Preterm delivery among female survivors of childhood, adolescent and young adulthood cancer
Article first published online: 6 JAN 2010
Copyright © 2010 UICC
International Journal of Cancer
Volume 127, Issue 7, pages 1669–1679, 1 October 2010
How to Cite
Madanat-Harjuoja, L.-M., Malila, N., Lähteenmäki, P. M., Boice, J. D., Gissler, M. and Dyba, T. (2010), Preterm delivery among female survivors of childhood, adolescent and young adulthood cancer. Int. J. Cancer, 127: 1669–1679. doi: 10.1002/ijc.25157
- Issue published online: 4 AUG 2010
- Article first published online: 6 JAN 2010
- Manuscript Accepted: 23 DEC 2009
- Manuscript Received: 14 OCT 2009
- National Institutes of Health
- National Cancer Institute through Vanderbilt University (VU). Grant Number: 1 R01 CA104666
- Finnish Cancer Organizations
- cancer survivors;
- late effects
We studied the deliveries of female cancer survivors and female siblings in a population-based setting in Finland. Nationwide cancer and birth registries were merged to identify 1,309 first postdiagnosis deliveries of early-onset (diagnosed under age 35) female patients with cancer and 5,916 first deliveries of female siblings occurring in 1987–2006. Multiple logistic regression models were used to estimate risk of preterm (<37 weeks), low birth weight (<2500 g) and small-for-gestational-age deliveries. The risk of preterm delivery among cancer survivors compared with siblings was overall increased [odds ratio (OR) 1.46, 95% confidence interval (CI) 1.14–1.85], the increase in risk being visible in all diagnostic age groups. Risk of low birth weight (LBW) was also significantly increased (OR 1.68; 95% CI 1.29–2.18) but not after adjustment for duration of pregnancy (OR 1.11; 95% CI 0.76–1.64). Neither was the risk of small-for-gestational-age (SGA) increased. The risk of preterm delivery was most pronounced in survivors delivering 10 years or more after diagnosis. Site-specific analyses indicated that survivors of germ cell tumors and central nervous system (CNS) tumors were at increased risk of preterm delivery, although numbers were small. In childhood survivors, kidney tumors formed the main cause of preterm delivery. Pediatric, adolescent and young adult cancer survivors are at risk for preterm delivery. Heightened surveillance is recommended especially for Wilms', germ cell and CNS tumor survivors. Such adverse pregnancy outcomes can occur a decade or more after cancer diagnosis, indicating a continued need for obstetric awareness, surveillance and counseling in former patients with cancer.
With the improvement of diagnostics and treatment, over the past decades, more patients with early-onset cancer survive. According to some estimates, 5-year survival rates have reached up to 81% among pediatric patients (aged 0–14 years at diagnosis) and 87% among adolescent and young adult (aged 15–24 years) patients.1 Thus, parenthood is possible for an expanding number of cancer survivors.2, 3 This raises questions regarding the potential effects of cancer and its treatments on pregnancy, neonatal outcomes and health of offspring.
Cancer survivors face anxiety and fears relating to reduction in overall fertility, possible pregnancy risks, and health effects in offspring.4 Many previous studies on the health of offspring are hospital based and focus on a limited subgroup of patients with cancer.5, 6 Moreover, most studies are limited to the offspring of pediatric cancer survivors.7, 8
Cancer survivors represent the largest group of people of reproductive age exposed to a wide range of ionizing radiation doses to the gonads as well as to genotoxic chemotherapeutic agents.9 Although some adverse reproductive outcomes (e.g., stillbirths, spontaneous abortions and cytogenetic abnormalities) in female cancer survivors are hypothesised to result from germ cell toxicity caused by mutagenic exposures, others (e.g., preterm delivery and LBW) have been shown to be linked to direct radiation-induced damage to the vasculature and elastic properties of the uterus incurred by abdomino-pelvic radiation.10
One multi-institutional study on female childhood cancer survivors showed patients to be at an increased risk for preterm delivery and LBW in a manner directly related to dose to the uterus from radiotherapy.10 Other population-based studies in different countries have confirmed this result.8, 11–14 A Norwegian study12 examined parenthood among male and female cancer survivors aged 15–45 years at diagnosis and perinatal health of their offspring. This study found both the risk of preterm delivery and LBW of offspring to be increased among female cancer survivors compared with the risk among the general population retrieved from the Medical Birth Registry (MBR). A Scottish study11 identified all postdiagnosis first deliveries of former female patients with cancer aged 35 years or less at diagnosis. In addition to an increased risk of preterm delivery, the study found that cancer survivors were at an increased risk of operative delivery compared with a randomly selected comparison population.
