Cohort study of familial viral hepatitis and risks of paediatric cancers

Abstract

Background

Although viral hepatitis causes paediatric hepatocellular carcinoma and hepatic and extrahepatic cancers in adults, there are few epidemiologic studies on paediatric-cancer risks from parental viral hepatitis. In a nationwide study in a viral hepatitis endemic region and with confirmation in another population-based sample, we examined associations between parental hepatitis B (HBV) and C (HCV) infections and risks of cancers in offspring.

Methods

We included all children born in Taiwan in 2004–2014 (N = 2 079 037) with 2160 cancer cases ascertained from the Cancer Registry. We estimated risks for paediatric cancers using Cox proportional-hazard regressions. We checked these associations in a nationwide case–control study in Denmark (6422 cases, 160 522 controls).

Results

In Taiwan, paternal HBV was related to child’s hepatoblastoma [hazard ratio (HR) = 1.77, 95% confidence interval (CI) = 1.05, 2.97] when identified at any time in the medical record, and when analyses were limited to hepatitis diagnoses occurring before the child’s birth, risks increased (HR = 2.08, 95% CI = 1.13–3.80). Paternal HCV was related to child’s non-Hodgkin lymphoma (HR = 2.06, 95% CI = 1.13–3.74). Maternal HCV was weakly related to increased risks of all childhood cancers [all types combined; HR = 1.45, 95% CI = 0.95–2.22]. The population-attributable fraction of hepatoblastoma for maternal, paternal and child HBV was 2.6%, 6.8% and 2.8%, respectively.

Conclusions

Parental HBV and HCV may be risk factors for hepatic and non-hepatic cancers in children. If associations are causal, then parental screening and treatment with antivirals may prevent some paediatric cancers.

Key Messages.

• This was the first cohort study on this topic conducted in an Asian region endemic for viral hepatitis that also looked across multiple childhood-cancer types.

• Our results indicate that maternal hepatitis C (HCV) and both paternal hepatitis B (HBV) and HCV increased the risk of cancer in offspring.

• This adds weight to the evidence promoting universal screening for viral hepatitis in prenatal and primary care, as it may reduce the transmission and risk of cancer in offspring.

Background

Prior to a 1984–1990 nationwide vaccination campaign, hepatitis B virus (HBV) was endemic in Taiwan, with 10–20% of all residents being asymptomatic carriers of hepatitis B surface antigen (HBsAg), including 5% of infants and 11% of children aged <2 years.¹ Nearly all HBV carriers in Taiwan acquired the virus prenatally or in early childhood, developing a lifelong chronic infection.^1,2 Currently, all pregnant women in Taiwan are screened for HBsAg and additionally hepatitis B e antigen (HBeAg) if HBsAg is positive. Newborns of antigen-positive mothers are vaccinated and given HBV immunoglobulin within 24 hours of birth, whereas all other newborns receive their first HBV vaccination in the first week of life. However, among the children of HBV-positive mothers, an estimated 11% still develop chronic HBV, probably due at least in part to vertical transmission in utero.³

Taiwan has traditionally also been endemic for hepatitis C virus (HCV), although the prevalence has dropped substantially in recent decades related to prevention campaigns. When the mother has HCV, vertical transmission occurs in 5.8% of pregnancies and can occur from mid-gestation to delivery, whereas post-partum transmission can occur due to infant contact with blood or bodily fluids.⁴ Seroprevalence surveys in Taiwan in the 2000s indicated that 2–3% of preschool-aged children and 4–5% of adults were seropositive for HCV antibodies.^5–7 For both HBV and HCV, there is substantial geographic and birth-cohort variation in prevalence.⁷ Family-member infection and community prevalence are previously identified risk factors for childhood HCV in Taiwan, but as the population prevalence has declined, the importance of parenteral transmission has increased.⁸ Lower socio-economic status and Taiwanese indigenous ancestry also predict higher HCV prevalence.^5,9,10

