Author + information
- Mark Nolan, MBBS, PhDa,
- Evangelos K. Oikonomou, MDb,
- Candice K. Silversides, MDa,c@CandiceSilvers1,
- Melissa R. Hines, MDd,
- Kara A. Thompson, MDe,
- Belinda A. Campbell, MBBS, MMedf,
- Eitan Amir, MD, PhDg,
- Cynthia Maxwell, MDh and
- Paaladinesh Thavendiranathan, MD, SMa,i,∗ (, )@dineshpmcc1
- aDepartment of Medicine, Division of Cardiology, Ted Rogers Program in Cardiotoxicity Prevention, Peter Munk Cardiac Centre, University Health Network, University of Toronto, Toronto, Ontario, Canada
- bDepartment of Internal Medicine, Yale New Haven Hospital, Yale School of Medicine, New Haven, Connecticut
- cPregnancy and Heart Disease Program, Department of Medicine, Division of Cardiology, Mount Sinai and Toronto General Hospitals, Toronto, Ontario, Canada
- dDivision of Critical Care, St. Jude Children’s Research Hospital, Memphis, Tennessee
- eDepartment of Cardiology, The University of Texas, MD Anderson Cancer Center, Houston, Texas
- fDepartment of Radiation Oncology, Peter MacCallum Cancer Center, Melbourne, Victoria, Australia, and Department of Clinical Pathology, The University of Melbourne, Parkville, Victoria, Australia
- gDivision of Medical Oncology, Princess Margaret Cancer Centre, Toronto, Ontario, Canada
- hDepartment of Obstetrics and Gynecology, University Health Network and Mount Sinai Hospital, University of Toronto, Toronto, Ontario, Canada
- iJoint Division of Medical Imaging, University Health Network, University of Toronto, Toronto, Ontario, Canada
- ↵∗Address for correspondence:
Dr. Paaladinesh Thavendiranathan, Ted Rogers Program in Cardiotoxicity Prevention, Peter Munk Cardiac Center, Division of Cardiology, Toronto General Hospital, University of Toronto, 200 Elizabeth Street, Toronto, Ontario M5G 2C4, Canada.
Background Cancer treatment can lead to left ventricular (LV) dysfunction in female cancer survivors of reproductive age, and pregnancy-related hemodynamic stress may result in LV dysfunction or heart failure (HF).
Objectives We performed a systematic review and meta-analysis to determine the incidence of LV systolic dysfunction or HF during or soon after pregnancy in cancer survivors and evaluated the impact of history of cancer therapeutics-related cardiac dysfunction (CTRCD).
Methods We systematically searched electronic databases (MEDLINE and EMBASE) from inception to January 2020 to identify cohort studies that examined cardiac disease in pregnant cancer survivors. Meta-analysis was performed using the inverse-variance fixed effects method. Potential sources of heterogeneity were explored using subgroup analyses and meta-regression.
Results Of 13,782 identified articles, 6 studies consisting of 2,016 pregnancies, predominantly in childhood cancer survivors, were included. Overall, there were 33 cardiac events. The total weighted incidence of LV dysfunction or HF with pregnancy was 1.7% (95% confidence interval [CI]: 0.9% to 2.7%) overall; 28.4% (95% CI: 14.6% to 43.9%) in those with a history of CTRCD and 0.24% (95% CI: 0% to 0.81%) in those without, translating into an odds ratio of 47.4 (95% CI: 17.9 to 125.8). Interstudy heterogeneity was low (I2 = 17.5%). Metaregression did not reveal significant sources of heterogeneity.
Conclusions The incidence of LV dysfunction or HF during pregnancy in cancer survivors was low. Although risk estimates are limited by the small number of events, women with a history of CTRCD compared to those without had a 47.4-fold higher odds of experiencing pregnancy-related LV dysfunction or HF.