An early study of primarily Wilms' tumour survivors collected information on pregnancy outcomes by a postal survey and found an increased proportion of adverse pregnancy outcomes (spontaneous abortions and LBW) among abdominally irradiated patients compared with unexposed patients diagnosed with the same malignancy.15 Other similar studies have published consistent reports.6, 16–18 Recent studies on larger cohorts of childhood cancer survivors have found an increased risk of adverse pregnancy outcomes in patients treated with radiotherapy for a wide range of malignancies.8, 10, 13, 14 In a questionnaire-based study in Ontario, Canada, Chiarelli et al.14 reported that in addition to preterm birth and LBW, infants born to patients who had received abdominal-pelvic irradiation were also at higher risk of perinatal death compared with offspring of those treated with surgery only. Signorello et al.10 found an increased risk of preterm delivery and LBW among children of patients treated with high-dose radiotherapy to the uterus (>500 cGy) compared with the children of survivors who did not receive radiotherapy. Mueller et al.,13 using a registry-based approach, reported an increased risk of preterm delivery and LBW not only in irradiated patients but also in those receiving chemotherapy only. Reulen et al.,8 in a population-based questionnaire study in the United Kingdom, found female survivors exposed to abdominal irradiation to be at a 3-fold risk of delivering prematurely, a 2-fold risk of delivery of a LBW infant and a slightly increased risk of miscarriage. Green et al.19 explored, in addition to LBW, the risk of various other pregnancy outcomes, including live births, stillbirths, miscarriages, abortions after treatment with radiotherapy and a wide range of chemotherapeutic agents. Although no increased risk of adverse pregnancy outcomes was associated with any chemotherapeutic agent, risk of LBW was associated with pelvic irradiation. In Denmark, Winther et al.20 reported a significant increased risk of spontaneous abortions associated with high-dose radiotherapy to the ovaries and uterus among female survivors of childhood cancer.
Using a nationwide population-based approach in Finland, we explored whether the risk of adverse pregnancy outcomes was increased among female cancer survivors diagnosed in childhood (0–14 years), adolescence (15–19 years) or early adulthood (20–34 years) in Finland. Using a registry-based approach for obtaining pregnancy information and obstetric histories of patients with cancer and a comparison group of siblings, our specific goals were to estimate the risk of preterm birth, LBW and restricted fetal growth in early-onset cancer survivors (groups that until recently are not often included in health outcome research) and to evaluate the effect of time from treatment to delivery.
Material and Methods
Each individual living in Finland has a unique personal identification number (PIN) since the late 1960s, which enables merging of data for individuals alive in 1969 or born thereafter. This retrospective cohort analysis used data derived from linkage between the population-based, nationwide Finnish Cancer Registry (FCR), Central Population Registry (CPR) and the MBR.
The FCR began systematic registration in 1953 and data are almost complete (100% for solid tumors, >90% for hematologic malignancies, and 100% for childhood cancers).21 In the FCR, information on treatment, when available, is based on clinical notifications and includes data on radiotherapy, chemotherapy and surgery. Details of the anatomical region irradiated, dose or in the case of chemotherapy dose and agent administered are not included. Details on coded treatment data are as follows: palliative/radical/radicality unknown and during the first 4 months/after the first 4 months/time point unknown.
