The incidence of childhood cancer is increasing worldwide (1). Leukemia and central nervous system (CNS) tumors are the most common childhood cancers and some of the leading causes of death among children in the western world (2). Survivors face increased morbidity and mortality into adulthood (3) and causes remain almost unknown. Occupational exposure to several chemicals and complex mixtures are well-documented causes of cancer in adults, and parental exposure to potential carcinogens may contribute to several different types of childhood cancer (4). The potential impact of maternal exposure might occur either preconceptionally, during gestation, or after birth either through mutations to the gametes, direct or indirect exposure of the womb, or direct transmission of substances to the child (5). Research also suggests that paternal preconceptional environmental exposures could affect the offspring’s risk of childhood cancer through epigenetic influences on sperm (6).
Exposures encountered in painting and printing industries are somewhat similar, and the International Agency for Research on Cancer (IARC) have classified them as group 1 and group 2B, respectively (7, 8).
In 2007, the IARC found positive associations between maternal exposure to painting and childhood leukemia, but this was based on a few studies (7). Further, positive associations between parental exposure to benzene, which has previously been a solvent or pollutant in some paints prior to the 1980s, has recently been associated with childhood acute myeloid leukemia (AML) (9). Somewhat inconsistent associations have been reported between occupation as a painter and in printing processes and childhood CNS cancers (4, 10–15). Thus, the small number of studies and inconsistent findings suggest further studies are warranted.
In epidemiologic studies, it is methodologically difficult to retrospectively assess prenatal occupational exposures in parents. A major common limitation in most existing studies of parental exposures and childhood cancer is the potential for recall bias when collecting exposure information after diagnosis and using proxy-informants, eg, mothers reporting on father’s exposure. Further, most studies have focused on paternal occupational exposure, but increasingly maternal exposures are being studied. The aim of this study is to examine whether perinatal paternal and maternal employment in painting or printing industries, as well as the two industries combined, is associated with childhood leukemia, CNS tumors, and prenatal cancers by using reliable nationwide register-based employment data.
Methods
Cases ascertainment and control selection
In total, 8011 children with cancer, including leukemia (N=1999), CNS cancers (N=1111) and prenatal cancers (N=2704) were identified in the Danish Cancer Registry. Prenatal cancers covered several cancers thought to have originated prenatally [ie, acute lymphoblastic leukemia, Wilms tumor (nephroblastoma), medulloblastoma, neuroblastoma, retinoblastoma, and hepatoblastoma] (16, 17). Founded in 1942, the Danish Cancer Registry contains information on name, sex, date of birth, and date of diagnosis. It also holds information on topography and morphology based on the original notification forms using a conversion system from a Danish modification of International Classification of Diseases, Revision 7 (ICD-7) before 1978, a converted edition of ICD-7 to ICD-10 until 2004, and subsequently ICD-10 afterward (18). Further, the International Classification of Childhood Cancer (ICCC-1 and ICCC-3) also classified morphology and topography.
Children aged ≤19 years and diagnosed with cancer between 1968 and 2015 were included in the study, with the requirement that they were born in Denmark in order to have access to birth-related information and parental employment history around the time of birth. Information on place of birth was retrieved from the Civil Registration System, established in 1968, thus initiating the study period, by using the unique personal 10-digit identifier (CPR number) given to all permanent residents of Denmark. Information in this registry is kept regardless of changes to the individual’s status (eg, death or emigration) providing a registry of very high quality (19). From this register, 20 controls per case were randomly chosen and matched on age and sex, who were free of cancer at the time of diagnosis of the index child.
Parental employment history
Family relationships can be ascertained by using the CPR number in the Civil Registration System and through record linkages, thus we were able to identify parents of both cases and controls. However, fathers of 0.65% of cases and 0.51% of controls could not be identified.
