Associations of Health-Related Quality of Life and Sleep Disturbance with Cardiovascular Disease Risk in Post-Menopausal Breast Cancer Survivors

Abstract

Background:

Breast cancer (BC) survivors are at an increased risk of long-term cardiovascular disease (CVD), often attributed to cancer treatment. However, cancer treatment may also negatively impact health-related quality of life (HRQoL), a risk factor of CVD in the general population.

Objective:

We examined whether sleep disturbance, and physical or mental HRQoL were associated with CVD risk in BC survivors.

Methods:

We conducted a longitudinal analysis in the Women’s Health Initiative (WHI) of post-menopausal women diagnosed with invasive BC during follow-up through 2010 with no history of CVD prior to BC. The primary outcome was incident CVD, defined as physician-adjudicated coronary heart disease or stroke, after BC. Physical and mental HRQoL, measured by the SF-36 Physical and Mental Component Scores (PCS and MCS, respectively), and sleep disturbance, measured by the WHI Insomnia Rating Scale (IRS), were recorded post-BC. Time-dependent Cox proportional hazards models were used starting at BC diagnosis until 2010 or censoring and adjusted for relevant confounders.

Results:

In 2,884 BC survivors, 157 developed CVD during a median follow-up of 9.5 years. After adjustment, higher PCS scores were significantly associated with a lower risk of CVD [HR=0.90 (95%CI: 0.81, 0.99); per 5-point increment in PCS]. No associations with CVD were found for MCS or IRS.

Conclusions:

In BC survivors, poor physical HRQoL is a significant predictor of CVD.

Implications for Practice:

Our findings highlight the importance for nurses to assess and promote physical HRQoL as part of a holistic approach to mitigating the risk of CVD in BC survivors.

Introduction

Survival rates for breast cancer (BC) continue to improve in the United States and, as of 2021, the 5-year survival rate is nearly 100% for early stage BC and 90% for all stages of BC.¹ While advancements in cancer treatment have contributed to improvements in BC mortality, BC survivors are at an increased risk of long-term cardiovascular disease (CVD) morbidity and mortality.² Cancer survivors also experience numerous other adverse effects and symptoms such as increased fatigue, pain, sleep disturbance, depression, and reduced physical function, all of which can negatively impact health-related quality of life (HRQoL).^{3, 4} HRQoL is a broad, multidimensional concept used to describe a person’s self-perceived health status, which includes an assessment of physical and emotional well-being.⁵

HRQoL is increasingly being recognized as a predictor of CVD risk in the general population.^6–8 Previous studies have shown that symptoms and adverse effects that are typically associated with HRQOL, such as fatigue, pain, sleep disturbance, depression, and physical function, are linked with an increased risk of a variety of CVD events.^9–12 In particular, one study within the Women’s Health Initiative (WHI) Observational Study found that participants with poorer baseline self-reported physical HRQoL, as identified by lower scores on the Rand Short-Form (SF)-36 Physical Component Summary (PCS), were at a 2-fold higher risk of coronary heart disease and stroke compared to those with higher scores. This study also examined associations with mental HRQoL (using the SF-36 Mental Component Summary [MCS]), though no significant associations were found.¹² In a separate study within the WHI, similar results were found for physical HRQoL when examining three-year change scores from baseline.¹³

However, these associations have not been adequately explored in BC survivors. Although the risk of CVD in breast cancer survivors is typically attributed to the direct cardiotoxic effects of cancer treatment, these therapies may also contribute to CVD risk indirectly by impacting health-related quality of life (HRQoL). Mechanistically, cancer treatments contribute to inflammation and oxidative stress, which are known, in part, to lead to the development of CVD. Likewise, inflammation and oxidative stress are associated with components of HRQoL, such as fatigue, sleep disturbance, depression, pain, and reduced physical function.^{14, 15} Additionally, cancer itself leads to a higher inflammatory state with BC patients reporting high levels of symptoms prior to ever receiving treatment.¹⁶ Therefore, HRQoL could partly contribute to the risk of CVD in BC survivors. However, it is unknown whether HRQoL is an independent predictor of CVD risk in a cancer survivorship population, who is more vulnerable to poor HRQoL and often has a greater number of cardiovascular risk factors than the general population.¹⁷

The purpose of this analysis was to examine the association of post-cancer physical and mental HRQoL, as measured by SF-36 PCS and MCS scores, and sleep disturbance, as measured by the WHI Insomnia Rating Scale, with CVD incidence in post-menopausal women diagnosed with BC.

