Maternal use of hormonal contraception and risk of childhood leukaemia: a nationwide, population-based cohort study

Articles

Maternal use of hormonal contraception and risk of childhood leukaemia: a nationwide, population-based cohort study Marie Hargreave, Lina S Mørch, Klaus K Andersen, Jeanette F Winther, Kjeld Schmiegelow, Susanne K Kjaer

Summary

Background Maternal hormonal contraception has been suspected of being linked to an increased risk of childhood cancer. The aim of this study was to assess the association between maternal use of hormonal contraception and diagnosis of leukaemia in their children. Methods In this cohort study, we followed a nationwide cohort of 1 185 157 liveborn children between 1996 and 2014 listed in the Danish Medical Birth Registry and identified those diagnosed with leukaemia in the Danish Cancer Registry. Redeemed prescriptions from the Danish National Prescription Registry provided information about maternal hormonal contraceptive use, categorised as: no use (never used contraception before birth; reference category), previous use (>3 months before start of pregnancy), and recent use (≤3 months before and during pregnancy). We also calculated risk estimates separately for maternal hormonal contraceptive use during pregnancy. The primary outcome of interest was a diagnosis of any leukaemia in the children. Secondary outcomes were diagnoses of lymphoid leukaemia and non-lymphoid leukaemia. We used Cox proportional hazards models to estimate hazard ratios (HRs) with 95% CIs for risk of leukaemia in children. The Data Protection Agency registration number for this study is 2017-41-5221. Findings Between Jan 1, 1996, and Dec 31, 2014, the 1 185 157 liveborn children accumulated 11 114 290 person-years of follow-up (median 9·3 years, IQR 4·6–14·2), during which 606 children were diagnosed with leukaemia (465 with lymphoid leukaemia and 141 with non-lymphoid leukaemia). Children born to women with recent use of any type of hormonal contraception were at higher risk for any leukaemia than children of women who never used contraception (HR 1·46, 95% CI 1·09–1·96; p=0·011); and for exposure during pregancy the risk was 1·78 (0·95–3·31; p=0·070). No association was found between timing of use and risk for lymphoid leukaemia (HR 1·23, 95% CI 0·97–1·57, p=0·089, for previous use and 1·27, 0·90–1·80, p=0·167, for recent use); however, the HRs for non-lymphoid leukaemia were 2·17 (1·22–3·87; p=0·008) for recent use and 3·87 (1·48–10·15; p=0·006) for use during pregnancy. Hormonal contraception use close to or during pregnancy might have resulted in one additional case of leukaemia per about 50 000 exposed children, or 25 cases during the 9-year study period. Interpretation Our findings suggest the maternal hormonal use affects non-lymphoid leukaemia development in children. Since almost no risk factors have been established for childhood leukaemia, these findings suggest an important direction for future research into its causes and prevention. Funding The Danish Cancer Research Foundation, the Arvid Nilssons Foundation, the Gangsted Foundation, the Harboe Foundation, and the Johannes Clemmesens Foundation. Copyright © 2018 Elsevier Ltd. All rights reserved.

Introduction Cancer is a major cause of mortality in children worldwide, with leukaemia being the most common type of cancer in children.1 Some reports have indicated that the incidence of leukaemia in children is increasing.2 However, its causes still remain largely unknown.3 Sex hormones are involved in the initiation and progression of tumours in humans,4 and exposure to exogenous hormones in utero has been shown to cause cancer in children.5 The first hormonal contraceptive pill came onto the market in the USA in 19606 and in Denmark in 1966.7 Today, an estimated 136 million married or cohabiting women worldwide use pills,

Lancet Oncol 2018 Published Online September 6, 2018 http://dx.doi.org/10.1016/ S1470-2045(18)30479-0 See Online/Comment http://dx.doi.org/10.1016/ S1470-2045(18)30509-6 Virus, Lifestyle and Genes (M Hargreave PhD, L S Mørch PhD, Prof S K Kjaer DMSc), Statistics and Pharmacoepidemiology (K K Andersen PhD), and Childhood Cancer Research Group (Prof J F Winther DMSc), Danish Cancer Society Research Center, Copenhagen, Denmark; Department of Clinical Medicine, Faculty of Health, Aarhus University, Aarhus, Denmark (Prof J F Winther); Department of Paediatrics and Adolescent Medicine, Juliane Marie Centre, Rigshospitalet, Copenhagen, Denmark (Prof K Schmiegelow DMSc); Institute of Clinical Medicine, Faculty of Health Sciences, University of Copenhagen, Copenhagen, Denmark (Prof K Schmiegelow); and Department of Gynaecology, Rigshospitalet, Copenhagen, Denmark (Prof S K Kjaer) Correpondence to: Dr Marie Hargreave, Virus, Lifestyle and Genes, Danish Cancer Society Research Center, DK-2100 Copenhagen, Denmark [email protected]

implants, or injectable contraceptives to prevent pregnancy, representing 14% of women aged 15–49 years worldwide.8,9 In Denmark, this proportion increased from 19% to 26% in the period 1994–2013.9 Including use of hormonal intrauterine devices, 40% of Danish women used hormonal contraception in 2013.10 Consequently, many women use hormonal contraception up to—or even inadvertently during—pregnancy. Few studies have been published on the association between maternal use of hormonal contraception and the risk of leukaemia in offspring, and those that were published have reported inconsistent results. The first epidemiological study on this subject,11 in 1985, reported

www.thelancet.com/oncology Published online September 6, 2018 http://dx.doi.org/10.1016/S1470-2045(18)30479-0