Given the high prevalence of viral hepatitis in Taiwan and other Asian countries, it is notable that Taiwan also has a higher incidence of paediatric hepatic cancers compared with Western countries.¹¹ The incidence of hepatocellular carcinoma in children declined in the years after universal HBV vaccination.¹² Yet the association between HBV and hepatoblastoma is not well characterized, and there are no etiologic epidemiology studies of HCV and paediatric cancer. Studies that examined associations between viral hepatitis and childhood cancers are mostly case series.¹³ Epidemiologic studies are largely limited by lack of statistical power or by lack of a proper control group.^14–16

Therefore, we conducted a nationwide study in Taiwan to examine associations between parental HBV and HCV, and the risks of cancers in offspring, hypothesizing that we would observe associations with hepatoblastoma. Our study encompasses a period of births after infant HBV vaccination had become widespread and when, in adults, HBV immunity through vaccination was beginning to increase.¹⁷

We attempted to corroborate these associations in a population-based case–control study of Danish children. Viral hepatitis is rare in Denmark: an estimated 0.24% of Danish residents are HBV carriers and 0.38% HCV carriers.^18,19 Unlike in Taiwan, where familial and community transmission have been relevant risk factors, in Denmark the highest prevalence is seen among high-risk groups such as intravenous drug users and healthcare workers, Greenland natives and immigrants from areas where viral hepatitis is endemic.^20–22

Methods

Taiwan

The Taiwan Maternal and Child Health Database (data available for births in the years 2004–2014, N = 2 079 037) links parents to their children and links to Taiwanese Birth Notification, the Death Registry and the Registry for Beneficiaries of the National Health Insurance Research Database. As previously described,²³ we linked this database to the Cancer Registry and ascertained all children with cancers (diagnoses 2004–2015). The present analysis includes 2160 cancer cases (diagnoses at age <11 years). We accessed all the databases from the Health and Welfare Data Science Center, a part of Taiwan’s Ministry of Health and Welfare.

In Taiwan, a universal healthcare programme began operations in 1995 that covers 99.9% of residents. Data on all enrollees’ healthcare visits are available in the National Health Insurance Research Database, including ICD-9 diagnoses of health conditions. We searched this database for HBV or HCV diagnoses [International Classification of Disease (ICD-9) (HBV: 070.2, 070.3; HCV: 070.41, 070.44, 070.51, 070.54, 070.70 and 070.71)] during the study period (2004–2015). Because in Taiwan, hepatitis transmission mostly occurs early in life,^1,2 we considered parents as having been HBV- or HCV-infected during or before the index pregnancy if they had been diagnosed at any time in the study period prior to 2015. We conducted sensitivity analyses restricting to parents with diagnoses of viral hepatitis prior to the child’s birth; a pre-pregnancy diagnosis in the medical record may indicate that the parent was experiencing adverse health effects from their viral hepatitis infection, possibly indicating more severe disease.

The mean number of prenatal visits attended by Taiwanese mothers is 9.2 [standard deviation (sd) 4.5].²⁴ Prior to 2017, HBV screening occurred at the fifth prenatal visit, at roughly the 30th week of gestation. In order to improve the accuracy of HBV diagnoses in the current analysis, we limited the main HBV analyses to children born after 30 weeks’ gestation. Secondary analyses were performed using the entire study population data. There is no standard screening for HBV or HCV in men or in children.

Although hepatocellular carcinoma is the most widely recognized malignancy resulting from chronic viral hepatitis, elevated incidence of extrahepatic cancers has been seen among adults with chronic HBV and HCV.^25–27 Thus, we examined a variety of paediatric-cancer types. We estimated the risk of cancer in offspring who had been exposed to parental HBV or HCV using Cox proportional-hazard models with t₀ at time of delivery to estimate hazard ratios (HRs) and their 95% confidence intervals (CIs). The proportional-hazards assumption was verified by testing the significance of an interaction term between covariates and log(time). We selected variables for confounding adjustment based on the literature and subject-matter knowledge.^23,28–34 We adjusted for birth year, parental age and child’s viral hepatitis status in final models. We considered additional adjustment for family income, urban or rural area of residence, parity, child’s sex, parental HIV infection, other parent’s viral hepatitis status and maternal country of birth (Taiwan/other) but the inclusion of these variables did not change effect estimates by >10% and they were left out of final models.

