Cardiac events among a cohort of 17,389 patients receiving cancer chemotherapy: short and long term implications

Abstract

Background

The association between cardiovascular disease and carcinogenesis is bidirectional and well-established. Furthermore, cancer treatment improves overall patient survival, potentially at the cost of incremental and fatal cardiovascular disease (CVD).

Aim

To evaluate (a) In a real-world cohort, the proportion of patients offered cancer chemotherapy who have antecedent CVD (CVD_A); (b) The rates of patient admission with subsequent development of CVD (CVD_S) requiring hospital admission post assignment to chemotherapy; (c) The impact of CVD_A and CVD_S on mortality rates relative to those seen in patients without overt CVD (CVD⁻) and (d) The time course of mortality in CVD⁻ versus CVD_S patients.

Methods

Retrospective analysis was performed in deidentified linked health data sets. Correlates of mortality were evaluated by Cox proportional hazards evaluation. Relative and absolute time-variability of CVD as a primary cause of death were determined.

Results

Of the total 17,389 patients, there were 2,159 with CVD_A. Over a median follow-up time of 4.6 years, CVD_S admissions (n = 8,529) occurred more commonly in the presence of CVD_A (70.0% vs. 46.1%, p < 0.001), and more than 50% of CVD_S cases occurred in the first 12 months of follow-up. The 5-year mortality rates were 71.5% for CVD_A, 64.7% for CVD_S, and 40.8% for CVD⁻ (p < 0.001). Development of CVD_S was associated with a substantially increased risk of mortality in the next 12 months. The development of CVDs was also associated with an increased risk of cardiovascular, as against non-cardiovascular, mortality (7.1% vs. 1.6%, p < 0.001).

Conclusions

Approximately 50% of patients assigned to cancer chemotherapy developed CVD_S, heralding a particularly high risk of mortality over the next 12 months. Both CVD_A and CVD_S are associated with substantial increases in mortality rates relative to those in CVD⁻ patients. This increased risk merits close individual monitoring.

Introduction

The past 20 years have been marked by extraordinary progress in the diagnosis and management of many forms of cancer and cardiovascular diseases, resulting in increased survivorship in both diseases. Cancer and cardiovascular disease (CVD) are the most prevalent diseases in the developed world. Moreover, these conditions are linked both epidemiologically and mechanistically through common clinical and biochemical risk factors. Anticancer therapies can cause a wide spectrum of short- and long-term non-fatal and fatal cardiotoxic effects [1–3]. Therefore, the link between cancers and cardiovascular disease is differentially manifest at the level of outcome, depending on frequency of iatrogenic development of life-threatening cardiovascular complications of cancer treatment.

Globally, cancer survival rates are continuing to improve:- not only have the rates at which people are being diagnosed with cancer been declining since 2008 [4] while for example 5-year survival rates for breast cancer have been reported to have increased from 77% in 1989–1993 to 92% in 2013–2017 [5]. Analogously, for heart failure in isolation, one-, five-, and 10-year survival rates increased progressively: from 74.2% in 2000 to 80.8% in 2016, from 41.0% in 2000 to 48.2% in 2012, and from 19.8% in 2000 to 26.2% in 2007, respectively [6].

However, more effective cancer therapies have introduced unexpected CVD complications. Most anticancer therapies are associated with some CVD toxicity (cardiotoxicity), including significant, long-lasting and often fatal cardiac events [1]. Cardiotoxicity can present as myocardial dysfunction, coronary artery or peripheral vascular disease [7, 8], , valvular disease, arrhythmias, arterial hypertension, and thromboembolism [9–11]. The risk of such developments tends to constrain both initial cancer chemotherapy in patients with antecedent heart disease, and the extent/duration of “high risk” chemotherapy [12–15].

In the last 20 years, there have also been a number of changes in the demographics of patients treated with cancer chemotherapy. Specifically, although cardiac risk factors such as untreated hypertension or smoking are rarer, the population is generally older [16, 17]:- advanced age is an independent risk factor for microvascular disease, which may precipitate both myocardial ischaemia and heart failure, with the latter usually characterized as heart failure with preserved ejection fraction (HFpEF). Furthermore, survival rates for patients with previous ischaemia and/or heart failure have improved considerably in recent years, leading to a higher than previous proportion of such patients becoming candidates for cancer chemotherapy.