1

Articles

Research in context Evidence before this study We systematically searched PubMed on April 1, 2018, using the keywords “maternal oral contraceptive” OR “maternal hormonal contraception” OR “maternal drug use” AND “cancer” OR “leukaemia”, AND “children” OR “offspring”. We also checked citation indices, bibliographies of the articles, and review papers in every paper retrieved to complete the search. We included only articles published in English. Unpublished studies were not considered. We thus compiled a set of epidemiological studies published between Jan 1, 1985, and April 1, 2018, on the impact of maternal use of hormonal contraceptives for developing leukaemia in offspring. In total, only seven studies on maternal hormonal contraception use and the risk of their children for leukaemia have been published, and with inconsistent results. Five studies reported increased risks in children of mothers who used hormonal contraception, while two reported no association. All studies published to date are case-control studies and based on self-reported information on hormonal contraception.

See Online for appendix

an increased risk of leukaemia in children born to women who used oral contraceptives in the year preceding pregnancy. Four subsequent studies12–15 reported increased risks in children of mothers who used hormonal contraception, whereas two16,17 reported no association. All these studies are, however, case-control studies, with potential shortcomings such as recall and selection bias, and all are based on self-reported information on hormonal contraception. We used nationwide population-based registries to follow a cohort of children born in Denmark between 1996 and 2014, to determine whether their risk of leukaemia was associated with their mothers’ use of hormonal contraception.

Methods

Study design and participants Using the Danish Medical Birth Registry,18 we identified all liveborn children in Denmark in the period from Jan 1, 1995, to Dec 31, 2014 (n=1 276 858) and their parents. Because information about hormonal contraception use was available only from 1995 onwards, we excluded children born before 1996 (n=68 910) and children for whom length of gestation was missing or implausible (ie, <157 days or >315 days; n=22  791), leaving 1 185 157 children. The study protocol was approved by the Danish Data Protection Agency (registration number: 2017-41-5221). Since the study is based on record linkage and we had no direct contact with participants, no approval from the National Committee on Health Research Ethics was needed.

Procedures The unique personal identification number (PIN), given to all residents in Denmark, allowed us to link individual2

Added value of this study To our knowledge, this is the first nationwide register-based cohort study with prospectively collected data and a median follow-up of 9·3 years (IQR 4·6–14·2), studying maternal use of hormonal contraception and the risk of their offspring for leukaemia. We found that maternal use of hormonal contraception up to or during pregnancy increases the risk of leukaemia in their children, notably non-lymphoid types of the disease. Implications of all the available evidence Since the absolute risk of childhood leukaemia remains low, the increased risk found is not a major health concern. However, our findings suggest that maternal hormonal use affects leukaemia development in children. Since almost no risk factors have been established for leukaemia, these findings suggest an important direction for future research into its cause and ultimately possible prevention.

level information about the children and their respective parents (appendix p 3). The Danish National Prescription Registry provided information about maternal pre­ scriptions for hormonal contraceptives (appendix p 4).19 We categorised exposure as: no use (women who had never used contraception; this was the reference category), previous use (>3 months before start of pregnancy), and recent use (≤3 months before start and during pregnancy), since most contraceptives are prescribed for 3 months; we altered the categorisation for the few contraceptives that were used for a different timeframe accordingly (eg, implants; appendix p 5). The primary exposure of interest was recent use. When five cases or more were available for analysis, we also calculated risk estimates for contraceptive use 0–3 months before the start and use during pregnancy. Pregnancy onset was estimated by subtracting gestational length from date of birth. Information about gestational length is accurate within 1 week for 87% of records.20 Children with leukaemia (and other cancers) were identified from the Danish Cancer Registry, which is estimated to be 95–98% complete.21,22 Genetic subtypes of acute lymphoid leukaemia were identified from the Danish Registry of Childhood Cancer.23 We obtained information on vital status (date of migration or loss to follow-up and death), year of birth, sex, origin, birth order, parental age at birth, and parental education from Statistics Denmark, which is the institution that hosts social and demographic registries. We gathered information about mode of birth, birth­ weight, gestational length, multiplicity, and maternal smoking from the Medical Birth Registry. We identified diagnoses of Down’s syndrome (International Classification of Diseases, revision 10 [ICD10] code Q90), neurofibromatosis (ICD-10 code Q85), other

www.thelancet.com/oncology Published online September 6, 2018 http://dx.doi.org/10.1016/S1470-2045(18)30479-0

Articles

congenital disorders (ICD-10 codes Q00–Q99, excluding codes Q85 and Q90, which correspond to neurofibromatosis and Down’s syndrome), and maternal infertility (ICD-8 code 628 and ICD-10 code N97) from the Danish National Patient Registry, which contains information about all contacts with Danish hospitals since 1977 and outpatient contacts since 1995.24 Before 1995, all diagnoses were reported in this registry according to ICD-8, whereas, after 1994, they were reported according to ICD-10.

