Sex Differences in Long-Term Cause-Specific Mortality After Percutaneous Coronary Intervention

Introduction

Women have higher long-term mortality compared with men after percutaneous coronary intervention (PCI).^1–3 Whether this sex survival gap is because of the older age and greater comorbidity burden of women undergoing PCI or to undefined sex-specific factors remains uncertain. In addition, the impact of changing demographics on mortality trends in women compared with men after PCI is incompletely understood.^1,4,5

After adjustment for known baseline characteristics, registry studies have variously reported women to have higher, similar, or lower mortality after PCI than men.^1,5–9 Methodologic aspects may explain the discrepancy between studies. These include studies undertaken in different eras of PCI technology,^6,9 the exclusion of subjects <65 years old (Medicare),⁷ and the presence of unmeasured confounding factors that influence mortality risk. A major limitation of previous studies has been the use of all-cause mortality, rather than cause-specific mortality as an end point. This is particularly relevant in the contemporary era, where noncardiac mortality is more common than death from cardiac causes during long-term follow-up after PCI.^10–12

Cause-specific mortality represents a more useful synthesis of baseline risk than all-cause mortality because it enables separation of the influence of cardiac and noncardiac morbidities. However, cause of death is difficult to ascertain. Death certificates and International Classification of Diseases–coded cause of death are often inaccurate, and more accurate establishment of cause of death requires time and labor-intensive resources.^13,14 Cause determination is, therefore, infrequently undertaken in national registry studies of PCI.

We performed a single-center study, using rigorously ascertained cause-specific mortality to (1) evaluate long-term trends in cause of death in women compared with men after PCI, (2) examine the effect of baseline risk on final cause of death to determine whether there is a sex-specific biological factor, and (3) determine specific causes of death in women and men to facilitate future targeting of clinical care.

Methods

The data, analytic methods, and study materials will not be made available to other researchers for purposes of reproducing the results or replicating the procedure

Patient Population

The study population was derived from the Mayo Clinic PCI registry, a prospective registry of consecutive patients who underwent index PCI at Mayo Clinic, Rochester, MN, between 1991 and 2012. A total of 1062 patients (4.3%) did not provide consent for use of their records for research purposes and were excluded as required by state of Minnesota statute. The study was approved by the Mayo Clinic Institutional Review Board.

Because demographics and treatment strategies have significantly changed for the past 20 years, patients were analyzed in 3 time periods: early (1991–1997), middle (1998–2005), and contemporary (2006–2012).

Analysis of Demographic Data

Detailed baseline demographics were recorded at the time of admission according to standardized definitions, as previously described.¹¹ Acute coronary syndrome was defined as a myocardial infarction (MI) or unstable angina in the 7 days before PCI. Multivessel disease was defined as a major epicardial coronary artery of ≥70% stenosis assessed visually, with a further major artery with stenosis of ≥50%, or ≥50% stenosis of the left main stem. Coronary artery lesions were classified according to American Heart Association guidelines.¹⁵ Comorbidities were defined using the Mayo Diagnosis Index database for all time points before index PCI. The Charlson index, a measure of the burden of comorbidity, was calculated as previously described¹⁶ and divided into cardiac and noncardiac components. The cardiac component was defined as the weights for MI and congestive heart failure whereas the noncardiac component was defined as the total score minus the cardiac component.

Determination of Vital Status and Cause of Death

Scheduled yearly surveillance was performed routinely after PCI by trained personnel as previously described.¹¹ If death occurred, details were obtained through review of local and external medical records, death certification, telephone contact with patient family, and external providers. In ≈10% of deaths, death certificate was the only available source for classification. Experienced data analysts recorded the details of each death and performed initial classification. Three physicians reviewed each death and performed final classification after reaching consensus.

Cause of death was subdivided as follows: (1) cardiac death: sudden cardiac death (SCD), confirmed MI, congestive heart failure, structural heart disease, cardiac procedure related (intervention or cardiac surgery); (2) noncardiac death: vascular (cerebrovascular accident, rupture of abdominal aortic aneurysm, other vascular fatality), cancer, infectious disease, chronic disease (pulmonary, renal failure, liver failure, neurological and natural causes), hemorrhage, pulmonary embolus, iatrogenic (excluding cardiac iatrogenic), trauma (including fatal motor vehicle accidents), acute surgical non cardiac cause, other noncardiac; and (3) unknown or unobtainable.

SCD was defined as a documented arrhythmogenic death or an unexpected out of hospital presumed-pulseless death in the absence of a noncardiac explanation. When the primary cause of death was listed on the death certificate as ischemic heart disease and no other information was available, the death was classified as MI. Death from valvular heart disease was classified as structural heart disease. In the situation of equally competing noncardiac and cardiac causes of death, we favored cardiac classification. Fatal motor vehicle accidents were classified as accident/trauma, with the recognition that SCD resulting in a collision could not be reliably excluded.