A further key component of the rationale for undertaking this study was that the population spectrum of coronary risk factors in most Western societies has changed in recent years, with a decline in rates of smoking and of untreated hypertension and hyperlipidaemia, but increased prevalence of an aged population [17, 18]. In this context, understanding of the pathogenetic nexus between old age and risk of ischaemic heart disease is only beginning to emerge, as indeed is the condition of heart failure with preserved ejection fraction (HFpEF), which occurs especially in older women and is also incompletely understood [19, 20]. Despite this change in demographics of candidate populations for cancer chemotherapy, there has been no decline in utilization of this treatment modality.

Finally, while there is a well-described risk of precipitation of cardiovascular events in patients undergoing cancer chemotherapy, we are not aware of any recent studies quantitating this risk, nor its timing after onset of treatment.

The purpose of the current evaluation was to update and expand available information regarding the current magnitude of the cancer chemotherapy-cardiac event relationship, both regarding magnitude and temporal relationships.

Aims

Quantitative data on the extent of cardiotoxicity risk and its clinical implications continue to emerge, building a body of knowledge to support the implementation of specialist cardio-oncology services for patient care [21]. In this study, we have utilized real-world linked administrative health datasets of patients with cancer who have been assigned to chemotherapy and examined the predictors of the time-related risk of subsequent mortality. Specifically, we evaluated.

• i.The proportion of patients who have antecedent CVD (CVD_A) when commencing chemotherapy.
• ii.The rates of admission in patients with subsequently development of CVD (CVD_S) after assignment of chemotherapy.
• iii.The impact of CVD_A and CVD_S on mortality rates relative to those seen in patients without CVD (CVD^-) and.
• iv.The time course of mortality in CVD^- versus CVD_S patients.

Materials and methods

Study design

This epidemiological review was a retrospective, multicenter, cross sectional cohort study of de-identified South Australia/Northern Territory (SA/NT) linked health administration data including the Integrated SA Activity Collection (ISAAC) of Hospital Separations and Outpatient episodes, South Australian Cancer Registry (SACR), Northern Territory Cancer Registry (NTCR) and mortality data from the SA/NT Births Deaths and Marriage Registry Data.

Data linkage

All data for this project was provided by SA/NT Datalink. SA/NT Datalink is the operational body for a consortium formed to link and administer seven core health datasets covering birth records, midwifery notifications, cancer registrations, hospital separation records, in-patient and outpatient mental health services, electoral records and death registrations (and cause of death) are available for linkage. Linkage records can be accessed from 1980 to 1 year prior to the current date. However, to ensure a uniform timeframe and the availability of a complete cancer incidence and survival record for all datasets, we included cancer registry data from August 1991 to December 2014. The linkage process generates a set of indices sometimes called “linkage keys” that are stored by Datalink in a Master Links File. These “linkage keys” are held separately from any personal demographic information and they are used to enable related health records for individuals to be joined for approved research projects.

Data were extracted from the Cancer Registry (SACR and NTCR) records from 1991 to 2014 with a unique patient ID (encrypted root number) assigned during the linkage process. All patients with cancer who were assigned for chemotherapeutic treatment were included.

Cancer Registry data were then linked with SA/NT hospital separations, outpatient episodes of chemotherapy or radiation therapy and index admission for cardiotoxicity or other chronic disease (these hospital admissions must post date cancer diagnosis).

Extracted Cancer Registry data were finally linked to SA/NT mortality data through linkage with Births Death and Marriages Registry Data.

Data and definitions

Variables from the Cancer Registries, ISAAC, and Births Deaths and Marriage Registry, including demographics, age, sex, country of birth, marital status, postcode and occupation were linked.