We examined whether the results differed after exclusion of children with other or unknown types of leukaemia (n=16) and restriction of the analyses to acute types of leukaemia. Furthermore, we examined asso­ ciations according to immunophenotype (B-cell precursor portedly or T-cell) and to karyotypes that were pur­ frequently initiated prenatally (high-hyperdiploid and ETV6/RUNX1-positive). We did sensitivity analyses for a

Outcomes The primary outcome of interest was a diagnosis of any leukaemia in the children. Secondary outcomes were diagnoses of lymphoid leukaemia and non-lymphoid leukaemia (appendix p 6). For sensitivity analyses, further endpoints included the acute leukaemias exclusively (any acute leukaemia, acute lymphoid leukaemia, and acute myeloid leukaemia) and genetic subtypes of lymphoid leukaemia (B-cell precursor leukaemia, T-cell leukaemia, and high-hyperdiploid or ETV6/RUNX1positive leukaemia).

Statistical analysis The study cohort was followed from the date of birth until a diagnosis of leukaemia or censoring, whichever occurred first. Censoring was done on the date of death, emigration, other cancer, or on Dec 31, 2014 (end of study). We used Cox proportional hazards models to estimate hazard ratios (HRs) and 95% CIs for any type of leukaemia and for the two main types: lymphoid leukaemia and non-lymphoid leukaemia. To account for children who were siblings, we calculated the variance of the estimates by a robust variance estimate adjusting for within-cluster correlation.25 We analysed maternal use of hormonal contraception according to any type and hormone content type (oestrogen and progestin combined or progestin only) and route of administration (oral or non-oral). The age of the children was the underlying timescale, and year of birth was included a priori in all models. Other potential confounders (origin, birth order, parental education, maternal infertility, and parental age at birth) were investigated for their effect on the association between maternal use of any hormonal contraception and any leukaemia, by complete case analysis. Only maternal age and maternal infertility changed the risk estimate for leukaemia (>10%) and were therefore included in all analyses. Year of birth and parental age at birth were included as restricted cubic splines and tested for linearity; deviation from linearity was found only for year of birth, which was therefore included in 5-year categories. Sex and selected perinatal factors (mode of birth, birthweight, gestational length, multiplicity, Down’s syndrome, neurofibro­ matosis, and other congenital disorders) were included in separate models, as possible intermediate factors a priori.

No use

Previous use

Recent use

270 291 (22·8%)

778 844 (65·7%)

136 022 (11·5%)

1996–99

132 404 (49·0%)

101 356 (13·0%)

23 928 (17·6%)

2000–04

72 331 (26·8%)

214 164 (27·5%)

34 729 (25·5%)

2005–09

34 730 (12·9%)

243 467 (31·3%)

38 906 (28·6%)

2010–14

30 826 (11·4%)

219 857 (28·2%)

38 459 (28·3%)

Total Number of children Baseline characteristics Year of birth

Median (IQR)

2000 (1997–2004)

2006 (2002–2010)

2006 (2001–2010)

Sex Male

139 093 (51·5%)

398 897 (51·2%)

70 212 (51·6%)

Female

131 198 (48·5%)

379 947 (48·8%)

65 810 (48·4%)

Danish

200 354 (74·1%)

742 527 (95·3%)

128 242 (94·3%)

Descendant of immigrant

69 403 (25·7%)

35 126 (4·5%)

7564 (5·6%)

Origin

Birth order First

97 360 (36·0%)

354 275 (45·5%)

62 090 (45·7%)

172 894 (64·0%)

424 438 (54·5%)

73 903 (54·3%)

Basic

61 404 (22·7%)

142 447 (18·3%)

34 860 (25·6%)

Vocational

95 673 (35·4%)

335 102 (43·0%)

59 602 (43·8%)

Higher

72 138 (26·7%)

291 295 (37·4%)

38 712 (28·5%)

Basic

56 969 (21·1%)

144 358 (18·5%)

31 804 (23·4%)

Vocational

109 665 (40·6%)

373 634 (48·0%)

66 682 (49·0%)

70 371 (26·0%)

231 622 (29·7%)

32 112 (23·6%)

<28

72 889 (27·0%)

230 399 (29·6%)

55 991 (41·2%)

28–31

74 297 (27·5%)

263 435 (33·8%)

41 514 (30·5%)

123 012 (45·5%)

285 009 (36·6%)

38 517 (28·3%)

Second or higher Parental factors Maternal education

Paternal education

Higher Maternal age at birth (years)

> 31 Median (IQR)

31 (27–35)

30 (27–33)

29 (25–32)

Paternal age at birth (years) <28

37 563 (13·9%)

131 809 (16·9%)

34 400 (25·3%)

28–31

59 121 (21·9%)

222 885 (28·6%)

38 854 (28·6%)

166 356 (61·6%)

413 490 (53·1%)

61 078 (44·9%)

33 (30–38)

32 (29–36)

31 (27–35)

>31 Median (IQR) Maternal smoking No

184 862 (68·4%)

612 304 (78·6%)

100 637 (74·0%)

Yes

38 072 (14·1%)

121 030 (15·5%)

25 558 (18·8%)

No

243 064 (89·9%)

712 617 (91·5%)

129 965 (95·6%)

Yes

27 227 (10·1%)

66 227 (8·5%)

Maternal infertility

www.thelancet.com/oncology Published online September 6, 2018 http://dx.doi.org/10.1016/S1470-2045(18)30479-0