Materials & Methods

Study population

The WHI is a population-based prospective cohort study of 161,808 post-menopausal women enrolled between 1993–1998 and followed initially through 2005.¹⁸ Women aged 50 to 79 were enrolled at 40 clinical centers nationwide in one or more Clinical Trials or an Observational Study. At the end of the main study in 2005, women were asked to participate in an extension study with follow-up through 2010. The WHI project was reviewed and approved by the Fred Hutchinson Cancer Research Center Institutional Review Board. All women provided written informed consent prior to data collection.

The Life and Longevity After Cancer (LILAC) study is a cancer survivorship sub-cohort within the WHI.¹⁹ A primary goal of LILAC is to retrospectively collect information on cancer treatment in women free of cancer at WHI baseline but with a confirmed cancer diagnosis of one of eight cancers (breast, endometrial, ovarian, lung, and colorectal cancers, and melanoma, lymphoma, and leukemia) during WHI follow-up in the Clinical Trials or Observational Study. Beginning in 2013, the LILAC study consented eligible participants and retrospectively abstracted available treatment data.¹⁹

We conducted a secondary longitudinal analysis using data from the Clinical Trials and Observational Study cohorts through 2010. Women diagnosed with an incident, invasive stage I-III BC diagnosis through 2010 were eligible for this analysis (n = 8,066). Women were excluded if they 1) had an adjudicated CVD outcome prior to BC (n = 178) or 2) did not have a documented MCS, PCS, or WHI Insomnia Rating Scale score between BC and follow-up (n = 5,004) resulting in a final sample size of 2,884 BC survivors (Figure 1). A sensitivity analysis was conducted in LILAC to account for cancer treatments. In this subsample, we included 1,662 BC survivors who were enrolled in LILAC with treatment data available (Figure 1).

Figure 1. Study Flow Chart of Included Participants.

Participants were eligible for this analysis if they were diagnosed with an incident, invasive stage I-III breast cancer through 2010. Women were excluded if they 1) had an adjudicated cardiovascular disease outcome prior to breast cancer or 2) did not have a documented exposure between breast cancer and follow-up resulting in a final sample size of 2,884 women. A sensitivity analysis was conducted in the Life and Longevity After Cancer Cohort, which further restricted the sample to 1,622 participants.

Outcome

The primary outcome of interest was an incident, physician-adjudicated CVD event defined as having coronary heart disease, which includes myocardial infarction and coronary death, or stroke, after BC diagnosis. CVD events were adjudicated on all participants through 2010.

WHI cardiac adjudication methods have been in described in detail elsewhere.²⁰ In summary, potential outcomes were identified through self-reported semi-annual or annual medical history forms. If an event was self-reported, medical records were requested and events were physician-adjudicated using standardized criteria. Cause of death, used in the definition for coronary heart disease, was determined through linkage with the National Death Index.

Coronary heart disease was defined as having an acute myocardial infarction requiring hospitalization, a silent myocardial infarction, or coronary heart disease death. Both definite and probable myocardial infarctions were included and were classified using an algorithm that consisted of a combination of data including medical history, electrocardiogram readings, and cardiac biomarkers. Coronary heart disease death was defined as death with an underlying cause of coronary heart disease with one or more of the following: hospitalization for myocardial infarction within 28 days before death, previous MI or angina, death coronary heart disease from a procedure related to coronary artery disease, or a death certificate indicating coronary heart disease as the underlying cause of death.²⁰ Stroke was defined as having a rapid onset of persistent neurologic deficit attributed to an obstruction or rupture of the brain arterial system without evidence for other cause and supported by imaging studies.²⁰

Exposures

Three primary exposures were of interest: PCS (physical HRQoL), MCS (mental HRQoL), and WHI Insomnia Rating Scale (sleep disturbance) scores. The SF-36 and WHI Insomnia Rating Scale were measured at multiple time points throughout the WHI main study. In the Observational Study, both the SF-36 and WHI Insomnia Rating Scale were measured at baseline and year 3 of follow-up. In the Clinical Trial, they were measured at baseline and years 3, 6, & 9 of follow-up in a subset of participants.