Advances in cancer treatment in the last several decades have led to an increasing number of young adult cancer survivors. In the United States alone, 1 in 640 young adults between the ages of 20 and 39 years is a cancer survivor (1). Cardiac toxicity from cancer treatment, including chemotherapy and radiation, is the leading nonmalignant cause of death among adult cancer survivors (2). Manifestations of adverse cardiovascular outcomes for both chemotherapy and radiation treatment include cancer therapeutics-related cardiac dysfunction (CTRCD) (ie, reduction in left ventricular ejection fraction [LVEF] or heart failure [HF]), premature coronary artery disease, valvular disorders, pericardial injury, and arrhythmia (3–5). Onset of cardiovascular complications can be delayed by up to 30 years after cancer treatment (6) and may be accelerated by either conventional cardiovascular risk factors or hemodynamic stress (7).
The increasing number of female cancer survivors with reproductive potential raises concern about cardiac complications associated with pregnancy. Pregnancy results in plasma volume expansion, increased heart rate, and higher cardiac output resulting in increased myocardial wall stress. Furthermore, there is a rise in oxidative stress during normal pregnancy, culminating in the last trimester (8). These changes may be tolerated poorly in patients with a history of CTRCD who have a vulnerable myocardium (9,10). Furthermore, up to one-third of patients with prior cancer treatment, regardless of a CTRCD history, can have subclinical LV systolic dysfunction which can be unmasked by the hemodynamic stress of pregnancy (11). Prior studies examining maternal adverse cardiac events in cancer survivors are individually limited by small sample size and single-center recruitment, and some were dependent on self-reported events (12–17). Drawing definitive conclusions from each study in isolation to guide management may be challenging.
The present objective was to perform a systematic review and meta-analysis of prior published studies to quantify the impact of a pre-pregnancy history of CTRCD on the risk of maternal LV systolic dysfunction or HF during pregnancy or within 12 months of delivery. The present authors hypothesized that the incidence of maternal LV systolic dysfunction or HF would be low and that a history of CTRCD would be associated with a higher risk.
A systematic review was performed, adhering to Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (18) (Supplemental Table 1) and registered prospectively in the PROSPERO database (Cardiac Outcomes in Pregnant Women with Prior Treated Cancer: a Systematic Review and Meta-Analysis; CRD42019123849). Electronic searches of MEDLINE and EMBASE (OvidSP) were performed from inception to January 2020. Due to multiple potential sources of variability, including different cardiomyopathy definitions and sources of CTRCD, a liberal search strategy was used to improve sensitivity (Supplemental Appendix). In addition, references of relevant publications were searched, and experts in the field were contacted to improve the sensitivity of the search criteria. Inclusion criteria consisted of publications in peer-reviewed, English-language journals reporting data for: 1) study type: either prospective or retrospective cohort studies with ≥10 patients who underwent ≥1 pregnancy during study duration; 2) study population: females of reproductive age with prior treatment of cancer using potentially cardiotoxic therapeutics, including any chemotherapy or radiotherapy with fields including the thorax and reported cardiac function or HF status prior to pregnancy; 3) follow-up: from cancer diagnosis until at least first pregnancy; and 4) outcome: LV systolic dysfunction (reduction in LVEF or fractional shortening) or HF during pregnancy or within 12 months of delivery. Exclusion criteria were: 1) nonhuman studies; and 2) abstracts or conference proceedings. All papers were reviewed for inclusion and exclusion factors by 2 authors (M.N. and E.K.O.), who also extracted data from the primary publications. Any discrepancies were resolved by consensus. Information for year of publication, sample size, follow-up duration, type and dose of cancer therapy drugs, number of patients with CTRCD prior to pregnancy, and number of patients with pregnancy-related LV dysfunction and HF were extracted independently from each study. When required data were missing, the corresponding authors were contacted. Quality of individual studies was assessed using the Newcastle-Ottawa Scale (19).
