Propensity Score-Matched Analysis of Effects of Clinical Characteristics and Treatment on Gender Difference in Outcomes After Acute Myocardial Infarction




The greater mortality observed in women compared to men after acute myocardial infarction remains unexplained. Using an analysis of pairs, matched on a conditional probability of being male (propensity score), we assessed the effect of the baseline characteristics and management on 30-day mortality. Consecutive patients were included from January 2006 to December 2007. Two propensity scores (for being male) were calculated, 1 from the baseline characteristics and 1 from both the baseline characteristics and treatment. Two matched cohorts were composed using 1:1 matching and computed using the best 8 digits of the propensity score. Paired analyses were performed using conditional regression analysis. During the study period, 3,510 patients were included in the registry; 1,119 (32%) were women. Compared to the men, the women were 10 years older, had more co-morbidities, less often underwent angiography and reperfusion, and received less medical treatment. The 30-day mortality rate was 12.3% (130 of 1,060) for the women and 7.2% (167 of 2,324) for the men (p <0.001). The 2 matched populations represented 1,298 and 1,168 patients. After matching using the baseline characteristics, the only difference in treatment was a lower rate of angiography and reperfusion, with a trend toward greater 30-day mortality in women. After matching using both baseline characteristics and treatment, the 30-day mortality was similar for the men and women, suggesting that the increased use of invasive procedures in women could potentially be beneficial. In conclusion, compared to men, the 30-day mortality is greater in women and explained primarily by differences in baseline characteristics and to a lesser degree by differences in management. The difference in the use of invasive procedures persisted after matching by characteristics. In contrast, after matching using the baseline characteristics and treatment, the 30-day mortality was comparable across the genders.


Gender differences in mortality after acute myocardial infarction (AMI) were first reported >20 years ago and have since been extensively studied. Bias in the comparison between the genders is mostly induced by differences in age, co-morbidities, and the clinical presentation or extent of coronary disease. Differences in management between men and women have also been observed, particularly in terms of the use of reperfusion and invasive procedures, even though these treatments have similar efficacy in both men and women. The main difference in management has been the underuse of invasive procedures in women, although it remains unclear whether wider use of percutaneous coronary intervention (PCI) in women would offset the difference in mortality observed between the genders after AMI. Interactions have been observed, for example with the type of AMI and age, but with conflicting results, potentially caused by bias or inadequate control for covariates. The use of a propensity score-matched analysis is an alternative approach that selects patients “matched” on the conditional probability of being male (or female). This is likely a more powerful method of controlling for bias, because adjustment often takes into account only a limited number of variables. The aim of the present study was to determine the effect of the difference in baseline characteristics and management on 30-day mortality between men and women after AMI, using an analysis of pairs, matched on the conditional probability of being male.


Methods


All participants in the registry provided oral informed consent to record their vital status at 1 month. The ethics committee of the University Hospital of Besançon approved this procedure.


The study population was taken from the “Registre Franc Comtois des Syndromes Coronariens Aigus,” an ongoing registry involving all 10 cardiology centers in the region of Franche-Comté, a region in Eastern France with a population of 1.2 million inhabitants. All consecutive patients admitted from January 2006 to December 2007 for AMI, with or without ST-segment elevation myocardial infarction (STEMI and NSTEMI), and who gave informed consent were eligible. A dedicated team of data managers was responsible for completing and verifying the data collection. Computerized checks were performed to verify the coherence of the data, queries were generated in case of inconsistencies, and a sample of the medical records was reviewed in each center.


The variables prospectively recorded included the type of AMI, demographic data, previous diseases, clinical presentation, and treatment actually given during hospitalization (i.e., aspirin, clopidogrel, angiotensin-converting enzyme inhibitors or angiotensin receptor antagonists, β blockers, statins, coronary angiography, and glycoprotein IIb/IIIa receptor inhibitors in patients with NSTEMI and reperfusion by primary angioplasty or thrombolysis in patients with STEMI).


The primary end point was all-cause mortality at 1 month. The secondary end point was in-hospital mortality.


