Cancer during pregnancy is a rare condition that occurs in 1 of 1000 pregnancies.1 However, because age is the most significant risk factor for cancer,2 the rate of cancer during pregnancy may be increasing as more women delay childbearing. Reproductive cancers (cancer of vulva, vagina, cervix uteri, uterine corpus, ovary, and fallopian tube) accounted for 19% of the estimated 5.1 million new cancer cases in 2002 worldwide.3 In the United States, an estimated 83,750 new cases of gynecologic malignancies (cancers of the uterus, ovaries, cervix, fallopian tubes, vagina, and vulva) occur per year, with a mortality rate of more than 27,000 women per year.4 In 2002, the incidence rate of breast cancer (which is defined as gynecologic cancer in this study) in the United States was 124.9 cases per 100,000.5
The diagnostic and therapeutic management of the pregnant mother with cancer brings to the fore considerations between the best possible maternal treatment and fetal safety. Although treatment modalities and timing should be individualized, health professionals are challenged to offer optimal maternal therapy while ensuring fetal well-being.6–10 Information on impact of reproductive cancers during or before pregnancy on birth outcomes is often contradictory and evidence on the real influence of gynecologic cancers on birth outcomes is weak.11 Hence, health professionals may have difficulty with providing evidence-based education to parents on the risk of adverse fetal outcomes before, during, or after treatment of reproductive cancers, as there is no strong evidence on which to base recommendations, and data are limited on the potential long-term effects of reproductive cancers on birth outcomes.
To fully understand the effects of cancer diagnosis on birth outcomes, studies with subset analysis directed to individual anatomic or histologic cancer type is vital. The Florida-linked database provides a powerful tool to study adverse birth outcomes for infants born to women affected by cancer before and during pregnancy. This database has previously been validated in other studies including at least one by a team from the Centers for Disease Control.12–14 The purpose of this study is to obtain population-based estimates of adverse birth outcomes among women with a diagnosis of reproductive cancer, with subset analysis by race/ethnicity. To the authors’ knowledge, the present study represents the first report on the association between the diagnosis of all reproductive cancers (breast, cervical, uterine, ovarian, vaginal, and vulvar) before and during pregnancy or within 30 days of birth and the risk of adverse birth outcomes [low birth weight (LBW), preterm birth (PTB), and small for gestational age (SGA)], using a large population-based sample.
MATERIALS AND METHODS
We conducted a retrospective population-based cohort study of women having a singleton live birth in the state of Florida from 1998 to 2007 inclusive. The Hospital Inpatient Discharge data from the Florida Agency for Health Care Administration linked with statewide vital records data from the Florida Department of Health were used for this study. The de-identified linked data had a total of 2,441,336 births to women in the state of Florida. After eliminating missing and out of range gestational age values including fetal viability (<20 and >44 weeks of gestation), a total of 1,573,971 records became available for final analyses. We defined reproductive cancer as cancers of the uterus, ovaries, cervix, fallopian tubes, vagina, vulva, and breast. We used the International Classification of Diseases, Ninth Revision (ICD-9) codes (“1740” “1741” “1742” “1743” “1744” “1745” “1746” “1748” “1749” “179” “1800” “1801” “1808” “1809” “181” “1820” “1821” “1828” “1830” “1832” “1833” “1834” “1835” “1838” “1839” “1840” “1841” “1842” “1843” “1844” “1848” “1849” “V103” “V1040” “V1041” “V1042” “V1043” “V1044” “2330” “2331” “2332” “23330” “23331” “23332” “23339”) to identify mothers with diagnosis of reproductive cancer before pregnancy, during pregnancy, or within 30 days after birth (n = 3212; 0.21%). We decided to include women diagnosed with reproductive cancer within 30 days after giving birth for the following 2 reasons. First, it is more likely that women diagnosed to have reproductive cancer within 30 days of giving birth had had the cancer at the time of pregnancy. Second, we wanted to capture women who gave birth at home and visited a health facility during the postpartum period.