The combined weighted incidence of LV dysfunction or HF from all studies was calculated using Freeman-Tukey double-arcsine transformation to weigh by inverse variance and a fixed-effects model. Between-study heterogeneity was assessed by calculation of the Cochran Q and I2 statistics. Meta-analysis for the odds ratio (OR) for LV dysfunction or HF in pregnant women with or without CTRCD was performed by first calculating the log odds ratio (OR) and 95% confidence interval (CI) for each study identified by the systematic review. The log (OR) was used as the summary statistic, as this estimate has a better normal approximation than that of the OR. A continuity correction, adding 0.05 to each cell, was done in the presence of any 0 outcome cell counts. The combined log (OR) was calculated using the inverse-variance fixed-effect meta-analysis method. The overall OR with 95% CI was obtained by exponentiating the log (OR) results. A sensitivity analysis was performed using a random effects meta-analysis method (DerSimonian and Laird method).
The impact of baseline characteristics on the association between CTRCD and pregnancy-related cardiac events was explored using meta-regression techniques. Meta-regression was performed using a fixed-effects model with OR for CTRCD as the dependent variable and using previously defined study level characteristics including year of publication, number of patients enrolled, and cumulative anthracycline dose as the independent variables.
Publication bias was assessed visually by funnel plots of effect estimates and by the Egger regression statistical test. Because the number of studies assessed was small, further assessment of possible publication bias was undertaken by using the Duval and Tweedie Trim and Fill test. Statistical analysis was performed using the “metafor” application in R statistical software (Vienna, Austria) (20). Statistical tests were 2-sided, and significance was defined as a p value of <0.05.
Systematic review identified 13,782 potential studies (Supplemental Figure 1). After review, 6 studies met eligibility criteria (Table 1). All studies were of good quality according to the Newcastle-Ottawa Scale (Supplemental Table 2). A total of 5 of 6 studies were retrospective, and most focused on pediatric, adolescent, and young adult cancer survivors. Three studies used cardiac dysfunction defined by echocardiography as the outcome, 2 studies used symptomatic cardiac disease as the preferred outcome, and 1 study used both. A standard previous definition of CTRCD was not used in all studies, and most studies did not provide a formal definition (Table 1).
The 6 studies included a cumulative total of 1,137 women with 2,016 pregnancies. Median age at cancer diagnosis was 12.1 years (range: 0.02 to 41 years). Median age at first pregnancy was 21.8 years, and median follow-up since the cancer diagnosis was 24.8 years for the 5 studies that provided data. Cancer treatment spanned from 1963 to 2015. Among the patients, 66.5% received anthracyclines with a mean cumulative anthracycline dose of 234.0 mg/m2. For the 5 studies that provided data for cancer subtypes, 60.6% were hematological malignancies.
Incidence of LV dysfunction or HF
Using data from all studies, 33 cancer survivors (2.9%) had LV dysfunction or HF during pregnancy or in the 12 months after delivery. Seventeen of the 33 survivors (51.5%) had CTRCD previously, 16 (48.5%) did not; 32 (97.0%) had prior anthracycline treatment, and 1 (3.0%) did not. The total weighted incidence of LV dysfunction or HF during pregnancy was 1.7% (95% CI: 0.9% to 2.7%). In patients with a history of CTRCD, the weighted incidence was 28.4% (95% CI: 14.6% to 43.9%), whereas in patients without a history of CTRCD, the weighted incidence was 0.24% (95% CI: 0% to 0.81%). This translated to a number-needed-to-harm of 4. There were no maternal or fetal pregnancy-associated mortalities reported in the studies, although 3 studies reported incidence rates of spontaneous abortion of 15.1%, 10.8%, and 6.0% (12,13,17).
