Nationwide population-based cohort study on the
association
of acute coronary syndrome in patients with malignancies
Yen-Nien Lin & Yen-Jung Chang & Yin-Huei Chen &
Po-Yen Ko & Cheng-Li Lin & Fung-Chang Sung &
Kuan-Cheng Chang & Chia-Hung Kao
Introduction
Certain cancer types are inherent in prothrombotic states
because of platelet activation and aggregation and, in
procoagulant factors, induce the sharing of similar
atherothrombotic mechanisms for the development of coronary
artery diseases [1–4]. The prolongation of life expectancy
in cancer survivors through chemotherapies, hormone therapies,
and/or radiotherapies is accompanied by an increased
risk of coronary artery disease (CAD) [5, 6]. The commonly
observed comorbid depression in patients with malignancy
was recently reported as being associated with the progression
of underlying coronary atherosclerosis and clinical events
such as acute coronary syndrome (ACS) [7]. Moreover,
certain types of cancer were reported to trigger chronic
inflammations that may further impose on atherosclerotic
mechanisms [8, 9] because patients with malignancies are usually older and have a tendency toward
associations
with more cardiovascular risk factors.
The associations between ACS risk and cancers are largely
derived fromobservations of case series or single-center studies.
Moreover, concluding the universally accepted viewpoints for
clinical practice is difficult for patients with relatively short
survival spans and/or a limited eligible study population in
patients with selected cancer diseases. Therefore, we attempted
to define the association between malignancies and ACS with
the nationwide population cohort study, which includes approximately
a decade of data from Taiwan.
Materials and methods
Study design
A longitudinal cohort study from the Taiwan National Health
Institute Research Database (NHIRD) and its claims data from
one million randomly selected insurants from 1996 to 2009
were examined. The database contains comprehensive inpatient
health care data from more than 96 % of the entire
population and covers 97 % of outpatient clinics. The data
were linked to the anonymized identifications of people, and
the National Health Institute reimbursement data were kept
suitable for public research. The claims data included the
registries of sociodemographic information, health care visits,
diagnostic codes, prescriptions, and expenditure amounts of
insured residents, as described in detail in previous studies
[10, 11]. Moreover, the accuracy and high validity of CVD
diagnosis in the database has been demonstrated [12]. Diagnoses
were coded with the International Classification of
Diseases, 9th Revision (ICD-9).
Study population
We identified 22,522 patients with newly diagnosed malignancy
(ICD-9 codes 140–208) as the malignancy cohort from
1997 to 2006 and a comparison pool of subjects without
diagnosis record of malignancy. From the eligible comparison
pool, we randomly selected control subjects who were
matched with malignancy cases on sex, age (each 5-year
span), and year and month of the index year. Comparison
subjects with ACS diagnosis prior to the index date were
excluded. The index date from the Registry for Catastrophic
Illness Patient Database (RCIPD) was used as the date of the
national malignancy registration. These malignancy cases
were categorized into 11 groups based on the hematological
malignancy (ICD-9 codes 200–203), colorectal cancer (ICD-9
codes 153 and 154), liver cancer (ICD-9 code 155), lung
cancer (ICD-9 code 162), breast cancer (ICD-9 codes 174
and 175), female genital organ cancer (uterus, cervical, ovary,
and vagina) (ICD-9 codes 179–184), prostate cancer (ICD-9
code 185), stomach cancer (ICD-9 code 151), bladder cancer
(ICD-9 code 188), and head and neck cancer (ICD-9 codes
146–148). Patients with a medical history of ACS (ICD-9
code 410) diagnosed before the index date, or with incomplete
age or sex information, were excluded. Four controls for each
case were frequency-matched for age 1 year each, sex, and
year and month of the index date. Both cohorts were followed
up until the first ACS development identified by the RCIPD,
censored due to death or withdrawal from the national health
insurance program, or December 31, 2009.
Variables of interest
The study included sociodemographic information as well as
comorbidities, such as hypertension (ICD-9 codes 401–405),
diabetes (ICD-9 code 250), hyperlipidemia (ICD-9-CM 272),
cerebral vascular accident (CVA, ICD-9 codes 430 to 438),
chronic kidney disease (CKD, ICD-9 codes 490–496, 585),
rheumatic heart disease (RHD, ICD-9 codes 393–398), and
heart failure (HF, ICD-9 code 428). All comorbidities were
determined before the index date.