An index CVD hospitalization was defined as the first hospital CVD admission post cancer diagnosis date. The term “CVD admission” included all admissions primarily related to symptomatic myocardial ischaemia, heart failure and/or cardiac arrhythmias. Cardiovascular disease patients were identified from CVD-related principal or other discharge ICD-10 diagnostic codes. Patients with an index hospital admission following chemotherapy with a diagnosis (primary or other) of CVD were included in the study.

Cancer site/morphology was identified using ICD 10 site codes. When identifying cancer type the primary cancer diagnosis was used. All chemotherapy procedure codes within hospital and outpatient episodes were defined according to ICD-10-AM and morphology codes, within hospital episodes was defined according to uniform ICD-O.

SA/NT Births Deaths and Marriages coded ICD-10-AM mortality data were requested for all recorded causes of death for each patient by 2017, so that each patient within the study cohort had a minimum of 3 years’ follow-up time.

Statistical methods

Linked Health Data were extracted and coded using STATA 17 (STATA, 17.0 April 20, 2021; College Station, TX, USA) and statistical analysis was conducted using IBM SPSS for Windows version 22.0 (IBM, Armonk, NY, USA).

The total population was first divided into two groups according to subsequent development of CVD after cancer diagnosis, namely Cancer with subsequently developing CVD (CVD_S) and Cancer without CVD_S (CVD^-) to quantitate the between-group differences relative to demographic factors such as age, sex, marital status, country of birth, Aboriginal or Torres Strait Islander status, medical history, and cancer morphology/site. In this analysis, antecedent CVD (CVD_A) was considered a potential predictor of risk of CVD_S.

Categorical data were compared using Fisher’s exact test and Chi-squared test according to compartmental sizes to determine between-group (CVD_S vs. CVD^-) significance. Continuous data, if not normally distributed, were presented as median (interquartile range) and compared using appropriate non-parametric statistical tests otherwise parametric tests were performed. Data were considered as significant at p ≤ 0.05.

All-cause mortality and age-standardized mortality rates (ASMRs) were derived for CVD_S and CVD^- cancer patients. ASMRs were calculated by using the age-specific all-cause combined death rates of this cohort relative to the observed deaths in each group of our study population.

Then, Kaplan-Meier survival curves were derived to compare the time to death from cancer diagnosis for different categories of cancer patients according to presence and/or absence of either CVD_A or CVD_S.

A Cox proportional hazards model was fitted to evaluate the significant factors influencing (i) development of CVD_S; (ii) mortality of the cancer patients with/without CVD_S. As the Cox regression model assumes that the hazard ratios are proportional over time, which means the effect of the covariates on the hazard is constant throughout the study period [22]. The proportional assumption was assessed using SPSS with log minus log plotting of the covariates. The proportional assumption was not violated in this study as results partially shown in Fig. 1.

Fig. 1.

Log minus log plotting of the Cox regression proportional hazard assumption. (A) CVD status; (B) Age groups

Ethics

The protocol was approved by the South Australian Department for Health and Wellbeing Human Research Ethics Committee (HREC/15/SAH/63) and the Northern Territory Department of Health Human Research Ethics Committee (HREC 2015–2484). All members of the research team had been trained and certified in research Good Clinical Practice (GCP).

Results

Participants and outcomes

In this cohort of 17,389 patients (Fig. 2), with a median age of 62, diagnosed with cancer between 1991 and 2014 (median interval between CVD_A event and study entry 2.7 years), there were 2,159 (12.4%) patients with pre-existing CVD (CVD_A). Over a median follow up period of 4.6 years, 8529 developed clinically overt CVD requiring hospital admission subsequent to chemotherapy (CVD_S). On the other hand, 8,860 (51%) patients did not have an admission for CVD prior to their death or the completion of the date frame for the study (CVD^-).

Fig. 2.

