Introduction

The prevalence of pregnancy-associated cancers is estimated at 1 in 1000 to 1 in 1500 of pregnancies [1–4]. The most common cancer forms affecting pregnant women are cervical cancers, breast cancers, melanomas, lymphomas, and leukaemias [1]. Generally, cancers diagnosed during or shortly after pregnancy, within one or two years after childbirth are considered to be pregnancy-associated cancers [5–8]; but, long periods can pass between an exposure initiating the growth of an abnormal cell, invasion of the cancer, symptoms becoming apparent, and a diagnosis and treatment being made. The minimum latency periods to clinical diagnoses for cancers in adults, were determined by the World Trade Center Health Program to be 0.4 years for lymphoproliferative and haematopoietic cancers and 4 years for solid tumours (excluding mesotheliomas [11 years] and thyroid cancers [2.5 years]) [9]. Additionally, a study using data on over 1.6 million cancers from the National Cancer Institute’s Surveillance, Epidemiology, and End Results (SEER) program concluded that most types of cancer progressed for 10 years or more before being detected [10]. It is also possible that some cancers have even longer induction and latency times than these. Therefore, cancers diagnosed postpartum may have been initiated, but remained undetected, during pregnancy, possibly affecting the developing fetus. Pregnancy-associated cancers, in this paper, are defined as all cancers diagnosed from 2 years prior to pregnancy and up to 50 years of age in the mothers.

The risk of congenital anomalies has been suggested to be greater in children born to mothers who were diagnosed with a pregnancy-associated cancer [11], including lymphomas [7] and breast cancer [12]; however, the results have not been consistent [5–7;13;14] and more studies are warranted. Three of these studies, looking at different types of cancers, were carried out using Danish data [5–7], considering cancers diagnosed up to two years after pregnancy. Further Danish studies have found that cancers (specifically, lymphomas and leukaemias) were more common among parents who had had a child born with a congenital anomaly [15] and that cancers and birth defects tend to cluster in families [16]. A few mechanisms through which pregnancy-associated cancers may increase the risk of congenital anomalies in the offspring have been suggested, however knowledge in this area is limited. Firstly, for cancers that are diagnosed during the gestational period, treatment regimens or psychological stress could impact risk of congenital anomalies [17–19]. Secondly, cancer development presents the body with metabolic [20;21], hormonal [21] and nutritional challenges [21–24], which could affect the growing conceptus. Further, it has been suggested that hyperthermia is more common among cancer patients [25], which may impact risk of congenital anomalies [6;26;27]. If the presence of a cancer coincides with pregnancy, some of these fetal challenges could have occurred even if a diagnosis was not received for several years after the pregnancy. Alternatively, there may be shared environmental or genetic factors influencing development of cancer in the mother and congenital anomalies in the offspring [15;16;21]. Even, if these exposures were present shortly before or during pregnancy, due to the induction and latency periods, the cancer may not be diagnosed for a number of years.

Since results from previous studies investigating the risk of congenital anomalies in children of mothers with pregnancy-associated cancer, with or without treatment during pregnancy, have been mixed [5–7;11;13;14], further research is needed on this topic, in particular using a larger cohort with longer follow-up time.

This study thus aims to investigate the association between maternal cancer and congenital anomalies in the offspring. In our study, exposure to maternal cancer is indicated by a diagnosis of cancer before the age of 50 years. Timing of the cancer diagnosis, in relation to the in utero period, will be considered in an attempt to separate the side effects of treatment from other fetal challenges related to the cancer itself. We hypothesised that there is a higher risk of congenital malformations among children whose mothers are diagnosed with cancer before or during pregnancy due to treatments and psychological stress; additionally, we hypothesised that there are increased risks among children whose mothers are diagnosed in the years soon after pregnancy, as these cancers would be most likely to have been present during the pregnancy.

Methods

Results

The comprised of 2,123,228 children and 2.6% of the children (n = 56,016) were considered exposed (their mother received a diagnosis of cancer from between two years prior to pregnancy up to 50 years of age). There were 95,815 children (4.5%) who received any congenital anomaly diagnosis during their first year of life. Extending follow-up to cover the entire follow-up period (therefore including some more minor anomalies) resulted in 204,378 children (9.6%) with congenital anomalies. Of the 95,815 children with a congenital anomaly registered before 1 year of age, 9,408 (9.8%) had a sibling who also had a congenital anomaly registered prior 1 year of age; and 9,568 (10.0%) had more than one anomaly registered in their first year.

