A, Forest plots showing unadjusted (A) and adjusted (B) in-hospital all-cause mortality when women were compared with men. IV indicates inverse variance; M-H, Mantel-Haenszel; and RR, risk ratio.
Forest plots showing unadjusted (A) and adjusted (B) in-hospital all-cause mortality when women were compared with men. IV indicates inverse variance; M-H, Mantel-Haenszel; and RR, risk ratio.
eMethods. Search Terms, Data Sources and Detailed Search Strategy
eTable 1. Variables Adjusted in the Adjusted Analyses From the Included Studies
eTable 2. Extent and Character of Disease Reported in the Included Studies
eTable 3. Medication Use at the Time of Admission With STEMI
eTable 4. Subgroup and Sensitivity Analyses
eFigure 1. Flow Diagram Describing Study Selection Into the Meta-analysis
eFigure 2. Meta-regression (Difference in DM Prevalence and In-Hospital Mortality in Women)
eFigure 3. Meta Influence Analysis (In-Hospital Mortality)
eFigure 4. Meta Influence Analysis (1-Year Mortality)
eFigure 5. Meta-regression (Difference in Mean Age and In-Hospital Mortality)
eFigure 6. Meta-regression (Difference in Hypertension Prevalence and In-Hospital Mortality)
eFigure 7. Meta-regression (Differences in Diabetes, HTN, Smoking and Age and 1-Year Mortality)
eFigure 8. Funnel Plot for In-Hospital Mortality
eFigure 9. Trim and Fill Analysis for In-Hospital Mortality
eFigure 10. Funnel Plot for 1-Year Mortality
eFigure 11. Trim and Fill Analysis for 1-Year Mortality
Customize your JAMA Network experience by selecting one or more topics from the list below.
Get the latest research based on your areas of interest.
Identify all potential conflicts of interest that might be relevant to your comment.
Conflicts of interest comprise financial interests, activities, and relationships within the past 3 years including but not limited to employment, affiliation, grants or funding, consultancies, honoraria or payment, speaker's bureaus, stock ownership or options, expert testimony, royalties, donation of medical equipment, or patents planned, pending, or issued.
Err on the side of full disclosure.
If you have no conflicts of interest, check "No potential conflicts of interest" in the box below. The information will be posted with your response.
Not all submitted comments are published. Please see our commenting policy for details.
Pancholy SB, Shantha GPS, Patel T, Cheskin LJ. Sex Differences in Short-term and Long-term All-Cause Mortality Among Patients With ST-Segment Elevation Myocardial Infarction Treated by Primary Percutaneous Intervention: A Meta-analysis. JAMA Intern Med. 2014;174(11):1822–1830. doi:10.1001/jamainternmed.2014.4762
Although outcomes in patients with ST-segment elevation myocardial infarction (STEMI) have improved in the past 2 decades, a sex disparity exists in survival, with women having higher mortality than men.
To conduct a meta-analysis of observational studies that examined differences in mortality by sex in patients with STEMI treated with primary percutaneous coronary intervention (PPCI).
MEDLINE, EMBASE, Cochrane central, and electronic databases were searched for relevant studies in all languages and without time restriction.
Studies were included if (1) they studied patients who presented with STEMI, (2) primary percutaneous coronary intervention (PPCI) was the treatment for STEMI, (3) PPCI was performed within 12 hours of symptom onset, and (4) sex-specific in-hospital and/or 1-year mortality were reported.
Data Extraction and Synthesis
Two investigators independently reviewed retrieved citations and assessed eligibility. Discrepancies were resolved by consensus. Quality of included studies was assessed using Newcastle-Ottawa Quality Assessment Scale for cohort studies. Data were pooled using a random-effects model.
Main Outcomes and Measures
Sex-specific in-hospital and 1-year all-cause mortality. Risk ratios (RRs) of mortality were used for these 2 time points, if reported.