Statistical Analyses

Normally distributed continuous variables were presented as the mean and SD, categorical variables as the frequency, and group percentage and survival variables as the cumulative number of events and Kaplan–Meier estimated cumulative percentage. T tests were used to compare continuous variables, Pearson χ² test for categorical variables, Kruskal–Wallis for ordinal variables, and the log-rank test for survival variables. Linear contrasts of means were used to test differences in continuous variables across eras. The Cochran–Armitage trend test was used similarly for discrete variables.

Competing risks methods were used to estimate cumulative incidence for cause-specific mortality. Cox proportional hazards models were used to test differences in cause-specific mortality between groups. To investigate whether the differences in cause-specific incidence between sexes could be attributable to differences in comorbidity burden, we used strata-weighted estimation. A strata-weighted estimation approach was selected rather than adjustment through modeling because it is assumption free, does not impose a proportional hazards difference between men and women, and allows for differential covariate effects as well as covariate-only interactions without requiring modeling. We stratified all subjects by age (<50, 50–69, and ≥70 years), era (1991–1997, 1998–2005, and 2006–2012), acute coronary syndrome indication, number of diseased vessels, Charlson score for cardiac diagnoses, Charlson score for noncardiac diagnoses, and diagnoses for congestive heart failure, ulcer, moderate/severe renal disease, and metastatic solid tumor. To estimates adjusted survival curves, subjects were weighted such that each strata was balanced in both sexes. For example, if 1 strata comprised 8% of the entire sample, but was 6% in women and 10% in men, that strata would be upweighted in women and downweighted in men such that it comprised 8% in both. Tests for differences were conducted using Cox proportional hazards regression with the separate baseline hazards for the factors used to define the strata. SAS 9.4 (SAS Institute, Cary, NC) was used for all analyses.

Results

Clinical and Procedural Characteristics

Index PCI was performed in 16 280 men and 6847 women between 1991 and 2012. Clinical, angiographic, and procedural characteristics are shown in Table 1. Women were older than men, had a higher Charlson Index of noncardiac and cardiac comorbidities, and greater prevalence of hypertension and diabetes mellitus. The mean age, body mass index, and Charlson index increased for the 3 eras (P<0.0001) for both sexes while the incidence of previous MI decreased significantly.

Unadjusted Analysis

In-hospital death occurred in 525 patients (2.3%). There were 3213 (47%) deaths in the female and 6387 (39%) deaths in the male cohort during follow-up. Cause of death was established in 9149 of 9600 (95%) deaths.

All-cause mortality was higher in women compared with men for all eras. Cause-specific analysis indicated this excess was because of noncardiac deaths. Across 3 time periods, there was a temporal shift from predominantly cardiac causes of death to predominantly noncardiac deaths in both sexes (Figure 1). Similar trends were observed in the subset of patients presenting with acute coronary syndrome.

• Download figure
• Download PowerPoint

To exclude a change in tertiary referral patterns as a potential bias underlying the observations, a separate analysis was performed confined to residents within 50 miles of the institution. This indicated similar trends in causes of death in women and men (data not shown).

Adjusted Analysis

After weighted adjustment for age and comorbidities, the shift from predominantly cardiac to predominantly noncardiac causes remained. However, there was no statistically significant difference in adjusted cardiac and noncardiac mortalities between the sexes. In the contemporary era, the adjusted absolute cardiac mortality at 5 years was similar in both sexes (4.7% versus 6.4%; P=0.29) and so was the adjusted noncardiac mortality (10.4% versus 10.8%; P=0.34; Figure 2).

• Download figure
• Download PowerPoint

Subtypes of Cardiac and Noncardiac Deaths

Proportionate subtypes of death across 3 eras in both sexes are shown in Table 2. In the contemporary era, the cumulative 5-year incidence of MI/SCD after PCI was similar in women and men. However, nominally more women died of heart failure. Women were most likely to die of chronic diseases, whereas more men died from cancer (Figure 3). After adjustment for age and baseline comorbidities, there was no difference in heart failure or chronic disease deaths between the sexes; however, the higher rates of cancer in males compared with females remained. Further analysis indicated that hypertension was associated with a higher risk of heart failure deaths in both sexes in the contemporary era and that this was more so in women than men.

• Download figure
• Download PowerPoint

Effect of Age on Cause of Death After PCI

Comparison of unadjusted 5-year cardiac mortality showed a significant interaction between age and sex. To maximize sample size for comparison, the total population was studied. Women <50 years had significantly higher cardiac mortality compared with men (hazard ratio, 1.49; 95% confidence interval, 1.09–2.04), and women >70 years exhibited nominally toward lower cardiac mortality (hazard ratio, 0.94; 95% confidence interval, 0.86–1.02; Figure 4).

• Download figure
• Download PowerPoint

Subgroup analysis of patients under the age of 50 years revealed a higher unadjusted cardiac mortality in women (6.4% versus 3.3%; P=0.01) but with similar rates after adjustment (1.0% versus 2.3%; P=0.45).