6057 (4·5%) (Table 1 continues on next page)

3

Articles

No use

Previous use

Recent use

(Continued from previous page) Perinatal factors Caesarean section No

221 682 (82·0%)

617 889 (79·3%)

110 842 (81·5%)

Yes

48 609 (18·0%)

160 955 (20·7%)

25 180 (18·5%)

Birthweight (g) <2500

14 427 (5·3%)

39 960 (5·1%)

6891 (5·1%)

245 359 (90·8%)

711 618 (91·4%)

124 245 (91·3%)

>4500

8699 (3·2%)

23 298 (3·0%)

4194 (3·1%)

Median (IQR)

3500 (3140–3860)

2500–4500

3510 (3160–3860)

3500 (3150–3860)

37–42 >42 Median (IQR)

17 665 (6·5%)

53 439 (6·9%)

8892 (6·5%)

251 437 (93·0%)

724 250 (93·0%)

126 801 (93·2%)

1189 (0·4%)

1155 (0·2%)

329 (0·2%)

40 (39–42)

40 (39–41)

40 (39–42)

No

257 522 (95·3%)

746 739 (95·9%)

131 237 (96·5%)

Yes

12 769 (4·7%)

32 105 (4·1%)

4785 (3·5%)

269 973 (99·9%)

778 315 (99·9%)

135 931 (99·9%)

Multiplicity

Down’s syndrome No Yes

318 (0·1%)

529 (0·1%)

No

270 146 (100·0%)

778 504 (100·0%)

Yes

145 (0·1%)

91 (0·1%)

Neurofibromatosis 340 (<0·1%)

135 948 (100·0%) 74 (0·1%)

Other congenital disorders No

239 817 (88·7%)

692 069 (88·9%)

120 866 (88·9%)

Yes

30 474 (11·3%)

86 775 (11·1%)

15 156 (11·1%)

Data are n (%) or median (IQR). Previous use refers to contraception use more than 3 months before pregnancy start. Recent use refers to use 3 months or less before pregnancy start or during pregnancy. *Frequencies within each section in each column do not always add up to the total or percentages to 100% because of missing data or, for percentages, because of rounding.

Table 1: Characteristics of the study cohort, by maternal use of hormonal contraception*

subset of children born from 1998 onwards to allow for a longer exposure window before birth and to investigate whether maternal smoking affected the results, since information on maternal smoking was available only from 1997 onwards. Exploratory analyses of three exposure windows for previous use (3–6 months, >6–12 months, and >1 year before beginning of pregnancy) and risk for any type of leukaemia were done to clarify risks according to the proximity of exposure to pregnancy. These analyses were done in children born from 1997 onwards to allow for an exposure window of more than 1 year. Additionally, we also did analyses excluding all children with Down’s syndrome and age-stratified analyses (<1 year old, 1–5 years old, 6–10 years old, and 11–18 years old) of the association between maternal use of any type of hormonal contraceptive and any type of leukaemia. We also calculated incidence rates, absolute risk differences, and numbers needed to harm. Proportional hazards tests were based on weighed residuals and fulfilled for all models. All p values were two-sided, and a 5% significance level accepted in all analyses. We did the analyses using 4

Role of the funding source The study sponsors had no role in study design, data analysis, data interpretation, writing of the report, or the decision to submit the paper for publication. The corresponding author had full access to all data in the study and had final responsibility for the decision to submit for publication.

Results

Gestational length (weeks) <37

the statistical software Stata version 11.2. The Data Protection Agency registration number for the study is 2017-41-5221.

Between Jan 1, 1996, and Dec 31, 2014, 1 185 157 liveborn children accumulated 11 114 290 person-years of followup, with a median follow-up of 9·3 years (IQR 4·6–14·2). Of the 1 185 157 liveborn children, 778 844 (65·7%) were born to mothers who had stopped using hormonal contraception more than 3 months before pregnancy start (previous use), while 136 022 (11·5%) were born to mothers who had used hormonal contraception 3 months or less before pregnancy or during pregnancy (recent use); and 270 291 (22·8%) were born to mothers who had never used hormonal contraceptives (no use). See appendix (p 7) for a detailed overview of the timing of the exposure. During the study period, a total of 606 children were diagnosed with leukaemia (465 with lymphoid leukaemia and 141 with non-lymphoid leukaemia). Table 1 shows the characteristics of the study cohort by maternal contraceptive use. In comparison with no use, recent use was more common for children born in later calendar years, of Danish origin, firstborn, with younger parental age and lower parental educational level, with mothers who smoked during pregnancy, and those whose mother had no fertility problems. The risk of the children developing any leukaemia increased progressively from previous maternal use of any type of hormonal contraceptive (HR 1·25, 95% CI 1·01–1·55; p=0·039), to recent use (1·46, 1·09–1·96; p=0·011) compared with no use (table 2). For exposure during pregnancy, the risk was 1·78 (0·95–3·31; p=0·070). Neither previous use (HR 1·23, 95% CI 0·97–1·57), recent use (1·27, 0·90–1·80), nor use during pregnancy (1·22, 0·53–2·81) was associated with lymphoid leukaemia. The risk for non-lymphoid leukaemia was increased with recent use (HR 2·17, 95% CI 1·22–3·87) and with use during pregnancy (3·87, 1·48–10·15). The risk estimate for non-lymphoid leukaemia was still significantly increased with recent use (p=0·03), when previous use was set as the reference instead of no use (exploratory analysis to investigate the risk of selection bias). Analyses of three exposure windows for previous use showed that, in comparison with no use, only the use 3–6 months before pregnancy start was associated with increased risk for any leukaemia (HR 1·38, 95% CI 1·03–1·85; p=0·031); use 6–12 months (1·22, 0·90–1·65; p=0·159) or more than 1 year (1·24, 0·96–1·60; p=0·108)

www.thelancet.com/oncology Published online September 6, 2018 http://dx.doi.org/10.1016/S1470-2045(18)30479-0