We additionally calculated the population-attributable fraction of hepatoblastoma due to HBV. The population-attributable fraction was calculated as [(incidence in total population) – (incidence in unexposed)]/[incidence in total population]*100%.

Denmark

This case–control study, which has been previously described,²⁹ included all children aged <20 years at cancer diagnosis and listed in the Danish Cancer Registry (diagnoses 1968–2016). Controls, alive and free of cancer at the date of diagnosis of the cases, were matched (25:1) on birth date and sex, and randomly selected from the Central Population Register. The matching ratio was selected to allow the examination of rarer exposures. These data sets were additionally linked to the Medical Births Registry for gestational information and to the National Patient Register. We ascertained diagnoses of viral hepatitis from the National Patient Register for the mother, father and child (ICD-8 070.0x; ICD-10 B15.0–19.9; O98.4, P35.3). In order to have thorough information on gestational exposures, in the parent study, we only included children who were born in Denmark. We included children in this analysis who were born after the establishment of the National Patient Register (births 1978–2013).

Greenland, like other polar regions, has an elevated prevalence of viral hepatitis.^20,35 In our parent study, we had excluded children born in Greenland, although our sample includes parents born in Greenland who immigrated to Denmark (0.29% of mothers and 0.16% of fathers in our sample).

Here we examined the risks of cancer with parental viral hepatitis (either maternal or paternal diagnosis). In Denmark, viral hepatitis is more often acquired later in life compared with in Taiwan, thus we considered parents as having viral hepatitis when the diagnoses occurred in the medical record prior to the date of the child’s cancer diagnosis. Due to underdiagnosis of HBV and HCV in Denmark,^18,19 we additionally examined risks of cancers when a parental viral hepatitis diagnosis occurred at any point in the parents’ medical record, through to the end of the study period in 2016. Fewer than five case children had a diagnosis of viral hepatitis prior to their cancer diagnosis, so we did not estimate the risk of cancer when the child had viral hepatitis.

Using conditional logistic regression, we estimated association measures relating offspring cancers to parental viral hepatitis. Given the large number of unspecified types of viral hepatitis in the data set (the virus responsible for HCV was not identified until 1989), we report all types together including type unspecified and for HBV and HCV separately. In the regressions, we adjusted for mother’s age, father’s age and number of siblings in the household (minimally adjusted model). Due to the greater association between drug use and hepatitis in Denmark, we created a fully adjusted model that accounted for any parental diagnosis of drug-use disorder prior to the child’s birth (ICD codes shown in Supplementary Table S1, available as Supplementary data at IJE online). Additional adjustment for parents’ countries of birth, parental HIV infection and maternal smoking did not change estimates by 10% and these variables were left out of final models. We conducted all analyses using SAS 9.4 (SAS Institute, Cary, NC).

The studies in Taiwan and Denmark were approved by the Human Subjects Protection boards of the University of California, Los Angeles (UCLA), the Taipei City Hospital Research Ethics Committee and the Danish Data Protection Agency.

Results

Taiwan

There was a declining prevalence of maternal viral hepatitis across the years of our study (Supplementary Table S2, available as Supplementary data at IJE online). Mothers with viral hepatitis were more often ≥35 years old at the time of the child’s birth, and a larger proportion of the case families belonged to the lowest quartile of family income and were living in small towns.

In 18.3% of families, at least one parent had HBV, and at least one had HCV in 6.9% of families, although, typically, both parents were not diagnosed. With regard to HBV, in 0.9% of families, both parents had been diagnosed, 8.0% had only the father diagnosed, 4.4% had only the mother diagnosed and 81.7% had neither ever diagnosed. With regard to HCV, in only 0.1% of families were both parents diagnosed, 1.2% had only the father diagnosed, 0.6% had only the mother diagnosed and 93.1% had neither parent ever diagnosed.