Predictors of LV dysfunction or HF
The association between pre-pregnancy CTRCD and development of LV dysfunction or HF during pregnancy or within 12 months of delivery based on the meta-analysis is shown in Figure 1. Low and nonsignificant interstudy heterogeneity were observed (Cochran Q: 4.2; p = 0.53, and I2 = 17.5%). Patients with a history of pre-pregnancy CTRCD had significantly higher odds of LV dysfunction or HF during pregnancy or within 12 months of delivery (OR: 47.4; 95% CI: 17.9 to 125.8; p < 0.001). Analysis repeated with a random effects model demonstrated a similar OR (OR: 40.4; 95% CI: 12.8 to 126.5). A sensitivity analysis that excluded the largest single study, which accounted for 74.5% of all patients, still found that pre-pregnancy CTRCD was associated with higher odds of LV dysfunction or HF (OR: 19.9; 95% CI: 5.1 to 77.0; p < 0.001).
Meta-regression did not demonstrate any significant contribution toward interstudy heterogeneity using study level variables of publication year (r = −0.06; p = 0.96), cumulative anthracycline dose (r = −0.01; p = 0.18), or number of patients enrolled (r = 0.002; p = 0.07) (Figure 2).
Assessing publication bias
Risk of publication bias as assessed by visual inspection of funnel plot did not reveal any asymmetry (Figure 3). The Egger regression test did not demonstrate evidence of bias (p = 0.12), although it was underpowered due to the small number of studies. The Duval and Tweedie Trim and Fill test results did not significantly predict missing studies, nor did they alter the effect size.
In this systematic review and meta-analysis of 1,137 women previously treated for cancer who underwent at least 2,106 pregnancies, the incidence of LV dysfunction or HF with pregnancy was 1.7% in the cancer survivor population. There were significant differences in the incidence of LV dysfunction or HF in patients with a history of CTRCD compared to patients without CTRCD (28.4% vs. 0.24%, respectively). The number needed to harm was 4, meaning that among every 4 women with a history of CTRCD, one would develop LV dysfunction or HF during pregnancy or within 12 months after delivery. These findings add to existing reviews of this subject (21) by providing data that can be used for patient counseling, including a weighted risk estimate of LV dysfunction or HF in cancer survivors going through pregnancy, an OR for risk based on a history of CTRCD, and a number-needed-to-harm estimate.
Incidence of LV dysfunction or HF associated with pregnancy in cancer survivors
The overall reported incidence of LV dysfunction or HF associated with pregnancy in survivors of adult and pediatric cancer survivors has varied between 0.0% and 7.8% (Table 1). The studies that reported the lowest incidence either did not have pre- and post-pregnancy cardiac assessment (13) or relied on self-reporting of peripartum HF, which can be associated with recall and survival bias (14). In the present meta-analysis of 6 published studies, the weighted risk of pregnancy-associated LV systolic dysfunction or HF in cancer survivors was 1.7% with no reported maternal cardiac deaths. It is possible, but we believe less likely, that the cases of HF in the published studies represent superimposed conditions such as peripartum cardiomyopathy or viral myocarditis, as opposed to decompensation of subclinical dysfunction in patients with a history of CTRCD. However, the reported incidence of peripartum cardiomyopathy is approximately 0.03% in the general population (22), which is much lower than the incidence seen in our meta-analysis.
Predictors of LV dysfunction or HF during pregnancy in cancer survivors
Potential predictors of LV dysfunction or HF during pregnancy or in the post-partum period based on results presented in the individual studies but not from our primary analysis are summarized in Table 1. These predictors include lower pre-pregnancy LV systolic function (12,16), younger age at cancer diagnosis (15,17), longer time from cancer treatment to first pregnancy (15), and cumulative anthracycline dose (14,15,17). The younger age at cancer diagnosis and longer time from cancer treatment to pregnancy may reflect patients who received older treatment regimens with higher doses of anthracycline, subclinical cardiomyopathy, or the development of additional cardiovascular risk factors. In the present meta-analysis, the weighted incidence of pregnancy associated HF or worsening LV function in cancer survivors restricted to those with a history of CTRCD was 28.4%. A history of CTRCD was associated with 47.4-fold higher odds of LV dysfunction (95% CI: 17.9 to 125.8) or HF with pregnancy. This incidence of cardiac events in those with a history of CTRCD is similar to the reported incidence of new onset HF in women with a history of peripartum cardiomyopathy (29%) going through a second pregnancy (23). These scenarios demonstrate that persistent myocardial injury can be unmasked by the stress of pregnancy. Interestingly, the majority of the reported LV dysfunctions or HF events (32 of 33) occurred in those who had previously undergone anthracycline therapy. However, the number of events in the nonanthracycline subgroup was inadequate to accurately and precisely perform subgroup analysis to examine the impact of cancer treatment regimen on LV dysfunction or HF related to pregnancy.
