Incident Myocardial Infarction, Heart Failure, and Oncologic Outcomes in Breast Cancer Survivors

Abstract

Background

Cardiovascular disease (CVD) is associated with higher rates of incident cancer. Data are scarce regarding the association of incident CVD with oncologic outcomes after a cancer diagnosis.

Objectives

This study sought to determine whether incident myocardial infarction (MI) or heart failure (HF) in breast cancer survivors is associated with oncologic outcomes.

Methods

This was a population-based cohort study in Ontario, Canada, using linked administrative data sets of women diagnosed with first breast cancer between April 1, 2007, and March 31, 2015. A landmark analysis was conducted of women alive 2 years after breast cancer diagnosis, aged ≥40 years, and with available staging data and without recurrent/distant disease or preceding CVD. The exposure was a composite of MI and/or HF after the landmark date. The outcomes were cancer mortality, new non-breast malignancy diagnosis, and new chemotherapy initiation. Multivariable cause-specific hazards regression was used to determine the association of incident MI/HF (time-varying exposure) with outcomes.

Results

A total of 30,694 women (median age of 60 years) were included, of whom 1,346 developed incident MI/HF at a median of 3.9 years after the landmark date. At 5 years, the cumulative incidence was 5.9% (95% CI: 5.6%-6.1%) for cancer death, 4.3% (95% CI: 4.1%-4.6%) for non-breast malignancy, and 25.7% (95% CI: 25.2%-26.2%) for new chemotherapy. Incident MI/HF was associated with a higher hazard of cancer death (HR: 3.94; 95% CI: 3.38-4.59), non-breast malignancy (HR: 1.39; 95% CI: 1.06-1.82), and new chemotherapy (HR: 1.25; 95% CI: 1.02-1.53).

Conclusions

Incident MI and/or HF after breast cancer treatment are associated with higher hazards of adverse oncologic outcomes, highlighting the need to prioritize care for these patients.

Central Illustration

It is well established that chemotherapy, radiotherapy, HER2-targeted therapies, and certain endocrine therapies increase cardiovascular disease (CVD) risk in people with cancer.¹ Intriguingly, multiple studies have suggested that heart failure (HF)² and myocardial infarction (MI)^3,4 may increase the risk of subsequent cancer.⁵ Given this association, it is plausible that patients with cancer who develop CVD after completing cancer therapy may be at increased likelihood of cancer-related mortality and the development of secondary malignancies.

Conversely, it is possible that higher cancer incidence in people with CVD may reflect surveillance bias leading to greater cancer detection in the setting of more frequent imaging or unmasking of malignancy because of complications of antithrombotic therapy. This source of bias is less likely in cancer survivors for whom cancer surveillance is usually protocolized. Our objectives were to determine whether incident MI or HF after the completion of cancer therapy in breast cancer survivors is associated with 1) cancer death, 2) non-breast malignancy diagnosis, and 3) new chemotherapy use. We hypothesized that incident MI/HF among breast cancer survivors free of CVD at baseline is associated with cancer-related death, diagnosis of non-breast primary malignancies, and new chemotherapy use.

Methods

Data sources

Ontario residents receive universal health coverage through the Ontario Health Insurance Plan, enabling ascertainment of physician and hospital-based care through linked administrative databases. A description of these data sources is provided in the Supplemental Appendix. We used validated algorithms (where available) to identify patients, baseline characteristics, exposure, and outcomes.6, 7, 8, 9, 10, 11, 12, 13, 14 Cause of death ascertained using the Ontario Cancer Registry in breast cancer patients has been previously validated against cause of death from a prospective study with 90% agreement for cause of death between both sources.¹³ These datasets were linked using unique encoded identifiers and analyzed at ICES (formerly the Institute for Clinical Evaluative Sciences) using predefined protocols. The use of data in this project was authorized under section 45 of Ontario’s Personal Health Information Protection Act and did not require review by a research ethics board.