Of the 149 studies identified, 35 met inclusion criteria, representing 18 555 women and 49 981 men. In the unadjusted analyses, women were at a higher risk for in-hospital (RR, 1.93; 95% CI, 1.75-2.14 [P < .001, I2 = 14%]) and 1-year all-cause mortality (RR, 1.58; 95% CI, 1.36-1.84 [P < .001, I2 = 51%]) compared with men. However, when adjusted RRs were used, the association between women and higher risk of all-cause mortality was attenuated but still significantly elevated for in-hospital mortality (RR, 1.48; 95% CI, 1.07-2.05 [P = .02, I2 = 56%]), but the higher risk for 1-year mortality in women was no longer significant (RR, 0.90; 95% CI, 0.69-1.17 [P = .42, I2 = 58%]).
Conclusions and Relevance
An increased mortality in women with STEMI treated with PPCI was detected in this large meta-analysis but is likely confounded by baseline cardiovascular risk factors and the differences in clinical profile of male and female patients with STEMI. Intensive cardiovascular risk modification efforts in women may help to reduce this sex disparity.
Coronary artery disease (CAD) is a global public health problem.1,2 In 2010, 379 559 Americans died of CAD, which accounts for 1 of every 6 deaths and averages 1 death every 40 seconds.1 Every year, nearly 620 000 Americans have a first acute coronary event, and 295 000 have a recurrent event.1 Despite this high incidence, survival trends over the last decade (2000-2010) indicate a 31% decline in deaths attributable to cardiovascular disease in the United States, mainly due to cardiovascular risk modification and improvement in acute cardiac care.1
ST-segment elevation myocardial infarction (STEMI) constitutes 25% to 40% of all acute coronary events; primary percutaneous coronary intervention (PPCI) is the established acute reperfusion therapy for STEMI.3 While nearly 2 decades have passed since the first observations, controversy remains as to why short-term and long-term mortality after STEMI is reported to be higher among women than among men.4-40 This was true in the pre-PPCI era29 and remains true today.5,7,19,23 Though studies from across the globe largely support this observation, some do not.5,11-13 It remains unclear if a sex difference in outcomes exists and if this observed difference in mortality between sexes represents a true difference in the biology of the disease or if this is a confounded observation due to baseline differences in cardiovascular risk profile and/or health care utilization between sexes. A pooled analysis combining all available evidence in this topic is lacking in the literature. In this systematic review and meta-analysis, we reviewed the literature and pooled all available evidence reporting sex specific short-term and long-term mortality among patients presenting with STEMI and treated with PPCI.
Meta-analysis of Observational Studies in Epidemiology (MOOSE) guidelines were followed for of the conduct of the current systematic review and meta-analysis.41 The search strategy was developed by the librarian of Johns Hopkins University Bloomberg School of Public Health. We searched MEDLINE, EMBASE, CINAHL, OVID, Cochrane library database, the Web of Science, and Google Scholar for studies assessing sex differences in mortality among patients who underwent PPCI for STEMI. We did not restrict the search by language or date of publication. The last search was performed on April 13, 2014. Literature search details are explained in the eMethods in the Supplement.
We initially reviewed the title and abstracts of retrieved citations. Then full texts of those citations considered relevant were assessed for eligibility for inclusion. Inclusion criteria included the following:
1. Observational studies or randomized clinical trials (RCTs) that involved patients who presented with STEMI. We included studies in which the study population included, in addition to patients with STEMI, patients with non-STEMI, but reported relevant data specific to the STEMI population.
2. PPCI was the reperfusion strategy for STEMI. We included studies in which patients, in addition to those treated exclusively with PPCI, included other subgroups who received other reperfusion strategies such as rescue PCI, as long as they clearly reported sex-specific data with regard to patients with STEMI treated exclusively with PPCI.
3. Studies that explicitly mentioned that PPCI was performed within 12 hours of symptom onset or studies in which PPCI was the intervention in the study but did not specify the time frame within which PPCI was performed. We excluded studies that mentioned that percutaneous coronary intervention (PCI) was performed within 24 hours but did not explicitly say that the PCI received by all or some of their patients with STEMI was PPCI, as we suspected that they reported pooled results for all types of PCIs and did not separately report results for patients who were treated with PPCI.