Articles

Number of children

Person-years of follow-up

Any leukaemia Number of cases

Lymphoid leukaemia

HR (95% CI)†

p value

Number of cases

HR (95% CI)†

Non-lymphoid leukaemia* p value

Number of cases

HR (95% CI)†

p value

No use

270 198

3 368 750

146

1 (ref)

··

115

1 (ref)

··

31

1 (ref)

Previous use

778 843

6 552 250

383

1·25 (1·01–1·55)

0·039

297

1·23 (0·97–1·57)

0·089

86

1·33 (0·83–2·11)

0·232

··

Recent use

136 022

1 193 290

77

1·46 (1·09–1·96)

0·011

53

1·27 (0·90–1·80)

0·167

24

2·17 (1·22–3·87)

0·008

0–3 months before pregnancy start

120 533

1 048 300

66

1·42 (1·05–1·93)

0·025

47

1·28 (0·90–1·83)

0·173

19

1·95 (1·05–3·60)

0·033

During pregnancy

15 489

144 990

11

1·78 (0·95–3·31)

0·070

6

1·22 (0·53–2·81)

0·635

5

3·87 (1·48–10·15)

0·006

Previous use refers to contraception use more than 3 months before pregnancy start. Recent use refers to use 3 months or less before pregnancy start or during pregnancy. HR=hazard ratio. *Including 16 other or unspecified leukaemias. †Adjusted for year of birth (5-year categories), maternal age, and maternal infertility.

Table 2: Relative risk of childhood leukaemia according to maternal use of any type of hormonal contraception

before pregnancy start did not show associations. Further, age-stratified analyses showed that the association between recent use and any leukaemia was strongest in children aged 6–10 years (appendix p 8). Table 3 shows the relative risks for childhood leukaemia according to maternal use of specific types of hormonal contraception. Previous and recent use of combined (oestrogen and progestin) oral products increased the risk of any type of leukaemia when compared with no use (table 3). The risk of non-lymphoid leukaemia in children of recent users of combined oral products was increased compared with children of women who never used contraception, and use during pregnancy further increased the risk (table 3). Maternal use of combined non-oral, progestin only or emergency contraception did not affect the risk of children for any type of leukaemia (table 3). No specific type of hormonal contraception was associated with the risk for lymphoid leukaemia (table 3). The results were almost unchanged when sex and perinatal factors were included in the analyses, when children with unknown or other leukaemia (n=16) were excluded, when children with Down’s syndrome (n=938) were excluded, when analyses were restricted to children born from 1998 onwards. Also, with adjustment for maternal smoking estimates were unchanged, although statistical precision for the association between recent use and any type of leukemia was lost (appendix pp 9–13). Restricting the analyses to acute leukaemias did not change the results for lymphoid or non-lymphoid leukaemias (appendix p 14). Analyses according to B-cell precursor, T-cell, or acute lymphoid leukaemia subsets with frequent prenatal initiation (ETV6-RUNX1/highhyperdiploid) showed no association with maternal hormonal contraceptive use (appendix p 15). The crude incidence of leukaemia per 100 000 personyears was 4·33 cases (95% CI 3·69–5·10) for no use of maternal hormonal contraception and 6·45 cases (5·16–8·07) for recent use, for an absolute risk difference of 2·12 cases (0·52–3·72) per 100 000 person-years. Thus, maternal use of hormonal contraception up to or during pregnancy resulted in about one additional case of

leukaemia per 47 170 exposed children. If these numbers are applied to contraceptive use in Denmark between 1996 and 2014, use of hormonal contraception resulted in about 25 additional cases of leukaemia during the median follow-up of 9·3 years (IQR 4·6–14·2).

Discussion In this large Danish nationwide cohort study, we found that maternal use of hormonal contraception up to or during pregnancy increased the risk of leukaemia in their children, especially the non-lymphoid types of leukaemia. The associations seemed to be driven by oral combination contraceptives, the most commonly used hormonal contraceptives today. Although the results of some previous studies are in accordance with our findings,11–15 whereas others are not,16,17 all previous studies were case-control studies that were potentially biased by self-reported information on contraception use. Comparison with other studies is challenging, since the window of exposure varies widely across studies. Maternal use of hormonal contraception was associated mainly with an increased risk of non-lymphoid leukaemia in their children. As the incidence of non-lymphoid leukaemia increases after the age of 6 years, it is not surprising that the association for recent use of contraceptives and any leukaemia was strongest for children aged 6–10 years. Epidemiological studies of childhood leukaemia have examined several potential risk factors, including infectious, genetic, and environmental factors. Only one environmental risk factor (ionising radiation) has so far been significantly linked with both lymphoid and non-lymphoid leukaemias.26,27 A recent review by Greaves,26 provides evidence for a multifactorial causation of childhood acute lymphoblastic leukaemia, involving two discrete steps: first, in-utero initiation, caused by an unknown ubiquitous or very common factors, since about 1% of children are born with a preleukaemic clone; and second, postnatal acquisition of secondary genetic changes, purportedly caused mainly by an abnormal immune response to common infections (ie, the delayed infection hypothesis). Although equivalent

www.thelancet.com/oncology Published online September 6, 2018 http://dx.doi.org/10.1016/S1470-2045(18)30479-0