Study design and patients

We identified women diagnosed with first breast cancer in Ontario between April 1, 1997, and March 31, 2015, who were treated with a mastectomy or lumpectomy within 6 months of breast cancer diagnosis. We conducted a landmark analysis in which the analytical sample was restricted to breast cancer survivors who were alive 2 years after the breast cancer diagnosis date and aged ≥40 years at the index date without evidence of recurrent or distant disease. The index date for the study was set at 2 years after the date of breast cancer diagnosis to select women who completed any cancer chemotherapy or radiation therapy.

The exclusion criteria included non-Ontario residents and women without Ontario Health Insurance Plan eligibility in the year preceding breast cancer diagnosis (to allow for the assessment of baseline comorbidities). We also excluded women with invalid death data. To create a cohort free of baseline CVD, we excluded women who had been diagnosed with heart failure,¹⁵ ischemic heart disease,¹⁶ atrial fibrillation,⁷ stroke/transient ischemic attack,¹⁷ or peripheral artery disease⁷ or who received any valve intervention¹⁸ before the landmark date. To select women who had invasive breast cancer but were likely in remission, we excluded women with metastatic disease, carcinoma in situ, and multiple breast tumors as well as women who received chemotherapy, radiotherapy, or trastuzumab within the 6 months preceding the landmark date (ie, months 19-24 after breast cancer diagnosis) because this may have indicated cancer recurrence. Additionally, we excluded women with prior malignancy because it may confound the analysis of future oncologic outcomes. Because cancer stage is strongly associated with oncologic outcomes, we excluded women with missing American Joint Committee of Cancer (AJCC) stage data at the time of diagnosis (data available for women diagnosed after 2007). Application of these criteria resulted in a cohort of women diagnosed between April 1, 2007, and March 31, 2015, aged ≥40 years who had AJCC stage I to III cancer at diagnosis and survived 2 years afterward without prior cancer or prior CVD and who had not received cancer treatment in the preceding 6 months except for endocrine therapies.

The key exposure was incident MI/HF after the landmark date (ie, >2 years after breast cancer diagnosis), defined as a composite of MI¹⁹ and/or HF.¹⁵ The outcomes studied were: 1) cancer death; 2) new non-breast malignancy; and 3) new chemotherapy (definitions provided in Supplemental Table 1). Patients were followed until December 31, 2019, to avoid bias introduced by reduced health care coverage during the coronavirus disease-2019 pandemic.

Statistical analysis

We stratified women based on the development of incident MI/HF to compare baseline characteristics using medians with 25th to 75th percentiles (Q1-Q3) or counts (percentages) for continuous and categoric variables, respectively. Standardized mean differences (SMDs) were used to determine the magnitude of differences between groups, with SMD >0.10 indicative of potentially meaningful differences. We used the cumulative incidence function with 95% CIs to estimate the unadjusted risk of cancer death in the overall cohort while accounting for noncancer deaths as competing events. Similarly, we used the cumulative incidence function to estimate the risk of incident non–breast cancer diagnosis and new chemotherapy initiation while treating death of any cause as a competing risk.

Because we wanted to determine the potential etiologic association between incident MI/HF and outcomes of interest, we used multivariable cause-specific hazards regression adjusting for baseline characteristics.²⁰ The following covariates were included in the regression models: age, neighborhood income quintile, rural residence, Charlson Comorbidity Index, chemotherapy, radiotherapy, trastuzumab, left-sided breast cancer, diabetes, hypertension, AJCC stage at breast cancer diagnosis, chronic obstructive pulmonary artery disease, chronic kidney disease, and year of cohort entry. Incident MI/HF was modeled as a time-varying exposure in all models.