The primary outcomes of interest in this study were LBW (birth weight <2500 g), PTB (<37 gestational weeks), and SGA (infants weighing below the 10th percentile of birth weight for their gestational age using normalized growth curves).15 Gestational age was computed in weeks by calculating the number of weeks between the first day of the last menstrual period and the date of delivery of the fetus. Where menstrual estimate of gestational age was inconsistent with birth weight (eg, very low birth weight at term), a clinical estimate of gestational age on the vital records was used instead.16 All demographic variables including racial/ethnic status (white non-Hispanic, black non-Hispanic, white Hispanic, black Hispanic, or other), maternal age (<35 or ≥35 years old at the time of delivery), marital status (married or not married), education (<12 or ≥12 years), parity (nulliparous or multiparous), cigarette smoking during pregnancy (yes or no), and adequacy of prenatal care (adequate or inadequate) were extracted from birth certificate data. Adequacy of prenatal care utilization was measured using the revised graduated index algorithm,17 which is based on the prenatal care recommendation by the American Congress of Obstetricians and Gynecologists.18 The revised graduated index algorithm has been found to be more accurate than other available indexes in describing the level of prenatal care utilization among groups that are at high risk.19,20 This index assesses the adequacy of care based on the trimester prenatal care began, number of visits, and the gestational age of the infant at birth. Maternal pregnancy complications obtained from the hospital discharge data based on ICD-9 principal and other diagnostic codes included anemia (280, 2800, 2801, 2808, 2809, 2810, 2811, 2812, 2813, 2818, 2819, 2820, 2821, 2822, 2823, 2824, 28241, 28242, 28249, 2825, 2826, 28260, 28261, 28262, 28263, 28264, 28268, 28269, 2827, 2828, 2829, 2830, 28310, 28311, 28319, 2832, 2839, 2840, 28401, 28409, 2841, 2842, 2848, 28481, 28489, 2849, 2850, 2851, 2852, 28521, 28522, 28529, 2853, 2858, 648.2), gestational diabetes (648.8), diabetes mellitus (250, 648.0), gestational hypertension (642.3) and chronic hypertension (642.0, 401.0, 401.1, 401.9, 642.1, 642.2, 742.7), preeclampsia (642.4, 642.5, 642.7, 642.9), eclampsia (642.6), placental abruption (641.2), and placental previa [maternal (641.0, 641.1), infant (762.0)].
We compared baseline characteristics (demographic characteristics and pregnancy-related clinical or medical conditions, see Tables 1 and 2) between mothers who had no reproductive cancer and those who have or had reproductive cancer using χ2 test for categorical variables in the matched sample. Because these baseline characteristics of mothers with reproductive cancers (cases) differed from mothers with no reproductive cancer diagnosis (controls), we used weighted propensity score methods to adjust for these differences.21 Use of propensity score technique allowed us to weight both the experiment and control groups with a single composite measure computed from all covariates of interest. This technique equated the observed covariates listed in Tables 1 and 2 to minimize their potential influence on group assignment and magnitude of the effect of the exposure.
Multivariate logistic regression model was applied to calculate the predicted probability of the dependent variable (probability of exposure to reproductive cancers) and the propensity score for each observation in the data set. We calculated the propensity score by weighting for both demographic and pregnancy-related clinical/medical conditions (listed in Tables 1 and 2). Then we applied the propensity score to represent the relationship between all the covariates (Tables 1 and 2) and our dependent variable (reproductive cancer diagnosis) by weighing each patient’s data based on the inverse propensity of being in one of the 2 groups (reproductive cancer diagnosis or no reproductive cancer diagnosis) when computing the estimated effect of exposure to reproductive cancers on birth outcomes.21 Each mother with a diagnosis of reproductive cancer (case) was matched to 2 mothers with no diagnosis of reproductive cancers (controls) whose propensity scores were sufficiently close to that of the case. In the presence of more than 1 control that could be matched to a case, a match was selected at random from all identical controls using a random number with the RANUNI function using SAS version 9.2 (SAS Institute, Inc, Cary, NC).22 We decided to use a 1 to 2 matching without replacement to maximize the accuracy of our estimates.
After the application of propensity score weighted matching, exposed and nonexposed matched subjects were found to be more similar than randomly selected exposed and unexposed subjects, thereby violating the assumption of independence.23–25 Because matching was performed postexposure, statistical analysis that takes into account the matched nature of the data should be used to avoid bias in estimating the outcomes. We therefore, applied the generalized estimating equations (GEE) method to account for intracluster correlation23–26 using the GENMOD procedure in SAS (SAS Institute, Inc, version 9.2). All tests of hypothesis were 2-tailed with a type 1 error rate fixed at 5%. This study was approved by the University of South Florida institutional review board.