Discussion

The major findings of this study were as follows: (1) There has been a temporal decline in absolute incidence of cardiac deaths after PCI in both sexes, with noncardiac deaths predominating in the contemporary era. (2) The excess mortality in women compared with men after PCI was because of noncardiac causes and related to higher baseline risk in women. (3) There was no evidence of a sex-specific factor influencing cardiac mortality. (4) The most common causes of death in the contemporary era were chronic diseases and heart failure in women but cancer and MI/SCD in men.

Trends in Cause of Death in Women and Men

The current study demonstrated a major temporal shift in cause of death after PCI in both women and men such that in the current era of PCI, both women and men are twice as likely to die from noncardiac compared with cardiac disease. The decrease in cardiac mortality observed after PCI for the 3 eras is consistent with the temporal decline in cardiac deaths observed in women and men in the post-PCI and general population^8,17,18 and may, in part, be related to the increase in use of secondary prevention medications observed in both sexes across eras.

Because cardiac mortality has decreased and the demographics of patients presenting for PCI have changed over time, the likelihood of death from the competing risks of noncardiac disease has become proportionately greater. Of note, the current study demonstrated differences in detailed classification of cause of death between women and men, with women being at higher risk for chronic disease and heart failure death and men for cancer and MI. Although these findings likely result from differences in baseline risk in women and men, they do suggest sex-specific differences in long-term healthcare needs after PCI.

We demonstrated that the excess mortality observed in women after PCI was because of noncardiac, rather than cardiac deaths. Although there are biologically plausible reasons for sex differences in cardiac disease, including cyclic estrogen levels¹⁹ and anatomic factors,^20,21 adjustment for baseline risk demonstrated no difference in long-term cardiac death rates between the sexes.

Mechanisms for Sex-Specific Differences in Mortality

Prior studies have also demonstrated a higher unadjusted all-cause mortality in women compared with men.^6–8 However, after adjustment, US registries reported lower rates of all-cause mortality in women,⁷ but European registries of similar size in the contemporary era still reported female sex to be an independent risk factor for all-cause mortality.²²

Despite multicenter design and large sample sizes, the capacity of adjusted analyses in these studies to determine whether sex differences in long-term mortality risk could be accounted for by baseline risk has been limited. Variables collected and used in statistical models have tended to be cardiac centric and procedural in type. Given that the dominant cause of death in the contemporary post-PCI population is noncardiac, these factors have diminishing or no influence on long-term all-cause mortality after PCI compared with parameters of noncardiac risk.^4,23 More so, differences in type and number of variables collected, duration of follow-up, and era of study may contribute to discrepant findings between prior registry studies.

The current study distinguishes itself from prior registry studies through the use of meticulously ascertained cause-specific mortality as an end point for study. This has enabled a direct examination of the impact of cardiac and noncardiac variables on mortality after PCI. The study determined that the excess in mortality in women is because of noncardiac deaths and that post-PCI, women have similar rates of adjusted cardiac mortality compared with men. These findings suggest that increased long-term chronic disease and health surveillance in both sexes may be warranted to reduce overall mortality after PCI.

Effect of Age of Presentation on Sex Differences in Mortality Post-PCI

Women <50 years of age had significantly higher cardiac death rates compared their male counterparts. Despite the relatively small sample size, these observations are consistent with other studies in large populations indicating higher mortality risk in younger women with coronary artery disease.^20,24,25 Reasons behind the elevated risk could relate to an increased burden of risk factors, differences in disease extent and nature,²⁶ differences in timing and pathogenesis of acute presentation,²⁷ and selection biases for revascularization. Although further study is required to understand the increased risk, the current findings underscore the critical importance of close attention to secondary preventative measures and follow-up in young women after PCI.

Study Limitations

This is a single-center cohort study at a tertiary institution and therefore may not be generalizable to the US population as a whole. Follow-up data were collected prospectively, and cause of death was typically determined through prospective scheduled surveillance. However, the final ascertainment of cause of death was performed retrospectively. We studied patient records extensively to determine cause of death rather than relying on death certificates alone because these are known to have limited accuracy.¹³ Despite this methodology, it is possible that misclassification of causes of death occurred in some. Given overlap in clinical presentations, conditions such as pulmonary embolus and acute stroke or other conditions occurring during sleep could have been misclassified as SCD. More so, the risk of misclassification might be expected to be greater in populations with multiple competing comorbidities, such as, in the current study.

Our study addressed the post-PCI population alone. Therefore, differences in treatment utilization between sexes and differences resulting from differing clinical presentation and investigation were not assessed. These may have resulted in selection bias in the population studied. We did not collect data on menopausal status, hormone replacement therapy, or oral contraceptive use and so were unable to estimate their effects on outcomes in women. Data on ejection fraction at baseline were incomplete, and follow-up ejection fraction was not a collected variable.

Conclusions

In women and men, there has been a shift from predominantly cardiac deaths to predominantly noncardiac death during long-term follow-up after PCI. Men are now most likely to die from cancer and women from chronic disease and heart failure. Although women had higher unadjusted rates of all-cause mortality compared with men, this was because of an excess of noncardiac, rather than cardiac deaths with no evidence of a sex-specific biological factor.

Footnotes