4. Studies that reported in-hospital and/or 1-year mortality as one of the study outcomes. In-hospital and 1-year mortality were chosen as study outcomes because, though arbitrary, in-hospital mortality will represent a reliable time point to assess short-term or immediate cardiovascular mortality and 1-year mortality, a reliable long-term time point where cardiovascular causes will play a predominant role in mortality. We included conference abstracts that reported data relevant to our research question. We excluded case reports. Two coauthors (S.B.P. and G.P.S.S.) independently assessed studies for eligibility. Discrepancy was resolved by consensus.
Two coauthors (S.B.P. and G.P.S.S.) independently extracted data from the included full-text citations. The following information was abstracted: the last name of the first author, publication year, country where the study was performed, study design (observational or RCTs), total participants in the study, number of male participants, number of female participants, number of deaths (in-hospital and/or 1-year mortality) among male and female participants, baseline cardiovascular risk profile (mean age for both sexes, percentage with diabetes mellitus [DM], hypertension [HTN], dyslipidemia, smoking [ever smoked]), left ventricular ejection fraction (LVEF), door-to-balloon time (DTB) in minutes, angiography findings (number of diseased vessels, infarct-related arteries), medication use at the time of presentation with STEMI (aspirin, statins, β-blockers, angiotensin-converting enzyme inhibitors [ACEIs]/angiotensin receptor blockers [ARB]), and use of stents (bare metal stents or drug-eluting stents) specific to each sex. For studies that reported adjusted measures of association with mortality (women compared with men), the variables that were adjusted in these analyses were abstracted. We assessed quality of included studies using the Newcastle-Ottawa Quality Assessment Scale for cohort studies.42
Abstracted data from the included studies were entered into RevMan 5.1 (Nordic Cochrane Center) statistical software.43 We sought to perform 4 analyses: (1) unadjusted risk ratios (RRs) for in-hospital all-cause mortality (women vs men), using raw numbers of death and total participants at risk for death specific to each sex; (2) adjusted RRs for in-hospital all-cause mortality (women and men), using adjusted RRs if described in the included studies; (3) unadjusted 1-year all-cause mortality (women and men), using raw numbers of death and total participants at risk for death specific to each sex; and (4) adjusted 1-year all-cause mortality (women and men), using adjusted RRs for mortality at 1-year if described in the reports. We used RRs instead of odds ratios because if the event of interest is rare, odds ratios tend to overestimate the strength of association.43 Hazard ratios were used interchangeably with RRs. Considering the clinical and statistical heterogeneity between studies, data were combined using the DerSimonian and Laird random-effects model with inverse variance weighting.44 Estimates were reported as RRs comparing women with men, with 95% CIs. Differences were considered statistically significant at P < .05 (2-sided).
Heterogeneity across studies was assessed with the Cochran Q statistic (χ2), with P < .10 considered significant, and with the I2 test.45 An I2 value of less than 25% was considered low heterogeneity; 25% to 50%, moderate; and greater than 50%, substantial heterogeneity. Also, we performed meta-regression investigating the sources of heterogeneity in the included studies. The potential sources assessed were differences in age, DM prevalence, HTN prevalence, smoking prevalence, dyslipidemia prevalence, LVEF, DBT, and number of diseased coronary vessels. Further, we performed 2 subgroup analyses: (1) studies performed in the United States vs studies performed outside of United States and (2) studies published before vs after 2004 (2004 was the year of US Food and Drug Administration approval for drug-eluting stents); this analysis was to account for temporal trends in the STEMI care. Also, we performed 3 sensitivity analyses: (1) restricting to studies with a Newcastle-Ottawa quality assessment score of 6 points or more; (2) using a 1-study-removed analysis, where 1 study was removed at a time from the pool analysis to assess the effect of each study on the combined effect; and (3) using data from RCTs. Because only 1 RCT30 reported sex-specific in-hospital mortality data, we could not report meta-analytic estimate for in-hospital mortality for this analysis. However, 3 RCTs30,33,39 and 2 RCTs30,33 respectively reported sex-specific 1 year all-cause mortality with unadjusted and adjusted rick estimates, and hence we have reported the meta-analytic estimate for 1-year all-cause mortality for the unadjusted and adjusted analyses. Further, we performed a meta-influence analysis to assess if 1 or more studies had greater impact on the strength of association, and we evaluated the effect of removal of such studies on the pooled estimate.