Information on paternal and maternal employment history was gathered by linkage to the Supplementary Pension Fund Register (in Danish: Arbejdsmarkedets Tillægspension, ATP), founded in 1964, which keeps records of mandatory pension contributions for all employees in Denmark aged 18–66 years; from 1978 also aged 16–17 years (20). This register keeps information on the name and CPR number of the employee as well as start and end dates of each employment. Information is preserved even in the event of closure of a company, retirement, emigration or death of employees. Each employer is identified through a unique 8-digit company number.
Each company is assigned an industry code (in Danish: Danmarks Statistiks Erhvervsgrupperingskode, DSE) that, until 1993, is a 5-digit Danish extension of UN International Standard Industrial Classification (ISIC) of all economic activities, revision 2, based on the economically most important activities of the company. A Danish 6-digit extension (in Danish: Dansk Branchekode, DB) of the four-digit European Union NACE classification was used from 2003–2007 and subsequently a revised version, DB-07, was introduced (20). All industry codes were transferred back to the original version, namely, DSE-77.
An expert on the working environment industry evaluated all codes, and employment was separated into painting and printing industries as classified by the DSE coding system. Painting industries included automobile paintings (95133), paint and varnish manufacture (35210), industrial painting (38196), building and home painting (50150), and paint and wallpaper shops (62131). Printing industries and corresponding to the DSE-77 code included various printing occupations (34220, 34221, 34222, 34223, 34224 and 34229). We examined parental employment in painting and printing industries separately and together. In order to capture the biologically relevant periods of preconception and the postnatal period when breastfeeding would occur, employment history for mothers were restricted to one year before and one year after birth. For fathers, the study period of interest was restricted to one year before birth to represent the period of spermatogenesis (21). Study subjects of parents without any employment during the time period of interest were excluded in order to reduce selection bias by making cases and controls more socio-economically comparable, ie, all employed. Results from initial analyses where non-working parents where included in the analyses did, however, only marginally change final results. After exclusion of mothers, 1643 (23.2%) leukemia cases, 905 (13.0%) cases of CNS cancer, and 2218 (35.0%) mothers of prenatal cancers remained in the study. The same exclusion criteria left 1736 (23.1%) fathers of leukemia cases, 956 (12.7%) fathers of CNS cancer cases, and 2362 (35.0%) fathers of prenatal cancers for analysis.
Covariates
Selection of covariates was based upon the literature (22–26) as well as our observations of associations in our data and included information on socioeconomic status (SES) at the family level, parental age at birth of index child, birth order and parity, and maternal smoking status. This information was obtained from the Civil Registration System apart from maternal smoking status, which was obtained from the Medical Birth Registry. This computerized registry was established in 1973, and midwives gathered information until 1996 when hospitals took over. This registry has undergone a major revision, which is why the accessibility of variables has changed over time (27).
SES was based on the highest level of education for either parent from the last known job title, taken from the latest available income tax form, and was categorized into six groups corresponding to the definition by the Danish Institute of Social Sciences. This categorization included academics or executive managers reflecting highest status group, higher education of intermediate duration, higher education of shorter duration, skilled work, unskilled work, and unknown (28).
Categories for covariates were defined as follows: parental age at birth of index child (≤25, 26–28, 29–33, ≥34 years), birth order (first, second, third or later), parity (1, 2, ≥ 3). Maternal smoking status was dichotomized from more detailed information (ever, never).
Statistical analyses
We used conditional logistic regression to estimate relative risks using odds ratios (OR) and their corresponding 95% confidence interval (95% CI) for paternal and maternal data separately. We considered adjustment for SES at the family level, parental age at birth of index child, birth order, and parity. The inclusion of these variables changed estimates by <10% and was therefore left out of final models.
Sensitivity analysis was performed for maternal smoking status as this variable was not available for the entire study period. Further, we examined the associations of parental employment from 1968–1980 compared to 1981–2015 to see whether the former period carried an increased risk due to different components of paints, dyes, and pigments in that period (7).
Stata version 14.2 (Stata Corp, College Station, TX, USA) was used for statistical analyses.