Table 1 displays the characteristics of the study population. Fewer exposed children were seen in the lowest tertile for maternal income; however there were more missing data among the exposed, especially for socioeconomic variables.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Descriptive statistics of the study population by exposure status.

https://doi.org/10.1371/journal.pone.0173355.t001

The associations between maternal cancer and congenital anomalies in the offspring, diagnosed before 1 year of age, are shown in Table 2. When we compared the exposed group to the unexposed group, we observed an adjusted hazard ratio (HR) of 1.04 (95% confidence interval [CI]: 1.00, 1.09). The additional adjustment for socioeconomic variables resulted in a lower HR of 1.02 (95% CI: 0.98, 1.07). When considering timing of cancer diagnosis in relation to the in utero period, we observed a higher risk of congenital anomalies in children exposed to cancers diagnosed before or during pregnancy (HR: 1.37, 95% CI: 1.07, 1.75). Splitting this further to look at the period 2 years pre-pregnancy and the pregnancy period separately, we found HRs of 1.41 (95% CI 1.03, 1.93) and 1.33 respectively (95% CI 0.91, 1.94) (data not displayed). When considering exposure during the postpartum period, a statistically significant HR was seen only for children who were exposed to cancers diagnosed more than 10 years postpartum (HR: 1.06, 95% CI: 1.01, 1.16).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Hazard ratios (HR) for all congenital anomalies diagnosed before 1 year of age according to exposure status.

https://doi.org/10.1371/journal.pone.0173355.t002

Including all congenital anomalies diagnosed throughout the entire follow-up period resulted in 204,378 children with diagnoses (9.6% of the study population) and gave a HR of 1.03 (95% CI: 0.99, 1.05) (data not displayed). The sensitivity analysis using 45 years of age as the maximum cut off for maternal diagnosis of cancer resulted in a similar HR of 1.05 (95% CI: 1.00, 1.10) (data not displayed).

Using paternal cancer diagnosed during the same window as the exposure gave HR results close to that of the main analysis (HR: 1.08, 95% CI: 1.02, 1.14). We further assessed timing of cancer diagnosis in relation to the in utero period and we observed a similar pattern for HRs with regards to timing: pre-pregnancy or during pregnancy HR 1.26 (95% CI: 1.02, 1.55); in the 5 years postpartum HR 0.97 (95% CI: 0.86, 1.09); >5 years to 10 years postpartum HR 1.14 (95% CI: 1.04, 1.26); >10 years postpartum HR 1.07 (95% CI: 0.99, 1.16); or in multiple time points HR 1.07 (95% CI: 0.61, 1.87).

In Table 3 results are shown for specific groups of congenital anomalies among the exposed during the entire follow-up period compared with the unexposed. The largest point estimates were seen for congenital anomalies of the musculoskeletal system (HR: 1.13, 95% CI: 1.02, 1.25) and congenital anomalies of the respiratory system (HR: 1.11, 95% CI: 0.88, 1.41),. Among the exposed, the most common congenital anomalies in these groups were congenital anomalies of the feet (39%) and of the larynx, trachea and bronchus (67%) respectively. The lowest point estimates were seen for congenital anomalies of the nervous system (HR: 0.80, 95% CI: 0.61, 1.04), orofacial clefts (HR: 0.82, 95% CI: 0.55, 1.23) and congenital anomalies of the genital organs (HR: 0.87, 95% CI: 0.68, 1.1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Hazard ratios (HR) for congenital anomaly groupings according to exposure.

https://doi.org/10.1371/journal.pone.0173355.t003

Table 4 displays results for all congenital malformations, according to the type of cancer the mother was diagnosed with. Statistical significance was reached for children whose mothers were diagnosed with leukaemia (HR: 1.31, 95% CI: 1.01, 1.69). The largest, insignificant point estimates were seen for cancer of the lip, oral cavity or pharynx (HR: 1.15, 95% CI: 0.83, 1.59) and cancer of the respiratory or intrathoracic organs (HR: 1.18, 95% CI: 0.99, 1.39). Among cases, the most common type of lip, oral or pharynx cancer they were exposed to was cancer of the tonsil (24%) and the most common type of respiratory or intrathoracic cancer they were exposed to was cancer of the bronchus and lung (88%). Largest, but statistically insignificant negative HRs were seen for cancer or the bone or articular cartilage (HR: 0.71, 95% CI 0.29, 1.76) and cancer of the endocrine glands (HR: 0.83, 95% CI: 0.66, 1.04).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Hazard ratios (HR) for any congenital anomaly diagnosis before 1 year of age according to type of cancer.

https://doi.org/10.1371/journal.pone.0173355.t004

Discussion

Maternal cancer diagnosis in the two years before pregnancy and the years after pregnancy was associated with a slightly higher risk of congenital anomalies in children. The highest risk was seen children of mothers who were diagnosed before or during pregnancy, with smaller risk estimates in those whose mothers were diagnosed within 5 years postpartum or after 10 years postpartum. Similar associations were seen for paternal cancers which indicate that maternal cancer does not influence risk of offspring congenital anomalies, and genetic factors or other time stable factors, like socio-economic factors, play a role instead.

Immortal observation time bias and reverse causation should also be considered. To be diagnosed up to the age of 50 years, the mother needs to stay alive. Some of the cancers may have been diagnosed following increased monitoring because of the presence of a congenital anomaly in the offspring (although it could be argued that this is unlikely). The first type of bias may lead to an underestimation of effects, and the second type to an overestimation; but both types of bias would only have small effects.