The categorical variables are presented as the number of cases (percentage), continuous non-normally distributed variables as the median (interquartile range), and continuous normally distributed variables as mean ± standard deviation. To account for potential confounding and selection biases, a propensity score-matched comparison was used. Although matched analyses can result in the selection of patients, they provide a valid estimation of the difference between genders. To assess the effects of the difference in both baseline characteristics and treatment, a 2-step approach was used. First, a propensity score for being male was calculated using logistic regression analysis, including age, body weight, AMI type, hospital center, cardiovascular risk factors (i.e., hypertension, diabetes, hypercholesterolemia, and smoking status), history (i.e., previous infarction, percutaneous or surgical revascularization, stroke, peripheral artery disease, or heart failure), hemodynamic conditions at admission (i.e., Killip class, heart rate, systolic blood pressure, troponin release), and biologic variables (i.e., hemoglobin and serum glucose and creatinine levels). A 1:1 matching was computed on the best 8 digits of propensity score, matching on a large number of variables simultaneously. The maximum difference in propensity score allowing a match was 0.015. A second propensity score was calculated using the same variables as for the first, plus the use of aspirin, clopidogrel, β blockers, angiotensin-converting enzyme inhibitor, statins, coronary angiography, and reperfusion for patients with STEMI or use of glycoprotein IIb/IIIa inhibitors for those with NSTEMI and coronary revascularization. A second matched population was thus obtained after matching using the second propensity score.


To assess the efficacy of propensity score matching to generate 2 comparable groups, we compared the average values in baseline characteristics between groups (not the difference within the matched pairs), using t statistics for continuous variables, and chi-square statistics for categorical variables.


A second analysis was performed in the whole cohort adjusting for the 2 propensity scores (used as covariates). Kaplan-Meier cumulative 30-day mortality curves were constructed for the whole population and in each of the 2 matched populations, and the differences between genders were tested using the log-rank test. To compare the matched pairs, we used the generalized estimation equation, the results are expressed as the odds ratios, with the 95% confidence interval. Interactions were tested using the Breslow-Day test between gender and age (by comparing the 2 groups categorized according to the median age) and between gender and the type of AMI. All tests were 2-sided, and p <0.05 was considered significant. All analyses were performed using SAS software, version 9.2 (SAS Institute, Cary, North Carolina).




Results


During the study period, 3,510 patients were included in the registry, 1,119 women (32%) and 2,391 men (68%). The 2 matched populations represented 1,298 and 1,196 patients, respectively ( Figure 1 ).




Figure 1


Flow chart of study population.


Compared to men, the women admitted for AMI were 10 years older and more frequently had diabetes, hypertension, and renal dysfunction. Women presented with a longer interval since the onset of symptoms and with worse hemodynamic conditions (i.e., greater heart rate, more patients with Killip class >2), and greater Global Registry of Acute Coronary Events risk score, glucose, brain natriuretic peptide and high-sensitivity C-reactive protein levels. In contrast, female patients were less likely to have hypercholesterolemia or a history of myocardial infarction and had less severe coronary disease. Women less often received aspirin, clopidogrel, β blockers, angiotensin-converting enzyme inhibitors, and statins than men and less often underwent coronary angiography. In patients with STEMI, the men more often underwent reperfusion therapy and more often received primary angioplasty or thrombolysis than women. Similarly, in patients with NSTEMI, glycoprotein IIb/IIIa inhibitors were less frequently used in women ( Table 1 ). Finally, coronary revascularization was less frequently performed in women. Compared to women, the in-hospital and 30-day mortality was significantly lower in men (p <0.0001, log-rank test; Figure 2 ).