The Danish Data Protection Agency approved this study (No. 2008-41-2639, 2014-41-3174). No personal consent was needed as only register-based information was used.
Results
A larger proportion of children with cancer were boys than girls. Mothers of children with leukemia tended to have fewer children than mothers of healthy children. Fewer fathers of children with CNS cancer were in the oldest age category, while mothers of children with CNS cancer tended to be younger, and less likely to have unknown smoking status. Compared to control children, children with CNS cancer were more likely to be firstborn (table 1).
Table 1
Characteristics of the population.
Offspring of fathers employed in painting and printing industries (combined analysis) had a twofold increased risk for AML (table 2). Further, similar-sized OR were observed when employment in painting and printing industries were separated (table 3).
Table 2
Odds ratios (OR) for offspring of mothers and fathers with employment in painting and printing industries.
| Mothers | Fathers | |||
|---|---|---|---|---|
|
|
|
|||
| Exposed cases/ controls | OR (95% CI) | Exposed cases/ controls | OR (95% CI) | |
| Leukemia | 16/308 | 0.96 (0.58–1.61) | 36/671 | 1.03 (0.73–1.45) |
| Acute lympho-blastic leukemia | 12/246 | 0.89 (0.49–1.60) | 19/513 | 0.70 (0.44–1.12) |
| Acute myeloid leukemia | * a | 12/110 | 2.10 (1.14–3.87) | |
| Central nervous system cancer | 12/116 | 1.97 (1.08–3.61) | 26/343 | 1.54 (1.02–2.31) |
| Astrocytoma | 6/67 | 1.79 (0.76–4.17) | 13/191 | 1.38 (0.78–2.46) |
| Medulloblastoma | * a | 7/78 | 1.63 (0.74–3.56) | |
| Prenatal cancer b | 19/391 | 0.89 (0.56–1.41) | 43/893 | 0.92 (0.67–1.25) |
Table 3
Odds ratios (OR) for offspring of mothers and fathers with employment in painting versus printing industries.
| Mothers | Fathers | |||||||
|---|---|---|---|---|---|---|---|---|
|
|
|
|||||||
| Painting industries | Printing industries | Painting industries | Printing industries | |||||
|
|
|
|
|
|||||
| Exposed cases / controls | OR (95% CI) | Exposed cases / controls | OR (95% CI) | Exposed cases / controls | OR (95% CI) | Exposed cases / controls | OR (95% CI) | |
| Leukemia | 9/156 | 1.05 (0.53–2.08) | 7/153 | 0.83 (0.39–1.79) | 29/416 | 1.34 (0.91–1.96) | 8/255 | 0.60 (0.29–1.21) |
| Acute lymphoblastic leukemia | 6/128 | 0.83 (0.36–1.89) | 6/119 | 0.92 (0.40–2.10) | 17/325 | 1.00 (0.61–1.64) | * a | |
| Acute myeloid leukemia | * a | * a | 8/69 | 2.26 (1.07–4.80) | 5/41 | 2.43 (0.94–6.23) | ||
| Central nervous system cancer | 8/59 | 2.34 (1.10–4.95) | * a | 13/219 | 1.20 (0.68–2.12) | 13/124 | 2.09 (1.17–3.75) | |
| Astrocytoma | *a | * a | 7/120 | 1.20 (0.55–2.60) | 6/71 | 1.64 (0.70–3.83) | ||
| Prenatal cancer b | 12/205 | 1.03 (0.57–1.86) | 7/187 | 0.69 (0.32–1.47) | 30/546 | 1.05 (0.72–1.52) | 13/347 | 0.72 (0.41–1.25) |
We observed no increased risk for leukemia but did find a significant twofold increased relative risk for CNS in children of mothers employed in combined painting and printing industries (table 2).
A significantly increased relative risk of CNS cancers in offspring was seen for fathers in the combined industries (table 2) and for the subgroup employed in the printing industries (table 3).