To study the association of the individual components of the exposure variable with outcomes, we developed a separate model in which incident MI/HF was modeled as a 4-level time-varying exposure with 4 possible values: no MI/HF, MI alone, HF alone, or both MI and HF. With this categorization, a woman would start as no MI/HF and could then progress to MI or HF alone and then progress to both MI and HF. Models are presented with HRs and their respective 95% CIs. The proportional hazards assumption for non–time-varying covariates was assessed by including an interaction term of each covariate with logarithmic transformation of time in the models (covariate∗log[t]). Whenever the P value for the interaction term was <0.05, the interaction term was retained in the model. Women who were event free were censored on December 31, 2019.

To assess for residual confounding, appendectomy,²¹ cholecystitis,²² and hospitalization with pneumonia²³ or urinary tract infection (UTI) were used as falsification endpoints. These outcomes were selected given their association with the following main unmeasured confounders: obesity^24,25 and smoking.^26,27 In order to assess the strength of the exposure-outcome associations in the presence of possible residual confounding, we report the E value along with each statistically significant HR as measured by an online calculator.²⁸ SAS Enterprise Guide Version 7.15 (2017, SAS Institute Inc) was used for all statistical analyses (except for E values). Statistical significance was determined by a 2-sided P value <0.05 for outcome analyses. Where applicable, small cells (n = <6) were suppressed according to ICES policies to minimize the risk of individual reidentification.

Results

A total of 30,694 women were alive 24 months after breast cancer diagnosis without documentation of chemotherapy in the prior 6 months (ie, chemotherapy completed if received), evidence of recurrent or distant disease, or any diagnoses of CVD as of the index date (Figure 1). Their baseline characteristics are provided in Table 1. The median age at the index date was 60 years (Q1-Q3: 52-69 years). Chemotherapy was used in 15,526 (50.6%) patients, and 23,640 (77.0%) women received radiotherapy. A total of 2,710 women (8.8%) received trastuzumab; this proportion increased to 11.8% among women diagnosed after 2015 (Supplemental Figure 1).

Figure 1.

Study Flowchart

This figure lists the number of patients who met inclusion criteria for the study and numbers excluded at each step to generate the final study cohort. OHIP = Ontario Health Insurance Plan.

Table 1.

Baseline Characteristics

	All Patients (N = 30,694)	No Incident MI/HF (n = 29,348)	Incident MI/HF (n = 1,346)	Standardized Difference
Median age	60 (52-69)	60 (52-68)	71 (62-79)	0.83
Median neighborhood income quintilea
Quintile 1b	4,945 (16.1)	4,661-4,665	280-284	0.13
Quintile 2	5,801 (18.9)	5,491 (18.7)	310 (23.0)	0.11
Quintile 3	6,121 (19.9)	5,864 (20.0)	257 (19.1)	0.02
Quintile 4	6,565 (21.4)	6,301 (21.5)	264 (19.6)	0.05
Quintile 5	7,186 (23.4)	6,956 (23.7)	230 (17.1)	0.17
Rural residencea	2,225 (7.2)	2,122 (7.2)	103 (7.7)	0.02
Charlson Comorbidity index
0	3,020 (9.8)	2,872 (9.8)	148 (11.0)	0.04
1	385 (1.3)	339 (1.2)	46 (3.4)	0.15
2	6,632 (21.6)	6,328 (21.6)	304 (22.6)	0.03
≥3	3,904 (12.7)	3,627 (12.4)	277 (20.6)	0.22
No hospitalization	16,753 (54.6)	16,182 (55.1)	571 (42.4)	0.26
Breast cancer stage
I	15,411 (50.2)	14,844 (50.6)	567 (42.1)	0.17
II	11,776 (38.4)	11,192 (38.1)	584 (43.4)	0.11
III	3,507 (11.4)	3,312 (11.3)	195 (14.5)	0.10
Chemotherapy	15,526 (50.6)	14,967 (51.0)	559 (41.5)	0.19
Radiotherapy	23,640 (77.0)	22,693 (77.3)	947 (70.4)	0.16
Trastuzumab	2,710 (8.8)	2,619 (8.9)	91 (6.8)	0.08
Left-sided breast cancer	15,441 (50.3)	14,726 (50.2)	715 (53.1)	0.06
Diabetes	4,397 (14.3)	4,025 (13.7)	372 (27.6)	0.35
Hypertension	13,025 (42.4)	12,135 (41.3)	890 (66.1)	0.51
COPD	3,440 (11.2)	3,161 (10.8)	279 (20.7)	0.28
Chronic kidney disease	493 (1.6)	430 (1.5)	63 (4.7)	0.19
Anthracyclinec	8,758 (61.7)	8,458 (61.7)	300 (61.5)	<0.01
Medicationsd
ACEI/ARB	900 (8.5)	781 (8.1)	119 (13.4)	0.17
Beta-blocker	252 (2.4)	216 (2.2)	36 (4.0)	0.10
Calcium-channel blocker	470 (4.5)	412 (4.3)	58 (6.5)	0.10
Loop diuretic	120 (1.1)	97 (1.0)	23 (2.6)	0.12
Mineralocorticoid antagonistb	27 (0.3)	22-26	1-5	0.02
Statin	645 (6.1)	574 (5.9)	71 (8.0)	0.08
Thiazide	531 (5.0)	455 (4.7)	76 (8.5)	0.15
Tamoxifen	454 (4.3)	413 (4.3)	41 (4.6)	0.02
Aromatase inhibitor	1,146 (10.9)	1,048 (10.9)	98 (11.0)	0.01