Table 1

Baseline characteristics of unmatched population

























































































































































































































































Variable Women Men p Value
Patients (n) 1,119 (32%) 2,391 (68%) <0.0001
ST-segment elevation myocardial infarction 461 (41%) 1,117 (47%) 0.0008
Non–ST-segment elevation myocardial infarction 658 (59%) 1,274 (53%) <0.0001
Age (years) 74 ± 13 64 ± 13 <0.0001
Body weight (kg) 66 ± 14 79 ± 14 <0.0001
Diabetes mellitus 301 (27%) 495 (21%) <0.0001
History of hypertension 762 (68%) 1,143 (48%) <0.0001
History of hypercholesterolemia 451 (40%) 1,132 (47%) <0.0001
Smoker 258 (23%) 1,583 (66%) <0.0001
Previous myocardial infarction 151 (13%) 449 (19%) <0.0001
Previous coronary angioplasty 94 (8%) 349 (15%) <0.0001
Previous coronary bypass 37 (3%) 105 (4%) 0.12
Previous stroke 79 (7%) 121 (5%) 0.002
Peripheral vessel disease 86 (8%) 286 (12%) <0.0001
Creatinine (mmol/L) 104 ± 69 76 ± 51 0.0012
Creatinine clearance (Cockcroft-Gault formula) 61 ± 41 76 ± 51 <0.0001
Admission glucose (mmol/L) 7 ± 4.4 6.7 ± 3.5 0.0035
Hemoglobin (g/dl) 12.7 ± 1.7 13.9 ± 1.9 <0.0001
Admission heart rate (beats/min) 80 ± 20 77 ± 20 <0.0001
Admission systolic blood pressure (mm Hg) 134 ± 29 130 ± 28 0.79
Killip class >2 94 (8%) 127 (5%) <0.0001
Cardiogenic shock 48 (4%) 94 (4%) 0.61
Global Registry of Acute Coronary Events risk score 147 (124–167) 132 (110–152) <0.0001
Left ventricular ejection fraction 56 ± 13 55 ± 13 0.56
Troponin I at 24 hours (μg/L) 4.4 (1–20) 6.4 (2–22) 0.45
Brain natriuretic peptide (pg/ml) 480 (180–1,202) 217 (83–544) <0.0001
High-sensitivity C-reactive protein (mg/L) 8 (3–27) 6 (2–22) <0.0001
Time to admission ST-segment elevation myocardial infarction (hours) 4 (2–15) 3 (2–9) <0.0001
Patients with angiography (n) 856 (76%) 2,007 (84%) <0.001
Number of coronary lesions
0 112 (13%) 93 (5%) <0.001
1 384 (45%) 909 (45%) 0.65
2 190 (22%) 559 (28%) 0.001
>2 170 (24%) 446 (28%) 0.04
Percutaneous coronary intervention of infarct-related artery 552 (49%) 1,662 (69%) <0.0001
Coronary artery bypass 47 (4%) 141 (6%) 0.04
Non–ST-segment elevation myocardial infarction: glycoprotein IIb/IIIa inhibitors 271 (41%) 682 (53%) <0.0001
ST-segment elevation myocardial infarction: any reperfusion 281 (61%) 843 (75%) <0.001
Primary percutaneous coronary intervention 215/461 (47%) 615/1,117 (55%) <0.0001
Thrombolysis 66/461 (14%) 228/1,117 (20%) <0.009
Aspirin 1,089 (97%) 2,360 (99%) <0.0001
Clopidogrel 1,075 (92%) 2,290 (96%) <0.0001
Aspirin and clopidogrel 1,035 (90%) 2,290 (96%) <0.0001
Angiotensin-converting enzyme inhibitors/angiotensin-receptor blockers 667 (60%) 2,052 (86%) <0.0001
β Blockers 791 (71%) 847 (77%) <0.0001
Statins 982 (88%) 2,296 (96%) <0.0001
In-hospital death 108 (9.7%) 118 (5%) <0.0001
30-Day death 130 (12%) 167 (7%) <0.0001

Data in parentheses are interquartile range, unless otherwise noted.



Figure 2


Kaplan-Meier cumulative curves for unadjusted 30-day survival probability in whole population (black lines) , in population matched for characteristics (blue lines) , and in population matched for characteristics and treatment (red lines). p Values from log-rank test.


The median value of the first propensity score (using baseline characteristics) was 0.142 (interquartile range 0.068 to 0.304) in women and 0.558 (interquartile range 0.313 to 0.740) in men, and the area under the curve was 0.83. The matched data set consisted of 649 pairs, matched using 8 digits of the propensity score for 1 pair, 6 digits for 2 pairs, 5 digits for 27 pairs, 4 digits for 136 pairs, 3 digits for 338 pairs, 2 digits for 135 pairs, and 1 digit for 10 pairs. Matching was effective because the baseline characteristics were comparable between the men and women, except for a lower creatinine clearance and a lower hemoglobin level in women ( Table 2 ). The Kaplan-Meier curves showed a nonsignificant trend toward greater 30-day mortality in women, as assessed using the log-rank test ( Figure 2 ). The paired comparison ( Figure 3 ) showed that, despite similar baseline characteristics, men more frequently underwent coronary angiography (+57%) and reperfusion in STEMI, either by thrombolytics (+72%) or primary PCI (+24%). The in-hospital death rate was 48% lower in men than in women, and the 30-day mortality difference was borderline significant (hazard ratio 0.70, 95% confidence interval 0.46 to 1.01).