Schematic of key patient events during study period. Note: * 12.4% (n = 2,159) of the study population had antecedent CVD prior assignment to chemotherapy. Note increased probability of death (73% vs. 46.3%) and of cardiovascular death (7.1% vs. 1.6%) in patients with CVD_S

Demographic and survival data for CVD_S and CVD^- patients are summarized in Table 1. Patients with CVD_S were older, more likely to be male, and had greater age-standardized mortality rates during follow-up. The median follow-up time was shorter (3.6 years) for patients with CVD_S compared to 5.4 years for patients without (p < 0.001). Furthermore, while CVD_A was present in 7.3% (n = 648) of CVD^- cases, it was much more common among patients with CVD_S (17.7%, n = 1511; p < 0.001). Overall, 49% (n = 8529) of the cohort developed CVD_S over the course of follow-up. Approximately 66% (n = 8697) of the deaths were attributed primarily to cancer. However, the CVD_S group had a higher prevalence of primarily CVD deaths (Fig. 2: 7.1% vs. 1.6% in CVD^- cohort). The impacts of CVD_A and CVD_S on overall mortality rates over the first 5 years and over the first 15 years of the study are shown in Fig. 3A and B. CVD_A was associated with increased risk of both medium- and long-term mortality, which was evident within the first year of follow-up (Fig. 4A). This mortality increment was similar to that seen with CVD_S (p < 0.001 for both CVD_A vs. CVD^- and CVD_S vs. CVD^-).

Table 1.

Demographics and mortality rates, among chemotherapy-receiving cancer patients

Characteristics	CVDS Cohort N = 8,529 (49.0%)	CVD−Cohort N = 8,860 (51.0%)	P-value
Age (at cancer diagnosis), years Median (IQR)	65.0 (56–73)	58.0 (47–68)	< 0.001*
Follow-up time, years Median (IQR)	3.6 (1.4–8.7)	5.4 (1.6–10.9)	< 0.001*
Age Group, years, n (%) < 20 20–29 30–39 40–49 50–59 60–69 ≥ 70	190 (2.2) 115 (1.3) 238 (2.8) 700 (8.2) 1,569 (18.4) 2,533 (29.7) 3,185 (37.3)	427 (4.8) 290 (3.3) 620 (7.0) 1,342 (15.1) 2,119 (23.9) 2,178 (24.6) 1,883 (21.3)	< 0.001*
Sex, n (%) Female Male	3,545 (41.6) 4,985 (58.4)	4,105 (46.3) 4,754 (53.7)	< 0.001*
Indigenous Status, n (%) Indigenous Non-Indigenous Status not stated	160 (1.9) 8,059 (94.5) 311 (3.6)	489 (5.5) 7,882 (89.0) 488 (5.5)	< 0.001*
Cancer Morphology/Site, n (%) Breast Prostate Bowel Others	619 (7.2) 603 (7.1) 569 (6.7) 6,739 (79.0)	1,208 (13.6) 888 (10.0) 486 (5.5) 6,277 (70.9)	< 0.001*
#Antecedent CVD, n(%)	1,511 (17.7)	648 (7.3)	< 0.001*
All-cause Mortality Crude, n (%) ASMR	6,231 (73.0) 702	4,100 (46.3) 499	< 0.001*
CVD as primary COD Crude, n (%)	445 (7.1)	66 (1.6)	< 0.001*
Cancer as primary COD Crude, n (%)	5021 (80.6)	3676 (89.6)	< 0.001*

CVD: cardiovascular diseases; IQR indicates interquartile range (25th–75th percentile); ASMR, Age-standardized mortality rate; COD, cause of death

Note # antecedent CVD cohort was included in both CVD_S and CVD⁻Cohort depends on with/without subsequent development of CVD after cancer diagnosis

* Significantly different at p ≤ 0.05

Fig. 3.

(A) Time to index CVD_S; (B) Time to death after CVD_S event. The highest mortality was in the first-year post cancer diagnosis

Fig. 4.

(A) Overall survival of CVD_S, CVD_A and CVD_S/A vs. CVD⁻; patient number data are truncated to the first 15 years of follow-up. (B) Survival within 5 years post cancer diagnosis: comparison of CVD_S, CVD_A and CVD_S/A vs. CVD⁻. Note the combination of CVD_A and CVD_S (CVD_A/S) had a significantly incremental mortality impact (p < 0.001). CVD_S: patients with only subsequent CVD admission after assignment to chemotherapy; CVD_A: patients with only antecedent CVD (history of CVD before cancer diagnosis date); CVD_A/S: patients with both CVD_A and CVD_S; CVD⁻: patients without CVD (neither CVD_S nor CVD_A)

Furthermore, the combination of CVD_A and CVD_S (CVD_A/S) was associated with a significantly incremental mortality relative to that in CVD⁻ patients (p < 0.001).