Publication bias was assessed using the Egger linear regression test,46 visual inspection of funnel plots, and the Begg-Mazumdar test. The trim-and-fill method was used to adjust for publication bias. The trim-and-fill method determines where missing studies are likely to fall, adds them to the analysis, and then recomputes the combined effect. These analyses were performed using STATA version 11 (StataCorp).47P < .05 was considered statistically significant.
Study selection details are described in eFigure 1 in the Supplement. From a total of 2344 citations identified, 35 studies5-28,30-40 were included for qualitative synthesis. Of these, 22 studies,5-26 12 studies,6,9,10,16-18,22,24-28 12 studies,9,23,30-39 and 5 studies9,30,33,38,40 were included for meta-analysis for unadjusted in-hospital mortality, adjusted in-hospital mortality, unadjusted 1-year mortality, and adjusted 1-year mortality, respectively (eFigure 1 in the Supplement).
Of the 35 included studies, 6 were conducted in the United States (Table 1).6,15,17,24,32,37 Except for 4 included studies,13,30,33,39 which were post hoc analyses of data from RCTs, the remaining 31 followed an observational study design. Furthermore, 16 of the 35 included studies reported adjusted analyses.6,9,10,16-18,22,24-28,30,33,38,40 Most studies adjusted for age, HTN, smoking, dyslipidemia, and diabetes, while some adjusted for prior vascular disease (ie, prior myocardial infarction, stroke, coronary artery bypass graft, peripheral vascular disease)6,9,16,40 and patient’s hemodynamic status6,9,22,24 (eTable 1 in the Supplement). Few performed propensity-matched analysis to control for confounders.17,33
The 35 included studies involved 68 536 patients with STEMI treated by PPCI (18 555 women and 49 981 men). Most studies reported a higher age and higher prevalence of DM, HTN, and dyslipidemia in women compared with men, while greater proportions of men were smokers compared with women (Table 2). Most studies did not report significant differences between sexes with variables with regard to LVEF, DBT, number of diseased arteries, infarct-related arteries, and medication use at the time of admission (Table 2 and eTables 2 and 3 in the Supplement).
Twenty-two studies reported unadjusted in-hospital all-cause mortality comparing women with men,5-26 and involved 11 894 women and 29 872 men, a total of 41 766 patients. There were 887 (7.5% rate) and 1165 (3.9% rate) in-hospital deaths in women and men, respectively. In the unadjusted analysis, women had a significantly higher risk of in-hospital all-cause mortality compared with men (RR, 1.93; 95% CI, 1.75-2.14 [P < .001, I2 = 14%]) (Figure 1A). With the 2 subgroup analyses and the 2 sensitivity analyses described in the Statistical Analysis subsection of the Methods section, the strength of association for women with in-hospital mortality remained robust (eTable 4 in the Supplement). However, in the adjusted analysis, using adjusted RRs from 12 studies, the strength of association for all-cause mortality in women compared with men, though remaining significant, was significantly attenuated (RR, 1.48; 95% CI, 1.07-2.05 [P = .02, I2 = 56%]) (Figure 1B). Meta-influence analysis showed a possibly higher influence on the effect estimate attributable to 3 studies in the analysis for unadjusted in-hospital all-cause mortality17,18,23 (eFigure 3 in the Supplement). Removal of these 3 improved the strength of association (RR, 2.21; 95% CI, 1.93-2.53 [P < .001]) of higher unadjusted in-hospital all-cause mortality in women. However, in the adjusted in-hospital all-cause mortality analysis, none of these studies exerted higher influence, since the removal of these 3 studies did not alter the effect estimate (RR, 1.54; 95% CI, 1.07-2.23 [P = .02]).