In the main analysis (limited to children born after 30 weeks’ gestation), HBV was diagnosed in 112 458 (5.4%) of non-case mothers, in 123 (5.8%) of all case mothers and in 8 (7.9%) mothers with offspring who developed hepatoblastoma (Table 1). HCV was diagnosed in 25 (1.2%) mothers whose children developed cancer in childhood and 15 158 (0.7%) non-case mothers. Maternal HBV infection was related to an increased risk of hepatic cancers but the estimates had wide CIs. No increased risks were seen for other cancer types. In sensitivity analyses restricting to mothers with hepatitis diagnoses in the medical record occurring before the child’s birth, results were similar to overall findings. Results were also similar in an analysis that included cases born at any gestational age (Supplementary Table S3, available as Supplementary data at IJE online). The risk of cancer (all types combined) in the population was increased with maternal HCV, with point estimates increased across cancer types.

Table 1.

Hazard ratios and 95% confidence intervals showing the relation between mother’s hepatitis B (HBV) and hepatitis C (HCV) diagnosis and risk of cancer in the child in Taiwan, 2004–2015

	Hepatitis diagnosis at any time in the medical record			Hepatitis diagnosis prior to child’s birth
	N (%)	Crude model HR (95% CI)	Adjusted model HR (95% CI)	N (%)	Crude model HR (95% CI)	Adjusted model HR (95% CI)
Mother with HBVa,b
Controls	112 458 (5.4)	1.00 (ref.)	1.00 (ref.)	62 279 (3.0)	1.00 (ref.)	1.00 (ref.)
All cancers	123 (5.8)	1.02 (0.85–1.22)	0.99 (0.82–1.20)	56 (2.6)	0.95 (0.73–1.24)	0.91 (0.69–1.20)
All leukaemias	43 (5.3)	0.93 (0.68–1.26)	0.95 (0.70–1.30)	17 (2.1)	0.76 (0.47–1.24)	0.79 (0.49–1.28)
Acute lymphoblastic leukaemia	32 (5.3)	0.92 (0.64–1.31)	0.95 (0.67–1.36)	12 (2.0)	0.72 (0.41–1.28)	0.75 (0.42–1.32)
Acute myeloid leukaemia	10 (6.5)	1.16 (0.61–2.21)	1.18 (0.61–2.26)	5 (3.2)	1.17 (0.48–2.86)	1.24 (0.50–3.05)
Non-Hodgkin lymphoma	16 (3.7)	0.63 (0.38–1.04)	0.65 (0.39–1.07)	5 (1.2)	0.43 (0.18–1.04)	0.38 (0.16–0.91)
Central nervous system tumours	18 (6.1)	1.09 (0.68–1.75)	1.08 (0.66–1.78)	8 (2.7)	1.02 (0.50–2.06)	1.09 (0.54–2.20)
Neuroblastoma	9 (4.0)	0.71 (0.36–1.38)	0.74 (0.38–1.45)	8 (3.1)	1.22 (0.60–2.48)	1.22 (0.60–2.48)
Retinoblastoma	9 (7.0)	1.28 (0.65–2.52)	1.34 (0.68–2.65)	<5	***d	***d
All hepatic tumours	11 (8.3)	1.54 (0.83–2.86)	1.06 (0.55–2.07)	6 (4.5)	1.62 (0.71–3.68)	1.20 (0.51–2.80)
Hepatoblastoma	8 (7.9)	1.47 (0.71–3.02)	1.18 (0.55–2.54)	5 (5.0)	1.70 (0.69–4.18)	1.40 (0.56–3.52)
All soft-tissue sarcomas	6 (4.7)	0.81 (0.36–1.84)	0.82 (0.36–1.89)	<5	***d	***d
Germ-cell tumours	11 (5.3)	0.94 (0.51–1.73)	0.63 (0.30–1.34)	6 (2.9)	1.07 (0.47–2.41)	0.58 (0.19–1.82)
Mother with HCVb,c
Controls	15 158 (0.7)	1.00 (ref.)	1.00 (ref.)	7661 (0.4)	1.00 (ref.)	1.00 (ref.)
All cancers	25 (1.2)	1.50 (1.01–2.22)	1.45 (0.95–2.22)	9 (0.4)	1.24 (0.64–2.38)	1.17 (0.58–2.34)
All leukaemias	9 (1.1)	1.42 (0.74–2.73)	1.61 (0.83–3.10)	<5	***d	***d
Acute lymphoblastic leukaemia	7 (1.1)	1.46 (0.70–3.08)	1.68 (0.80–3.54)	<5	***d	***d
Non-Hodgkin lymphoma	5 (1.1)	1.44 (0.60–3.47)	1.76 (0.71–4.15)	<5	***d	***d
Central nervous system tumours	6 (2.0)	2.64 (1.18–5.92)	2.05 (0.76–5.50)	<5	***d	***d