The CPR, found in 1969, is nationwide and includes the name and former names, PIN, municipality of birth and residence, date of emigration or date of death of all Finnish citizens and permanent residents in Finland. The MBR contains, from 1987 onward, data on all mothers giving birth and on all children born in Finland. All live- and stillbirths at a birth weight of at least 500 g or a gestational age of at least 22 weeks are included in the registry. The data are received from the hospitals of delivery or from the midwife or physician assisting in the delivery. Data for <0.5% of infants are missing in the MBR,22 but information on these cases with missing MBR data is routinely collected within the CPR.
Preterm delivery was defined as a birth occurring at <37 full weeks (<37 weeks) of gestation and early preterm delivery as <34 weeks (<34 weeks) of gestation. In the MBR, the best clinical estimate of gestational age at birth is reported in weeks and days, but in this study, only full weeks were used. A LBW infant was defined as a neonate weighing <2,500 g at birth. SGA was defined as having a birth weight on the −2SD (standard deviation) curve or below compared with infants of the same sex born during the same gestational week using national birth weight statistics.23 Information on the potential confounding variables of maternal age, maternal smoking, maternal hypertension, maternal infection, maternal diabetes, pre-eclampsia, placental problems, delivery year, child sex, malpresentation, caesarean delivery and use of artificial reproductive technology were available from the MBR for survivors and siblings alike.
The cancer survivor cohort was identified from the records of the FCR. All patients diagnosed with a malignant neoplasm between January 1, 1953, and December 31, 2004, and aged 0–34 years at diagnosis were identified.3 By linkage to the CPR, female full- and half-siblings of the cancer patients were identified. By further linkage to the MBR, we identified all offspring of female patients with cancer and of female siblings. For both patients and siblings, only singleton live births occurring from 1987 to the end of 2006 were included. In the study, the majority of patients, 97.4%, were diagnosed with cancer after 1970 and 85.5% after 1980. The MBR contains information on the birth order of the child in the family for mothers registered since 1987, even for births occurring before 1987. Children born before 1987 are excluded because details of birth weight, gestational age and other characteristics could not be obtained; however, birth order of all children is known for both survivors and siblings and is adjusted for in analyses including all postdiagnosis offspring for patients and all offspring for siblings. Moreover, for patients, only births occurring at least 9 months after the parent's diagnosis were included to exclude those women whose cancer was diagnosed during pregnancy and offspring born before diagnosis.
As parity is expected to have an influence on birth weight and primiparity is a known risk factor for preterm delivery,24 only first deliveries of cancer survivors and siblings were included in the main analyses, thus eliminating any influence of previous obstetric history on the end points studied. Furthermore, because twin and triplet deliveries are strongly associated with the outcomes studied, only singleton deliveries were included.
Multiple logistic regression modeling was used to calculate odds rations (ORs) as measures of relative risk for the dichotomous outcomes of preterm birth, LBW and SGA among first live-born offspring. Additional analyses including stillbirths were also performed. Because data were available on a large number of exposures during pregnancy that are potential risk factors for an adverse pregnancy outcome and therefore possible confounders (Table 1), the log-likelihood ratio test was used to identify those explanatory variables to be included in the final model. Despite the a priori decision to include maternal infection, maternal diabetes or impaired glucose tolerance, pre-eclampsia and decade of diagnosis, in our data, these did not prove to have an effect on the outcomes studied. Models for assessing LBW were also adjusted for full gestational weeks at delivery as a continuous variable. Maternal age, year of delivery and socioeconomic status were defined as categorical variables. All other variables adjusted for were dichotomous.
By combining the available information on the site of the tumor and whether the initial treatment included radiotherapy, survivors were classified into 4 mutually exclusive groups: no radiotherapy, abdomino-pelvic radiation, cranial radiation, radiotherapy other than to the brain or the abdomino-pelvic region. To study the possible effect of treatment other than radiotherapy, a separate analysis of the patients who had not received ionizing radiation as part of their therapeutic exposure resulted in the following groups: chemotherapy with or without surgery and surgery only.