The SF-36 consists of 8 subscales, each ranging from 0 to 100 with higher scores indicating a more favorable health state: (a) physical functioning, (b) general health perceptions, (c) bodily pain, (d) vitality (i.e., energy/fatigue), (e) role limitations due to physical health, (f) mental health (i.e. depression), (g) role limitations due to emotional problems, and (h) social functioning. As described elsewhere, the MCS and PCS were created using principal components methods using scores from all 8 subscales (Supplemental Material).²¹ MCS and PCS were considered missing if a participant was missing any of the 8 subscales. Convergent validity of the SF-36 with the Functional Assessment of Cancer Therapy-Breast, a HRQoL measure designed specifically for breast cancer, shows strong correlations.²² Lastly, in studies of cancer survivors, the SF-36 has demonstrated high internal consistency with alpha coefficients ranging between 0.76–0.91 across subscales.²²

Sleep disturbance was measured using the WHI Insomnia Rating Scale, which is composed of five questions. These questions assessed the frequency of insomnia symptoms (sleep latency, sleep maintenance, early morning awakening, and average sleep quality) over the past four weeks prior to each assessment. Total Insomnia Rating Scale scores range from 0–20 with higher scores indicating greater sleep disturbance. A score of 9 has been validated to classify clinically meaningful sleep disturbance.²³ The Insomnia Rating Scale shows acceptable internal consistency in a large sample of WHI participants with an alpha coefficient of 0.79 (95% CI: 0.70 – 0.85) and high construct validity when compared with measures known to be related to sleep.²³

Additional variables

Demographic information on age, self-identified race/ethnicity, and education were collected on WHI baseline questionnaires. Additional information on potential confounders, such as smoking, physical activity, and alcohol consumption were collected via self-report questionnaires. Physical activity was measured as total metabolic equivalent-minutes per week (MET-minutes/week) calculated from questionnaires as previously described.²⁴ Alcohol consumption was recorded as the number of alcoholic servings of beer, wine, and liquor per week. BMI was measured at clinic visits by trained WHI staff. Medication use data were collected as part of the WHI clinic visits. For this analysis, we were interested in statin use and antihypertensive medications, which included angiotensin-converting enzyme inhibitors, angiotensin receptor antagonists, calcium channel blockers, beta blockers, and diuretics. Current smoking status, physical activity, alcohol consumption, body mass index, and medication use were measured at multiple timepoints throughout WHI follow-up. As such, data from the time point most proximal, but prior to BC diagnosis, was used. Lastly, cancer characteristics, including age of diagnosis, stage at diagnosis, laterality, hormone receptor status, and cancer treatments, including receipt of chemotherapy and radiation, were collected from medical history records. Chemotherapy use was dichotomized based on documentation of chemotherapy or other targeted therapies, including Trastuzumab, based on medical record review.

Statistical Methods

Normality of continuous variables was assessed visually. Pre-BC characteristics of the sample are reported using means and standard deviations or frequencies and proportions for continuous and categorical variables, respectively. Bivariate statistics, including t-tests and chi-square tests for continuous and categorical variables, respectively, were used to compare tertiles of PCS and MCS, and compare those with a WHI Insomnia Rating Scale score with a cut point of 9 as previously described. Median (25^th to 75^5th quartiles) were calculated for follow-up time.

Incidence rates (cases/1,000 person-years) and 95% confidence intervals (CI) were calculated for CVD for the entire cohort and by categories of PCS, MCS, and Insomnia Rating Scale scores. The cumulative incidence of CVD for each exposure (i.e., post-cancer MCS, PCS, and WHI Insomnia Rating Scale) was estimated using time-dependent cumulative incidence curves to account for differences in time to each exposure from BC diagnosis. Time to incident CVD was defined as the number of days since BC diagnosis. Participants without a CVD outcome were censored at time of last follow-up through 2010.