Correlates of variable mortality rates within groups

As shown in Table 2, mortality rates, adjusted for age, sex, Aboriginal or Torres Strait Islander status, tumour grade, cancer site and CVD_A, were increased in both CVD_S and CVD⁻ groups with increasing age, male sex, non-indigenous status, higher tumour grade, and the presence of CVD_A.

Table 2.

Adjusted time-varying Cox Proportional Hazard models for all-cause mortality for CVD_S and CVD⁻

	CVDS Cohort		CVD−Cohort
Parameter	HR (95% CI)	P-value	HR (95% CI)	P-value
Age group, years < 20 20–29 30–39 40–49 50–59 60–69 ≥ 70	Referent 1.68 (1.12–2.52) 2.89 (2.09–4.01) 3.35 (2.49–4.49) 3.91 (2.94–5.20) 4.47 (3.37–5.93) 6.07 (4.58–8.05)	0.012* < 0.001* < 0.001* < 0.001* < 0.001* < 0.001*	Referent 1.57 (1.09–2.24) 2.12 (1.57–2.86) 3.22 (2.45–4.22) 4.72 (3.62–6.14) 6.66 (5.12–8.66) 9.54 (7.34–12.40)	0.015* < 0.001* < 0.001* < 0.001* < 0.001* < 0.001*
Sex Male Female	Referent 0.94 (0.89–0.99)	0.017*	Referent 0.89 (0.83–0.95)	< 0.001*
Indigenous status Indigenous Non-Indigenous Status not stated	Referent 1.81 (1.45–2.27) 0.65 (0.48–0.87)	< 0.001* 0.004*	Referent 1.31 (1.13–1.52) 0.36 (0.27–0.48)	< 0.001* < 0.001*
Tumour grade Grade I Grade II Grade III Grade IV T-cell B-cell Others	Referent 1.02 (0.87–1.20) 1.52 (1.31–1.78) 2.73 (2.26–3.29) 1.58 (1.20–2.09) 0.76 (0.64–0.91) 1.356 (1.17–1.58)	0.776 < 0.001* < 0.001* 0.001* 0.002* < 0.001*	Referent 1.33 (1.08–1.62) 2.18 (1.74–2.58) 3.88 (3.09–4.88) 1.58 (1.07–2.34) 0.63 (0.49–0.80) 1.63 (1.34–1.99)	0.006* < 0.001* < 0.001* 0.02* < 0.001* < 0.001*
Cancer site Breast Cancer Prostate Cancer Bowel Cancer All Others	Referent 0.79 (0.68–0.93) 1.23 (1.06–1.43) 1.84 (1.64–2.07)	0.005* 0.006* < 0.001*	Referent 0.45 (0.37–0.54) 1.77 (1.50–2.09) 2.11 (1.86–2.38)	< 0.001* < 0.001* < 0.001*
#Antecedent CVD No Yes	Referent 1.26 (1.18–1.34)	< 0.001*	Referent 1.65 (1.49–1.83)	< 0.001*

CVD: cardiovascular diseases; HR, hazard ratio; and CI, confidence interval

Data were adjusted for age, sex, indigenous status, tumour grade, cancer site and antecedent CVD

Note # antecedent CVD cohort was included in both CVD_S and CVD⁻Cohort depends on with/without subsequent development of CVD after cancer diagnosis

* Statistically significant at p ≤ 0.05

Comparison of these trends in the whole data set (Table 3), confirmed the importance of increasing age, male sex, CVD_S, non-indigenous status and tumour grade as markers of overall mortality risk. Mortality risk, relative to that of breast cancer, was lower in patients with prostate cancer (HR: 0.62, 95% CI: 0.55–0.69, p < 0.001), but higher in those with bowel cancer (HR: 1.43, 95% CI: 1.28–1.59, p < 0.001). Both CVD_A and VCD_S were independent predictors of increased mortality risk.