^aModels adjusted for birth year, parental age and child’s HBV status. During the study period, pregnancy HBV testing occurred at the fifth prenatal care visit, which occurred at the ∼30th week of gestation. Children born at ≥31 weeks’ gestation (n = 2 066 807) are included in the HBV analyses.

^bResults for cancer types with fewer than five exposed mothers not shown.

^cModels adjusted for birth year, parental age and child’s HCV status.

^dSample size was too small to estimate associations.

Paternal HBV was related to child’s risk of hepatoblastoma and weakly related to non-Hodgkin lymphoma (NHL), and HCV was related to NHL (Table 2). Among the small numbers of children who had any diagnosis of HBV, we observed increased risks of all cancers (combined) and hepatic tumours (Table 3).

Table 2.

Hazard ratios and 95% confidence intervals for the relation between father’s hepatitis B (HBV) and hepatitis C (HCV) diagnosis and risk of cancers in the child in Taiwan, 2004–2015 (N = 2 079 037)

	Hepatitis diagnosis at any time in the medical record			Hepatitis diagnosis prior to child’s birth
	N (%)	Crude model HR (95% CI)	Adjusted modela HR (95% CI)	N (%)	Crude model HR (95% CI)	Adjusted model HR (95% CI)
Father with HBVa,b
Controls	183 624 (8.8)	1.00 (ref.)	1.00 (ref.)	109 862 (5.3)	1.00 (ref.)	1.00 (ref.)
All cancers	207 (9.6)	1.05 (0.91–1.21)	1.06 (0.92–1.22)	118 (5.5)	1.12 (0.93–1.34)	1.13 (0.94–1.37)
All leukaemias	76 (9.3)	1.01 (0.80–1.28)	1.02 (0.80–1.29)	41 (5.0)	1.03 (0.75–1.40)	1.04 (0.76–1.43)
Acute lymphoblastic leukaemia	62 (10.1)	1.11 (0.85–1.44)	1.12 (0.86–1.46)	33 (5.4)	1.11 (0.78–1.58)	1.12 (0.79–1.60)
Acute myeloid leukaemia	10 (6.5)	0.70 (0.37–1.32)	0.68 (0.36–1.30)	5 (3.2)	0.64 (0.26–1.57)	0.66 (0.27–1.61)
Non-Hodgkin lymphoma	50 (11.4)	1.26 (0.94–1.69)	1.28 (0.95–1.72)	24 (5.5)	1.17 (0.77–1.76)	1.04 (0.69–1.57)
Central nervous system tumours	25 (8.5)	0.93 (0.61–1.40)	0.95 (0.63–1.44)	12 (4.1)	0.87 (0.49–1.55)	0.90 (0.51–1.62)
Neuroblastoma	20 (8.9)	0.98 (0.62–1.54)	0.98 (0.62–1.56)	13 (5.8)	1.12 (0.64–1.97)	1.09 (0.62–1.91)
Retinoblastoma	9 (7.0)	0.75 (0.38–1.47)	0.75 (0.38–1.49)	<5	***e	***e
Hepatoblastomac	17 (15.0)	1.81 (1.08–3.05)	1.77 (1.05–2.97)	12 (10.6)	2.17 (1.19–3.97)	2.08 (1.13–3.80)
All soft-tissue sarcomas	13 (9.9)	1.11 (0.62–1.97)	1.11 (0.63–1.98)	8 (6.1)	1.35 (0.66–2.77)	1.42 (0.69–2.92)
Germ-cell tumours	18 (8.6)	0.94 (0.58–1.53)	0.97 (0.60–1.58)	12 (5.7)	1.19 (0.66–2.14)	1.26 (0.70–2.27)
Father with HCVb,d
Controls	25 681 (1.2)	1.00 (ref.)	1.00 (ref.)	13 261 (0.6)	1.00 (ref.)	1.00 (ref.)
All cancers	23 (1.1)	0.81 (0.54–1.22)	0.83 (0.55–1.26)	10 (0.5)	0.77 (0.42–1.44)	0.79 (0.43–1.48)
All leukaemias	9 (1.1)	0.83 (0.43–1.60)	0.85 (0.44–1.64)	<5	***e	***e
Acute lymphoblastic leukaemia	8 (1.3)	0.98 (0.49–1.97)	1.02 (0.51–2.04)	<5	***e	***e
Non-Hodgkin lymphoma	11 (2.5)	1.88 (1.03–3.42)	2.06 (1.13–3.74)	<5	***e	***e