Multivariable-adjusted time-dependent Cox proportional hazards models were used to examine the association of time to PCS, MCS, or WHI Insomnia Rating Scale scores as time-varying exposures with risk of incident CVD in separate models. To reduce the possibility of reverse causation, exposures (PCS, MCS, or WHI Insomnia Rating Scale scores) within 6 months of CVD or censoring were excluded. MCS, PCS, and SF-36 subscale scores were modeled as continuous variables, whereas Insomnia Rating Scale scores were dichotomized using a cut point of 9 as described previously. For MCS, PCS, and SF-36 subscale scores, hazard ratios (HRs) correspond to a 5-unit difference as a 5-point change in SF-36 scores has been considered a minimally clinically important difference in prior studies in BC cohorts.^{25, 26} Covariates were decided a priori and included age at diagnosis (years), education (high school or General Educational Development Test, > High school – Bachelor’s degree, > Bachelor’s degree), race/ethnicity (non-Hispanic White, non-Hispanic Black, other), cancer stage (local, regional, distant), body mass index (kg/m²), physical activity (MET-min/week), alcohol (servings/week), current smoking, antihypertensive medications, and statin use. For models examining PCS, MCS was additionally adjusted for and vice versa. Models for WHI Insomnia Rating Scale were adjusted for both PCS and MCS. A complete case analysis was performed (n = 2,686). Overall, rates of missing data are minimal with no variable having missing data in more than 1% of participants.

We conducted an exploratory analysis to see which, if any, individual SF-36 subscales scores, were associated with CVD risk. Separate models were created for each SF-36 subscale as the independent variable and were additionally adjusted for the most proximal pre-BC SF-36 subscale of interest. We additionally conducted a sensitivity analysis within LILAC to determine if BC cancer treatment was a significant confounder. The analysis was repeated in the sub-cohort of 1,622 BC survivors in LILAC with and without chemotherapy and radiation treatment in the models to determine if the addition of treatment influenced the hazard ratio estimates.

The proportional hazards assumption was confirmed using Schoenfeld residuals. Multicollinearity was evaluated using variance inflation factors with no violations observed. A two-sided p-value of 0.05 was used to determine statistical significance. All analyses were performed using R Version 3.6.1 (R Foundation for Statistical Computing, Vienna, Austria).

Results

Characteristics of study sample

A total of 2,884 women with a mean (SD) age at diagnosis of 67.3 (6.94) were followed for a median (25^th to 75^th quartiles) of 9.5 (7.0, 11.7) years after BC diagnosis. Overall, a majority of women identified as non-Hispanic White, had at least a bachelor’s degree, and were non-smokers. Most women were diagnosed with local BC with estrogen/progesterone receptor positive tumors (Table 1). The mean (SD) baseline PCS, MCS, and WHI Insomnia Rating Scale scores were 48.0 (9.9), 53.7 (8.5), and 6.5 (4.3), respectively. The median (25^th to 75^th quartiles) time from BC to next measurement of PCS, MCS, or WHI Insomnia Rating Scale was 2.3 (1.0, 4.2) years, whereas the median (25^th to 75^th quartiles) time from PCS, MCS, or WHI Insomnia Rating Scale to CVD or censoring was 5.8 (5.5, 9.2) years.

Table 1.

Baseline Characteristics of Overall Sample and by Physical Component Scale Tertiles