Table 3.

Multivariate analysis (cox proportional hazards model) of potential correlates of mortality in the study population

Variable	HR (95% CI)	P value
Age	1.03 (1.03–1.03)	< 0.001*
Sex Male Female	Referent 0.90 (0.87–0.94)	< 0.001*
Antecedent CVD	1.29 (1.22–1.36)	< 0.001*
Subsequent CVD	1.27 (1.22–1.33)	< 0.001*
Indigenous status Indigenous Non-Indigenous Status not stated	Referent 1.39 (1.23–1.58) 0.45 (0.34–0.54)	< 0.001* < 0.001*
Tumour grade Grade I Grade II Grade III Grade IV T-cell B-cell Others#	Referent 1.13 (1.00-1.28) 1.74 (1.54–1.97) 3.17 (2.75–3.66) 1.60 (1.28–2.01) 0.75 (0.65–0.86) 1.46 (1.29–1.65)	0.051 < 0.001* < 0.001* < 0.001* < 0.001* < 0.001*
Cancer primary sites Breast cancer Prostate cancer Bowel cancer Others	Referent 0.62 (0.55–0.69) 1.43 (1.28–1.59) 1.92 (1.77–2.09)	< 0.001* < 0.001* < 0.001*

CVD: cardiovascular diseases; HR, hazard ratio; and CI, confidence interval

# NK-cell, Null-cell or Grade or differentiation not determined, not stated or not applicable

Data were adjusted for age, sex, indigenous status, tumour grade, cancer site, antecedent CVD and subsequent CVD

* Statistically significant at p ≤ 0.05

Time course of onset of CVD_S and of its mortality impact

More than 50% (n = 4803) of cases of CVD_S were diagnosed in the first year following treatment assignment (Fig. 3A), with approximately 90% (n = 7634) of CVD_S cases appearing within 5 years.

Among patients diagnosed with CVD_S, mortality, when it occurred, was also mainly in the first 12 months (Fig. 3B), with approximately 88% (n = 5672) of deaths within the first 5 years post onset of CVD_S. Patients with CVD_A/CVD_S exhibited substantially incremental mortality rates, including early appearance of this trend (Fig. 4).

Impact of follow-up time on probability of primarily CVD-related deaths

With the passage of time, the proportion of deaths attributed to cancer fell from > 85% to approximately 50% (Fig. 5A), while the proportion of primarily CVD-related deaths increased from approximately 4% to approximately 20%. The majority of CVD-attributed deaths occurred (a) in patients with CVD_S (n = 275 vs. 51 in CVD⁻ cohort); (b) within the first 2 years of follow-up (Fig. 5B).

Fig. 5.

(A) Designated primary cause of death over follow-up time (p < 0.001 for both cancer death and CVD death); (B) Absolute number of CVD-related fatalities per year for the first 5 years

The ratio of cardiovascular to cancer deaths rose from approximately 1:20 to approximately 1:4.5 during the course of the study (Fig. 5A). There were also 1123 deaths (6.5% of the patient population) independent of cardiovascular disease or cancer:- this proportion did not fluctuate markedly during the study (data not shown).

Discussion

The current study was designed to evaluate both demographics and outcomes among a large cohort of patients in whom cancer chemotherapy was to be initiated for a variety of neoplasms. Approximately 50% of patients were aged > 60 years at entry, 12% had antecedent cardiovascular disease and breast and prostate malignancies were the most commonly treated. Next, over the follow-up period, more than half of these patients died. Finally, and most importantly, both antecedent cardiovascular disease and the subsequent emergence of cardiovascular disease were both strong predictors of mortality.

