In this study, the main results displayed no association between hyperemesis and cancer risk in offspring. This is reassuring news for women suffering from hyperemesis, which is the most common cause of hospital admissions in early pregnancy. However, in the age group 10–20 years we observed a significant positive association between hyperemesis and lymphoma. As we performed multiple analyses, we explored the association with stricter criteria for statistical significance. Our main finding was not significant at an alpha-level of 0.01. Further investigation on the impact of offspring’s age, revealed that only the oldest adolescents in the highest age-group were at increased risk. The potential effect of adverse perinatal exposure becomes more difficult to isolate from later environmental influences with increasing age of the offspring. This warrants caution in the interpretation of the findings. However, according to the hypothesis of fetal programming, adverse exposure in-utero may increase an individual’s vulnerability for disease in adulthood, co-acting with environmental exposures. Despite the pooling of data from Scandinavia, the numbers of cases exposed to hyperemesis was still low, limiting the ability to detect significant associations. Stratifying by birth registry, the same positive tendency regarding lymphoma risk was observed both in Sweden and Norway. In Denmark there were not enough cases to perform the analyses. Since current knowledge on the long-term consequences of hyperemesis for the offspring is limited, these findings warrant further research on the topic.
There is increasing evidence that several sub-types of hematological malignancies can originate in-utero [19,20]. Single studies have also reported increased risk of cancer in offspring following hyperemesis exposure. In the United Kingdom Childhood Cancer Study (UKCCS), a positive association was reported between severe hyperemesis and acute lymphatic leukemia and acute myeloid leukemia with an OR of 3.6 (95 % CI: 1.3-10.1). For non-Hodgkin’s lymphoma an OR of 6.8 was reported, but the association did not reach the level of statistical significance [21]. The UKCCS was based on high-quality data and specifically designed to explore perinatal risk factors for childhood cancer. However, the number of exposed cases was low, with only eight cases in total. Based on 28 exposed leukemia cases, we did not observe any such association.
How hyperemesis may increase the risk of lymphoma is a matter of speculation. Lymphoma has been linked to fetal growth and low birth weight [22]. The association of hyperemesis with low birth weight has been inconsistent [3,5,6], possibly because the maternal hunger-period is short, causing any weight loss early in pregnancy to be compensated for in the remaining weeks. Also, efficient treatment may secure fetal growth. However, the general environment in utero could still be adversely affected [11]. Previous studies exploring the effect of famine exposure confined to early pregnancy have reported negative outcomes for long-term health regardless of birth weight [11,23,24].
The underlying biology behind the programming of cancer susceptibility in-utero is unknown but is likely to involve epigenetic mechanisms. Epigenetics refer to any change to the genome which does not include alterations in the nucleotide sequence. DNA methylation and histone modification are two important epigenetic mechanisms by which the gene expression may be modified [25]. The epigenetic changes may affect different regulatory pathways such as the production of stem cells or hormones, which may alter organogenesis [13]. DNA methylations have been observed in several steps of carcinogenesis [26]. It has also been suggested that nutritional restriction may cause changes to the fetal blood circulation, sparing the brain at the expense of other organs and tissues during a “window of vulnerability” in fetal development. Altered perfusion patterns may result in long-term increased disease susceptibility.
Although the etiology of hyperemesis is unknown, several studies have suggested that elevated levels of estrogen and human chorionic gonadotropin (hCG) are important risk factors [27,28]. hCG can act as a growth factor and is associated with placental- and germ cell-cancers in particular, but subtypes of the molecule are believed to be produced in most advanced malignancies [29,30]. During pregnancy, estrogen levels are more than ten times higher than normal, and can be even higher among women with hyperemesis. Pregnancies with a female or multiple fetuses both have been associated with higher levels of estrogen as well as higher risk of hyperemesis [27,31,32]. Estrogens may be oncogenic to hematopoietic cells, and some studies have shown an association between estrogen exposure and leukemia [33]. It is not known whether the same association exists between hyperemesis and lymphoma.
As risk of breast cancer has been linked to estrogen exposure in-utero [34], offspring born to hyperemetic mothers may also be at increased risk. While we did not find such an association, we only followed offspring to age 21. Given that breast cancer has a median - age of incidence of about 60 years in western countries [35], our dataset was not appropriate for studying a possible association between in-utero exposure to hyperemesis and subsequent breast cancer. At the same time, with age it becomes more difficult to distinguish the biological from the environmental impacts on cancer risk. In addition, as the MBRs were founded in the 1960s, the majority of offspring have yet to enter higher risk age-group.
In contrast to findings of an American study of 131 men with testicular cancer, we found no association between hyperemesis and risk of testicular cancer in offspring [36]. However, this study dating back to 1979, was small and included only eight cases of testicular cancer following a hyperemetic pregnancy. To our knowledge, such an association has not been reported since. Testicular cancer in childhood is rare and differs histologically, genetically and etiologically from that observed in adolescence and adulthood [37]. The condition has been associated with both a high ponderal index and high birth weight, suggesting links between childhood testicular cancer, the intrauterine environment and fetal growth. As in the case of breast cancer, inclusion of older age-groups of cases might have yielded interesting findings. Still, it would have been difficult to isolate the effect of the intrauterine environment on risk of these cancers.
The major strength of this study is its large sample size, resulting from collaboration between the three Scandinavian countries. As well, merging population-based registries provided a relatively high number of cases making selection bias therefore unlikely and increasing the generalizability of our results. Registration of hyperemesis was performed prior to development of cancer in the offspring, which eliminated the risk of recall-bias. The MBRs offer extensive information on maternal and fetal variables making it possible to control for more potential confounders than earlier studies.
Differences in the MBRs pose potential limitations. An important example is high numbers of missing values for several variables, including maternal smoking. However, a previous study on hyperemesis and risk of short-term adverse outcomes for offspring, using the Norwegian MBR, included sub-analysis specifically exploring the possible impact of smoking on risks associated with hyperemesis, with no change in the observed associations [5]. Although different ICD-codes were used at different time-points to register hyperemesis, clinically relevant cases are most likely to have been included, regardless of ICD version. Moreover, the prevalence of hyperemesis among controls was relatively low compared to earlier estimates, possibly due to underreporting of mild hyperemesis to the birth- and patient registries, the latter requiring hospital admission. It could also reflect the lack of a general consensus regarding the correct definition of the condition.
Another possible limitation is that information on severity and onset of hyperemesis was unavailable. However, widely used definitions of hyperemesis, including the one applied by the ICD-10, pinpoints time of onset before the end of 22nd week of gestation. Previous research has shown that the majority of women experience hyperemesis during the first and second trimesters [38]. The validity of hyperemesis registration in the MBR has been explored in Norway, but not in Denmark or Sweden. In Norway, investigators reported a sensitivity and specificity of 83.9 % and 96.0 % respectively - satisfactory validity for large-scale epidemiological studies. A risk of misclassification of hyperemesis was also reported, which could result in registration of fewer cases of severe hyperemesis. This may in turn weaken observed associations of exposure and outcome [39].