Table 2

Comparison between men and women in population matched by baseline characteristics

























































































































































































































































Variable Women Men p Value
Patients (n) 649 (50%) 649 (50%)
ST-segment elevation myocardial infarction 264 (41%) 270 (42%) 0.71
Non–ST-segment elevation myocardial infarction 385 (59%) 379 (58%) 0.71
Age (years) 70 ± 14 70 ± 113 0.85
Body weight (kg) 72 ± 14 72 ± 13 0.77
Diabetes mellitus 160 (25%) 156 (24%) 0.79
History of hypertension 390 (60%) 383 (59%) 0.69
History of hypercholesterolemia 285 (44%) 287 (44%) 0.91
Smoker 240 (37%) 231 (36%) 0.6
Previous myocardial infarction 102 (16%) 97 (15%) 0.7
Previous coronary angioplasty 80 (12%) 75 (12%) 0.67
Previous coronary bypass 25 (4%) 21 (3%) 0.55
Previous stroke 37 (6%) 31 (5%) 0.45
Peripheral vessel disease 60 (9%) 65 (10%) 0.64
Creatinine (mmol/L) 107 ± 20 107 ± 14 0.90
Creatinine clearance (Cockcroft-Gault formula) 69 ± 37 69 ± 42 0.95
Admission glucose (mmol/L) 7.0 ± 5.5 7.0 ± 3.6 0.96
Hemoglobin (g/dl) 12.8 ± 1.8 13.7 ± 2.1 <0.0001
Admission heart rate (beats/min) 79 ± 18 78 ± 19 0.49
Admission systolic blood pressure (mm Hg) 135 ± 29 135 ± 26 0.91
Killip class >2 40 (6%) 41 (6%) 0.82
Cardiogenic shock 23 (3.5%) 26 (4%) 0.66
Global Registry of Acute Coronary Events risk score 138 (119–160) 138 (120–158) 0.85
Left ventricular ejection fraction 57 ± 13 56 ± 12 0.27
Troponin I at 24 hours (μg/L) 5 (1–26) 5 (2–37) 0.03
Brain natriuretic peptide (pg/ml) 376 (151–890) 227 (96–646) 0.001
High-sensitivity C-reactive protein (mg/L) 8 (3.5–22) 5 (2–17) 0.003
Interval to admission ST-segment elevation myocardial infarction (hours) 4 (2–12) 4 (2–10) 0.57
Patients with angiography 495 (85%) 534 (90%) 0.01
Number of coronary lesions
0 59 (12%) 21 (4%) <0.001
1 238 (48%) 214 (40%) 0.01
2 104 (21%) 160 (30%) 0.0015
>2 94 (19%) 139 (26%) 0.009
Percutaneous coronary intervention of infarct-related artery 337 (52%) 407 (69%) <0.001
Coronary artery bypass 22 (3.4%) 44 (7%) 0.008
Non–ST-segment elevation myocardial infarction: glycoprotein IIb/IIIa inhibitors 158 (48%) 178 (54%) 0.15
ST-segment elevation myocardial infarction: any reperfusion 178 (67%) 216 (80%) 0.001
Primary percutaneous coronary intervention 143 (54%) 160 (59%) 0.75
Thrombolysis 35/264 (13%) 56/270 (21%) 0.02
Aspirin 638 (98%) 639 (98%) 0.82
Clopidogrel 623 (96%) 624 (96%) 0.88
Aspirin and clopidogrel 614 (95%) 617 (95%) 0.7
Angiotensin-converting enzyme inhibitors/angiotensin-receptor blockers 529 (81%) 552 (85%) 0.08
β Blockers 486 (75%) 488 (75%) 0.89
Statins 606 (93%) 619 (95%) 0.13
In-hospital death 50 (7.7%) 27 (4.2%) 0.006
30-Day mortality 61 (9.4%) 44 (6.8%) 0.07

Data in parentheses are interquartile range, unless otherwise noted.



Figure 3


Forest plot for use of treatment and 30-day mortality. Odds ratio and 95% confidence intervals from conditional logistic regression for paired cohorts. *Unadjusted odds ratio; **odds ratio matched by baseline characteristics; ***odds ratio matched by characteristics and treatment.


The median value of the second propensity score (using baseline characteristics plus treatment) was 0.137 (interquartile range 0.066 to 0.301) in women and 0.554 (interquartile range 0.310 to 0.752) in men. The area under the curve was 0.84. The matched data set consisted of 584 pairs, matched on 6 digits for 2 pairs, 5 digits for 16 pairs, 4 digits for 144 pairs, 3 digits for 315 pairs, and 2 digits for 107 pairs.


After matching using the second propensity score, the only variables that differed significantly between genders were hemoglobin level and creatinine clearance, and the difference in the rate of treatment use was no longer significant ( Table 3 ). The p values for the univariate comparison between men and women are presented in Figure 4 , and the cumulative curves of the Global Registry of Acute Coronary Events risk score in the unmatched data set and in the 2 matched cohorts are presented in Figure 5 . The Kaplan-Meier cumulative curves for 30-day mortality in these 2 cohorts showed no significant difference between genders (p = 0.95, log-rank test; Figure 2 ). The paired comparison confirmed the absence of a significant difference according to gender, for either in-hospital or 30-day mortality ( Figure 3 ).