^aModels adjusted for birth year, parental age and child’s HBV status.

^bResults for cancer types with fewer than five exposed mothers not shown.

^cModels adjusted for birth year, parental age and child’s HCV status.

^dEvery exposed liver-cancer case was hepatoblastoma, with no other liver-cancer types having a father with HBV.

^eSample size was too small to estimate associations.

Table 3.

Hazard ratios and 95% confidence intervals for the associations between child’s hepatitis B diagnosis^a and cancer risk, Taiwan, 2004–2015 (N = 2 079 037)

	Children with HBV	Crude model	Adjusted modelb
	N (%)	HR (95% CI)	HR (95% CI)
Controls	3200 (0.2)	1.00 (ref.)	1.00 (ref.)
All cancers	12 (0.6)	3.62 (2.05–6.38)	3.85 (2.18–6.79)
All hepatic tumours	7 (4.8)	36.38 (16.99–77.89)	37.47 (17.46–80.42)
Non-hepatic cancers	5 (0.2)	1.61 (0.67–3.88)	1.72 (0.72–4.14)

^aDiagnoses occurring in medical records at any time period to 2015. HBV, hepatitis B virus. There were too few children with hepatitis C to estimate risks.

^bModels adjusted for birth year and parental age.

The population-attributable fraction of hepatoblastoma for maternal, paternal and child HBV was 2.6%, 6.8%, and 2.8%, respectively.

Denmark

Elevated risk estimates were seen for parental HBV and HCV (either maternal or paternal) based on small numbers (Table 4). When examining the cancer types occurring with parental HBV or HCV, no cancer type predominated. With HBV, cases of melanoma, germ-cell tumours, retinoblastoma, nephroblastoma, NHL, acute lymphoblastic leukaemia (ALL) and unspecified cancers occurred. With familial HCV, there were cases of germ-cell tumours, melanoma, brain, ALL, lymphoma and other carcinoma.

Table 4.

Odds ratios and 95% confidence intervals for the associations between parental viral hepatitis (either maternal or paternal)^a and risks of offspring cancer: Denmark

Either parent diagnosed with:	Cases (n = 6422)	Controls (n = 160 522)	Crude model	Adjusted Model Ab	Adjusted Model Bc	Adjusted Model Cd
	N (%)	N (%)	OR (95% CI)	OR (95% CI)	OR (95% CI)	OR (95% CI)
Any type of viral hepatitis, including unspecified	31 (0.48)	810 (0.50)	0.92 (0.63, 1.33)	0.94 (0.65, 1.36)	0.90 (0.62, 1.31)	0.90 (0.62, 1.31)
Hepatitis B	11 (0.17)	207 (0.13)	1.34 (0.73, 2.45)	1.37 (0.75, 2.51)	1.32 (0.72, 2.43)	1.35 (0.74, 2.50)
Hepatitis C	10 (0.16)	218 (0.14)	1.16 (0.62, 2.20)	1.20 (0.64, 2.27)	1.11 (0.59, 2.11)	1.11 (0.59, 2.11)

^aDiagnoses occurring in medical records prior to the date of the child’s cancer diagnosis.

^bModels adjusted for parental age and number of siblings in the household.

^cModels adjusted for parental age, number of siblings in the household and parental ICD code indicating drug use prior to or during the pregnancy.