Regarding CNS cancers, we found an increased relative risk of CNS for mothers working in painting industries, while an analysis was not feasible for printing industries due to the low number of exposed cases (table 3).
Analyses of cancers with a prenatal origin were not associated with increased relative risk for either parent (table 3).
In sensitivity analyses, results showed maternal smoking status was not associated with higher risk of childhood leukemia or CNS cancers. Further, we found that employment from 1968–1980 was not associated with higher relative risk of childhood leukemia or CNS cancer compared to employment from 1980–2015 (table 4), with similar point estimates and overlapping CI.
Table 4
Odds ratios (OR) for offspring born before and after 1980 of mothers and fathers with employment in painting and printing industries.
| Mothers | Fathers | |||||||
|---|---|---|---|---|---|---|---|---|
|
|
|
|||||||
| Born before 1980 | Born after 1980 | Born before 1980 | Born after 1980 | |||||
|
|
|
|
|
|||||
| Exposed cases / controls | OR (95% CI) | Exposed cases / controls | OR (95% CI) | Exposed cases / controls | OR (95% CI) | Exposed cases / controls | OR (95% CI) | |
| Leukemia | 7/100 | 1.15 (0.53–2.52) | 9/208 | 0.83 (0.42–1.64) | 16/278 | 1.05 (0.63–1.75) | 20/393 | 1.01 (0.64–1.59) |
| Central nervous system cancer | * a | 9/86 | 1.96 (0.97–3.93) | 12/146 | 1.65 (0.90–3.01) | 14/ 197 | 1.45 (0.83–2.52) | |
For covariates used in analyses, we did not find that these factors changed results in crude compared to adjusted analysis.
Discussion
We conducted a nationwide nested case–control study of childhood leukemia and CNS cancers in relation to parental employment in painting and printing industries, using objective register-based data showing that both paternal and maternal employment in these industries increased the risk of CNS tumors. Further, fathers had a two-fold increased risk of AML after employment in both painting and printing industries, but no increased risk was observed in the offspring of mothers working in these industries. While our sample size was limited for specific cancer subtypes, our study was among the first to rely upon prospectively collected employment records, with a study design that included population-based controls and thus was not subject to recall biases or selective participation.
Around half of all childhood cancers are suggested as having emerged in utero (16), and our finding may be related to epigenetic changes, which are suggested as a potential underlying biological mechanism for certain childhood cancers (29). These proposed mechanisms may differ for mothers and fathers. Over a lifetime, maternal exposures can accumulate in the fixed number of ova present at birth. Exposures can be transferred directly to the child through the placenta, or via lactation (30). For fathers, genetic changes are thought to be transferred in the sperm, and morphological changes in sperm have been seen after use of paint stripper, but not paint, but whether this has a clinical impact on the offspring is not yet known (31). Thus, preconceptional exposure could be a plausible biological mechanism. As prenatal cancers were not associated with increased risks in our study, perhaps postnatal exposure is of more relevance, but extension of the study period for fathers to also include one year after birth led to minor changes to most results.
Paternal employment as a painter and risk of childhood AML was investigated in Bailey et al’s meta- and pooled analysis (32) and a subsequent study (33), neither of which supported our positive association. For CNS cancers, the largest available study investigating risk used proportional mortality rate in offspring of fathers employed as painters. The study used information from death records to assess exposure and did not find a positive association (34). The use of death certificates may introduce bias in part because survival rates were lower during the study period than current rates. Yet, several smaller studies, where exposure was established mainly by interviews, often found increased OR and wide CI due to the small number of exposed cases (10, 14, 35–38). For printing industries, the largest study of 103 exposed fathers did not find an increased risk (13), and several smaller studies demonstrated inconsistent results (10, 12, 15, 36, 38). Two of the studies have grouped several hydrocarbon-related occupations (36, 38) and risk estimates are thus not restricted to occupation in our selected industries, which might obscure a possible positive effect. While interview-based studies may contain more detailed information on exposure than register-based studies, these studies may be subject to selective participation.