	Overall (N = 2,884)	PCS tertile 1 (N = 910) (<46.5)	PCS tertile 2 (N = 910) (46.5 – 53.9)	PCS tertile 3 (N = 909) (>53.9)	P-value
Demographics
Age at breast cancer diagnosis, mean (SD)	67.3 (6.94)	68.6 (6.8)	67.4 (6.9)	65.7 (6.8)	< .001
Ethnicity, n (%)					.740
Hispanic/Latino	65 (2.3)	19 (2.1)	23 (2.5)	17 (1.9)
Not Hispanic/Latino	2813 (97.5)	889 (97.7)	884 (97.1)	891 (98.0)
NA	6 (0.2)	2 (0.2)	3 (0.3)	1 (0.1)
Race, n (%)					.443
White	2,534 (87.9)	794 (87.3)	800 (87.9)	819 (90.1)
Black	192 (6.7)	69 (7.6)	57 (6.3)	41 (4.5)
Asian	59 (2.0)	15 (1.6)	21 (2.3)	21 (2.3)
American Indian/Alaskan Native	7 (0.2)	4 (0.4)	1 (0.1)	2 (0.2)
Native Hawaiian/Pacific Islander	6 (0.2)	2 (0.2)	2 (0.2)	2 (0.2)
Some other race	21 (0.7)	4 (0.4)	8 (0.9)	8 (0.9)
More than one race	59 (2.0)	20 (2.2)	18 (2.0)	15 (1.7)
NA	6 (0.2)	2 (0.2)	3 (0.3)	1 (0.1)
Education, n (%)					< .001
< High school	92 (3.2)	43 (4.7)	16 (1.8)	16 (1.8)
High school or GED	427 (14.8)	159 (17.5)	132 (14.5)	118 (13.0)
> High school – Bachelor’s degree	1,407 (48.8)	459 (50.4)	442 (48.6)	432 (47.5)
> Bachelor’s degree	936 (32.5)	246 (27.0)	310 (34.1)	337 (37.1)
NA	22 (0.8)	3 (0.3)	10 (1.1)	6 (0.7)
Cardiovascular risk factors
Current smoker, n (%)	161 (5.6)	48 (5.3)	54 (5.9)	48 (5.3)	.080
BMI (kg/m2), mean (SD)	28.5 (6.0)	30.3 (6.40)	28.5 (5.7)	26.7 (5.3)	< .001
Physical activity (MET-hours/week), mean (SD)	12.11 (13.5)	8.8 (11.1)	13.0 (13.6)	14.7 (14.8)	< .001
Alcohol (servings/week), mean (SD)	2.43 (4.9)	2.1 (5.5)	2.6 (4.8)	2.7 (4.6)	.029
Medications, n (%)
Antihypertensives	314 (10.9)	130 (14.3)	93 (10.2)	76 (8.4)	<.001
Statins	97 (3.4)	28 (3.1)	38 (4.2)	29 (3.2)	.372
Cancer characteristics, n (%)
Cancer stage					.399
Local	2181 (75.6)	685 (75.3)	708 (77.5)	669 (73.6)
Regional	659 (22.9)	212 (23.3)	189 (20.8)	223 (24.5)
Distant	17 (0.6)	4 (0.4)	6 (0.7)	7 (0.8)
NA	27 (0.9)	9 (1.0)	7 (0.8)	10 (1.1)
Laterality of primary tumor					.336
Right	1452 (50.3)	460 (50.5)	466 (51.2)	450 (49.5)
Left	1430 (49.6)	448 (49.2)	444 (48.8)	459 (50.5)
Tumor markers
Estrogen receptor +	2201 (76.3)	699 (76.8)	695 (76.4)	688 (75.5)	.347
Progesterone receptor +	1798 (62.3)	587 (64.5)	568 (62.4)	540 (59.4)	.346
HER2 neu +	284 (9.8)	84 (9.2)	93 (10.2)	94 (10.3)	.746

Abbreviations: BMI, body mass index; HER2, human epidermal growth factor receptor 2; MET, metabolic equivalents; NA, not available; PCS, Physical Component Score; SD, standard deviation

Comparison of baseline characteristics by categories of PCS, MCS, and WHI Insomnia Rating Scale scores

When comparing tertiles of pre-BC PCS, those in the lowest tertile were more likely to be older, have a higher body mass index, be less physically active, consume less alcohol per week, and report worse mental HRQoL and greater sleep disturbance (Table 1, Supplemental Figure 1). Similar results were found when comparing tertiles of MCS; however, those in the lowest MCS tertile were more likely to be younger, which is in contrast with PCS (Supplemental Table 1, Supplemental Figure 1). Approximately 28.3% of the sample had an Insomnia Rating Scale score ≥ 9 pre-BC, indicating clinically relevant sleep disturbance. Those with sleep disturbance were also more likely to be older, have higher body mass index, report less physical activity, and have lower MCS and PCS scores (Supplemental Table 2, Supplemental Figure 1).

Cumulative incidence curves and incidence of CVD

During follow-up, there were 157 CVD outcomes with an overall incidence rate of 5.9 (95% CI: 5.0, 6.8) events per 1,000 person-years. The incidence rate of CVD decreased with higher PCS tertiles, with a significant difference between the 1^st and 3^rd PCS tertiles (Supplemental Table 3). When examining MCS tertiles, participants in the 3^rd tertile had a higher incidence rate of CVD; however, this was not significantly different when compared to the 1^st or 2^nd tertiles (Supplemental Table 3). Lastly, participants with an Insomnia Rating Scale score ≥ 9 had a higher incidence of CVD compared to those with a score < 9; however, this difference was not statistically significant (Supplemental Table 3). Similar trends were seen when examining the unadjusted cumulative incidence curves (Figure 2).