5

Articles

Number of Person-years of children follow-up

Any leukaemia Number of cases

HR (95% CI)†

Lymphoid leukaemia p value

Number HR (95% CI)† of cases

Non-lymphoid leukaemia* p value

Number HR (95% CI)† of cases

p value

Any type 270 198

3 368 750

146

1 (ref)

Previous use

781 833

6 485 790

382

Recent use

110 013

1 036 230

69

0–3 months before pregnancy start

96 781

910 330

During pregnancy

13 232

No use

··

115

1·28 (1·03–1·58)

0·026

296

1·55 (1·14–2·10)

0·005

46

60

1·53 (1·11–2·10)

0·009

42

125 900

9

1·70 (0·86–3·35)

0·127

<5

29 996

107 350

<10

1·10 (0·52–2·32)

0·802

7

2930

14 090

<5

0·93 (0·13–6·80)

0·946

<5

1 (ref)

··

31

1·26 (0·98–1·60)

0·067

86

1·35 (0·85–2·16)

0·202

1·30 (0·91–1·86)

0·152

23

2·49 (1·39–4·45)

0·002

1·35 (0·93–1·95)

0·114

18

2·21 (1·18–4·12)

0·013

0·95 (0·35–2·59)

0·920

<10

4·58 (1·75–12·02)

0·002

1·45 (0·62–3·40)

0·393

<5

0·56 (0·14–2·28)

0·416

1·44 (0·20–10·50)

0·721

0

1 (ref)

··

Combined products Oral contraception

Non-oral contraception Previous use Recent use

··

··

Progestin-only contraception Oral contraception Previous use Recent use

77 152

494 590

25

0·88 (0·54–1·42)

0·593

16

0·75 (0·43–1·32)

0·319

9

6134

48 280

<5

1·25 (0·39–4·00)

0·703

<5

1·64 (0·51–5·24)

0·407

0

1·26 (0·47–3·42) ··

0·645 ··

Non-oral contraception Previous use

10 906

47 380

<5

1·28 (0·44–3·71)

0·654

<5

1·95 (0·66–5·75)

0·227

0

Recent use

17 129

83 440

5

0·98 (0·38–2·57)

0·975

<5

0·89 (0·27–2·93)

0·849

<5

1·12 (0·21–5·97)

··

0·898

··

131 503

929 740

62

1·14 (0·78–1·66)

0·501

44

1·09 (0·71–1·67)

0·701

18

1·31 (0·59–2·90)

0·510

1791

24 320

0

··

0

··

0

Emergency contraception Previous use Recent use

··

··

··

··

Previous use refers to contraception use more than 3 months before pregnancy start. Recent use refers to use 3 months or less before pregnancy start or during pregnancy. HR=hazard ratio. *Including 16 other or unspecified leukaemias. †Adjusted for year of birth (5-year categories), maternal age, and maternal infertility.

Table 3: Relative risk of childhood leukaemia according to maternal use of specific types of hormonal contraception

pre-leukaemic cells for non-lymphoid leukaemia have also been identified in patients with acute myeloid leukemia and healthy adults,26 Greaves surmises that non-lymphoid leukaemia is caused by other mechanisms than lymphoid leukaemia. Knowledge about risk factors for nonlymphoid leukaemia is poor because of the small sample sizes and the grouping of non-lymphoid leukaemia with lymphoid leukaemia in most studies, but associations have been found for in-utero exposure to benzene, pesticides, and antibiotics.27 This finding might point to the potentially different causes of lymphoid and nonlymphoid leukemias, where lymphoid leukaemia appears to be mainly linked to an infectious origin and nonlymphoid leukaemia to environmental risk factors. Only two other studies have investigated the risks for nonlymphoid leukaemia and hormonal contraception. In a study16 of 331 children with Down’s syndrome, no association was found between maternal use of oral contraceptives at any time before pregnancy or 2 years before pregnancy and lymphoid or myeloid leukaemia. By contrast, a recent French study12 of 747 cases of childhood leukaemia and 1421 controls reported increased risks for both lymphoid and myeloid leukaemia associated with maternal use of third-generation oral contraceptives during pregnancy.12 We found no association between 6

maternal use of hormonal contraception and risk of lymphoid leukaemia. This finding contrasts with those of four other studies11,12,14,15 reporting an increased risk of lymphoid leukaemia associated with maternal use of hormonal contraceptive. The discrepancy could be due to differences in the measurement of exposure (eg, different exposure windows and use of self-reported information), in addition to methodological limitations of case-control studies such as the inherent risk of recall bias. The increased risk for leukaemia was associated mainly with use of oral combined products containing oestrogen; we found no effect of progestin-only products. Another study looking at the risks associated with progestin-only products and combined products separately had similar results.12 We found no effect of non-oral combined products, including those containing oestrogen; however, the analyses were based on few exposed cases and therefore these results should be interpreted with caution. Sex hormones are considered to be potent carcinogens and the causal association between in-utero exposure to the oestrogen analogue diethylstilboestrol and subsequent risk for adenocarcinoma of the vagina is firmly established.28 The mechanism by which maternal use of hormones increases cancer risk in children is, however, still not clear. Hormones might affect offspring

www.thelancet.com/oncology Published online September 6, 2018 http://dx.doi.org/10.1016/S1470-2045(18)30479-0

