We evaluated whether gender exerts a similar influence on the likelihood of additional noninvasive or invasive testing after an initial abnormal cardiac stress study in a group of patients referred for clinically suspected coronary artery disease. Subsequent coronary revascularization procedures and cardiac event rates were compared during a 2-year follow-up.

Methods

Top Methods Follow-up Statistical Analyses Results Discussion Author & Article Info References

Patients

A consecutive series of 3975 patients 45 to 65 years of age referred to St. Louis University Health Science Center for exercise stress testing or intravenous dipyridamole thallium-201 myocardial imaging from 1 January 1988 to 31 December 1989 was screened; 872 patients met criteria for study entry with clinically suspected coronary artery disease. Patients were included if they had typical or atypical chest pain, had symptoms of congestive heart failure, or were asymptomatic with known cardiac risk factors. The duration of symptoms was defined as the length of time from symptom onset to referral for outpatient diagnostic testing.

Patients with confirmed or clinically suspected myocardial infarction, coronary artery revascularization, or previous cardiac evaluation were excluded from study entry (n = 2503). Patients with known myocardial or valvular heart disease or those referred for ventricular arrhythmia evaluation were also excluded (n = 369). Clinical history was confirmed in each patient at the time of the pre-stress test interview. Patients whose myocardial perfusion studies were degraded by substantial soft tissue attenuation were also excluded from this analysis (44 women; 7 men); no cardiac events were noted in this latter group. Our investigation was limited to middle-aged patients, decreasing potential covariation because of age, as previously reported [7-11]. We used a cutoff point of 65 years of age to be consistent with previous studies [12, 13]. Of the 872 patients, follow-up information was available for 840 (96.3%) during 24 ±7 months.

Exercise Testing

Symptom-limited treadmill exercise was done using the Bruce or modified Bruce protocol [14]. The occurrence and character (typical, atypical) of exercise-induced chest pain was noted. The exercise electrocardiogram was classified as abnormal if horizontal or downsloping ST-segment depression of 1 mm or more occurred during exercise or recovery. Thallium-201 results were given precedence over the exercise ST-segment response because of their enhanced diagnostic accuracy [15]. Only thallium-201 results were used in the presence of electrocardiographic left ventricular hypertrophy or left bundle-branch block. Leads V5 and V6 were used in patients with a right bundle-branch block.

Infusion Protocol, Thallium-201 Imaging, and Coronary Arteriography

We used the standard dipyridamole infusion protocol [16-18]. Dipyridamole was infused at a rate of 0.56 mg/kg of body weight over 4 minutes using an intravenous catheter [16-18].

Planar myocardial imaging was initiated within 5 minutes after injection. Sequential and 3- to 4-hour delayed post-stress images were acquired in the left anterior oblique (anterior, 45 degree) and left lateral views at the 80-kilovolt photopeak (20% window) stored on a 256 x 256 byte computer matrix (approximately 500 000 counts) [16-18]. Thallium-201 myocardial perfusion studies were interpreted by two expert observers without previous knowledge of the patients' clinical history. Studies were not reread for this analysis. The final test interpretation sent to each referring physician was used for data analyses. Tests were scored as normal or as having a fixed or reversible thallium-201 defect (with 1 or more reversible thallium-201 defects considered abnormal).

Selective coronary arteriography among patients with an initially abnormal noninvasive test was done using the brachial or femoral approach. A coronary stenosis was prospectively defined as a luminal diameter vessel narrowing of 70% or more in one of the major coronary arteries or side branches. All angiographic data were analyzed by independent observers blinded to the clinical history and noninvasive test results.

