informed consent was obtained from all the participants.

We used the “jecs-ta-20190930” fixed data set that was released in October 2019. The data set included prospectively collected data for children up to 3 years old as well as data on their parents. The data for 104 062 fetuses from 103 060 pregnancies were linked to the respective maternal data. These were also linked to the respective paternal data when their fathers were registered (n = 51 897). Among the 104 062 fetuses, we excluded 1636 pregnancies that resulted in miscarriage or stillbirth and another 2495 children who had missing information concerning sex or birth weight. We further excluded 2705 children with missing information concerning their mothers’ exposure to medical agents during pregnancy. The remaining 97 226 children were defined as participants. Because of the nature of the survey, the actual dropout cases were challenging to estimate precisely, as participants could resume responding in subsequent surveys after a period of nonresponse. We thus excluded infants and children lacking responses concerning neoplasms (n = 4019) from this analysis to avoid influencing the study outcomes. Ultimately, the data from 93 207 children were included in this analysis (Figure 1). Information on paternal exposure was available to almost half of the participants.

In the questionnaires, parents were asked whether or not they had handled medical agents, including anticancer drugs, ionizing radiation, and anesthetics, for over half a day at least once per month.²⁵ Exposure periods were defined as “during pregnancy (first, second, and third trimesters)” for mothers, and “3 months before awareness of partner’s pregnancy” for fathers, considering the susceptible period of spermatogenesis.^26,27 We reported an association between maternal exposure and neuroblastoma in the early cohort of the JECS.¹⁸ A separate study of ours has recently found the high levels of anticancer drug metabolites in the urine and saliva of caregivers and the detectable levels in some of the medical staff as well.²⁸ These findings prompted us to consider that the threshold in the JECS is significant for occupational exposure. Ionizing radiation includes radiation from radioactive rays, materials, and isotopes. Parental exposure to these hazardous agents for their own medical treatment, such as during the course of cancer treatment, was recorded separately. The primary outcome was leukemia or brain tumors diagnosed by community physicians during the first 3 years of life. This information was collected using questionnaires at 6, 12, 24, and 36 months old, although the exact onset was not identified. When the diagnosis was obtained via questionnaires, the physician who diagnosed the neoplasms was asked to fill in the medical certificate determined by the JECS. The certificate includes detailed data on fusion genes and/or karyotypes and was released as the “jecs-qa-20210401” fixed data set in April 2021.

Two-sided Fisher exact tests were performed with cross-tabulation tables of parental exposure and neoplasms in the offspring. Effect sizes were calculated as φ/Cramer V and r using χ² and Student t-tests for categorical and numerical variables, respectively.²⁹ Furthermore, we determined leukemia incidence rates per 100 000 population and their 95% confidence intervals (CIs) using the Clopper-Pearson method, further substantiating the identified significant associations. Because of the unavailability of the exact onset ages of neoplasms, we used a Poisson regression model instead of the Cox proportional hazards model. A complete case analysis using the Poisson regression model and robust SEs was performed.³⁰ As this analysis decreased the sample size because of the missing data, we also employed multiple imputation by chained equations (mice) in R, followed by applying a Poisson regression model to minimize potential selection bias.³¹ Covariates included the following: birth weight,³² maternal age,³³ family income, highest maternal education, maternal alcohol consumption,³⁴ and maternal smoking history³⁵ during any period of pregnancy. A directed acyclic graph was used to elucidate variable interactions (supplemental Figure 1, available on the Blood website). In the main analysis, birth weight was included as a covariate in our model, not as a confounder. This differentiation was important to avoid overadjustment and to maintain the integrity of our findings. We also conducted a subanalysis excluding birth weight from the covariate set. All statistical analyses were performed using the R software program, version 4.1.1.

A total of 93 207 children were included in the analysis, and 4019 were excluded (Table 1). Most of the characteristics of the 4019 children who were not analyzed were similar to those of the children included in the analysis (effect sizes ≤ 0.05). However, parents of these 4019 children tended to have less education and a higher rate of maternal tobacco use than those included in the analysis.