^dModels adjusted for parental age, number of siblings in the household, parental ICD code indicating drug use prior to or during the pregnancy and child’s hepatitis status.

When we included parental diagnoses of viral hepatitis at any time in the medical record (to 2015), effect estimates were slightly higher, with overlapping CIs (Supplementary Table S4, available as Supplementary data at IJE online).

Discussion

We observed increased cancer risks in offspring with parental and child HBV and HCV. We trust our maternal HBV results due to the widespread, standard testing for HBV in pregnancy in Taiwan throughout the study period. In Denmark, universal HBV screening of all pregnant women was implemented in 2005, whereas prior to that date, selective pregnancy screening for HBV occurred only among women in high-risk groups. HBV is under-ascertained in Denmark,¹⁹ hence women with HBV would have been missed in our pre-2005 sample (small numbers precluded us limiting analyses only to children born after universal screening was introduced). We observed no risk increase for most cancers with maternal HBV, except that in the Taiwanese sample, elevated point estimates were observed for retinoblastoma and hepatoblastoma, based on nine exposed retinoblastoma and eight exposed hepatoblastoma cases. While paternal HBV was related to child’s risk for hepatoblastoma, it must be noted that there is no universal screening for HBV among Taiwanese men; likewise there is no universal screening in paediatric care. Misclassification of HBV diagnoses would be expected to have biased results to the null.

Similarly, with regard to HCV, our results must be considered in light of the fact that there is no standard HCV screening in primary care in either Taiwan or Denmark, in either children or adults. Parents with HCV in their medical record would likely have been screened due to manifestations of disease that led to testing or to being in a high-risk group. High-risk groups include having family members with HCV, behavioural or occupational risk factors (healthcare workers) or certain health conditions (hemodialysis recipients, abnormal alanine aminotransferase levels, blood transfusions, organ transplant) that prompted the HCV testing. Of these risk factors, there are few independently associated with childhood cancers, although parental employment in healthcare has sporadically been related to offspring-cancer risk.³⁶ While parents’ illicit drug use has been related to some cancer types in a small number of studies,^28,37–39 to our knowledge, no study has examined childhood-cancer risk specifically from injection drug use.

Several biologic mechanisms are possible. The parent may have transmitted HBV or HCV to the child either vertically or in early childhood, but the child’s infection was not diagnosed. HBV and HCV DNA can integrate into sperm chromosomes, inducing aberrations and altered methylation.^40–44 Cancer may have also occurred due to maternal immune dysregulation. In addition, maternal HCV has adverse impacts in pregnancy, including preterm labour, intrauterine growth retardation, low birthweight, congenital anomalies and possibly gestational diabetes.⁴ These are also risk factors for certain paediatric cancers,^23,28,32,33 particularly low birthweight and preterm birth, which are known risk factors for hepatoblastoma.^23,30

Chronic viral hepatitis contributes to hepatocellular-carcinoma risk through the direct infection and dysregulation of cells, and systemic inflammation. Overexpression of the HCV core protein can generate reactive oxygen species as well as inhibit tumour-suppressor genes including retinoblastoma 1 gene (RB1) and TP53,⁴⁵ which is notable given that both HBV and HCV alter RB1 expression.^46,47 HCV crosses the blood–brain barrier and is related to activation of brain macrophages and higher levels of pro-inflammatory cytokines in microglia cells, possibly increasing brain-cancer risk.⁴⁸

A previous study examined cancer cases diagnosed in Moscow from 1986 to 1988 (e.g. prior to Russia’s universal HBV-vaccination campaign) and reported that the child’s diagnosis with viral hepatitis was associated with both leukaemia (OR = 3.8) and soft-tissue sarcomas (OR = 5.2).¹⁶ This article did not specify the type of viral hepatitis, suggesting that they likely grouped all types of viral hepatitis together for analyses. Seroprevalence surveys occurring near the time of that study showed that HBV (4.2%) had a higher prevalence in Moscow than HCV (<1%).^49,50