Follow-up

Top Methods Follow-up Statistical Analyses Results Discussion Author & Article Info References

Clinical follow-up information was obtained through either a telephone interview or medical clinic visit. The follow-up protocol was approved by the St. Louis University Institutional Review Board; written informed consent was obtained at patient contact. The following clinical events and procedures were ascertained: coronary arteriography, coronary artery bypass surgery, percutaneous transluminal coronary angioplasty, death, or nonfatal myocardial infarction. Cardiac death was defined as fatal myocardial infarction (within 24 hours of the peak creatine kinase level), sudden cardiac death, or autopsy-confirmed death from cardiovascular complications. Acute myocardial infarction was defined as two of the following: the appearance of new Q waves, an increase in the peak creatine kinase level of two times the normal level, or 5% or more of the creatine kinase MB isoenzyme. The cardiac event rate was predefined as the occurrence of cardiac death or nonfatal myocardial infarction. All events were confirmed by medical chart review or telephone contact with the referring physician.

A modification of the Rose questionnaire was used to assess patient symptom frequency and cardiac event rates [19]. This follow-up questionnaire is being used in the Bypass Angioplasty Revascularization Investigation trial [20]. Questionnaire reliability was tested using a randomly selected subsample [21]. A 10% random sample (enriched 25% with abnormal events) of interviews were repeated to assess inter-rater reliability. Kappa coefficients were more than 0.80 for all items.

Statistical Analyses

Top Methods Follow-up Statistical Analyses Results Discussion Author & Article Info References

Frequencies were compared using chi-square analyses. Continuous variables were compared with categorical variables using a one-way analysis of variance. A P value of less than 0.05 was considered statistically significant. Actuarial event-free and follow-up testing rates were analyzed using Kaplan-Meier survival curves [22]. Differences in follow-up testing frequencies and event rates were calculated using the log-rank statistic for the duration of the follow-up period.

Regression Modeling

A proportional hazards regression model was used for prediction of two dependent variables: subsequent follow-up cardiac diagnostic testing and cardiac death or nonfatal myocardial infarction (time to subsequent follow-up diagnosis or cardiac event) [23-28]. Relative risk ratios and 95% CIs were calculated [23]. Follow-up testing after an intercurrent cardiac event was not coded as the outcome measure, a follow-up cardiac diagnostic test. Demographic and clinical variables were entered as categorical (dummy) variables [29]. Variables with P < 0.20 were included in the multivariate models. Model overfitting procedures were examined, as were plots for the effect of influential cases and the ratio of outcome events to number of independent variables [30-32]. The proportional hazards assumption was satisfied for each regression model. The following variables were included: 1) affecting access to health care (race, insurance status, age of the referring physician, level of physician training [resident or fellow or attending physician] and type of physician [cardiologist or noncardiologist]); 2) clinical need for further medical care [history of typical or atypical chest pain, symptoms of congestive heart failure, length of symptoms, cardiac risk factors, medication use, and the resting electrocardiogram interpretation]; 3) indicators of illness severity (the presence of a positive test, type and number [1 and >1] of thallium-201 perfusion defects); and 4) the primary variable, gender. No statistically significant interaction effects were found.

Results

Top Methods Follow-up Statistical Analyses Results Discussion Author & Article Info References

Clinical Characteristics

The clinical and electrocardiographic characteristics by gender are presented in Table 1. Women were on average 1.3 years older, had a greater frequency of atypical chest pain, and had a greater symptom duration before testing. Hypertension, hypercholesterolemia, and the mean number of cardiac risk factors were greater in women. Nonspecific ST-T changes on the resting electrocardiogram were more common in women.

Electrocardiography and Myocardial Perfusion Imaging

The percentage of patients with an initially abnormal exercise electrocardiogram was similar in women (16.5%) and men (19.4%). A similar number of initial thallium-201 radionuclide perfusion studies were interpreted as normal in women and men (76.8% compared with 73.3%; P > 0.2, Table 2. The presence of an abnormal 4-hour post-stress reversible perfusion defect was similar in women and men [17.0% compared with 21.3%; P = 0.19].

When the number of thallium-201 perfusion defects was examined, more than 1 myocardial perfusion defect was noted in 75.0% (42 of 56) of men compared with only 47.0% (21 of 44) of women (P = 0.001). Increased lung uptake of thallium-201, noted in a small number of patients (four men and two women), was always associated with a reversible myocardial perfusion defect and was not analyzed further because of the small number of patients.