For the preterm birth outcomes, analyses including all postdiagnosis singleton births of patients with cancer and all singleton births of siblings were also performed. In addition to the previously mentioned explanatory variables, these models were also adjusted for birth order, previous history of an early preterm delivery at <34 weeks, previous history of stillbirths, and spontaneous or induced abortions. Because more than 1 pregnancy per subject were included in these analyses, conditional fixed-effects logit models25 were applied to take into account the dependent nature of the data for children born to the same subject.
Defining socioeconomic status for young women is difficult, because many of them are not in the labour market, therefore data on socioeconomic status were missing in >30% of cases. We checked models adjusting for this variable and because the results did not differ materially, the final models are presented without adjustment for socioeconomic status.
In a separate analysis, female patients were also matched to their female siblings and conditional logistic regression modeling was used to produce ORs for the main outcomes. The results were in agreement with those using nonmatched data, although, most ORs lost their significance because of reduced sample size in matched analyses, i.e., matching limited the sample size to 12% (392 patients and 504 sibling contributed to the 392 matched sets) of the entire data available. Results are presented for the nonmatched data.
Checking for possible interactions between the variables in each model was based on the likelihood ratio test. All interactions among the variables in the final model were checked and none were found to be significant. Estimates of model parameters and 95% CI were computed by the maximum likelihood technique.
Analyses on first born children were conducted on the entire data and separately on the subcohort that excluded the 2.6% diagnosed before 1970. Because no major differences were observed, final analyses included only deliveries of mothers diagnosed after 1970. Because the MBR came into existence in 1987, we recognize that not all first born children are included in these analyses.
Descriptive characteristics of pregnancies of patients and siblings are shown in Table 1. A total of 9,079 women with first deliveries were identified (Fig. 1). Of these, 2,880 were cancer survivors and 6,199 were siblings. After excluding multiple deliveries, pregnancies that resulted in a stillbirth and those occurring before or within 9 months of a patient's diagnosis, 1,309 deliveries of patients and 5,916 deliveries of siblings, with information on possible explanatory variables, were eligible for the analyses.
Of the eligible cancer survivors, 297 (22.7%) were diagnosed with cancer in childhood (0–14 years), 249 (19.0%) in adolescence (15–19 years) and 763 (58.3%) in young adulthood (20–34 years). The distribution of primary site of cancer by age at diagnosis of women experiencing a first postdiagnosis delivery is shown in Table 2. Most childhood cancer survivors were leukemia patients, whereas malignant epithelial neoplasm patients formed the majority of adolescent and young adulthood cancer survivors.
Overall, the risk of preterm delivery and early preterm delivery was increased among female cancer survivors compared with female siblings (Table 3). This increase was still significant after adjustment for the main confounders. The crude risk of delivering a LBW infant was significantly increased but not after adjustment for duration of pregnancy. The risk for delivery of a small-for-gestational age infant was not increased among female cancer survivors.
In analyses grouped by age at cancer diagnosis, mothers diagnosed in childhood or as young adults were at a significantly increased risk for preterm delivery, this being more pronounced among survivors of pediatric cancer (Table 4). Among survivors of adolescent cancer, the risk for delivering at <37 weeks was also increased, although not significantly (Table 4). Risk of having an early preterm delivery was increased in all cancer diagnostic age groups; after adjustment, the risk was significantly increased only among childhood cancer survivors. A significantly increased risk of LBW was observed among pediatric and young adult cancer survivors in crude analyses. After adjusting for gestational age, a nonsignificant increase in risk among childhood survivors remained, whereas the association disappeared for young adults. Results of SGA risk did not vary by diagnostic age group (Table 4).