In this study, we utilized real-world linked administrative health datasets of patients with cancer who received chemotherapy and examined the predictors of the time-related risk of subsequent mortality. Specifically, in a cohort of 17,389 patients, there were 2,159 patients with CVD_A. Furthermore, almost of half of the cohort of patients (n = 8,529) were diagnosed with at least one major cardiovascular event subsequent to entry, and this occurred more commonly in the presence of CVD_A. More than 50% of these CVD_S cases developed in the first 12 months of follow-up. The 5-year mortality rates were 71.5% for CVD_A, 64.7% for CVD_S, and 40.8% for CVD⁻:multivariate analyses confirmed that CVD_A (HR = 1.289, p < 0.001) and CVD_S (HR = 1.274, p < 0.001) were independent markers of increased mortality risk. Development of CVD_S was associated with a substantially increased risk of mortality in the next 12 months. The development of CVD_S was associated both with a substantially increased risk of mortality in the next 12 months, and with an increased proportion of cardiovascular deaths.

The major finding from this component of the study was that new admission with CVD (CVD_S) occurred in almost 50% of patients over the course of the study. Multivariate correlates of increased risk of CVD admission during the study were age, sex, and CVD_A.

We next sought to evaluate the short and long-term impact on patient mortality. Overall, either CVD_S or (especially) CVD_A imposed at least an incremental mortality risk of 20%. When we considered the temporal nexus between diagnosis of CVD_S and mortality risk, there was a significantly great risk of death within the first 12 months post appearance of CVD_S. As far as we can determine, this finding has not previously been reported in the literature. However, overall, in the study, only approximately 5–10% of total deaths, mainly in CVD_S patients, were primarily ascribed to CVD: this proportion raise markedly from the 10th year of follow-up onwards.

The results raise the question of whether cardiovascular risk can be categorized prospectively in patients assigned to cancer chemotherapy. Beyond the incremental risk inherent in such treatment in patients with known heart failure [23], it has even been shown that presence of coronary calcification is positively correlated with risk of cardiac events [23]. Most interestingly, more than half of the deaths from cardiovascular disease (and also cardiovascular admissions) occurred within the first 12 months following introduction of chemotherapy. Similarly, there is some evidence that utilization of conventional anti-failure medication might reduce cardiovascular risk in chemotherapy patients [24, 25]. However, specific details are lacking as regards extent and specifics of the benefit. State of cardiovascular fitness in individual patients, which may modulate outcomes [26], could not be determined. Moreover, it is not clear to what extent echocardiographic monitoring, which may facilitate early adjustment of chemotherapeutic regiment, was utilized [27].

Limitations

Although the findings of the study are potentially very important in decision-making regarding cancer therapeutics, the current study has several important limitations. First, it was not designed for evaluation of CVD/cancer nexus for particular cancers: the possibility of substantial heterogeneity in this regard has not been excluded. Second, no information is available as to the prevalence of coronary risk factors among the patients studied, the smoking status, the precise form of CVD_A (where applicable) or the nature of the index admission in CVD_S patients. At the time of this study, data on pharmacotherapy doses and type of chemotherapy/immunotherapy, cardioprotective therapy and doses and type of radiotherapy were not available for linkage. This would be a very important consideration in explaining increased death rates immediately following onset of CVD_S.

In the absence of exhaustive investigation, the designated cause of death in this group of patients is unlikely to be completely accurate and may have been biased in favour of the most recent clinical diagnosis.

We also do not know the extent of selection bias which was applicable in the current data set: it is possible that many patients with a diagnosis of potentially treatable cancer were not offered chemotherapy because of general debility and/or severity of pre-existent cardiovascular disease.

Elimination of these various potentially confounding factors would require a prospective study with standardized decision-making and preferably a high rate of post-mortem detailed evaluation.

In particular, lack of access to individual data regarding pre-diagnosis smoking patterns and their potential variability during subsequent chemotherapy is an important limitation of the study. While it is well-established that cessation of smoking may reduce risk of clinical emergence of cardiovascular disease [28], there is also evidence that smoking cessation may favorably affect patient survival post cancer diagnosis [29].

The fact that the study was conducted over the time period between 1991 and 2014 of course has both advantages and disadvantages. The main disadvantage is that both patient selection criteria and choice of chemotherapeutic (and cardioprotective) agents have advanced in the period since 2014: the main advantage is that the study was of sufficient size to identify the early peak in cardiovascular events and conversely to exclude late hysteresis between onset of chemotherapy and that of cardiovascular events and to account for causes of death in the majority of the patient cohort.