Rates of Follow-up Diagnostic Studies and Medical Therapy

Additional diagnostic testing was done in 62.3% of men but in only 38.0% of women, even though initial test positivity results were similar Table 3, Figure 1; P 0.002). Subsequent initiation of medical therapy in those patients with an abnormal test result who did not subsequently have coronary revascularization was greater in men than in women with the exception of nitroglycerin use. Individual gender differences were noted for the initiation of aspirin (men, 31.3%; women, 20%; P = 0.08), calcium channel blockers (men, 31.3%; women, 13.4%; P = 0.005), and nitroglycerin (men, 6.3%; women, 13.4%; P = 0.11). Selective coronary arteriography, after an initially abnormal noninvasive test, was done in 33.8% of women and 45.0% of men [P = 0.007]. The difference in the rate of follow-up angiography between women and men was detected 3 to 6 months after testing (Figure 1); P < 0.001).

View larger version (15K): [in this window] [in a new window]   Figure 1. A gender-based comparison of noninvasive test use rates over time. A greater percentage of men received follow-up noninvasive testing (NIVT) compared with women (P < 0.001). Similarly, a comparison of use of coronary arteriography during the 2-year follow-up period showed that a greater percentage of men than women subsequently had catheterization (Cath) (P < 0.001).

Thirty-five percent of women studied by coronary angiography had no luminal narrowing of 70% or more compared with 18.6% of men (P > 0.2; (Table 3). Minor arteriographic lesions [ 20% but <70%] were reported in five women and three men. Of the 5 women with minimal disease, 3 had a cardiac event (2 had myocardial infarction and 1 had cardiac death) during follow-up compared with none of the men. A trend toward a higher percentage of multivessel disease (for example, 2- or 3-vessel disease) was present in men compared with women (39.0% compared with 23.0%; P = 0.07).

Cardiac Events and Coronary Revascularization Rates

After 24 ±7 months of follow-up in the study group (n = 840), the overall rate of myocardial infarction was greater in women (5.0% compared with 2.0%; P = 0.02). Cardiac death was slightly greater in women compared with men [2.0% compared with 1.0%; P = 0.05]. The combined end point of cardiac death or nonfatal myocardial infarction was 6.9% in women and 2.5% in men (Table 4); P = 0.002).

View larger version (14K): [in this window] [in a new window]   Figure 2. Cardiac events by gender. No cardiac events were recorded in women or in men who subsequently had catheterization (were revascularized [REV]). Patients who were revascularized had greater cardiac event-free survival (P < 0.001).

A gender-based comparison of test use rates over time Figure 1 and of cardiac event rates Figure 2 showed an inverse relation between follow-up medical intervention and cardiac hospitalizations for cardiac events. A higher cardiac event rate was noted in those patients without follow-up testing. Coronary revascularization procedures, either percutaneous transluminal coronary angioplasty or coronary bypass surgery, were more commonly done in men (5.0% compared with 2.0%; P = 0.03). No cardiac events were observed in men or women who were revascularized during a mean of 2 years after surgery (Figure 2). For all patients, greater follow-up diagnostic testing was associated with fewer cardiac events (P < 0.001).

Multivariable Cox Regression Analysis

Prediction of Follow-up Diagnostic Study Use Rates

Univariate followed by multivariate Cox regression was done to elucidate the effect of gender on follow-up test use and on cardiac event rates (in relation to other predictive clinical and test variables). Statistically significant univariate variables entered into the analysis included body size and weight, age, race, symptom status, risk factor profiles, selected medications, resting electrocardiographic abnormalities, and stress testing results.