It seemed that time from diagnosis to delivery was an important determinant of risk for preterm delivery and LBW (Table 5). In pediatric and young adult cancer survivors delivering >10 years from diagnosis, the risk of preterm delivery was doubled compared with siblings. It is noteworthy, however, that in this subgroup only 8 of 22 of pediatric patients had received radiotherapy, whereas the equivalent proportion among young adult cancer survivors delivering prematurely was 8 of 10. Among pediatric and adolescent survivors, risk of early preterm delivery was increased among those delivering later than 10 years from diagnosis, though significantly only among pediatric patients. Similarly, the risk of delivering a LBW infant was increased for those pediatric and young adult cancer survivors delivering at 10 or more years after diagnosis, though nonsignificantly (Table 5).
Despite small numbers in most cancer sites (Table 6), risk of preterm delivery was significantly increased among survivors of kidney tumors (OR 5.50, 95% CI 2.39–12.64) and germ cell tumors (OR 2.94, 95% CI 1.38–6.25). Also, risk of early preterm delivery was increased in survivors of brain and CNS tumors (OR 2.67, 95% CI 1.04–6.87) as well as kidney tumors (OR 9.31, 95% CI 2.93–29.57). The risk of LBW was increased, though not significantly, in survivors of kidney tumors (OR 2.74, 95% CI 0.63–12.03). All kidney tumor patients delivering prematurely were diagnosed in childhood with Wilms' tumors, whereas 3 of 9 germ cell tumor survivors with a preterm delivery were diagnosed in adolescence and 6 in adulthood. Early preterm delivery occurred in a total of 5 CNS tumor survivors, 2 of which were diagnosed in childhood, 2 in adolescence and 1 in adulthood.
We also performed the analyses excluding the above mentioned high risk diagnostic subgroups 1 at a time. The risk of preterm delivery among pediatric cancer survivors was no longer significantly increased after excluding Wilms' tumor patients (OR 1.26, 95% CI 0.76–2.08). The same was true for adulthood survivors after excluding the high risk subgroup of germ cell tumor patients (OR 1.29, 95% CI 0.94–1.77).
Overall, patients receiving radiotherapy treatment were at an increased risk of preterm delivery compared with siblings (OR 1.49; 95% CI 1.03–2.16) because 36 of 434 survivors in this treatment subgroup delivered at <37 weeks (data not shown). Abdomino-pelvic irradiation increased the risk of preterm delivery because 13 of 72 survivors treated in this way delivered prematurely (OR 3.81; 95% CI 2.02–7.19). Cranial irradiation and radiation directed at other sites were not associated with an increased risk of the outcomes studied. The increase in risk of preterm delivery was visible in 7 of 37 pediatric (OR 4.01, 95% CI 1.71–9.40) and 3 of 14 adolescent cancer survivors (OR 5.44, 95% CI 1.45–20.48) who had received abdomino-pelvic irradiation, but not significantly in adult patients with cancer receiving the same exposure (3 of 21; OR 2.44, 95% CI 0.67–8.92). Interestingly, however, in the young adults age group, risk of preterm delivery was significantly increased among 37 of 452 patients whose treatment regimens did not include radiotherapy (OR 1.49; 95% CI 1.03–2.15).
Among patients who did not receive radiotherapy, chemotherapy was associated with a significantly increased risk of preterm delivery because 19 of 155 receiving chemotherapy had a preterm delivery (OR 2.42, 95% CI 1.45–4.05). Of 598 patients receiving surgery alone, however, only 43 had a preterm delivery and were not at a significantly increased risk of preterm delivery (OR 1.33, 95% CI 0.95–1.87).
Although overall results did not change substantially, in analyses including stillbirths, the risk of early preterm delivery was significantly increased among young adulthood cancer survivors, and specifically among those delivering within 10 years of diagnosis. It is noteworthy that of the 5 stillbirths among cancer survivors, 4 were preterm deliveries occurring at <34 weeks (3 among young adult survivors and 1 in an adolescent cancer survivor).
After matching patients with their biological half- or full siblings, the risk of preterm delivery was still significantly increased among former patients with cancer.