A total of 1,573,971 births were retained for the final analyses; 3212 (0.21%) of them had reproductive cancer either before or during pregnancy or 30 days after giving birth. The remaining 1,570,759 (99.80%) had no reproductive cancer. The distribution of selected demographic characteristics and clinical conditions before and after applying propensity score weighted matching are presented in Tables 1 and 2, respectively. In the unmatched sample (left section of Table 1), mothers with diagnosis of reproductive cancer were more likely to be younger, married, educated, multiparous, and to have used alcohol, drug, and tobacco during pregnancy (P < 0.001). Women with reproductive cancer in the unmatched sample (left section of Table 2) were also more likely to be anemic, hypertensive, and diabetic than their counterparts with no diagnosis of reproductive cancer (P < 0.001). However, after the application of the propensity score weighted matching, differences in demographic and clinical conditions between cases (women with reproductive cancer) and controls (women with no diagnosis of reproductive cancers) disappeared (right section, Tables 1 and 2, respectively). Reproductive cancer types encountered in the study population were breast cancer (n = 1032; 0.07%), cervical (n = 1846; 0.12%), ovarian (n = 201, 0.01%), and uterine (n = 82; 0.01%).
In the unmatched sample, women with diagnosis of reproductive cancer had a statistically significant increased risk of giving birth to LBW [adjusted odds ratio (AOR), 1.33; 95% confidence interval (CI), 1.17–1.52] and PTB (AOR, 1.39; 95% CI, 1.24–1.55) infants, after adjustment for demographic characteristics and clinical conditions listed in Tables 1 and 2, respectively. The risk of SGA remained the same for the 2 groups. After applying propensity score weighted matching based on all the demographic and behavioral variables listed in Table 1 and a composite variable complication (absence or presence of at least one of the pregnancy-related clinical conditions listed in Table 2), the risk of adverse birth outcomes decreased from 33% to 24% for LBW and from 39% to 33% for PTB.
Results of risk comparisons across racial/ethnic groups showed that black women with reproductive cancer had an increased risk of LBW (93% more), SGA (125% more), and PTB (25% more, but statistically nonsignificant) when compared to white women with reproductive cancers (Table 3). After matching, the differences remained similar, albeit slightly attenuated to 83% for LBW and to 64% for SGA, and amplified to 47% for PTB. Comparison of Hispanic to white women did not depict any significant differences both in the matched and unmatched samples.
We further conducted analyses to compare fetal outcomes among women with diagnosis of reproductive cancer with women who had no reproductive cancer within their respective racial groups (Table 3). Among white women, those with reproductive cancer diagnosis had an increased risk for LBW (33% more) and PTB (44% more). This difference in risk estimates was more prominent among Hispanic women; whom those with reproductive cancer were 75% and 68% more likely to have infants with LBW and PTB, respectively, than their peers without reproductive cancer. These risks were not statistically significant for black women in the unmatched sample. However, after the application of propensity score weighted score matching, black women with reproductive cancers demonstrated an increased risk of LBW (80%), PTB (45%), and SGA (64%) when compared to black women with no diagnosis of reproductive cancers. The elevated risk that was observed among white and Hispanic women in the unmatched sample was not evident in the matched sample, except for PTB in white women.
In an effort to understand whether the association between reproductive cancer diagnosis (either before or during pregnancy) and fetal birth outcomes differs across cancers of specific organs of the reproductive system, we conducted additional analyses stratified by the following 3 major organ-specific cancers of the female reproductive system: (1) breast cancer, (2) cervical cancer, (3) ovarian cancer including the fallopian tubes, and (4) uterine cancer. Table 4 summarizes the findings of the stratified analysis, and Table 5 provides findings of the stratified analysis by racial groups. In the matched sample, women with cervical and ovarian cancers demonstrated 38% and 58% increased risk of giving birth to a preterm infant, respectively. Although women with breast cancer and uterine cancer did not show statistically significant elevated risk of PTB when compared to cancer-free women. An increased risk of LBW infants was also observed among women with cervical cancer (AOR, 1.28; 95% CI, 1.06–1.56). Stratification of the analysis by race showed that black women with breast cancer had 131% and 68% greater risk of giving birth to LBW and SGA infants, respectively, when compared to their white counterparts. Furthermore, black women with cervical cancer are at an increased risks of LBW (56%) and PTB (44%) than white women with cervical cancer. Hispanic women with breast cancer had lower risk of LBW, PTB, and SGA when compared to their white counterparts, but these findings did not reach statistical significance.
We also investigated the association between cancers of specific organs of the female reproductive system and adverse fetal outcomes within racial groups (Table 5). Black women with breast cancer demonstrated significantly higher risks of LBW (AOR, 2.36; 95% CI, 1.53–3.65) and SGA (AOR, 1.71; 95% CI, 1.11–2.64) when compared to women of their racial group with no reproductive cancer. Black women with cervical cancer also had 59% and 45% elevated risks of giving birth to LBW and preterm infants when compared to black women with no diagnosis of reproductive cancer. Risks of LBW and PTB were also higher among Hispanic and white women with diagnosis of cervical and ovarian cancers without reaching statistically significant levels.