In opposition to the IARC’s evaluation on maternal employment as a painter, we did not find an increased relative risk of leukemia cancers in offspring. However, the evaluation was based on four studies investigating maternal exposure, including one study on household painting. One study did not mention the number of exposed cases, while the remaining three included 53–160 exposed case mothers. To our knowledge, no studies have had a sufficient number of exposed cases to conduct analyses of the potential effects of maternal employment as a painter in relation to CNS cancers. This may be explained by the lower frequency of women of the childbearing age in the workforce in many other countries and of women working in this industry in particular. Since the 1960s, Danish women have had a high rate of employment, and, in 2015, 70.1% of women worked outside the home (39) with around 1% employed in the industries of interest in this study. This represents among the higher female employment rates internationally.
Parental leave in Denmark, and therefore exposure time, changed across the study period, which might have induced exposure misclassification. Additionally, job adjustment or absence due to working conditions not being acceptable during pregnancy or pregnancy-related sickness could also influence exposure encountered during pregnancy; for example, a mother who was not physically able to paint might be assigned a different job task during her pregnancy, however, this would not change her job industry, and therefore she would still be listed as a painter and thereby exposed in our study. No registries directly record what proportion of mothers are reassigned during pregnancy, but estimates vary from 35 to 48% (38). Hence, during the later study period (2002 and onwards) maternal early postnatal exposure would have been rare; however, excluding the postnatal period changed risk estimates only minimally. Thus, the maternal exposures that we examined were likely to represent first and second trimester and possibly early third trimester exposures; this is also presumably the most vulnerable period for the fetus due to the timing of organogenesis. For fathers, we expect the biologically important period to be preconception. Since paternity leave is of short duration, we do not expect this to have affected our results.
Both industries involve complex exposures to several chemicals, including in particular volatile organic solvents, pigments and dyes, which makes separating between specific exposures impossible (7, 8). Parental exposure to volatile organic solvents has previously been associated with childhood leukemia and CNS cancers (40, 41). Moreover, household exposure to paints has previously been shown to be associated with specific cytogenic subtypes of all children (42). Thus, many different components could potentially impact relative risk estimates.
Further, we initially created two exposure periods to reflect changes in exposure before 1980 compared to after. Yet only minor changes in risk estimates were found, and further exploration of potential changes was not feasible due to the relatively few exposed cases and controls. Identification of individual tasks performed was not possible using register-based data.
Despite including all nationwide cases over a 47-year period, limitations of this study also include the relatively low number of exposed cases due to the disease being rare. Even though employment rates remained stable throughout the period, a relatively small proportion of parents were employed in these industries. Additionally, not all potential confounding factors were available for study for the entire study period. Further, our study used pension fund records to determine exposure thereby eliminating recall bias, but possibly introducing misclassification bias as a uniform exposure at the industrial level is assumed. Also, as industry codes are used for exposure classification, a few white-collar workers might be misclassified as exposed to paints and printing. This may have led to an underestimation of potential effects of parental employment in relation to the selected childhood cancers. We did not compare exposed cases to other blue-collar workers not exposed to solvents as controls, as this could have introduced unpredictable confounding from unknown exposures potentially associated with childhood cancer, eg, dusts, air pollution, social contacts, pesticides, etc.
Major strengths of the study are use of national register-based data and the objective and prospective assessment of employment history (20), thereby eliminating participation and recall bias. In addition, the registries used have a high degree of completeness and are generally of high quality (18–20, 27). The personal identifier allows correct linkages between the registers and excludes counting the same case more than once, and further secures access to various potential confounders.
Concluding remarks
Our results showed an increased risk for CNS cancers in offspring of mothers and fathers employed in painting as well as painting and printing industries. For fathers, also a positive association was present between AML and employment in both painting and printing industries. In contrast to the IARC’s evaluation on maternal employment as a painter, we did not find an increased relative risk of leukemia cancers in offspring.