Among the cases included in the analysis, 1291 (1.4%), 2145 (2.3%), and 1005 (1.1%) mothers handled anticancer drugs, ionizing radiation, and anesthetics at work during pregnancy, respectively. By 3 years old, 36 children had developed leukemia (n = 29) or brain tumors (n = 7) (Table 2). None of the parents of the 36 children had received anticancer drugs or ionizing radiation for their own medical treatment. Four children developed leukemia of 1291 children whose mothers were exposed to anticancer drugs, resulting in an incidence rate of leukemia in this population (309.8 [95% CI, 84.5-791.4] per 100 000 population) significantly higher than that of children without maternal exposure to anticancer drugs (27.2 [95% CI, 17.6-40.2] per 100 000 population) (P = .000012). Two of the 2145 children whose mothers were exposed to ionizing radiation developed leukemia, resulting in a higher incidence rate (93.2 [95% CI, 11.3-336.4] per 100 000) than that of the children without the exposure (29.7 [95% CI, 19.5-43.1] per 100 000) (P = .0038). Similarly, 1 child of 1005 children whose mothers were exposed to anesthetics developed leukemia, resulting in a higher incidence rate of leukemia in this population (99.5 [95% CI, 2.52-553.1] per 100 000 population) than that of the children without maternal exposure to anesthetics (30.4 [95% CI, 20.2-43.9] per 100 000 population) (P = .0032).

Anticancer drugs, ionizing radiation, and anesthetics were handled occupationally by 278 (0.6%), 1457 (3.2%), and 328 (0.7%) fathers, respectively. None of the children developed leukemia (Table 3). Brain tumors were not observed in any exposed group in which either of the parents occupationally handled medical agents.

The Poisson regression model was used to analyze the association between maternal exposure to anticancer drugs or ionizing radiation and pediatric leukemia, with the following covariates: birth weight,³² maternal age,³³ family income, highest maternal education, maternal alcohol consumption,³⁴ and maternal smoking history³⁵ during any period of pregnancy. Maternal exposure to anesthetics was excluded from the statistical analysis because leukemia developed in only 1 offspring in this study. Paternal exposure to medical agents was not included in the multivariable-adjusted regression analysis because no pediatric neoplasms were observed in the offspring of these exposed fathers. In a complete case analysis, when the models including all possible predictors and interactions (full models) and models excluding interaction terms (main effect models)³⁶ were calculated and a likelihood ratio test based on the difference in deviance was performed on the goodness of fit of the 2 models, the higher fit of the full model was not evident (χ² [112] = 31.911; P = 1.00). The Akaike information criterion and Bayesian information criterion of the full model were 864.9 and 1178.0, respectively, and the Akaike information criterion and Bayesian information criterion of the main effect model were 478.8 and 491.1, respectively. Therefore, the relative goodness of fit of the main effect model was shown. In this model with a complete case analysis, only maternal exposure to anticancer drugs was indicated as a risk factor for pediatric leukemia in their offspring, with an adjusted relative risk (aRR) of 7.99 (95% CI, 1.80-35.5) (supplemental Table 1). A Poisson regression analysis with multiple imputation was also conducted, and maternal exposure to anticancer drugs was shown to have a high aRR of 7.99 (95% CI, 1.98-32.3) (Table 4). Maternal exposure to ionizing radiation was not associated with leukemia in offspring, with an aRR of 1.68 (95% CI, 0.33-8.71). In a subanalysis without birth weight from our covariate data set, the results were similar to those of the main analysis (supplemental Table 2). In a further analysis confined to mothers who were health care professionals (including physicians, dentists, veterinarians, pharmacists, public health nurses or health officers, midwives, nurses, medical technologists, and other health insurance medical workers), the findings were similar to our primary results (supplemental Table 3). The aRR and 95% CI for maternal exposure in health care provider position to anticancer drugs was 9.94 (95% CI, 1.49-66.2), and the aRR and 95% CI for exposure in health care provider position to ionizing radiation was 2.70 (95% CI, 0.41-17.8) (supplemental Table 4). These results support the findings of the present study.

Information on 19 offspring with leukemia was available, whereas the data for the other 10 were missing (Table 5). These 19 offspring consisted of 10 with ALL (9 with B-cell precursor leukemia and 1 with T-cell leukemia, including 2 infantile cases), 5 with Down syndrome–associated myeloid leukemia (DS-ML), 3 with acute myeloid leukemia, and 1 with mixed-phenotype acute leukemia (further information was missing). Five offspring with DS-ML were found only in the unexposed group, where parents tended to be older than others. In a secondary analysis excluding 5 cases of DS-ML, we found no significant difference from the original results (supplemental Table 5). No cases of infant leukemia were found in the exposed group. All 3 children in the exposed group were diagnosed with ALL between 1 and 3 years old.

This cohort study was the first to demonstrate an association between parental occupational exposure to hazardous medical agents during pregnancy and childhood cancer by 3 years old. Maternal exposure to anticancer drugs has been indicated as the sole risk factor for pediatric leukemia in offspring during early childhood. In particular, the associated leukemia might be confined to acute lymphoblastic leukemia that occurs after 1 year old. In contrast, paternal exposure to these medical agents is not associated with childhood neoplasms. This result provides a new perspective for the management of health care providers during pregnancy.