Figure 2. Unadjusted Time-Dependent Cumulative Incidence Curves of Cardiovascular Disease by A) PCS Tertiles, B) MCS Tertiles, and C) WHI Insomnia Rating Scale Categories.

Higher PCS and MCS tertiles correspond to poorer health-related quality of life. WHI Insomnia Rating Scale scores are categorized using a cut point of 9 with scores >= 9 corresponding to clinically relevant insomnia. The cumulative incidence of CVD increases with higher tertiles of PCS scores with a significant difference of CVD risk occurring between the 1^st and 3^rd tertiles. There are no significant differences in CVD risk when comparing MCS tertiles or WHI Insomnia Rating Scale scores. Abbreviations: CVD, cardiovascular disease; MCS, Mental Component Score; PCS, Physical Component Score; WHI, Women’s Health Initiative

Adjusted Cox PH analysis

In multivariable-adjusted Cox proportional hazards models, higher PCS scores were significantly associated with a lower risk of CVD. For each 5-point increment in PCS scores, the risk of CVD was 9% lower [HR = 0.90 (95% CI: 0.81, 0.99); P = .049] (Table 2). There was no association found between MCS scores or Insomnia Rating Scale scores and risk of CVD.

Table 2.

Cox Proportional Hazards Models for Association of Quality-of-Life Measures with Risk of Cardiovascular Disease

	Events	Hazard Ratioa (95% CI)	P-value
Physical Component Scoreb,c	117	0.90 (0.81, 0.99)	.038
Mental Component Scoreb,d	117	0.98 (0.87, 1.09)	.732
WHI Insomnia Rating Scaleb,e,f	121	1.25 (0.83, 1.89)	.287
SF-36 Subscales g
PCS
Physical function	131	0.95 (0.91, 0.99)	.023
Role limitations due to emotional physical health	131	0.99 (0.96, 1.01)	.229
Pain	133	0.98 (0.95, 1.01)	.214
General health	134	0.92 (0.87, 0.98)	.006
MCS
Vitality (i.e., fatigue)	131	0.96 (0.91, 0.99)	.020
Emotional well-being (i.e., depression)	130	0.96 (0.89, 1.02)	.213
Social functioning	134	0.97 (0.94, 1.01)	.106
Role limitations due to emotional well-being	132	1.00 (0.97, 1.03)	.952

Abbreviations: PCS, Physical Component Score; MCS, Mental Component Score; WHI, Women’s Health Initiative; IRS, Insomnia Rating Scale

^aHR represents a 5-unit difference

^bAdjusted for age at diagnosis (years), education (HS or GED, > HS – Bachelor’s, > Bachelor’s), race/ethnicity (Non-Hispanic White, Non-Hispanic Black, Other), baseline QoL measure⁵, cancer stage (local, regional, distant), BMI (kg/m²), physical activity (MET-min/week), alcohol (servings/week), smoking (yes/no), antihypertensive meds (yes/no), statin use (yes/no)

^cAdditionally adjusted for baseline MCS

^dAdditionally adjusted for baseline PCS

^eAdditionally adjusted for baseline PCS & MCS

^fEach model is adjusted for the baseline variable according to the exposure

^gWHI IRS dichotomized using cut point of 9; < 9 serves as the reference category

Exploratory and sensitivity analyses

When examining individual sub-scales of the SF-36, higher scores on vitality, physical functioning, and perceptions of general health were all significantly associated with a lower risk of CVD (Table 4). For each 5-point increment in vitality, physical functioning, and perceptions of general health subscale scores, the risk of CVD was lower by 4% [HR = 0.96 (95% CI: 0.91, 0.99)], 5% [HR = 0.95 (95% CI: 0.91, 0.99)] and 8% [HR = 0.92 (95% CI: 0.87, 0.98)], respectively. In the subset of 1,662 LILAC participants with treatment data available, the addition of chemotherapy and radiation treatment to the models did not alter the HRs, thus, cancer treatment did not appear to be a significant confounder (data not shown).