Values are median (Q1-Q3) or n (%).

ACEI = angiotensin-converting enzyme inhibitor; ARB = angiotensin receptor blocker; COPD = chronic obstructive pulmonary disease; HF = heart failure; MI = myocardial infarction.

^aMissing values for median neighborhood income quintile (n = 76) and rural residence (n = 250).

^bSmall cells are suppressed.

^cn = 14,199 with systemic chemotherapy data available.

^dn = 10,538 patients above the age of 65 years with medication data available. Standardized mean differences >0.1 were taken as indicative of a potentially meaningful difference.

In total, 1,150 (3.7%) patients developed HF during follow-up, 289 (0.9%) patients developed MI, and 93 (0.3%) patients developed both conditions, with MI/HF events occurring at a median of 3.9 years (Q1-Q3: 2.2-6.2 years) after the index date. The 5- and 10-year cumulative incidences of MI/HF were 3.0% (95% CI: 2.8%-3.2%) and 7.0% (95% CI: 6.6%-7.5%) respectively. Women who developed MI/HF during follow-up were older than women who did not develop MI/HF, had greater prevalence of diabetes and hypertension, and were less likely to have received chemotherapy or radiotherapy (SMD >0.1 for all comparisons). However, previous treatment with trastuzumab was not different between groups (SMD = 0.08).

A total of 3,522 women died during a median follow-up of 6.2 years (Q1-Q3: 4.3-8.2 years). There were 2,408 cancer deaths (68.4% of all deaths), 395 (11.2%) cardiovascular deaths, and 719 (20.4%) deaths from other causes. The 5- and 10-year cumulative incidences of cancer death were 5.9% (95% CI: 5.6%-6.1%) and 11.6% (95% CI: 11.0%-12.1%), respectively (Figure 2A). In univariable analyses, incident MI/HF was associated with higher hazards of cancer death (HR: 5.75; 95% CI: 4.96-6.67) (Supplemental Table 2). After adjusting for baseline differences, incident MI/HF was independently associated with a higher rate of cancer death (HR: 3.94; 95% CI: 3.38-4.59; E value = 7.34) (Table 2). When treating incident MI/HF as a 4-level time-varying variable, incident MI alone (HR: 2.14; 95% CI: 1.32-3.45; E value = 3.7) and incident HF alone (HR: 4.50; 95% CI: 3.84-5.29; E value = 8.47) were each associated with a higher hazard of cancer death (Table 2, Figure 3A).

Figure 2.

Cumulative Incidence Function Curves for Outcomes

This figure illustrates the cumulative incidence function curves for the outcomes of (A) cancer death, (B) non–breast cancer malignancy diagnosis, and (C) the initiation of new chemotherapy.