Articles

before and after conception by inducing epigenetic changes in the follicle and ovum or zygote and embryo, or by controlling cell division, cell differentiation, and the total number of susceptible cells.29,30 We found that the risk for leukaemia increased with maternal use of hormonal contraception up to and during pregnancy, indicating that the proximity of the exposure to pregnancy is relevant to risk. The magnitude of the risk for leukaemia associated with previous use was lower. However, exploratory analyses support the importance of proximity to conception; the risk decreased and became non-significant with exposure more than 6 months before pregnancy start. Important strengths of this first cohort study to address this topic include detailed individual-level information on exposure and outcome and the fact that this information was collected prospectively and inde­pen­dently, limiting concerns about temporality, differential misclassification, and recall and selection bias. Since information on hormonal contraception is automatically transferred from pharmacies by barcodes and because the Danish Cancer Registry is virtually complete, with correct diagnoses, we consider the validity of the exposure and outcome data to be high. The nationwide populationbased design increases the generalisability of the study findings, and the large numbers of person-years (>11 million) and outcomes (606 cases of leukaemia) increased the statistical precision. No loss to follow-up and the fact that we were able to account for several potential confounders and intermediate factors are other strengths of our study. The limitations of our study include that some women might not have actually used the contraceptives prescribed or used them but at a different time, which could have led to some non-differential misclassification of exposure, possibly resulting in underestimates of effect. Further­ more, the reference group, never users, might constitute a special group of women, which would lead to selection bias. However, the risk for non-lymphoid leukaemia after recent use compared with previous use was also increased, suggesting that selection bias does not explain our results. Although we examined a large number of potential confounders, we had no information about exposure to radiation in utero. One could speculate that women who use hormonal contraception up to or during pregnancy are more likely to receive diagnostic x-rays, because they are unaware of their pregnancy, thereby causing the increased risk of leukaemia in children of mothers using hormonal contraception. However, our data do not support this hypothesis, since radiation affects the risk of not only non-lymphoid leukaemia but also lymphoid leukaemia,26,27 for which our data showed inconsistent results. We had no information on perinatal infections, which is a possible mediating factor. However, since mainly lymphoid leukaemia is expected to have an infectious origin, perinatal infections are unlikely to explain the increased risk of non-lymphoid leukaemia,

which was the main contributor to the overall risk association. Furthermore, we cannot rule out the possibility of unknown or residual confounding. However, if unknown confounding explains our results, the factor would have to be associated with use of hormonal contraception and risk of non-lymphoid leukaemia exclusively (ie, not with the risk of lymphoid leukaemia), because our data did not show consistent risks for this type of leukaemia. We have no knowledge of any factor that fulfils these criteria. The absolute increase in risk of leukaemia in our study was 2·12 cases per 100 000 years. Maternal use of hormonal contraception could have resulted in approxi­ mately 25 additional cases of leukaemia in Denmark during the median follow-up of 9 years—ie, 4% of the leukaemia cases in this study. Therefore, it is not a major concern with regard to the safety of hormonal contraceptives. Although the absolute risk is low, our findings suggest the intrauterine hormonal environment affects leukaemia development in children. Since almost no risk factors have been established for leukaemia, these findings suggest an important direction for future research into its causes and prevention. Contributors MH initiated the study and obtained the funding. MH, LSM, and SKK designed the study. MH and KKA analysed the data and MH wrote the manuscript. All the authors interpreted the results, revised the paper, and approved the final version for publication. Declaration of interests MH reports grants from the Danish Cancer Research Foundation, the Arvid Nilssons Foundation, the Gangsted Foundation, the Harboe Foundation, and the Johannes Clemmesens Foundation. KS reports speaker fees from Jazz Pharmaceuticals and Shire Pharmaceuticals. The other authors declare no competing interests. Acknowledgments The study was supported by the Danish Cancer Research Foundation, the Arvid Nilssons Foundation, the Gangsted Foundation, the Harboe Foundation, and the Johannes Clemmesens Foundation. References 1 Steliarova-Foucher E, Colombet M, Ries LAG, et al. International incidence of childhood cancer, 2001–10: a population-based registry study. Lancet Oncol 2017; 18: 719–31. 2 Steliarova-Foucher E, Stiller C, Kaatsch P, et al. Geographical patterns and time trends of cancer incidence and survival among children and adolescents in Europe since the 1970s (the ACCISproject): an epidemiological study. Lancet 2004; 364: 2097–105. 3 Wiemels J. Perspectives on the causes of childhood leukemia. Chem Biol Interact 2012; 196: 59–67. 4 Sex hormones (II). IARC monographs on the evaluation of the carcinogenic risk of chemicals to humans, Vol 21. Lyon: International Agency for Research on Cancer, 1979. 5 Herbst AL, Ulfelder H, Poskanzer DC, Longo LD. Adenocarcinoma of the vagina. Association of maternal stilbestrol therapy with tumor appearance in young women. Am J Obstet Gynecol 1999; 181: 1574–75. 6 Christin-Maitre S. History of oral contraceptive drugs and their use worldwide. Best Pract Res Clin Endocrinol Metab 2013; 27: 3–12. 7 Chakravarty D, Mortensen H. The history of Danish women. Aarhus: Systime, 2014: 161 (in Danish). 8 Alkema L, Kantorova V, Menozzi C, Biddlecom A. National, regional, and global rates and trends in contraceptive prevalence and unmet need for family planning between 1990 and 2015: a systematic and comprehensive analysis. Lancet 2013; 381: 1642–52. 9 UN. Trends in contraceptive use worldwide 2015. New York (NY): United Nations, Population Division, 2015.