Discussion

This is the first study, to our knowledge, to investigate whether physical and mental HRQoL and sleep disturbance are independently associated with the risk of CVD in BC survivors. Overall, this study found that lower PCS scores (i.e., poorer physical HRQoL) were a significant predictor of CVD in BC survivors after adjustment for known cardiovascular risk factors. Additionally, this association appeared to be largely driven by physical function and perceptions of general health, which are two SF-36 subscales that are more heavily weighted in the calculation for PCS. By contrast, we found no association of either MCS scores or sleep disturbance with CVD risk. While associations with MCS scores were not significant, lower vitality sub scores (i.e., a measure of energy/fatigue) were a significant predictor of CVD despite being a component of MCS.

The findings from this study are supported by research in non-cancer populations showing that poor physical HRQoL is a significant predictor of CVD and CVD-related mortality.^{6–8, 27–29} Prior research has shown that BC survivors are at a greater risk of both CVD and poor physical HRQoL compared to the general population.³⁰ Additionally, BC survivors typically have a higher burden of cardiovascular risk factors than the general population that accounts for a large portion of the risk of CVD. Despite adjusting for cardiovascular risk factors and cancer characteristics, physical HRQoL remained an independent predictor of CVD in this study. When compared to a study in the overall WHI, the association between physical HRQoL and CVD risk appears to be stronger in BC survivors.²⁸ However, underlying differences in sample populations and definition of CVD may partially account for this difference. Further studies are needed to make definitive comparisons between the general population and BC survivors.

The mechanism by which PCS scores influence CVD risk is unclear. One possible explanation is the role of inflammation. The PCS score is largely calculated by four subscales of the SF-36 including physical functioning, perceptions of general health, physical limitations due to physical function, and pain. Prior studies have shown that poorer physical HRQoL, and these subscale measures, are associated with higher circulating level of inflammatory cytokines even in healthy individuals without evidence of clinical disease³¹. This is important for BC survivors as cancer treatments produce an inflammatory response that has persisted for up to 10 years after treatment in some individuals and is greater compared to the those without a history of cancer.^{32, 33} However, it is unclear whether low physical HRQoL is an indicator of an inflammatory state or if it is a driver of inflammation. Prior studies have found significant associations between self-rated physical health and CVD risk even after adjustment for inflammatory biomarkers.³⁴ This suggests low physical HRQoL may independently contribute to CVD risk above and beyond the role of inflammation. Future studies are needed to disentangle the mechanisms contributing to this association.

Another explanation is the possibility of reverse causation. It has been suggested that changes in PCS are related to the presence of undetected subclinical CVD. To minimize the risk of reverse causation, we excluded PCS scores within 6 months before CVD or censoring. We also excluded participants with an adjudicated CVD outcome prior to BC. To account for the possibility that participants had cardiac disease that did not develop into CHD or stroke, we adjusted for statins and antihypertensive medications. It has also been suggested that individuals with lower PCS scores are less likely to engage in healthy lifestyle habits, such as physical activity (as supported by our data), which is known to reduce the risk of CVD.³⁵ However, we found significant associations between PCS scores and CVD risk despite adjustment for physical activity.

In this study, we did not find evidence to support either MCS scores or sleep disturbance as risk factors for CVD in BC survivors. In non-cancer populations, studies have also found weak or null associations when examining MCS scores as a risk factor for CVD with similar mean MCS scores as reported in this study.^{6, 28, 29} While consistent with previous reports, this finding is interesting as mental health disorders, such as depression and anxiety, have been highly associated with CVD risk.³⁶ However, the MCS is a broad measure of mental wellbeing and is not designed to capture clinical depression or anxiety, which could explain the null findings. Future studies should use measurement tools designed specifically to measure depression and anxiety.

The lack of association between sleep disturbance and CVD risk found in this study is inconsistent with previous literature, which supports that sleep disturbances, including insomnia, are associated with CVD risk.^{37, 38} While this could indicate the WHI Insomnia Rating Scale is not sufficiently sensitive to accurately capture insomnia, this is unlikely as a previous study in the WHI OS found that participants with a WHI Insomnia Rating Scale score ≥ 9 had a significantly higher risk of incident CHD.³⁷ The lack of association in this study could be related to sample size. Although not statistically significant, the hazard ratio indicates a higher risk of CVD with higher WHI Insomnia Rating Scale scores. Thus, a larger sample size is likely needed to detect a smaller effect size related to sleep disturbance.