The diagnosis of cancer during pregnancy poses a dilemma between optimal maternal treatment and safety of the fetus.27 For women diagnosed with cancer, especially reproductive cancers during pregnancy, the primary issues of concern include retaining fertility, ability to conceive, carrying pregnancy to term, and giving birth to a healthy infant.28 In addition to the impact that the disease process might have on the fetus, cancer therapy including chemotherapy, radiation, and surgery may affect fetal outcomes by either directly affecting the reproductive organs or through mutation.29,30 Hence, examining the impact of cancer diagnosis before or during pregnancy is essential to establish the impact of cancer diagnosis and its treatment on birth outcomes. In this retrospective population-based observational study, we found an elevated risk of LBW (24%) and PTB (33%) among infants born to women with diagnosis of reproductive cancer, with remarkable variation by race. Our study adds to the current knowledge by exploring the impact of reproductive cancers diagnosed before and during pregnancy on multiple adverse birth outcomes. The application of propensity score weighted matching approach strengthened our study by minimizing the effect of bias in the assignment of subjects to groups (cases and controls).
The association between maternal sociodemographic characteristics and fetal outcomes has been established by multiple authors.31–35 Previously, it was shown that women with less education and low income tend to be smokers (known risk factor for intrauterine growth retardation).36–38 Therefore, one can argue that the increased risks of LBW and PTB observed in this study could be explained by sociodemographic characteristics of the mother. However, our analysis was based on a sample that was matched by these demographic variables, thus reducing the impact of sociodemographic factors. A second theory could be that mothers in the 2 groups (cases and control) could differ in their pregnancy-related clinical conditions. Then again, both groups were matched on selected pregnancy-related clinical conditions that are known to affect pregnancy outcomes. The association between maternal reproductive cancers diagnosed before or during pregnancy and increased risks of LBW and PTB that was not explained by either sociodemographic characteristics or pregnancy-related clinical conditions corroborates an independent effect of reproductive cancer on adverse birth outcomes. However, because of the nature of our data, we did not control for vital psychological factors (eg, stress,39,40 racism,41,42 and discrimination43) that could have impacted birth outcomes. An alternative explanation to the 33% increased odds of PTB for newborns of women who were diagnosed with reproductive cancer before or during their pregnancy could be because of frequent elective early delivery, probably to allow an earlier start of cancer therapy.
Further analysis of the association between reproductive cancer diagnosed before or during pregnancy and birth outcomes by race revealed disparity to the disadvantage of blacks. When compared to white women with diagnosis of reproductive cancer, black women with reproductive cancer demonstrated an increased risk of LBW, PTB, and SGA. Similar analysis comparing Hispanics mothers with reproductive cancers with whites who had reproductive cancer showed no difference (Table 3). In general, adverse birth outcomes are more prevalent among blacks and have been widely discussed in the literature.44–50 Analyses stratified by specific organs of the female reproductive system showed increased risks of LBW and PTB. However, some of these findings did not reach statistically significant levels (Table 4). This could be because of the relatively small sample size in each group after the stratification. Our findings of an increased risk of giving birth to a PTB and LBW among women who were diagnosed with cancer during pregnancy concur with findings of previous studies.7 ,51–53 The mechanism and reasons for why the risk is more pronounced among blacks is not clear. Further studies are warranted to delineate the pathways that explain the noted disparity.
In drawing conclusions from this study, there are several limitations that need to be addressed. First, we considered only viable pregnancies within a given range (20 ≤ gestational weeks ≤ 44). Women with reproductive cancer could have had more miscarriages or induced abortions because of fetal abnormalities leading to potential underestimation of effects. The second potential limitation of this study is the so-called “healthy mother effect.”54 Women who became pregnant after cancer treatment or while still having cancer diagnosis at the time of pregnancy could have more favorable characteristics than women who did not become pregnant. Those who feel well have children, whereas women who have had a recurrence after diagnosis do not. This could represent a bias even if cases and controls were well matched for other covariates that affect assignment of subjects. A third limitation is related to the retrospective nature of our data which resulted in limitations that include (1) lack of results stratified by treatment, time of diagnosis (pregestation or intragestation), (2) difficulty in controlling for biological factors and risk determinants, and (3) use of ICD codes to identify cases of reproductive cancer that formed the basis of the study sample in our study. In the absence of readily available clinical diagnosis data, ICD code provide a standard method for determining rates of diseases and it has been used by previous population-based observational studies that used linked data set. Our data do not have information on the exact time of initial diagnosis and we could not distinguish mothers who had cancer before pregnancy from those diagnosed during pregnancy. Women with active cancer during pregnancy might have differential risk of adverse fetal outcomes than partially or fully treated women.