We previously reported an association between maternal occupational exposure to ionizing radiation and infantile-onset neuroblastoma in offspring as part of this large prospective birth cohort in Japan.¹⁸ In the last cohort in infancy, no obvious associations between maternal exposure and infantile-onset leukemia or brain tumors were identified. Thereafter, this cohort was expanded to 3 years old, and at that age, an association between maternal exposure to anticancer drugs and pediatric leukemia in their offspring was identified. The discordance between our previous and present findings might be ascribed to the fact that infant leukemia is a unique entity of high-risk disease different from the other types of pediatric leukemia.³⁷ All 8 cases of leukemia during infancy in the last study were included in 29 leukemias that were evaluated in the present study (supplemental Figure 2). The origin and size of these study populations were similar, and the age distribution of leukemia onset was not biased when compared with previous reports.³⁸ In the present study, newly diagnosed cases of leukemia during 1 to 3 years old were added to those of infant leukemia reported in the previous study.¹⁸ A large proportion of infant leukemia cases have unique gene fusions involving the KMT2A gene and originate from a prenatally detectable clone.³⁹ The occurrence of DS-ML is also skewed toward the first year of life, and most cases are acute megaloblastic leukemia arising from GATA1-mutated clones in newborns.⁴⁰ In the present study, infant leukemia and DS-ML were exclusively found in the nonexposed group (Table 5). Down syndrome–associated ALL was not observed in the study population. The genetic landscapes of noninfant pediatric ALL, as well as acute myeloid leukemia, are distinct from those of infant leukemia or DS-ML.^41,42 In this line, maternal occupational exposure to anticancer drugs was associated with increasing net rates of “noninfant pediatric” leukemia in offspring.

No environmental factors have been established as major contributors to the global childhood leukemia burden. Maternal exposure remains of great interest for leukemogenesis in infants and children. No significant association with maternal exposure has been found in the JECS for infant leukemia.¹⁸ However, the current study demonstrates the association between maternal occupational exposure to anticancer drugs and pediatric leukemia occurring in patients older than 1 year. Occupational exposure to antineoplastic drugs has been suggested, but not proven, to exert a potential effect on childhood cancer in offspring.¹⁸ Anticancer drugs have the potential to induce breaks in double-stranded DNA, which results in gene mutations or translocations.^43,44 No specific chromosomal abnormalities or genetic alterations were identified in the exposed group, although the total number of patients with leukemia was small (Table 5). Genomic alterations accumulate and contribute to the development of leukemia, as represented by the evolution of clonal hematopoiesis of indeterminate potential into myelodysplastic syndrome in adults. In a recent study of myelofibrosis carrying CALR mutations in adult monozygotic twins, abnormal clones that occurred in utero were shown to have the potential to evolve into malignant diseases in adulthood.^45,ETV6-RUNX1 fusion is also known to occur prenatally and is the most common genotype of B-cell precursor ALL, occurring at a peak age of 2 to 3 years old, and its preceding somatic UBA2 deletion has been identified prenatally.⁴⁶ TCF3-PBX1 fusion, reportedly generated in utero and found in ≈0.6% of healthy newborns,⁴⁷ was observed in both the exposed and unexposed groups in our study. These findings indicate that the acquisition of mutations in utero has a significant potential to drive the development of leukemic clones during the first year of life. In our study, pediatric myelodysplastic syndrome was not included, and the clonal diversity of lymphoblasts was not assessed. Disparate mechanistic insights into the leukemogenesis between infants and children with maternal exposure to medical agents need to be clarified in another prospective cohort of pediatric leukemias in offspring.

The present cohort, including offspring up to 3 years old, revealed no association between pediatric brain tumors and parental occupational exposure to medical agents. Brain tumors are the most common solid neoplasms in infants and children. The types of histopathology and the genetic landscapes involved in oncogenesis vary depending on the age of the patients. Gliomas are the most common brain tumors in the pediatric population. Pediatric-type diffuse high-grade gliomas were divided into 4 subgroups according to the World Health Organization classification 2021.⁴⁸ One of these subgroups, infant-type hemispheric glioma, is defined as having specific alterations in the ALK, ROS1, NTRK, and MET genes that are uniquely found in newborn and infantile patients.²¹ In this context, similar to leukemia, the tumorigenic process in infantile-onset brain tumors may be distinguished from that in pediatric-onset ones. However, the mature integrity of the hematopoietic system precedes the development of the central nervous system in newborns. Given the vulnerability of fetal organs to chemical agents or irradiation, further studies are needed to clarify the effect of maternal occupational exposure for the gestational age on the development of brain tumors in the offspring.