For the outcomes of preterm and early preterm deliveries, additional analyses including all postdiagnosis deliveries resulted in similar results as those including only the first postdiagnosis delivery (data not shown).
Overall, previous history of cancer places females at an increased risk for preterm delivery.10–12 In our study, cancer survivors had a 50% increased risk of delivering before 37 full weeks of gestation. Age at cancer diagnosis was an important determinant of this risk; pediatric patients had a 62% increased risk and young adults a 36% increased risk compared with the sibling comparison group. Furthermore, we found the risk of preterm delivery to be high also in survivors delivering a long time from cancer diagnosis in all age groups. Results were similar when all postdiagnosis deliveries were included in analyses and adjustment for birth order was made.
A novel finding in our study was that the risk of preterm delivery was also increased among adolescent and young adulthood cancer survivors, and not associated solely with radiotherapy exposure in young adults. Previously, most reports have included survivors of cancer in childhood or in a restricted subgroup of survivors.6, 26 Two recent studies, however, extended the age range of cancer survivors by including patients aged 15–35 years12 and 0–43 years11 at diagnosis, but results for young adults were not reported separately. In our study, adult cancer survivors were not likely to deliver LBW infants after adjusting for gestational age.
A second finding, not previously reported, was that deliveries occurring >10 years after diagnosis were more likely to be preterm. Radiation induced fibroatrophy is a late effect of radiotherapy, which may take years to develop.27 This may explain the finding, because the elasticity of uterus is more likely to be restricted by fibroatrophic changes a decade or more after treatment. The extent to which radiotherapy explains the increased risk observed with increasing time lag from diagnosis to first delivery is not entirely clear because although the majority of young adult patients delivering prematurely had received radiotherapy, pediatric patients with the outcome were mainly nonirradiated according to FCR data. Another possible explanation could be the differential effect of age on obstetric risk factors. Although adjustment for age at delivery takes into account the observed higher age of patients at first delivery (Table 1), the possibility that increased maternal age poses a higher obstetric risk for patients than for siblings cannot be dismissed. This is implied by previous studies showing that cancer survivors suffer a wide range of metabolic problems over time.28 Another contributing factor may be that women in this subgroup, due to higher maternal age, experience more problems with conception and achieving pregnancy, possibly requiring more assistance from reproductive technologies, which as such have been associated with the outcomes studied.29
A third notable finding in our study relates to the site-specific risk estimates. We found an increased risk of preterm delivery among mothers who survived a germ cell tumour and a tumour of the CNS. With most CNS cancer survivors, the amount of scatter radiation from the treatments to the head and neck region to the reproductive organs cannot be considered substantial to cause meaningful uterine damage, and thus a possible effect on the hypothalamo-pituitary axis may contribute to the premature initiation of labor. In addition, in some CNS tumors, administered spinal irradiation may be responsible for direct uterine effects. Chance associated with small numbers, however, may play a role because a similar association was not seen in a study of pregnancies among childhood cancer survivors.10
Among the 56 cancer survivors with germ cell malignancies (45 with ovarian cancer), all 7 mothers with preterm deliveries received chemotherapy and none was irradiated. Furthermore, the overall result of an increased risk of preterm delivery among survivors receiving chemotherapy is consistent with this observation. The underlying pathophysiology for preterm delivery in this patient group remains unknown, but possible effects of treatments other than radiotherapy cannot be dismissed nor a possible effect of the malignancy being treated. Risk of preterm delivery was also increased among survivors of Wilms' tumors, a result in agreement with previous findings,6, 16 and possibly due to the development of fibrosis after pelvic irradiation which restricts the growth and elastic potential of the uterus.30
Pediatric patients had a high risk of both preterm and early preterm delivery. Most of the increased risk was explained by Wilms' tumor patients. All 25 Wilms tumor patients in our study were diagnosed under the age of 8 years and were most likely prepubertal when treated. Pediatric cancer survivors also had an increased risk of delivery earlier than 34 weeks, and there was a nonsignificant increased risk of LBW infants even after adjustment for duration of pregnancy. However, there was no increase in the risk of delivering an SGA infant, using the internationally accepted definition of SGA.31
Results of analyses by treatment were generally in agreement with previous studies,8, 10, 13, 14 as risk of preterm delivery was increased among pediatric and adolescent patients receiving abdomino-pelvic irradiation. In young adults, however, risk of preterm delivery was significantly increased in patients not receiving radiotherapy. Although previous results on the association of chemotherapy and adverse pregnancy outcomes in pediatric patients are not entirely clear (but generally negative),13, 19 the possibility of a differential effect on the adult reproductive axis or the possible influence of the adult cancers e.g. of the ovary, being treated may explain our finding. As this diagnostic age group has been overlooked in the past, further studies including information on chemotherapeutic agent and dose administered are needed to confirm this result.