Despite the limitations stated previously, our study has several strengths. Our results demonstrate consistencies in alignment with expected patterns. In addition, the number of cancer cases identified in this population-based study is much larger than in previously reported studies, with reasonable power for detection of group differences in the analysis. Our study also provides important information regarding existing racial disparities in birth outcomes with potential public health and clinical implications.
1. Wingo PA, Tong T, Bolden S. Cancer statistics, 1995. CA Cancer J Clin
. 1995; 45: 8–30.
2. DePinho RA. The age of cancer. Nature
. 2000; 408: 248–254.
3. Sankaranarayanan R, Ferlay J. Worldwide burden of gynaecological cancer: the size of the problem. Best Pract Res Clin Obstet Gynaecol
. 2006; 20: 207–225.
4. Jemal A, Siegel R, Xu J, et al.. Cancer Statistics, 2010. CA Cancer J Clin
. 2010; 60: 277–300.
5. U.S. Cancer Statistiscs Working Group. United States Cancer Statistics: 2002 Incidence and Mortality
. Atlanta, Ga: U.S. Department of Health and Human Services, Centers for Disease Control and Prevention and National Cancer Institute; 2005.
6. Carbonne B, Ansquer Y. [Pregnancy
after gynecologic or breast cancer]. Bull Acad Natl Med
. 2010; 194: 509–518; discussion 518–520, 529–530.
7. Cardonick E, Usmani A, Ghaffar S. Perinatal outcomes of a pregnancy
complicated by cancer, including neonatal follow-up after in utero exposure to chemotherapy: results of an international registry. Am J Clin Oncol
. 2010; 33: 221–228.
8. Chen YH, Lin HC, Chen SF. Increased risk of preterm births among women with uterine leiomyoma: a nationwide population-based study. Hum Reprod
. 2009; 24: 3049–3056.
9. Van Calsteren K, Heyns L, De Smet F, et al.. Cancer during pregnancy
: an analysis of 215 patients emphasizing the obstetrical and the neonatal outcomes. J Clin Oncol
. 2010; 28: 683–689.
10. Voulgaris E, Pentheroudakis G, Pavlidis N. Cancer and pregnancy
: a comprehensive review. Surg Oncol
. 2011; 20: e175–e185.
11. Amant F, Van Calsteren K, Halaska MJ, et al.. Gynecologic cancers in pregnancy
: guidelines of an international consensus meeting. Int J Gynecol Cancer
. 2009; 19 (suppl 1): S1–S12.
12. Clayton HB, Sappenfield WM, Gulitz E, et al.. The Florida Investigation of Primary Late Preterm and Cesarean Delivery: the accuracy of the birth certificate and hospital discharge records. Matern Child Health J
. 2012: 1–10. E-mail: http://dx.doi.org/10.1007/s10995-012-1065-0. Accessed 1 June 2012.
13. Kim SY, England L, Sappenfield W, et al.. Racial/ethnic differences in the percentage of gestational diabetes mellitus cases attributable to overweight and obesity, Florida, 2004–2007. Prev Chronic Dis
. 2012; 9: E88.
14. Salihu HM, Stanley KM, August EM, et al.. The association between HIV/AIDS during pregnancy
and fetal growth parameters in Florida: a population based study. Curr HIV Res
. 2012; 10: 539–545.
15. Bakketeig LS. Current growth standards, definitions, diagnosis and classification of fetal growth retardation. Eur J Clin Nutr
. 1998; 52 (suppl 1): S1–S4.
16. Taffel S, Johnson D, Heuser R. A method of imputing length of gestation on birth certificates. Vital Health Stat 2
. 1982; 93: 1–11.
17. Alexander GR, Kotelchuck M. Quantifying the adequacy of prenatal care: a comparison of indices. Public Health Rep
. 1996; 111: 408–418; discussion 419.
18. Heaman M, Newburn-Cook C, Green C, et al.. Inadequate prenatal care and its association with adverse pregnancy
outcomes: a comparison of indices. BMC Pregnancy Childbirth
. 2008; 8: 15.
19. Kotelchuck M. An evaluation of the Kessner Adequacy of Prenatal Care Index and a proposed Adequacy of Prenatal Care Utilization Index. Am J Public Health
. 1994; 84: 1414–1420.
20. Alexander GR, Allen MC. Conceptualization, measurement, and use of gestational age. I. Clinical and public health practice. J Perinatol
. 1996; 16: 53–59.
Keywords:Copyright © 2013 by IGCS and ESGO
Low birth weight; Pregnancy; Preterm birth; Reproductive Cancer; Small size for gestational age