After adjusting for several clinical and test variables (Appendix Table), the multivariate Cox regression results showed that a positive test result was an independent predictor of time to follow-up studies (relative risk [RR] = 9.9; P < 0.001). Male gender was also independently related to the rate of follow-up intervention (RR =1.9; P = 0.005). Additional predictors of follow-up testing are noted in Table 5 with their corresponding relative risks and 95% CIs. Adjustment for the size of the thallium-201 defect or the degree of ST-segment depression did not alter the effect of male gender within this multivariate model. When coronary arteriography was analyzed as an end point, male gender was also predictive within a multivariate Cox regression. Finally, when subgroup analysis was done by type of initial test (thallium-201 imaging or stress electrocardiography alone), the effect of gender on cardiac follow-up testing remained independent of clinical and testing variables.

When regression analysis was done for women and men separately, except for the presence of a positive initial test (P < 0.001) or for one or more reversible defects (P < 0.001), few other common predictors of follow-up testing were noted. For women, hypertension (P = 0.01), typical angina (P = 0.003), congestive heart failure (P = 0.001), diabetes (P = 0.009), and left bundle-branch block (P = 0.02) were predictive of subsequent follow-up testing. In contrast, for men, referral for initial noninvasive testing by a cardiologist (P < 0.001), a duration of symptoms of 6 months or more (P = 0.005), symptoms of congestive heart failure (P = 0.02), diabetes (P < 0.001), and the use of calcium channel blockers (P = 0.01) were predictive of subsequent follow-up evaluation.

Prediction of Cardiac Death or Myocardial Infarction

When cardiac death or nonfatal myocardial infarction was examined as an end point, the presence of thallium-201 defect redistribution was an independent predictor in men and women (RR = 14.5; P < 0.001). Female gender was also predictive within the multivariate Cox regression analysis (RR = 4.9; P < 0.001). Additional predictors of cardiac death or myocardial infarction are listed in Table 5, with their corresponding relative risks and CIs.

When the analysis was done on women and men separately, the presence of an abnormal electrocardiogram and a thallium-201 defect redistribution (P < 0.001) were shared predictors of a cardiac event in women and men. For women, age (55 years of age or more [P = 0.001]), insulin use (P < 0.001), a left bundle-branch block (P = 0.01), and nitrate therapy (P = 0.01) were independent predictors of a subsequent cardiac event. In men, diabetes (P < 0.001), electrocardiographic left ventricular hypertrophy (P < 0.001), and a fixed thallium-201 defect (P = 0.05) were predictive of an event during follow-up.

Discussion

Top Methods Follow-up Statistical Analyses Results Discussion Author & Article Info References

Statistically significant differences have been reported [1-6] in the use of diagnostic procedures and surgical interventions between women and men with known coronary artery disease. The adjusted odds of having coronary arteriography were 15% to 28% lower for women than for men [1, 2]. Women were 27% to 45% less likely to have coronary revascularization procedures compared with men [1]. Women had lower rates (statistically significant) of coronary revascularization procedures compared with men after adjusting for coronary anatomy or left ventricular ejection fraction [5].

Previous reports have focused exclusively on patients with known coronary artery disease with one exception [6]. In our report, we selected a cohort of patients referred for noninvasive testing who had a lower pretest likelihood of coronary artery disease and evaluated the likelihood of additional cardiac testing based on gender.

Our study group with a high percentage of women (47%) may be contrasted with other data using a smaller proportion of female study participants. A recent review [33] on cardiovascular multicenter trials noted that women were under-represented in most study groups, including previous gender-based comparisons [1-6]. In our study, women were more symptomatic (82%) compared with men (65%), but the character of chest pain was more often atypical and women had a longer duration of symptoms before noninvasive testing than did men. Chest pain symptomatology differs qualitatively between women and men presenting to emergency departments [34] and may, in part, explain differential use rates. In a group of patients referred for coronary bypass, women had more severe clinical manifestations of coronary disease than men [2, 4, 6]. Anxiolytic medications were also more frequently prescribed to our study cohort of women than men before their initial noninvasive testing.