Potentially, individualization of chemotherapy according to a priori patient risk of cardiotoxicity might have reduced mortality risk [30], but we could not determine to what extent such principles were applied especially as 12.4% of patients had CVD_A. As foreshadowed, this was a relatively elderly population, at risk of age-related ischaemic heart disease and heart failure, both of which emanate substantially from coronary microvascular disease [31, 32]. Overall, this makes the current findings more relevant, but does not include exploration of potential pathogenetic and therapeutic links between such microvascular disease and ideal therapy of both heart disease and cancer, we feel that this should be a key future priority.

Recommendations for clinical practice and further research

Antecedent cardiovascular disease identifies patients who will be at markedly increased risk of death during cancer chemotherapy/immunotherapy. About 50% of patients assigned to chemo/immunotherapy will eventually develop CVD_S (that is, requiring hospital admission) and these patients also have very high risks of death, especially in the first year after a cardiovascular event.

In view of these findings, it would now be appropriate to conduct a prospective, targeted mechanistic study, focusing on (i) types of CVD_S (for example: heart failure/myocardial ischaemia/arrhythmias) and their relationships to patient demographics and to specific cancer treatments; (ii) whether CVD_S was adequately treated with agents known to improve outcomes; (iii) the probable impact of cancer chemo/immunotherapy on survival and quality of life according to patients’ demographics, prior CVD and type of therapy. The issue of whether incrementally prophylaxis of cardiovascular disease is justified during the first 12 months of chemotherapy is also worthy of further investigation.

Conclusion

Patients with pre-existing CVD and those who developed CVD after initiation of cancer treatment had significantly increased mortality rates relative to those did not develop CVD. The presence of CVD in patients with cancer heralds a particularly high risk of mortality over the first 12 months after diagnosis (Fig. 6). This increased risk merits both prospective analysis and close individual monitoring. Beyond this generality, these data beg the case for routine consideration of initiation of cardioprotective therapies at the onset of chemotherapy.

Fig. 6.

Schematic: design of study and main outcomes at 1 year and at conclusion of follow-up. CVD_S: patients with only subsequent CVD admission after assignment to chemotherapy; CVD_A: patients with only antecedent CVD (history of CVD before cancer diagnosis date); CVD_A+S: patients with both CVD_A and CVD_S; CVD⁻: patients without CVD (neither CVD_S nor CVD_A)

Abbreviations

CVD

Cardiovascular diseases

CVD_A

Cardiovascular diseases: antecedent

CVD_S

Cardiovascular diseases: subsequent

CVD⁻

No cardiovascular diseases

SA/NT

South Australia/Northern Territory

ISAAC

Integrated SA Activity Collection

SACR

South Australian Cancer Registry

NTCR

Northern Territory Cancer Registry

Author contributions

Clark and Koczwara conceived the idea driving the analysis. Liu performed the data analysis and drafting of the manuscript. Horowitz verified the analysis methods. Clark and Horowitz checked the findings of this work. All authors discussed the results and contributed to the final manuscript. All authors have reviewed the manuscript and substantively revised it.

Funding

RA Clark was supported by an Australian Heart Foundation Future Leader Fellowship (ID100847). AL Sverdlov is supported by an Australian Heart Foundation Future Leader Fellowship (ID106025).

Data availability

The data that support the findings of this study are available from the authors upon reasonable request.

Declarations

Ethical approval

In accordance with the National Health and Medical Research Council (NHMRC)’s Australian Code for the Responsible Conduct of Research, the protocol was approved by the South Australian Department for Health and Wellbeing Human Research Ethics Committee (HREC/15/SAH/63) and the Northern Territory Department of Health Human Research Ethics Committee (HREC 2015–2484). All members of the research team have been trained and certified in research Good Clinical Practice (GCP).

Consent to participate

Not applicable as no potentially identifiable human images or data are presented in this study.

Competing interests

The authors declare no competing interests.