Previous studies have reported significantly increased risks of preterm delivery and LBW ranging from 1.3 to 3.6 and 1.3 to 2.1, respectively (Table 7). Overall our result on preterm delivery among pediatric patients was in agreement with those of previous studies.8, 10, 13, 14, 19 Other than chance, several methodological differences may explain small differences in risk estimates. Pediatric patients in the previous studies were aged <21 years at diagnosis, whereas our definition of the age group was restricted to patients diagnosed at <15 years of age. Comparison groups varied, and majority of studies used the general population.8, 11, 13 However, in 1 study the risks were compared with those patients treated with nonsterilizing surgery.14
Two studies based on data from the Childhood Cancer Survivor Study (CCSS) used siblings as a comparison group and are thus closest in study design.10, 19 In the CCSS, recruitment of siblings was based on participation of the survivor, as a random sample of CCSS participants was asked permission to contact their nearest-age full siblings. Furthermore, information on deliveries of both patients and a sample of siblings was self-reported in the form of questionnaires. Our study subjects were identified from population-based national registries, and we had access to information on all patients and siblings, producing a ratio of cancer survivors to siblings of about 1:4 (about 4:1 in the CCSS32). Our data were then less susceptible to biases associated with participation and response, although births before 1987 were excluded. Our data were also not influenced by recall or reporting bias as they were not based on self-reporting; instead information on deliveries was obtained from a nationwide registry which receives data directly from delivery hospitals. Information from national registries allowed reliable access to the important confounders, such as smoking and placental problems, both of which are well established risk factors for preterm delivery.33, 34 Although some of these variables (e.g. hypertension) may seem to be in the causal pathway for the outcomes of interest, a significant difference in risks remained in our study even after adjustment.
Familial factors did not explain the observed increase in risk for preterm delivery as even after matching patients with their biological half- and full siblings, the risk of preterm delivery was significantly increased among former patients with cancer.
Socioeconomic position has previously been shown to influence risk of preterm delivery.35 In our study results from models adjusting for socioeconomic status did not, however, differ materially from the results of nonadjusted models. Although final models did not adjust for socioeconomic position this cannot be considered to be a substantial source of bias as socioeconomic differences in perinatal health have been shown to be small and are diminishing in Finland.36 Other factors not easily accounted for are anxiety and depression, both of which have been associated with increased risk of preterm delivery37, 38 and former cancer survivors are known to suffer more psychosocial problems than the general population.39 Surveillance bias should also be considered; history of cancer may influence obstetric decisions and may place these individuals under increased surveillance.
In conclusion, our study indicates that women diagnosed with cancer in adolescence and young adulthood as well as in childhood were at increased risk for preterm delivery. Our findings underscore the necessity for continued prenatal follow-up of all former female patients with cancer as late as a decade or more after their diagnosis.
The article's contents are solely the responsibility of the authors and do not necessarily represent the official views of the National Institutes of Health, National Cancer Institute and VU. The authors are indebted to Lisa B. Signorello for her contributions to the manuscript.
- 25Fixed effects regression models. Newbury Park, CA: Sage, 2009..