When examining individual subscales, we found energy/fatigue, general health perception, and physical functioning to be significantly associated with incident CVD. These findings provide further support for an association between self-rated health and fatigue with CVD risk.³⁴ Like self-rated health, fatigue is hypothesized to contribute to CVD through inflammatory pathways in addition to numerous metabolic and hemodynamic processes.³⁹ Physical function has also been linked to CVD risk through physical activity and exercise capacity.⁴⁰ This may also be linked to inflammation as exercise interventions have been shown to reduce circulating inflammatory biomarkers.⁴¹ However, these results should be interpreted with caution as these subscales are highly correlated with one another. Future studies are needed to examine the effects of these constructs in relation to CVD risk using additional measurements.

Study Limitations

As this is an observational study, residual confounding is a potential limitation, although we attempted to minimize confounding by excluding participants with an CVD event prior to BC and adjusting for variety of CVD risk factors. While we were not able to adjust for inflammatory biomarkers, we controlled for variables that likely contribute to inflammation, such as cancer treatment.⁴² In our sensitivity analysis within the LILAC cohort, cancer treatment did not appear to be a confounder in the relationship between PCS and risk of CVD. However, treatment data was not available on all participants, and we were unable to examine different types of chemotherapy or include hormonal therapies. We also adjusted for physical activity, BMI, depression, and smoking, which are all predictors of inflammation in BC survivors.¹⁵ A large proportion of participants were excluded from this analysis based on the lack of HRQoL or sleep disturbance measures between BC diagnosis and CVD or censoring. However, the incidence of CVD in this analytic sample is comparable to the WHI BC cohort prior to study exclusions. Lastly, there was a relatively long gap between cancer diagnosis and the assessment of PCS, MCS, and Insomnia Rating Scale scores with a median time of 2.3 years. Thus, we likely had sufficient follow-up time to ensure participants completed their cancer treatment, which can take 12 months or more. Additionally, cancer treatment can have a long-lasting impact on many subscale measures, such as fatigue and depression.⁴³ Thus, these findings could provide insight into the effects of long-term effects of cancer and its treatment on CVD risk.

Despite these limitations, this study has many strengths. This is the first study, to our knowledge, to examine the role of physical and mental HRQoL and sleep disturbance in CVD risk in BC survivors. The WHI has a large sample size of BC survivors who were followed over 10 years of post-BC follow-up in this study. Given the large sample size and long-term follow-up, this study is adequately powered to examine cardiac outcomes that typically occur 10 or more years after cancer treatment. Lastly, CVD was physician-adjudicated, which reduces the possibility of misclassification.

Clinical implications

Our findings highlight the importance of assessing and promoting physical HRQoL as part of a holistic approach to mitigating the risk of CVD in BC survivors. Given the greater recognition of the cardiovascular effects of cancer treatments, it is important for clinicians and nurses to recognize and assess risk factors of CVD throughout survivorship to provide timely interventions. As cancer survivors transition from oncology to primary care settings, primary care nurses should be aware of potential long-term cardiovascular effects and adequately assess for CVD risk factors in women with a history of breast cancer during routine examinations.^{44, 45} In addition to traditional cardiac risk factors, physical HRQoL measurements could be implemented in clinical settings to aid in the CVD risk assessment. Second, our findings support the need for nurses to provide education and promote physical HRQoL. Cancer survivorship care plans should not only incorporate strategies to mitigate cardiovascular risk factors, such as smoking cessation, weight reduction, and heart healthy diets, but should also emphasize the importance of promoting management strategies, which have been shown to improve physical HRQoL and cardiovascular health in cancer survivors, such as physical activity interventions or cardiac rehabilitation programs.⁴⁶

Conclusion

Our findings suggest that physical HRQoL is associated with the risk of CVD in BC survivors, which align with previous reports in cancer-free populations. However, no associations were found for mental HRQoL or sleep disturbance. Further research should examine whether the addition of PCS scores improves the risk prediction models of CVD in BC survivors beyond that of traditional risk factors and examine the influence of inflammation on physical HRQoL and CVD risk.