Hypertension and hypercholesterolemia were more common in women than in men, supporting earlier studies [7-11] indicating that women have a greater frequency of cardiac risk factors than men in the post-myocardial infarction setting or when referred for coronary arteriography or coronary bypass surgery. In our study, mortality rates for women exceeded those for men even after statistical adjustment for cardiac risk factors.

Despite similarities in noninvasive testing results, 62% of women with a positive initial test had no further diagnostic testing compared with 38% of men. Statistical adjustment for the number of thallium-201 defects did not change the independent predictive value of male gender on follow-up coronary revascularization or diagnostic arteriography. Nearly 66% of women and greater than 80% of men in this study cohort had clinically significant angiographic coronary artery disease. Although the prevalence of multivessel disease in our sample of women was less than that observed for men, lack of follow-up testing may have prevented the detection of multivessel disease and may have inhibited intervention to decrease cardiovascular risk. Decreasing the use of further diagnostic studies may have placed these women at greater risk for a subsequent adverse outcome, which is suggested by the higher cardiac event rates in these women. Our cardiac death rates were 1.1% for women and 0.3% for men (Table 4). Annual cardiovascular death rates in the general population in this age group are 1.3% for men and 0.7% for women [35].

In a previous series of 390 patients referred for cardiovascular nuclear imaging (including thallium-201 and exercise-gated blood pool studies), only 4% of women compared with 40% of men with abnormal radionuclide scans were referred for diagnostic catheterization [6], including patients with a previous myocardial infarction (women, 16%; men, 35%). This result adds independently to a greater odds differential in the male-female follow-up intervention rate (10 times compared with 1.9 times). Also, nearly two thirds of men and one third of women in the former series [6] had an abnormal exercise nuclear scan, a statistically greater percentage than that observed in our group of patients.

Differences in referral to coronary bypass surgery may be partially related to differences in age and functional class [4]. Men with minimal or no symptoms were more frequently referred for cardiac catheterization and subsequent surgery than were women, whereas women were less likely to be referred for coronary artery bypass surgery after an abnormal exercise test [4]. Our study of 840 patients showed a relative likelihood for further diagnostic testing in men that was approximately two times greater than that observed in women. Women may represent a higher-risk group in the later stages of their illness, partly because of observed and previously reported [6] delays in medical diagnostic referral. Approaches designed to attenuate this adverse outcome for women should include early identification of clinical risk factors. Effective use of appropriate noninvasive testing (adjusted for cardiac risk factors, symptom status, and comorbid conditions) would aid in early detection of coronary artery disease and could theoretically improve identification of women at increased risk.

Data collection from a single institution is influenced by patient referral patterns, limiting the general applicability of study results to various medical institutions. Although we attempted to elucidate physician decision making, direct information was not included or considered in this report. St. Louis University is a private tertiary referral institution with a high percentage of medically insured patients (95% of patients in this study). Different outcomes could be anticipated in other health care or socioeconomic settings. Our data were collected from a large heterogeneous group, and, as such, the results should be applicable to an unselected group of patients referred for the evaluation of suspected coronary artery disease.

In light of recent reports [1-6] showing disparities in health care between women and men with known coronary artery disease, our goal was to examine post hoc medical resources use and cardiac event rates in relatively low-risk patients referred for the noninvasive stress evaluation of suspected coronary artery disease. Our report indicates that men and women share few common clinical predictors for increased rates of subsequent diagnostic testing and adverse cardiac events. However, for women and men, the single greatest outcome predictor was the presence of an abnormal noninvasive stress test or an abnormal myocardial imaging study. The performance and results of noninvasive testing cannot be related to subsequent medical care and outcome status in this retrospective study.

The observed gender variability in clinical predictors for follow-up and outcome should prompt further examination of clinical risk profiles between women and men with suspected coronary artery disease. If confirmed, these differences would argue in favor of the use of sex-specific guidelines to evaluate the appropriate medical care and the prognostic impact of diagnostic testing and medical-surgical treatment strategies. If prospectively validated, this approach could equalize medical care use rates in women and men across various risk strata.