While the majority of previous studies have not reported an association [7;13;14], some studies have suggested an increase in prevalence of congenital anomalies among offspring of mothers diagnosed with pregnancy-associated cancers [7;11;12]. Similar to our study, positive associations have been seen for all cancers [11],breast cancers [5] and lymphomas [12]. The comparisons that can be drawn between our results and those of previous studies are, however, limited due to limited power and differences in the exposure definitions: we were unable to distinguish between major and minor congenital anomalies, and other studies have limited the exposure window to pregnancy or from conception up to two years postpartum.

The associations seen with cancers diagnosed during pregnancy and the pre-pregnancy period (the time period in which we observed the highest HR) could be due to cancer treatments or stress following a diagnosis, rather than the cancer itself. Cancers diagnosed soon after pregnancy would have been more likely to be initiated before or during gestation than cancers initiated many years after; and, therefore, larger HRs would have been expected in the earlier postpartum years if the cancer itself and its metabolic consequences had an effect on offspring congenital anomaly risk. There was not, however, a clear pattern with regards to risk and timing of maternal cancer diagnosis.

The similar pattern of results seen when paternal cancer was substituted as the exposure suggests that any increase in congenital anomalies in the offspring is not specific to cancers in the mother; rather, it could be an indication that uncontrolled genetic or family-related confounding factors are causing the observed association. If a true association exists between maternal cancer and also paternal cancer, and congenital anomalies in the children, this lends support to the theory of shared genetic and pre-pregnancy environmental factors, as suggested by Zhu et al [15]. An alternative mechanism we discussed was the presence and development of maternal cancer resulting in nutritional and metabolic challenges which fetal morphogenesis or organogenesis; in this scenario a positive association would be unexpected in the analysis on paternal cancer.

The main strength of this study is the use of complete data from the Danish nationwide register system. The registers cover the whole of Denmark. This provides data on the entire population, with minimal selection bias. In the study of specific types of congenital anomalies or types of maternal cancers, numbers are small, resulting in a lack of power, which can be seen from the wide confidence intervals for some of our analyses. Cancer data are considered to be of high quality and completeness [28]. Data on congenital anomalies in the Danish National Patient Register is considered to be 90% complete; a level considered by some to be acceptable for some types of research [30]. Completeness would be expected to be poorer in the earlier years of the data and for minor congenital anomalies, especially if they are not visible. We were unable to distinguish between major and minor congenital anomalies in our data, however restricting the main analyses to congenital anomalies diagnosed in the first year of life will most likely limit these analyses to more major anomalies.

Although the inclusion of the entire population of Denmark minimised selection bias, the study only includes information on live births. As major congenital anomalies are associated with early fetal loss, stillbirth and termination [31–33], it is possible that bias occurs due to conditioning on live births [34]. Bias due to differences in terminations between the exposed and unexposed groups would be more likely to be an issue when considering children of mothers who received a diagnosis during pregnancy or the pre-pregnancy period, who may be more likely to terminate a pregnancy due to concerns about a fetotoxic treatment. It is less likely that there would be a link between postpartum diagnoses and termination. Bias due to early fetal loss and stillbirth could occur when considering any exposure category, if losses or prenatal screening for congenital anomalies are more common among mothers with cancer at any stage of development.

Another limitation of the study is the fact that it conditions on the future. Although maternal cancer susceptibility will exist and a tumour may already be initiated before a child's birth, the exposure status of the child is based on a future event: diagnosis of a maternal cancer. It is unlikely that cancer would be differentially diagnosed among mothers who have a child with a congenital malformation. However, in order for a diagnosis to be made and a child considered exposed, mothers have to survive long enough to receive a cancer diagnosis, not dying of alternative causes before this point.

There is potential for misclassification of exposure status due to incomplete exposure follow-up. Children were considered exposed if their mother was diagnosed with cancer from two years before conception, up until the mother reached 50 years of age. However, not all mothers in the cohort reached 50 years before December 31st 2010, the final date cancer data were available until, and some children will therefore have been incorrectly classified as unexposed in some analyses. However, the sensitivity analysis, reducing the maximum maternal age to 45 years, provided similar results.

There was some suggestion of an increased prevalence of congenital anomalies among children of mothers who were diagnosed with cancer between 2 years prior to pregnancy and in the years after pregnancy (up to 50 years of age). However, the finding of a positive association between paternal cancer and congenital anomalies suggest that genetic factors or environmental factors, in either the mother or the father, prior to conception influence risk. The presence of a developing maternal cancer whilst the offspring is in utero may not greatly alter congenital anomaly risk, at least for those who survive until birth. Our results should be interpreted in light of the limitations of the study. Additionally, it should be considered that a large number of analyses were carried out and some chance findings cannot be ruled out. With a larger sample size, it may be interesting to investigate associations between more specific types of cancer and specific congenital malformations to further consider some of our results.