The present analysis has some limitations worth discussing. Although the results are based on 6 single-center studies, there were a limited number of events. Still, this is the only meta-analysis of this subject, and it provides summary statistics that help guide clinical care. However, the overall small number of outcomes led to wide confidence intervals for the present estimates. Hence these estimates should be interpreted in the context of this limitation. Patients in the present study received cancer treatment spanning 52 years, and the treatment regimens have evolved. However, the absence of interstudy heterogeneity of studies published over a 16-year span and the lack of relationship between publication year and outcome in the present meta-regression suggest that the present findings are likely still relevant to current treatments. Furthermore, the main risk factor studied, CTRCD, remains an important consideration with current treatment regimens (24). Inconsistent definitions of CTRCD were used among studies, and cardiac imaging was not consistently performed before and after pregnancy in all patients. Despite this, there remained concordance of findings between studies with minimal interstudy heterogeneity, suggesting that the results of this study are relevant. Most studies focused on pediatric, adolescent, and young adult cancer survivors leading to under-representation of older adult cancer survivors. We were also unable to adjust the odds ratio for high-risk baseline demographics, such as extremes of patient age at pregnancy, due to lack of patient-level data. We were also unable to perform subgroup analysis examining the effects of radiotherapy or type of cancer therapy on outcomes due to incomplete reporting of these baseline characteristics and/or a small number of events. Future development of registries and prospective studies are needed to provide more accurate identification of risk factors and ideal diagnostic and therapeutic strategies for cancer survivors going through pregnancy. Several other risk factors beyond CTRCD have been outlined for LV dysfunction and HF during pregnancy or post-delivery in Discussion and in Table 1, based on prior publications. However, data from individual studies were not adequate to determine their independent and incremental predictive value over that of CTRCD history. Furthermore, important information regarding cardiac medications in patients with a history of CTRCD or their cessation during pregnancy was not consistently available in our studies (Supplemental Table 3). Therefore, whether cessation of cardiac medications during pregnancy contributed to the LV dysfunction or HF cannot be determined.
Current guidelines for cancer survivorship provide different recommendations, including consideration of cardiomyopathy surveillance before pregnancy or in the first trimester in those treated with anthracyclines or chest radiation without specific recommendations for ongoing surveillance if the LVEF is normal (25); cardiologist referral prior to pregnancy (26); or performing cardiac imaging in the third trimester in all cancer survivors (27). There appears to be a lack of consensus regarding patient selection, modality, and timeline of cardiac monitoring and use of cardioprotective strategies. The present meta-analysis provides important information for pre-conception counseling in female cancer survivors of reproductive age. Overall, the risk of developing LV dysfunction or HF is rare with pregnancy in women with prior exposure to cancer therapy but without a history of CTRCD. However, in women with a history of CTRCD, the risk of developing HF during or within 1 year of pregnancy is 28.4%. Therefore, although all women with prior potentially cardiotoxic cancer therapy could benefit from a baseline evaluation of cardiac function (25) and counseling pre-conception or early during pregnancy, the present authors believe those with a history of CTRCD should receive close cardiac surveillance during pregnancy at a center with expertise in cardiac disease in pregnancy (Central Illustration). Women without prior CTRCD may benefit from a baseline assessment of LV function, and in the absence of LV systolic dysfunction, these women can be reassured that they are at a low risk of developing LV dysfunction or HF during pregnancy with follow-up determined on an individual basis. Cardioprotection with β-blockers for patients with peri-partum cardiomyopathy going through subsequent pregnancies has demonstrated benefit in small observational studies (28) and could plausibly be used to manage high-risk cancer survivors during pregnancy. This will, however, have to be confirmed in prospective studies.