Table 2.

Cause-Specific Hazards Regression Models for Cancer Death, Non–Breast Cancer Malignancy, and New Chemotherapy

Exposure	Cancer Death			Non–Breast Cancer Malignancy			New Chemotherapy
	HR (95% CI)	P Value	E Value	HR (95% CI)	P Value	E Value	HR (95% CI)	P Value	E Value
Incident MI/HF	3.94 (3.38-4.59)	<0.001	7.34	1.39 (1.06-1.82)	0.016	2.13	1.25 (1.02-1.53)	0.029	1.61
Incident MI	2.14 (1.32-3.45)	0.002	3.7	1.42 (0.91-2.21)	0.12	—	0.99 (0.74-1.31)	0.93
Incident HF	4.50 (3.84-5.29)	<0.001	8.47	1.60 (1.32-1.92)	<0.001	2.58	1.21 (1.08-1.37)	0.002	1.54
Incident MI and HF	1.79 (0.80-4.00)	0.15	—	0.70 (0.31-1.56)	0.38	—	0.94 (0.60-1.47)	0.77

Covariates in all models include age, neighborhood income quintile, rural residence, Charlson Comorbidity Index, chemotherapy, radiotherapy, trastuzumab, left-sided breast cancer, diabetes, hypertension, breast cancer stage at diagnosis, year of cohort entry, chronic obstructive pulmonary disease, and chronic kidney disease.

Abbreviations as in Table 1.

Figure 3.

Forest Plots for Cancer Death, Non–Breast Cancer Malignancy, and New Chemotherapy

Forest plots with adjusted HRs (95% CIs) for (A) cancer death, (B) non–breast cancer malignancy, and (C) new chemotherapy. All models include age, neighborhood income quintile, rural residence, Charlson Comorbidity Index, chemotherapy, radiotherapy, trastuzumab, left-sided breast cancer, diabetes, hypertension, breast cancer stage at diagnosis, year of cohort entry, chronic obstructive pulmonary disease, and chronic kidney disease. HF = heart failure; MI = myocardial infarction.

A total of 1,773 patients developed a non–breast cancer malignancy during follow-up, with a cumulative incidence of 4.3% (95% CI: 4.1%-4.6%) at 5 years and 9.1% (95% CI: 8.6%-9.6%) at 10 years (Figure 2B). Lung (21%), colorectal (15%), and endometrial (13%) cancers and leukemia/lymphoma (9%) were the most common secondary malignancies. Incident MI/HF was associated with a higher hazard of being diagnosed with non-breast malignancy (univariable HR: 1.99; 95% CI: 1.53-2.60; multivariable HR: 1.39; 95% CI: 1.06-1.82; E value = 2.13; Table 2). Incident HF alone was associated with a higher hazard of non–breast cancer malignancy (HR: 1.60; 95% CI: 1.32-1.92; E value = 2.58), but the HR was not statistically significant for incident MI alone (HR: 1.42; 95% CI: 0.91-2.21) (Table 2, Figure 3B). The data describing univariable and multivariable associations are provided in Supplemental Table 3.

A total of 8,650 women were documented to receive new chemotherapy during follow-up. Of these, 7,743 women (89.5%) did not have a new cancer diagnosis (suggesting treatment was initiated for breast cancer recurrence). The cumulative incidence of new chemotherapy was 25.7% (95% CI: 25.2%-26.2%) and 33.6% (95% CI: 33.0%-34.3%) at 5 and 10 years, respectively (Figure 2C). Incident MI/HF was not associated with new chemotherapy upon univariable analysis (HR: 1.15; 95% CI: 0.94-1.40), but the association was significant after multivariable adjustment with a higher rate of new chemotherapy (HR: 1.25; 95% CI: 1.02-1.53; E value = 1.61). This was mainly driven by HF (HR: 1.21; 95% CI: 1.08-1.37; E value = 1.54) because there was no significant association between incident MI (HR: 0.99; 95% CI: 0.74-1.31) or MI and HF (HR: 0.94; 95% CI: 0.60-1.47) with new chemotherapy use (Table 2, Figure 3C).