Statistical analysis
Comparisons between malignancy and nonmalignancy groups
were performed using the chi-square test for categorical variables.
We calculated the person-years of follow-up times for
each person until ACS diagnosis or censorship. Cox proportional
hazards regression analysis was conducted to estimate
the hazard ratios (HR) and 95 % confidence intervals (CI). All
comparison tests were two sided. All p values less than 0.05
were considered statistically significant. All statistical calculations
were performed with the SAS software (version 9.2 for
Windows; SAS Institute, Inc., Cary, NC, USA).
Results
Demographic characteristics of the study patients
Table 1 shows the demographic characteristics of the
malignancy and comparison cohort groups. There were
more men than women (55.3 % against 44.7 %), and
43.5 % patients were over 65 years of age. Patients in
the malignancy cohort had more comorbidities regarding
hypertension, diabetes mellitus, cerebrovascular accident,
CKD, and chronic obstructive pulmonary disease than
the counterpart group. Conversely, hyperlipidemia occurred
significantly more in the malignancy group than
in the comparison cohort group. Risk and crude rate ratio of acute coronary syndrome (ACS)
Table 2 shows the incident densities and crude HR of ACS by
the baseline sociodemographic status. The malignancy cohort
group had a nonsignificantly lower incidence of ACS than the
nonmalignancy cohort group, but the incidence densities of
ACS were significantly higher in men with malignancies than
without malignancies.
A multivariate analysis by means of the Cox proportional
regression model shows that the hazard ratio of ACS was
marginally significantly greater in the malignancy cohort than
in the nonmalignancy cohort when adjusted for age, sex, and
comorbidity associations with malignancies (models 2 and 3,
HR=1.09, 95 % CI=0.99–1.20; HR=1.03, 95 % CI=0.93–
1.13), as shown in Table 3.
Furthermore, the specific analyses on hematological malignancy,
colorectal cancer, liver cancer, lung cancer, breast
cancer, female genital organ, prostate cancer, stomach cancer,
bladder cancer, and head and neck cancer are shown in Table 4
and Fig. 1. Cox proportional regression analysis of male
patients with prostate cancer against men without malignancies
showed a 1.3-fold adjusted HR, indicating slightly higher
ACS rates for male prostate cancer patients (95 % CI=1.01–
1.67). Moreover, the head and neck cancer group showed a
3.03-fold significantly higher ACS rate for patients with
malignancies (95 % CI=1.47–6.50).
Discussion
Patients with malignancies are believed to have an increased
ACS risk based on several reasonable mechanisms. However,
whether malignancy causes an innovative cardiovascular risk
factor remains unclear. Patients with malignancies usually
have a relatively short survival span, more comorbidities, and complex clinical statuses, which usually
render them
ineligible for clinical studies. However, current experience
from the real world derived mostly from small population
and regional studies of select cancer diseases is not devoid
of bias. The National Health Insurance (NHI) system in
Taiwan was sponsored by the National Health Insurance
Bureau, Department of Health, since March 1995. The
National Health Research Institute computerized the NHI data
into several data sets to facilitate research. This retrospective
cohort study was rendered possible with the NHIRD. To the
best of our knowledge, this is the first nationwide and largest
population study to associate the risk of ACS following
malignancy development.
Patients with malignancies in this study had more comorbidities
and cardiovascular risk factors, such as hypertension,
diabetes mellitus, CVAs, CKD, chronic obstructive lung
disease, RHD, and HF. There was a contradictory finding
regarding the fair balance of ACS incidence between
patients with and without malignancies in Taiwan within 1997
and 2009. In this study, patients with malignancies generally
had a worse life quality [13] and sought more medical attention,
which reduced the risk factors. The short median survival was
responsible for the balance of ACS incidence rather than the
reduction of risk factors in cancer patients.
Among the risk factors, sex difference and hyperlipidemia
have merit for discussion. This study shows that female patients were protected against ACS
development, but
male patients experienced an increased ACS hazard ratio.
Women experience a delayed atherosclerotic process because of
exposure to endogenous estrogen [14]. TheWomen’s Ischemia
Syndrome Evaluation study indicates that young women with
endogenous estrogen deficiency have a more than 7-fold
increase in coronary artery risk [15]. A study showed that
estrogen had a regulating effect on lipids, inflammatory markers, and the coagulant system [16], which
may modify
the atherothrombotic pathways of ACS in patients with
malignancy. Women have a lower rate of smoking in Taiwan
compared to men [17], which may have influenced their lower
ACS incidence.
Patients with malignancy had a lower ACS risk compared
with those without malignancy among the subjects with hyperlipidemia.
Patients who acquire malignancies usually became
cachexic because of anorexia and/or catabolism, chemotherapy
side effects, and medical intervention. Although
hyperlipidemia may occur in certain patients who undergo
anticancer therapy [18], the lipid levels usually became normal
or lower in clinical settings. The lower occurrence ofACS
development in this study may be attributable to patients’
better lipid profiles and short survival spans.
Several previous studies have suggested that breast cancer,
Hodgkin’s lymphoma, and prostate cancer increase
ACS risk [19–22]. Moreover, this analysis showed that
prostate cancer and head and neck cancer increased ACS
risk with 1.30 and 3.03 HRs, respectively. Although the
case number of the head and neck cancer group was small,
both were remarkably significant. In clinical practice, these
patients on hormone therapy and radiotherapy tended to
promote the progression of atherosclerosis, endovascular
proliferation, and increase ACS risk [23], especially if
concomitant with hyperlipidemia [24]. The antimetabolite
5-fluorouracil and the oral prodrug capecitabine are commonly
used in anticancer therapy for head and neck cancer,
but were shown to have associations with coronary spasm
and have resulted in serious myocardial infarction or malignant
ventricular arrhythmia complications [25]. Results
of these studies were mostly based on long-term observation,
which may suggest that ACS occurrence was
underestimated in their cancer “survivors.”
The reduction of arterial thrombosis risk with low-dose
aspirin and antiplatelet drugs that inactivate cyclooxygenase
are recommended for ACS prevention [26, 27]. All major
clinical trials of antithrombotic therapy and ACS have
generally excluded patients with cancer. Moreover, patients
receiving anticancer therapy are usually less prone to ACS
in a thrombocytopenia status. The administration of aspirin
remains controversial during clinical practice for patients
with cardiovascular risks and active cancer diseases.However,
aspirin is suggested for cancer patients with ACS, but requires
awareness of bleeding tendency, despite a thrombocytopenic
status [28].
Taiwan launched a national health insurance (NHI) in
1995, operated by a single buyer, the government. All insurance
claims should be scrutinized by medical reimbursement
specialists and peer review. The diagnoses of cancer and ACS
were based on the ICD-9 code determined by specialists or
experienced physicians; therefore, the diagnoses and codes for
cancer and ACS should be accurate and reliable. The accuracy
and high validity of diagnoses identified in the NHIRD have
been verified in the previous studies [29–32].
This study has several limitations. First, this study was
performed using NHI data sets and lacked specific social
and personal information and drug histories; thus, all other
potential influences, such as smoking, radiotherapy, antithrombotic
medication, and chemotherapy agents, were not
included in the comparisons. Second, information regarding
individual cancer stages and courses of progression were
unobtainable. In the comprehensive comparison, we experienced
difficulty when comparing ACS risks for cancer
patients with different survival periods. Third, this retrospective
cohort study is subject to biases related to possible confounders.
Despite our meticulous study design has been adequately
controlled for certain confounding factors, a key limitation
is the bias caused by unmeasured or unknown confounders,
such as smoking, BMI, medication used for risk
modification, drug compliance, etc. Moreover, the number of
cases for certain cancer diseases was small in our subgroup
analysis. Although ACS risks in patients with prostate and
HEENT cancers were significantly increased, large-scale
studies indicate that the real ACS risks for these two cancer
groups remain unknown.
In conclusion, this study showed that patients with malignancies
have marginally significantly greater ACS risk compared
to those without malignancies. Patients with prostate
and HEENTcancer had a significantly higher risk ofACS.We
suggest careful surveillance of ACS symptoms and regular
electrocardiography during follow-up of these patients. However,
further large-scale studies for patients with prostate and
HEENT cancer and cancer survivors (especially from posthormone
or radiotherapy) are required.