Table 3

Comparison between men and women in population matched by baseline characteristics and treatment

























































































































































































































































Variable Women Men p Value
Patients (n) 584 (50%) 584 (50%)
ST-segment elevation myocardial infarction 264 (45%) 271 (46%) 0.68
Non–ST-segment elevation myocardial infarction 320 (55%) 313 (54%) 0.52
Age (years) 69 ± 13 69 ± 13 1
Body weight (kg) 72 ± 14 72 ± 14 0.97
Diabetes mellitus 138 (24%) 145 (25%) 0.63
History of hypertension 357 (61%) 351 (60%) 0.81
History of hypercholesterolemia 257 (44%) 251 (43%) 0.70
Smoker 223 (38%) 211 (36%) 0.46
Previous myocardial infarction 92 (16%) 90 (15%) 0.87
Previous coronary angioplasty 63 (11%) 62 (11%) 0.92
Previous coronary bypass 22 (4%) 18 (3%) 0.50
Previous stroke 26 (4%) 27 (5%) 0.88
Peripheral vessel disease 53 (9%) 40 (7%) 0.16
Creatinine (mmol/L) 106 ± 83 106 ± 83 0.70
Creatinine clearance (Cockcroft-Gault formula) 69 ± 36 69 ± 46 0.70
Admission glucose (mmol/L) 7.1 ± 5 6.7 ± 3.8 0.12
Hemoglobin (g/dl) 12.9 ± 1.7 13.6 ± 2.1 <0.0001
Admission heart rate (beats/min) 79 ± 19 78 ± 20 0.48
Admission systolic blood pressure (mm Hg) 135 ± 29 134 ± 26 0.76
Killip class >2 43 (7%) 34 (6%) 0.28
Cardiogenic shock 20 (3%) 27 (5%) 0.37
Global Registry of Acute Coronary Events risk score 145 ± 38 147 ± 36 0.82
Left ventricular ejection fraction 57 ± 13 55 ± 14 0.73
Troponin I at 24 hours (μg/L) 4.1 (0.7; 22) 7.4 (0.9; 33) 0.02
Brain natriuretic peptide (pg/ml) 277 (150–886) 247 (92–525) 0.005
High-sensitivity C-reactive protein (mg/L) 6.8 (2.2–29) 4.9 (1.6–21) 0.13
Interval to admission ST-segment elevation myocardial infarction (hours) 4 (2–12) 4 (2–12) 0.76
Patients undergoing angiography 513 (84%) 510 (84%) 0.98
Number of coronary lesions
0 56 (11%) 26 (5%) 0.007
1 241 (47%) 229 (45%) 0.37
2 106 (21%) 127 (25%) 0.16
>2 101 (20%) 126 (25%) 0.08
Percutaneous coronary intervention of infarct-related artery 365 (62%) 373 (64%) 0.62
Coronary artery bypass 2 (4%) 44 (7%) 0.02
Non–ST-segment elevation myocardial infarction: glycoprotein IIb/IIIa inhibitors 161 (50%) 163 (52%) 0.87
ST-segment elevation myocardial infarction: any reperfusion 185 (70%) 204 (75%) 0.49
Primary percutaneous coronary intervention 143 (54%) 152 (56%) 0.37
Thrombolysis 42 (16%) 52 (19%) 0.10
Aspirin 576 (99%) 571 (98%) 0.27
Clopidogrel 558 (96%) 563 (96%) 0.45
Aspirin and clopidogrel 550 (94%) 551 (94%) 0.79
Angiotensin-converting enzyme inhibitors/angiotensin-receptor blockers 496 (85%) 499 (85%) 0.80
β Blockers 434 (74%) 450 (77%) 0.27
Statins 552 (95%) 552 (95%) 1
In-hospital death 41 (7%) 30 (5%) 0.18
30-day death 51 (8.7%) 46 (7.9%) 0.59

Only gold members can continue reading. Log In or Register to continue

Stay updated, free articles. Join our Telegram channel

Dec 16, 2016 | Posted by in CARDIOLOGY | Comments Off on Propensity Score-Matched Analysis of Effects of Clinical Characteristics and Treatment on Gender Difference in Outcomes After Acute Myocardial Infarction

Full access? Get Clinical Tree

Get Clinical Tree app for offline access