Appendectomy after the index date was performed on 178 (0.6%) women. There was no association between incident MI/HF and appendectomy (HR: 0.88; 95% CI: 0.22-3.58; P = 0.86). Hospitalization for acute cholecystitis occurred in 203 (0.7%) patients and was positively associated with incident MI/HF (HR: 2.22; 95% CI: 1.02-4.82; P = 0.043). There were 436 (1.4%) patients who required hospital admission for pneumonia and 329 (1.1%) for UTI. Both conditions were associated with incident MI/HF with an HR of 2.86 (95% CI: 2.03-4.02; P < 0.001) and 1.94 (95% CI: 1.23-3.05; P = 0.004), respectively.

Discussion

In this population-based cohort study, breast cancer survivors who developed incident MI or HF after the completion of cancer therapy had a 4-fold increased hazard of dying from cancer after accounting for baseline characteristics (Central Illustration). Incident MI/HF was also associated with an approximately 40% increase in the hazard of a second non-breast malignancy and approximately 25% higher hazard of new chemotherapy. However, incident MI/HF was also associated with higher hazards of pneumonia and UTI hospitalizations but not for appendectomy.

Central Illustration.

Association Between Incident Myocardial Infarction or Heart Failure and Oncologic Outcomes in Breast Cancer Survivors

Among women with treated stage I to III breast cancer who survived 2 years after diagnosis, the development of incident myocardial infarction and/or heart failure was associated with increased hazard of cancer death, incident non-breast malignancies, and the requirement of new chemotherapy. ∗Chemotherapy, radiotherapy, and/or trastuzumab treatment within the preceding 6 months. +Pre-existent cardiovascular disease including heart failure, ischemic heart disease, atrial fibrillation, stroke/transient ischemic attack, peripheral artery disease, and valve disease requiring intervention. CVD = cardiovascular disease; HF = heart failure; MI = myocardial infarction.

It is well established that cardiotoxic cancer treatments and shared risk factors between cancer and CVD increase the prevalence of CVD in cancer survivors.²⁹ The potential reverse association (ie, CVD promoting cancer) has been a recent focus of investigation. There are several potential biological mechanisms through which MI and HF can promote carcinogenesis.³⁰ Koelwyn et al³¹ showed that MI accelerates breast cancer progression in mice by innate immune reprogramming through epigenetic modifications of monocytes in the bone marrow. Myocardial remodeling in response to pressure overload has been shown to induce cell proliferation of breast and lung tumors.³² Similarly, Meijers et al³³ showed that circulating factors secreted by the failing heart after an induced MI are able to promote tumor growth.

This hypothesis is supported by several observational studies. In 2013, a case-control study reported a 68% increased hazard of incident cancer among patients with prevalent HF.³⁴ Subsequently, the same group demonstrated a 2-fold increase hazard of incident cancer in a cohort of 1,081 patients with ischemic HF.⁴ In a nationwide Danish study of 9,307 patients with HF, the incidence rate ratio for subsequent cancer was 1.24 (95% CI: 1.15-1.33).³⁵ Similarly, in 28,763 Norwegian individuals without prevalent CVD or cancer, incident MI was associated with a 46% increased hazard of cancer over a 15-year follow-up.³ A population-based Italian study showed that HF was associated with 76% increased hazard of incident cancer but no association with incident breast cancer specifically. However, HF was associated with a doubling of the hazard of breast cancer–specific mortality in that study.³⁶

We identified 1 prior study focusing on the association of CVD with oncologic outcomes in women already diagnosed with breast cancer. In 2 different cohorts with a total of 3,268 women with breast cancer, incident MI or stroke was associated with ∼60% increased hazard of breast cancer–specific mortality and breast cancer recurrence.³¹ We confirm the presence of this association in our cohort, although we observed a larger effect size of cancer mortality associated with incident MI. In the previous study, incident CVD was not handled as a time-varying covariate, which could have underestimated the effect size of the association. Moreover, we extend these observations beyond MI and demonstrate that HF might be a stronger risk factor for adverse cancer outcomes.³⁷ It is also possible that the competing risk of cardiovascular death might be higher in the short-term for MI than chronic HF (ie, patients who die from MI cannot have cancer progression, whereas those with HF can have cancer recurrence/new cancer).

Our study was motivated by the potential “double jeopardy” of CVD promoting cancer growth in a cohort of breast cancer survivors. By focusing on breast cancer survivors, our study was designed to reduce surveillance bias.^3,38 For example, cancer may be more likely to be diagnosed after investigations for bleeding from antithrombotic therapy or increased health care contact and imaging after incident CVD.³⁹ Every participant in our cohort had a prior cancer diagnosis, making them more likely to have regular contact with the health system. This is because breast cancer survivors undergo regular surveillance for cancer recurrence independent of an incident CVD event. Confounding by indication can also occur when cancer treatment is delivered less intensively for patients with CVD. Our study was designed to limit this source of bias by including only women who were free of CVD throughout the period before and during breast cancer treatment.

Despite this approach, 3 of the 4 falsification endpoint analyses raise the possibility of residual confounding. A limitation of falsification tests is that sources of bias can be different for the actual exposure-outcome and exposure-falsification associations; therefore, they cannot explain the degree to which the exposure-outcome association is confounded.⁴⁰ For instance, infection is the most common cause for hospitalization in HF patients requiring hospital admission,⁴¹ particularly during the first year following HF hospitalization.⁴² Similarly, gallbladder disease occurs more commonly in patients with coronary artery disease.⁴³ We must also consider the potential for reverse causation wherein an occult malignancy can provoke MI or HF or increase vulnerability to infections.

After establishing that residual confounding may exist, the next relevant question is to what degree does it affect the observed exposure-outcome associations. Based on the E value for the primary outcome of cancer death, the HR for the association between a potential confounder and incident MI/HF and between the confounder and cancer death must both be at least 7.34 for it to explain away the observed association. We suggest that such an unmeasured confounder is unlikely to exist. The adjusted HR for all-cancer death among current smoking women vs never smoking women is 3.2⁴⁴ and that for cancer death attributed to obesity among cancer survivors is 1.17.⁴⁵ The E value for the analysis of non–breast cancer malignancy is 2.13. A meta-analysis of 6 studies reported a relative risk of 1.89 for a second primary cancer attributed to smoking,⁴⁶ whereas a retrospective cohort of 6,481 breast cancer survivors reported an adjusted relative risk of 1.09 for elevated body mass index.⁴⁷ Hence, residual confounding from smoking or obesity may not necessarily explain away the observed association between incident MI/HF and new cancer. The association between incident MI/HF and new chemotherapy is weaker; the smaller E value indicates it is more likely to be affected by residual confounding. It may also indicate less aggressive treatments in someone with a second cancer in the setting of established CVD. To provide further context, we present the E values for the HRs in a recent meta-analysis studying the association between HF and incident cancer in Supplemental Table 4.² This suggests that our study might be less affected by residual confounding than other studies.

From a clinical perspective, the implications of the association of incident CVD with worse cancer outcome do not hinge on causality per se but on the fact that incident CVD is associated with worse prognosis in cancer survivors. Such patients should be triaged for closer follow-up regardless of whether it is a causal association. Our study adds to the body of literature highlighting the importance of understanding CVD in cancer survivors but from a novel angle. One of the key objectives of early-stage breast cancer treatment is to reduce cancer recurrence, secondary malignancies, and associated mortality. A second primary malignancy in breast cancer survivors is associated with higher mortality than a first primary cancer in the general population.⁴⁸ The overarching goal of cardio-oncology as a field is to reduce the incidence and impact of CVD, particularly HF, in cancer patients and survivors. The goals of the 2 disciplines become even more aligned if incident CVD is associated with a higher risk of adverse cancer outcomes, thus providing an additional recognition of the adverse prognosis associated with CVD in cancer survivorship. However, it should not be assumed that CVD prevention would result in better cancer outcomes because the association may not be causal.

Study limitations

Beyond the aforementioned limitations of potential residual confounding and reverse causation, other limitations related to our use of administrative data sets include the inability to detect breast cancer recurrence. Although accuracy of cause of death was validated against physician review, it remains a subjective determination. It is possible that some of the cancer deaths were misclassified, and the higher risk of cancer death in cancer survivors with CVD represents increased death from noncancer causes. Patients with breast cancers at higher risk of recurrence may have been more heavily treated and could have developed subclinical CVD before the landmark date. This scenario is associated with increased likelihood of both incident CVD and cancer recurrence. Less aggressive treatment of recurrent or second malignancies because of CVD after this scenario can increase cancer-related mortality.

Conclusions

Breast cancer survivors who first developed incident MI and/or HF more than 2 years after their cancer diagnosis are at increased risk of dying from cancer, being diagnosed with a second malignancy, and late chemotherapy initiation. The mechanisms underlying these associations are multifactorial and may include residual confounding. Our results support the potential utility of further studying cardiovascular risk reduction during cancer survivorship because it is unknown whether and to what extent it can reduce adverse cancer outcomes.

Perspectives.

COMPETENCY IN MEDICAL KNOWLEDGE: Among treated breast cancer survivors without prior CVD, incident MI and/or HF are associated with an increased risk of cancer death, a higher risk of being diagnosed with a second malignancy, and the requirement of new chemotherapy.

TRANSLATIONAL OUTLOOK: There is a need for further study of the shared mechanisms and factors between CVD and cancer associations and the impact of cardiovascular risk factor modification on oncologic outcomes in cancer survivors.

Funding Support and Author Disclosures

This work was funded by the Ted Rogers Program in Cardiotoxicity Prevention and the Early Career Women’s Heart and Brain Health Chair (from the Heart and Stroke Foundation of Canada and Canadian Institutes of Health Research) to Dr Abdel-Qadir. Dr Abdel-Qadir was funded by the National New Investigator Award from the Heart and Stroke Foundation of Canada. Dr Thavendiranathan (147814) was supported by a Canada Research Chair in Cardiooncology (CRC-2019-00097) and the Canadian Cancer Society/Canadian Institutes of Health Research’s W. David Hargraft Grant. This study was supported by ICES, which is funded by an annual grant from the Ontario Ministry of Health (MOH) and the Ministry of Long-Term Care (MLTC). This document used data adapted from the Statistics Canada Postal CodeOM Conversion File, which is based on data licensed from Canada Post Corporation, and/or data adapted from the Ontario Ministry of Health Postal Code Conversion File, which contains data copied under license from Canada Post Corporation and Statistics Canada. Parts of this material are based on data and/or information compiled and provided by the MOH, CIHI, and OH. Parts of this report are based on Ontario Registrar General (ORG) information on deaths, the original source of which is ServiceOntario. The views expressed therein are those of the author and do not necessarily reflect those of ORG or the Ministry of Public and Business Service Delivery. The analyses, conclusions, opinions and statements expressed herein are solely those of the authors and do not reflect those of the funding or data sources; no endorsement is intended or should be inferred. Dr Thavendiranathan has received consultation and speaker honoraria from Amgen, Boehringer Ingelheim, General Electric, and Takeda. Dr Amir has received honoraria from Sandoz and Seagen; and has provided consulting to AstraZeneca and Novartis. Dr Lee is the Ted Rogers Chair in Heart Function Outcomes, University Health Network, University of Toronto. Dr Abdel-Qadir has received consultation and speaker honoraria from Amgen, AstraZeneca and Jazz Pharmaceuticals. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.