According to the International Agency for Research on Cancer, HBV and HCV are established causes of hepatocellular carcinoma. With regard to NHL, HCV is an established cause, whereas HBV is a suspected cause, and risk appears stronger for specific NHL subtypes.^45,51 Apart from hepatocellular carcinoma and NHL, other cancers in adults have been infrequently studied despite suggestive associations.^25,27

There has not been compelling evidence that breastfeeding increases the risk of hepatitis transmission, although most studies had small samples or methodological limitations.⁵² Mothers with viral hepatitis are encouraged to breastfeed unless nipples are cracked or bleeding, or unless antiviral medications contraindicate it.^53,54 There have been reports that children who develop cancer were less likely to be breastfed,^28,55 perhaps due in part to early-life hospitalization, particularly for cancers diagnosed at an early age, e.g. germ-cell and embryonal tumours. Any potential protective effect from not breastfeeding would have negatively biased results.

Taiwan’s vaccination campaign for HBV began in 1984 with immunization of newborns of HBsAg-carrier mothers; it was extended to all newborns in 1986, to preschool children in 1987 and to school-age children and adults in 1988–1990. Given that we included children born in 2004–2014 and the mean age of mothers in our study was 30.2 years, the majority of mothers in our study were born in years when HBV was still endemic. The medical and registry data sources, the universal HBV testing in pregnancy and the nationwide scope are strengths of our study. Despite this, a weakness is its low statistical power due to the rarity of paediatric cancers and, increasingly, the rarity of viral hepatitis. Differences in point estimates between Taiwan and Denmark may have been due to the different age ranges of children in the studies and it is possible that early-life exposures such as viral hepatitis may be more relevant for earlier-diagnosed cancers. Another weakness is the use of medical claims to ascertain hepatitis diagnoses with no access to serologic test results. However, the adult HBV prevalence of 5.4% in our study is comparable to what has been seen in a large serosurvey of Taiwanese university entrants, which found a 9.0% prevalence among women born in 1974–1979, 6.6% among women born in 1979–1984 and 4.2% among women born in 1984–1986.⁵⁶ Although maternal alcohol use may be related to some cancers,²⁸ we could not adjust for it as this variable is not collected in Denmark and it appears underreported in the Taiwan Maternal and Child Health Database.

In summary, we observed increased risks of hepatoblastoma, NHL and all cancers in children born to parents with HBV or HCV infections. Point estimates were elevated in both of our studies, hence results may be generalizable to both Eastern and Western nations. Recent studies have suggested that the use of HCV antivirals in pregnancy may be safe for the fetus.⁴ Therefore, if the reported links turn out to be causal, implementing universal HCV screening and antiviral use in pregnant mothers and HBV/HCV screening among fathers and in offspring of infected parents may be a strategy to reduce cancer risk in offspring.

Ethics approval

The studies in Taiwan and Denmark were approved by the Human Subjects Protection boards of the University of California, Los Angeles (UCLA) (IRB#13–001904), the University of North Texas (IRB-21–469), the Taipei City Hospital Research Ethics Committee (TCHIRB-10703105-E) and the Danish Data Protection Agency.

Author contributions

J.E.H. and P.C.L. conceptualized the study, supervised the analysis, interpreted the results and drafted the manuscript. C.K.W. and X.H. analysed the data, and reviewed and edited the manuscript. K.W.C., M.T., N.F., B.R., O.A.A., C.Y.L., F.Y., J.O. and J.H. reviewed and edited the manuscript.

Data availability

Taiwanese data are owned by the Taiwan’s Ministry of Health and Welfare. Restrictions apply to the availability of these data, which were used under licence for our study. The National Health Insurance Database restricts the use of these data to Taiwanese citizens. Danish data availability is governed by EU regulations.

Supplementary data

Supplementary data are available at IJE online.

Funding

This work was supported by Alex’s Lemonade Stand Foundation [grant number 17–01882]; the US National Institutes of Health [grant numbers R21CA175959, R03ES021643]; the Taiwan Ministry of Science and Technology [grant number MOST 107–2314-B-227–009 -MY3 to P.C.L.]; and by the Taipei City Hospital [grant number 10901–62-014 to P.C.L.].

Conflict of interest

None declared.