www.thelancet.com/oncology Published online September 6, 2018 http://dx.doi.org/10.1016/S1470-2045(18)30479-0

7

Articles

10 Lindh I, Skjeldestad FE, Gemzell-Danielsson K, et al. Contraceptive use in the Nordic countries. Acta Obstet Gynecol Scand 2017; 96: 19–28. 11 Van Steensel-Moll HA, Valkenburg HA, Vandenbroucke JP, van Zanen GE. Are maternal fertility problems related to childhood leukaemia? Int J Epidemiol 1985; 14: 555–59. 12 Ajrouche R, Rudant J, Orsi L, et al. Maternal reproductive history, fertility treatments and folic acid supplementation in the risk of childhood acute leukemia: the ESTELLE study. Cancer Causes Control 2014; 25: 1283–93. 13 Pombo-de-Oliveira MS, Koifman S. Infant acute leukemia and maternal exposures during pregnancy. Cancer Epidemiol Biomarkers Prev 2006; 15: 2336–41. 14 Ou SX, Han D, Severson RK, et al. Birth characteristics, maternal reproductive history, hormone use during pregnancy, and risk of childhood acute lymphocytic leukemia by immunophenotype (United States). Cancer Causes Control 2002; 13: 15–25. 15 Shaw AK, Infante-Rivard C, Morrison HI. Use of medication during pregnancy and risk of childhood leukemia (Canada). Cancer Causes Control 2004; 15: 931–37. 16 Puumala SE, Ross JA, Olshan AF, Robison LL, Smith FO, Spector LG. Reproductive history, infertility treatment, and the risk of acute leukemia in children with Down syndrome: a report from the Children’s Oncology Group. Cancer 2007; 110: 2067–74. 17 Kwan ML, Metayer C, Crouse V, Buffler PA. Maternal illness and drug/medication use during the period surrounding pregnancy and risk of childhood leukemia among offspring. Am J Epidemiol 2007; 165: 27–35. 18 Knudsen LB, Olsen J. The Danish Medical Birth Registry. Dan Med Bull 1998; 45: 320–23. 19 Kildemoes HW, Sorensen HT, Hallas J. The Danish National Prescription Registry. Scand J Public Health 2011; 39: 38–41.

8

20 Kristensen J, Langhoff-Roos J, Skovgaard LT, Kristensen FB. Validation of the Danish Birth Registration. J Clin Epidemiol 1996; 49: 893–97. 21 Gjerstorff ML. The Danish Cancer Registry. Scand J Public Health 2011; 39: 42–45. 22 Storm HH, Michelsen EV, Clemmensen IH, Pihl J. The Danish Cancer Registry—history, content, quality and use. Dan Med Bull 1997; 44: 535–39. 23 Schroder H, Rechnitzer C, Wehner PS, et al. Danish Childhood Cancer Registry. Clin Epidemiol 2016; 8: 461–64. 24 Schmidt M, Schmidt SA, Sandegaard JL, Ehrenstein V, Pedersen L, Sorensen HT. The Danish National Patient Registry: a review of content, data quality, and research potential. Clin Epidemiol 2015; 7: 449–90. 25 Williams RL. A note on robust variance estimation for cluster-correlated data. Biometrics 2000; 56: 645–46. 26 Greaves M. A causal mechanism for childhood acute lymphoblastic leukaemia. Nat Rev Cancer 2018; 18: 471–84. 27 Puumala SE, Ross JA, Aplenc R, Spector LG. Epidemiology of childhood acute myeloid leukemia. Pediatr Blood Cancer 2013; 60: 728–33. 28 Diethylstilboestrol and diethylstilboestrol dipropionate. IARC monographs on the evaluation of the carcinogenic risk of chemicals to humans,Vol 21. Lyon: International Agency for Research on Cancer, 1979: 173–231. 29 Key TJ. Hormones and cancer in humans. Mut Res 1995; 333: 59–67. 30 Ellis L, Atadja PW, Johnstone RW. Epigenetics in cancer: targeting chromatin modifications. Mol Cancer Ther 2009; 8: 1409–20.

www.thelancet.com/oncology Published online September 6, 2018 http://dx.doi.org/10.1016/S1470-2045(18)30479-0