COMPETENCY IN MEDICAL KNOWLEDGE: The present systematic review and meta-analysis demonstrated that the pooled incidence of pregnancy-related cardiac dysfunction in cancer survivors was 1.7%. There was a 47.4-fold higher odds (95% CI: 17.9 to 125.8) of LV dysfunction or HF if there was a history of CTRCD. These findings indicate that patients going through pregnancy with a history of CTRCD represent a high-risk cohort that requires cardiac surveillance.
TRANSLATIONAL OUTLOOK: There is a need for large prospective studies to quantify the long-term risk of pregnancy-related cardiac dysfunction attributable to differences in cancer type, cancer treatment modalities, and clinical factors to facilitate individualized risk prediction. Furthermore, prospective studies to understand the role of cardiac medications (e.g., beta-blockers) to prevent LV dysfunction or HF during pregnancy in high risk cancer survivors are needed.
The authors acknowledge Michelle A.T. Hilderbrandt, PhD, and Lori Chait-Rubinek, MBBS, for providing additional data related to their publications.
Dr. Thavendiranathan is supported by the Canadian Institutes of Health Research New Investigator Award. Dr. Nolan was compensated as a speaker for Novartis. Dr. Hines had a relationship with Incyte. Dr. Amir presented expert testimony for Genentech and Roche; and is a consultant for Sandoz and Apobiologix. Dr. Maxwell is a consultant for Guidepoint Health. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
The authors attest they are in compliance with human studies committees and animal welfare regulations of the authors’ institutions and Food and Drug Administration guidelines, including patient consent where appropriate. For more information, visit the JACC: CardioOncology author instructions page.
- Abbreviations and Acronyms
- cancer therapeutics-related cardiac dysfunction
- heart failure
- left ventricle
- left ventricular ejection fraction
- Received November 5, 2019.
- Revision received April 8, 2020.
- Accepted April 11, 2020.
- 2020 The Authors
- Armstrong G.T.,
- Liu Q.,
- Yasui Y.,
- et al.
- Chen M.H.,
- Colan S.D.,
- Diller L.
- Lipshultz S.E.,
- Adams M.J.,
- Colan S.D.,
- et al.
- Mulrooney D.A.,
- Yeazel M.W.,
- Kawashima T.,
- et al.
- Khanna A.,
- Pequeno P.,
- Gupta S.,
- et al.
- Nolan M.T.,
- Marwick T.H.,
- Plana J.C.,
- et al.
- Siu S.C.,
- Colman J.M.,
- Sorensen S.,
- et al.
- Armstrong G.T.,
- Joshi V.M.,
- Ness K.K.,
- et al.
- Hines M.R.,
- Mulrooney D.A.,
- Hudson M.M.,
- et al.
- Thompson K.A.,
- Hildebrandt M.A.T.,
- Ater J.L.
- Liu S.,
- Aghel N.,
- Belford L.,
- et al.
- Chait-Rubinek L.,
- Mariani J.A.,
- Goroncy N.,
- et al.
- Liberati A.,
- Altman D.G.,
- Tetzlaff J.,
- et al.
- Wells G.A.,
- Shea B.,
- O'Connell D.,
- et al.
- Thavendiranathan P.,
- Abdel-Qadir H.,
- Fischer H.D.,
- et al.
- Carver J.R.,
- Szalda D.,
- Ky B.
- Codsi E.,
- Rose C.H.,
- Blauwet L.A.