The present study showed no association between parental occupational exposures to ionizing radiation or anesthetics and neoplasms in their offspring. However, whether preconception and intrauterine exposure to low-dose irradiation are carcinogenic remains controversial.^49-54 Health care workers, such as nurses, pharmacists, and physicians, may be exposed to not only ionizing radiation but also anticancer drugs, even if they do not directly handle or administer them.^55,56 We recently reported that the concentration of cyclophosphamide metabolites in urine and saliva obtained from family caregivers of pediatric patients hospitalized with cancer is significantly higher than that in health care workers who showed the detectable ranges.²⁶ These facts suggest a potential risk of exposure to anticancer drugs by health care workers as well as family members of patients who take care of individuals in need of intensive care, such as infants or elderly individuals. Considering the preliminary results, we may need to implement numerous precautions to control exposure to health care workers as well as “caregivers” to reduce the risk of leukemia in offspring.

Several limitations of the present study warrant mention. First, the occurrence of leukemia in our cohort (31.1 per 100 000 live births) may have been higher than the annual incidence of pediatric leukemia in previous reports (2.9-3.8 per 100 000 population).^41,57,58 However, these comparisons may not be fully appropriate because of differences in age groups and the populations studied. The statistics data appeared to reflect the true incidence in the real world, although we should consider that unrecognized confounding factors, including a potential family history of cancer, might have been at play in our population. Second, differences between the children included and those not included in the analysis may have led to a selection bias. Third, we were unable to assess the exact duration and dose of exposure or potential overlaps in the exposure to different medical agents. This limitation is evident in our study, where an individual was exposed to both anticancer agents and ionizing radiation (Table 2). Fourth, even with this large data set, the number of infants with leukemia remains small. As a result, the dose dependency of maternal occupational exposure to anticancer drugs and leukemia in their offspring could not be confirmed. Fifth, almost half of the fathers’ data were missing, and the effects of paternal exposure were more inconclusive than those of maternal exposure. Sixth, the nature of the survey allowed for some participants to resume responding after a period of nonresponse, making it challenging to accurately quantify attrition bias. Although we attempted to address this issue by excluding those with total nonresponse concerning neoplasms, this method does not provide a comprehensive picture of the potential attrition bias. Seventh, the effects of cancer predisposition, such as germline mutations in ETV6 that are associated with leukemogenesis, could not be determined.⁵⁹ This cohort study, which is scheduled to continue until adolescence, will elucidate the associations between paternal exposure and pediatric malignant neoplasm according to the type and age at the onset of the diseases if we can minimize the number of dropout cases. Finally, although we identified maternal exposure to anticancer agents as a potential risk factor, it was challenging to quantify its precise contribution to leukemia risk because of the observational nature of our study, the rarity of the exposure event, the lack of previous reports for estimating the increased risk linked to this exposure, and the limitation of follow-up data to 3 years in this analysis.

In conclusion, this prospective cohort study in Japan demonstrated that maternal occupational exposure to anticancer agents may increase the risk of leukemia in children up to 3 years old. Although this association needs to be assessed in other cohort studies, the present results suggest a need for effective protection to reduce maternal exposure to anticancer drugs not only for themselves but also for their offspring.

The authors thank all participants in the Japan Environment and Children’s Study (JECS) and all staff members involved in data collection; a complete membership list appears in the supplemental Appendix.

The JECS was funded by the Ministry of the Environment of Japan.

The findings and conclusions of this study are solely the responsibility of the authors and do not represent the official views of the Japanese government.

Contribution: S.Y. and Y.K. were responsible for the study design; K.K. and S.O. were responsible for the overall coordination and conduct; M. Sanefuji, Y.S., N.H., and M.O. supported the acquisition and administration of the data sets; S.Y. and M. Sanefuji participated in the data analysis; S.Y., Y.K., M. Sanefuji, M. Suzuki, W.K., H.O., U.O., and K.N. interpreted the data; S.Y., Y.K., M. Sanefuji, and S.O. drafted the manuscript; and all the authors had access to the data, critically revised the manuscript, approved the final version of the manuscript, and were accountable for all aspects of the work.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

A complete list of the members of the Japan Environment and Children’s Study Group appears in the supplemental Appendix.

Correspondence: Yuhki Koga, Department of Pediatrics, Graduate School of Medical Sciences, Kyushu University, 3-1-1 Maidashi, Higashi-ku, Fukuoka 812-8582, Japan; email: koga.yuhki.743@m.kyushu-u.ac.jp.