Twelve studies involving 7169 women and 21 767 men (28 936 patients total) reported unadjusted all-cause mortality numbers for women and men. They found 633 (8.8% rate) and 1191 (5.5% rate) deaths at 1 year in women and men, respectively. Women had significantly higher risk of 1-year all-cause mortality in the unadjusted analysis (RR, 1.58; 95% CI, 1.36-1.84 [P < .001, I2 = 51%]) (Figure 2A). In the 3 sensitivity analyses and the 2 subgroup analyses, the association of women with increased 1-year all-cause mortality remained significant (eTable 4 in the Supplement). Meta-influence analysis showed a possibly higher influence due to 1 study23 (eFigure 4 in the Supplement) and removal of this study did not alter the effect estimate (RR, 1.54; 95% CI, 1.31-1.82 [P = .002]). However, in the adjusted analysis (Figure 2B), the association between women and increased 1-year all-cause mortality lost significance (RR, 0.90; 95% CI, 0.69-1.17 [P = .42, I2 = 58%]). This was true in the sensitivity analysis restricted to data from 2 RCTs (RR, 0.81; 95% CI, 0.59-1.11 [P = .19, I2 = 0%]).
Difference in DM prevalence between sexes was identified to be the only significant source of heterogeneity (β coefficient, 0.036; P = .02) (eFigure 2 in the Supplement). Age difference (β coefficient, −0.106, P = .10) and difference in HTN prevalence (β coefficient, 0.031, P = .06) were not significant (eFigures 5 and 6 in the Supplement). Differences in smoking or dyslipidemia prevalence, LVEF difference, and DBT difference between sexes were not identified as significant sources of heterogeneity in in-hospital mortality. In the assessment for 1-year mortality, none of these variables (differences in age, DM or HTN prevalence, smoking, dyslipidemia prevalence, LVEF, DBT, and number of diseased vessels) were found to be sources for heterogeneity (eFigure 7 in the Supplement).
Visual inspection of funnel plots for in-hospital mortality showed evidence for some missing studies with possible weaker strengths of association ranging between log risk ratio 0 and 1 (eFigure 8 in the Supplement). This was confirmed by the Egger (P = .02) and Begg (P = .02) tests. The results of trim-and-fill analysis showed that 7 studies may have been missing, and the addition of these 7 studies to the pooled analysis for in-hospital mortality difference would make the RR 1.76 (95% CI, 1.55-1.99 [P < .001]) (eFigure 9 in the Supplement). Visual inspection of funnel plots for 1-year mortality showed evidence that smaller sample size studies were under represented (eFigure 10 in the Supplement), but the Egger (P = .93) and Begg (P = .54) tests did not support this observation. The results of trim-and-fill analysis showed that 1 study may have been missing, and its addition would make the RR 1.543 (95% CI, 1.316-1.810 [P < .001]) (eFigure 11 in the Supplement).
Our meta-analysis, using data from 35 studies involving 18 555 women and 49 981 men with STEMI treated with PPCI, found that women have nearly 2 times the risk for in-hospital all-cause mortality and 1.5 times the risk for 1-year all-cause mortality compared with men. However, this association was significantly attenuated when the analysis was adjusted for participants’ baseline cardiovascular risk factor and clinical profile.
Women with STEMI have a poorer baseline cardiovascular risk profile, tend to be older at presentation, and have a higher prevalence of DM, HTN, and dyslipidemia.5-28,48-52 Prior research has suggested that these baseline differences between sexes may be contributors to the sex differences in mortality after STEMI.5-28,48-52 Adding validity to this theory are our findings that the increased mortality in women attenuates after adjusting for cardiovascular risk factors and patient’s clinical and hemodynamic status at presentation. Although conclusive inferences are difficult, considering the variability in the factors adjusted for in the included studies (eTable 1 in the Supplement) and the availability of far fewer studies in the adjusted analyses compared with the unadjusted analyses, we hypothesize that the findings of an association of women with increased mortality are significantly confounded by differences in traditional cardiovascular risk factors, sex-specific differences in clinical presentation and response by the medical infrastructure in women compared with men with STEMI. Furthermore, with the available data we were limited in our ability to assess adequacy of risk adjustment with the variables used in the adjusted models of the included studies. Inadequate risk adjustment in the included studies may well explain the incomplete attenuation of higher risk for mortality in women in our adjusted analysis.
