King KM, Ghali WA, Faris PD, Curtis MJ, Galbraith PD, Graham MM, Knudtson ML. Sex Differences in Outcomes After Cardiac CatheterizationEffect Modification by Treatment Strategy and Time. JAMA. 2004;291(10):1220-1225. doi:10.1001/jama.291.10.1220
Author Affiliations: Faculty of Nursing (Dr King), Department of Medicine (Drs Ghali, Curtis, and Knudtson and Ms Galbraith), Department of Community Health Sciences (Drs King and Ghali), Centre for Health and Policy Studies (Dr Faris), University of Calgary, and Department of Medicine, University of Alberta (Dr Graham), Calgary.
Context Studies comparing outcomes of cardiac care in women vs men yield various
results, with some suggesting worse outcomes for women and others suggesting
Objective To determine whether extent of coronary disease, treatment strategy,
and follow-up time influence the risk of death in women vs men among patients
who have had cardiac catheterization.
Design, Setting, and Patients We studied a large inception cohort by using detailed clinical data
from a registry of 37 401 patients undergoing cardiac catheterization
in Alberta, Canada, from 1995-2000, with follow-up through December 31, 2001.
Main Outcome Measures The risk of death for women vs men was assessed for all patients combined
and then in analyses stratified by degree of coronary anatomic risk and by
treatment strategy (no revascularization, percutaneous coronary intervention
[PCI], coronary artery bypass graft [CABG] surgery). The latter analysis included
a graphic assessment of the changing relative risk over time for women vs
Results Women had higher 1-year mortality than men did (5.6% vs 4.6%; P<.001). However, stratified analyses demonstrated that
sex differences in risk occurred only early after catheterization and were
most apparent among patients undergoing revascularization. The early risk-adjusted
relative risks for women vs men were elevated at 3.49 (95% confidence interval
[CI], 1.95-6.24) for CABG surgery and 2.38 (95% CI, 1.48-3.83) for PCI on
day 1 after catheterization, with a subsequent decrease in relative risk over
time to equivalence in risk between sexes before 1 year.
Conclusions Sex-based differences in death rates after cardiac catheterization are
time- and treatment-specific. This finding may at least partially explain
the discrepancies in results from earlier studies on sex differences in outcomes
of cardiac care.
Sex differences in cardiac care and outcomes have been widely investigated
since Steingart et al1 stimulated clinicians
to consider sex-based biases in care practices. Investigators have examined
sex differences in access to cardiac procedures and outcomes after myocardial
infarction (MI) or the diagnosis of coronary artery disease (CAD).2- 18 Many
studies have suggested that women have less access to care or poorer outcomes,2- 9 whereas
others have concluded that there are few or no differences10- 17 or,
under particular circumstances, that women fare better than men.16 After
more than a decade of investigation, however, consistent findings are wanting,
perhaps because of unavailability of detailed clinical data, variability of
patient samples, and variability of follow-up times.
The Alberta Provincial Project for Outcome Assessment in Coronary Heart
Disease (APPROACH) is a database containing detailed clinical data for all
residents of Alberta, Canada, who undergo cardiac catheterization in the province.19 This database represents a resource for cardiovascular
outcomes research and has recently been used to study sex differences in access
to coronary revascularization.11
In this study, we extend that recent work to study sex differences in
survival after cardiac catheterization. We assessed survival outcomes for
all women vs men and then made outcome comparisons stratified for degree of
coronary anatomic risk and mode of treatment after catheterization (no revascularization,
percutaneous coronary intervention [PCI], or coronary artery bypass graft
[CABG] surgery). We also conducted a detailed analysis of time as a potential
modifier of the risk of death for women vs men.
Patients in APPROACH were followed up longitudinally for assessment
of long-term outcomes after cardiac catheterization. Clinical risk variables
recorded at cardiac catheterization included sex, age, congestive heart failure,
peripheral vascular disease, chronic pulmonary disease, cerebrovascular disease,
elevated creatinine level (≥2.26 mg/dL [≥200 µmol/L), dialysis
status, diabetes, hypertension, hyperlipidemia, liver or gastrointestinal
disease, malignancy or metastatic disease, previous MI, previous PCI, previous
CABG surgery, previous thrombolytic therapy for MI, and smoking status (categorized
as "never," "former," or "current"). The indication for catheterization was
recorded in 1 of 4 categories: MI within 8 weeks of catheterization, stable
angina, unstable angina, or other (eg, arrhythmias). Extent of coronary disease
was recorded and used to derive the weighted Duke Index and Duke Myocardial
Jeopardy Score.20- 23 Left
ventricular ejection fraction was graded into 5 categories: less than 30%,
30% to 50%, more than 50%, ventriculogram not done (usually because of renal
insufficiency or severely depressed cardiac function), and information missing.
The occurrence of revascularization procedures after catheterization was also
The outcome of interest for this study, all-cause mortality, was ascertained
through semiannual linkage to records from the Alberta Bureau of Vital Statistics.
We analyzed data from patients undergoing cardiac catheterization in 1995
through 2000, with follow-up of patients through December 31, 2001. The APPROACH
study protocol was approved by the ethics review boards of the Universities
of Calgary and Alberta. The requirement for informed consent was waived.
We used χ2 and 2-sample t tests
to compare the clinical characteristics of men and women. The distributions
for age of women and men met assumptions of normality and equal variances.
We used Kaplan-Meier plots and log-rank tests to compare unadjusted survival
of women vs men for 1 year after catheterization. We also compared unadjusted
survival of women vs men, stratified by coronary anatomic risk, with patients
categorized into high risk (left main coronary artery stenosis, 3-vessel disease,
or 2-vessel disease with proximal left anterior descending involvement) and
low risk (other 2-vessel disease, 1-vessel disease, lesions with <50% stenosis,
and normal) anatomy groups.
For our analysis stratified by initial treatment strategy, we grouped
patients according to first revascularization treatment received within a
year of catheterization: PCI, CABG, or no revascularization. We focused only
on the first revascularization procedure after catheterization because that
is the procedure most likely linked to the results of coronary angiography.
We used Cox proportional hazards models to model survival but found
that, regardless of treatment modality, the effect of sex violated the proportional
hazards assumption because the risk of events in women vs men changed over
time (traditional proportional hazards models assume that this relative risk
is fixed). This finding led us to confine our analysis of outcomes by sex
over time to the graphic examination of relative risks estimated by plotting
splines through residuals from Cox models that excluded the sex variable.
For these graphic methods, we relied on restricted cubic splines plotted through
rescaled Schoenfeld residuals.24,25
We also used logistic regression to examine the association between
sex and outcomes at 30 days and 1 year while controlling for severity of illness.
All potential risk variables were retained in the models regardless of statistical
significance because our objective was to focus primarily on the odds ratios
(ORs) for patient sex while controlling for all other potential confounders.
Analyses were stratified by treatment group (no revascularization, PCI, CABG
surgery) and coronary anatomy (low risk, high risk). We also performed a modified
propensity analysis for which we modeled propensity (ie, likelihood) to be
selected to CABG surgery or PCI (2 propensity models) and then assessed the
OR for death in women vs men across tertiles of propensity. All analyses were
performed with S-Plus (version 6.1 for Windows; Insightful Corp, Seattle,
Wash). The level of significance used for tests was .05.
Of 37 401 patients studied, 11 199 were women. Women were
significantly older than men and had more comorbid conditions, including congestive
heart failure, chronic lung disease, cerebrovascular disease, diabetes, hypertension,
liver disease, and malignancy (Table 1).
Women were less likely to have had an MI or to have had cardiac interventions.
Women tended to have a higher left ventricular ejection fraction, and their
coronary anatomy was generally of lower risk than was that of men. Correspondingly,
women had lower median weighted Duke Index values23 and
were less likely than men to have revascularization procedures after catheterization.
Women had higher 1-year mortality than men (626/11 199 [5.6%] vs
1203/26 202 [4.6%]; P<.001). The unadjusted
survival curves (Figure 1) for all
patients revealed a higher mortality in women vs men during 1 year after cardiac
catheterization (χ21 = 17.3, P<.001). Particularly in the early postcardiac catheterization period,
there was a significant decline in survival among women compared with men.
The unadjusted survival curves for patients with low-risk coronary anatomy
suggested little difference between low-risk women and men in survival rates
after cardiac catheterization (χ21 = 3.4; P = .06; Figure 1).
In contrast, the curves for patients with high-risk coronary anatomy revealed
that women had poorer survival early after cardiac catheterization, followed
by similar death rates beyond approximately 40 days once the period of early
risk had ended (χ21 = 83.2; P<.001).
To further investigate the early mortality risk in women with high-risk
anatomy, we stratified patients by treatment group (no revascularization,
PCI, CABG surgery). For PCI and CABG surgery treatment groups, there was a
markedly increased early risk for mortality for women vs men (Figure 2). For the CABG surgery group, the adjusted relative risk
for mortality for women vs men was 3.49 (95% confidence interval [CI], 1.95-6.24)
on day 1 after catheterization. The risk for women subsequently decreased
to a level equivalent to that for men (ie, relative risk of 1.0) at 157 days.
A similar pattern was seen in patients who underwent PCI: the relative risk
was 2.38 (95% CI, 1.48-3.83) on day 1 after catheterization and dropped to
1.0 at 342 days. The risk profiles of women and men in the "no revascularization"
group did not follow this pattern; the relative risk of mortality in women
vs men remained similar and near 1.0 throughout follow-up.
These findings of changing relative risks over time were confirmed by
logistic regression analyses of mortality at 2 points (Table 2). For patients who did not undergo revascularization, there
were more modest, statistically insignificant sex differences in odds of mortality.
For the analyses shown in Figure 2 and Table 2, we intentionally used time of
catheterization as a common "time zero" across treatment groups. For the PCI
group, this analysis accurately reflects time of actual treatment, given the
short median waiting time of 1 day for PCI. For CABG surgery, the median wait
was 22 days. As a result, the early hazard seen immediately after catheterization
was based on data from some patients who had their CABG immediately after
catheterization but also on data for others who had not yet undergone CABG.
Therefore, we performed sensitivity analyses for which plots were replicated
by using the time of revascularization (PCI or CABG surgery) as time zero.
These sensitivity analyses revealed an almost identical picture of elevated
early hazard in women vs men. The only difference with the main analyses presented
in Figure 2 is that the hazard ratio
for the CABG surgery group decreased to 1.0 at approximately 125 days rather
than 157 days.
To distinguish whether the revascularization procedures themselves imparted
higher early risk to women as opposed to high-risk coronary anatomy, which
revascularized patients tend to have, we performed an additional stratified
logistic regression analysis with separate analyses for high-risk and low-risk
anatomy patients in each of the treatment groups (Table 2). The early risk among nonrevascularized high-risk women
is less pronounced than the early risks in the PCI and CABG surgery analyses.
For nonrevascularized women with low-risk anatomy, the odds for mortality
were similar to that for men at 30 days and 1 year (Table 2).
The results of the analyses stratified by propensity to undergo PCI
or CABG surgery essentially replicate the results in Table 2. The propensity analysis revealed that the risk for women
vs men is highest when propensity (ie, likelihood) to be revascularized is
high. This finding is most notable for those who actually were revascularized
but is also present to some extent in the "no revascularization" group. The
similarity between the analyses stratified by propensity and by anatomic risk
(Table 2) is expected, given that
the strongest driver of propensity to undergo a revascularization procedure
is having high-risk anatomy.
We explored potential interaction effects between sex and other variables
such as age and comorbidities but did not find any other clinically or statistically
Our study extends current understanding of sex differences in outcomes
after cardiac catheterization by identifying a close link between anatomic
risk, treatment modality, and time in mortality outcomes. We prospectively
investigated sex differences in survival after cardiac catheterization in
a large (n = 37 401) unselected cohort of patients. Many previous studies
on sex differences in outcomes used data from single centers or hospital discharge
data, relied on short follow-up times, focused only on specific patient groups,
or used relatively small samples.2,14,16,18 Many
of these earlier studies thus provide a limited and inconsistent view, with
some reporting that women are at similar or even lower risk than men after
treatment for unstable angina2 or after PCI12 or CABG surgery.14,16
Malenka et al7 argued that a reason for
the lack of consistency in the findings across studies is that many were single-center
studies that were too small to find differences. Using a larger regional sample
of patients undergoing PCI (n = 13 061 procedures), they found that women
treated with PCI had higher in-hospital mortality than men, even after adjustment
for relevant clinical factors.7 Vaccarino et
al,6 also studying a large sample of patients
(n = 384 878) after MI, reported an increase in risk for women vs men,
most notably in an unadjusted analysis (in-hospital mortality 16.7% for women
vs 11.5% for men) but also after adjustment for clinical severity. Subsequent
work by this same group8,9 revealed
a similarly higher risk of in-hospital mortality for women vs men after PCI
and CABG surgery. Interestingly, each of these studies focusing on short-term
revealed that it is particularly younger women who are at high risk relative
to men and that the risk difference between sexes decreases with increasing
age. We did not find any such evidence of an interaction effect between age
and sex, perhaps because we were studying longer-term outcomes for a broad
spectrum of patients.
Our findings provide a potential explanation for the variability in
findings across studies. Outcomes for women vs men appear to be time-sensitive
and procedure-specific, so studies investigating only short-term outcomes
in specific clinical or treatment subgroups will tend to have different findings
than studies investigating longer-term outcomes in other patient subgroups.
Our results suggest that studies focusing on short-term outcomes after CABG
surgery will tend to find large differences in outcomes by sex, whereas a
study focusing on longer-term outcomes in medically treated patients will
perhaps find more modest sex differences.
We need to learn more about what places women at early risk when they
undergo revascularization after cardiac catheterization. Our data allow us
to describe these epidemiologic phenomena, but the APPROACH registry does
not permit us to identify underlying mechanisms. Although existing data demonstrate
potentially important sex differences in cardiac anatomy,26 women
may also have some as-yet unidentified physiologic risk factor or combinations
(interactions) of anatomic and physiologic risk factors.
Our findings of notable sex differences in outcomes early after PCI
and CABG surgery suggest that it is particularly in these areas that sex-based
technologic differences need to be investigated. We propose that special attention
be paid to early physiologic factors (eg, mediators of thrombosis for women),
technologic factors (eg, investigational technologies, PCI and CABG surgery
techniques/equipment tailored for women), and recovery variables that may
clarify women's risk profiles. Moreover, we need to continue to investigate
caregiver decision making for women vs men.
Our study has some limitations. First, it was limited to patients who
have had cardiac catheterization and thus does not account for the outcomes
of women who are not referred for this procedure. We cannot determine whether
the worse outcomes noted early after revascularization for women were due
to the procedures themselves or to the greater incidence of comorbidities
in women brought to the catheterization laboratory. Second, there could have
been different medication use between sexes in the year after catheterization
that could explain some of our findings. However, this explanation is relatively
unlikely because a recent study of pharmacotherapy after MI in Calgary, Alberta,27 revealed that medication therapy early after MI did
not differ between sexes. Third, we focused only on all-cause mortality as
the outcome because it is most readily and reliably captured in APPROACH and
has been widely studied by others and because our mortality analyses reveal
intriguing findings on early risk differences by sex. Fourth, our method for
ascertaining mortality (using data from the Bureau of Vital Statistics) leaves
the possibility of missing patients who had a catheterization procedure but
then moved out of the province. We anticipate, however, that such unmeasured
loss to follow-up is negligible because only Alberta residents' data were
used in these analyses and because Alberta is in a trend of remarkable inward
(rather than outward) migration. Furthermore, it would generally be atypical
for someone to decide to leave Alberta soon after catheterization while in
the midst of a CAD evaluation. A final caveat is that we focused on outcomes
within the first year after catheterization because we had complete ascertainment
of survival to 1 year and because we believed that the most notable finding
was the markedly elevated hazard early after catheterization. In sensitivity
analyses extending to 7 years, the risk for women vs men remained generally
stable and in fact drifted downward to a level slightly below that for men
but with wide confidence intervals that included equivalence of risk by sex.
Despite these limitations, our study extends current understanding of
sex-based differences in cardiac outcomes by demonstrating their time- and
treatment-sensitive nature, a finding that may at least partially explain
the discrepancy in results from earlier studies. Given that the mechanisms
underlying our findings are not explained, we propose a research agenda in
search of explanations for the sex-based outcome differences that we have
demonstrated. Such work will represent a crucial first step toward therapeutic