Health care utilization among women has often been found to be suboptimal compared with men presenting with STEMI.48,49 Women were less likely to receive thrombolysis in the thrombolysis era.48,49 Though acute reperfusion therapy confers a higher survival advantage in women than in men,32 women are less likely to receive primary reperfusion therapy.48 Women are sicker (worse Killip class), present later, and are less likely to get newer evidence-based therapies such as thienopyridines and ACEIs and/or ARBs compared with men.48 Also, women are more likely to have a prehospital missed diagnosis of STEMI, higher incidence of presentation to a non–PCI-capable facility, and a higher risk for interhospital transfer to a PCI-capable facility and hence are at risk for reperfusion delay.53 Melberg et al54 showed that women even have a lower priority for emergency ambulance service (78.7% vs 89.4% in women vs men; P = .04) when being transported presenting with possible STEMI, resulting in potentially longer ischemic time. While the reasons for these demonstrated disparities are unclear, they could easily contribute to increased mortality, both in-hospital and long term, in women presenting with STEMI. Furthermore, our pooled sample involved nearly two and a half times as many men (n = 49 981) as women (n = 18 555). This oversampling of men may possibly be due to selection bias. Our observation of poorer outcomes among women could be attributed to only the more severe cases in women reaching the health care system, while some less-severe cases among women did not present to the health care system and were underreported in these studies. Supporting or refuting this hypothesis is beyond the scope of this article but warrants further consideration and investigation.
Strengths of our systematic review include (1) its relatively large sample size and increased precision after pooling 35 studies; (2) our well-defined exposure measure (PPCI in patients with STEMI) and outcome measures (in-hospital mortality and 1-year mortality) minimizes the possibility of measurement error, and (3) the potential generalizability of our study findings to international populations, since our meta-analysis included studies from across the globe.
Limitations of our analysis include the following: (1) All 35 included studies used an observational design for examining sex differences. Even among the included RCTs, sex was not the randomization variable. Hence, confounding due to unknown factors is possible. (2) Because of the wide variability in the regions of the globe where these studies were conducted, differences in temporal trends in standards of care, lack of data on variables such as shock at presentation, PPCI-related bleeding complications, and time to PPCI (early vs late), which are known predictors of outcomes post-STEMI, could have all contributed to the high clinical heterogeneity in our analyses. (3) There was significant statistical heterogeneity between studies, which, except for DM in the in-hospital mortality analysis, could not be explained by other variables. (4) We could report only all-cause mortality and not specifically cardiovascular mortality owing to the lack of these data. (5) Adjusted analyses could be performed on far fewer studies compared with nonadjusted estimates; hence, the observed attenuation of strength of association may represent adjustment for confounders or a lack of sufficient power. (6) An absence of difference between sexes with important prognostic variables like DTB may also be due to sparse reporting, since only 10 of the 35 included studies reported sex-specific DTB times. Also, cardioprotective medication use, a factor known to differ between sexes, could not be assessed in the meta-regression because few studies (4 studies) reported sex-specific medication use data (eTable 3 in the Supplement).
Although women have been observed to have higher short-term and long-term mortality after STEMI when treated with PPCI, both in the present study and in previous studies, we report that this finding is possibly confounded by baseline differences in cardiovascular risk profile and clinical profile at the time of presentation with STEMI. Considering the limitations of the data from the included studies, our results should serve as hypothesis generating and stimulate further research with robust strategies for risk adjustment to elucidate this hypothesis more accurately. Because most of these risk factors are modifiable, appropriate measures to optimize health care utilization in women may reduce this gender gap.
Accepted for Publication: July 25, 2014.
Corresponding Author: Samir Bipin Pancholy, MD, The Commonwealth Medical College, 501 Madison Ave, Scranton, PA 18510 (email@example.com).
Published Online: September 29, 2014. doi:10.1001/jamainternmed.2014.4762.
Author Contributions: Drs Pancholy and Shantha had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Pancholy, Shantha, Patel.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Pancholy, Shantha, Patel.
Critical revision of the manuscript for important intellectual content: Pancholy, Shantha, Cheskin.
Statistical analysis: Pancholy, Shantha, Patel.
Study supervision: Pancholy, Shantha.
Conflict of Interest Disclosures: Dr Pancholy has received a speaker fee from Pfizer and Terumo and a research grant from Accumed Radial Systems Inc. No other disclosures are reported.
Create a personal account or sign in to: