Author + information
- Received November 21, 1997
- Revision received February 16, 1999
- Accepted March 19, 1999
- Published online July 1, 1999.
- Pierre Y Marie, MD∗,* (, )
- Nicolas Danchin, MD†,
- Fabrice Branly, MD∗,
- Michaël Angioı̈, MD†,
- Alain Grentzinger, MD†,
- Jean M Virion, MSc‡,
- Benoit Brouant, MD†,
- Pierre Olivier, MD∗,
- Gilles Karcher, MD∗,
- Yves Juillière, MD†,
- Faı̈ez Zannad, MD† and
- Alain Bertrand, MD∗
- ↵*Reprint requests and correspondence: Pr. Pierre-Yves Marie, Service de Médecine Nucléaire, CHU Nancy-Brabois, 54511 Vandoeuvre Cedex, France
The purpose of this study was to determine whether antianginal medications modify the prognostic significance of exercise single photon emission computed tomography (SPECT) ischemia.
Antianginal medications (especially beta-adrenergic blocking agents) limit exercise SPECT ischemia, but it is not known whether such medications also modify the prognostic effect of exercise SPECT ischemia.
We included 352 patients with coronary heart disease, who had exercise Tl-201 SPECT and coronary angiography, and who were initially treated medically. Survival Cox models were applied in patients for whom classes of antianginal medications taken at exercise SPECT were the same as those prescribed for follow-up (GI; n = 136), and in patients for whom new classes of antianginal medications, including beta-blockers (GII; n = 79) or not including beta-blockers (GIII; n = 113), were added for follow-up.
During a mean 5.3 ± 1.6 years of follow-up, 45 patients had cardiac death or myocardial infarction. Variables reflecting necrosis (irreversible defect extent, left ventricular ejection fraction) and those from coronary angiography provided equivalent prognostic information in the three groups. In contrast, the SPECT variable reflecting ischemia (reversible defect extent), which provided comparable prognostic information in GI (p = 0.005) and GIII (p = 0.004), lost its prognostic significance (p = 0.54) in GII, and was associated with a lower relative risk in GII than in GI or GIII (both p < 0.05).
In patients with coronary heart disease, the introduction of antianginal medications, when including beta-blockers, appears to have a favorable effect on the deleterious prognostic effect of exercise ischemia.
In patients with coronary heart disease (CHD), exercise single photon emission computed tomography (SPECT) provides strong prognostic information (1–9)and allows the detection of ischemic myocardium, which constitutes a potential therapeutic target (9–12).
Though a number of antianginal medications can decrease the amount of exercise ischemia evidenced on exercise scintigraphy (13–18), this effect appears particularly marked for beta-adrenergic blocking agents (13–15), which may normalize the test (15). In patients with a recent infarction, it has been observed that medical therapy including beta-blockers could lead to a reduction in the exercise (Tl-201 SPECT) ischemia which was equivalent to that provided by coronary angioplasty (19). In terms of primary or secondary prevention, beta-blockers have also proved beneficial in a number of trials (20–23), but it is not known whether this beneficial effect is mediated by their possible influence on stress ischemia.
The aim of this study was to determine whether the deleterious effect of exercise (Tl-201 SPECT) ischemia on the prognosis of patients with CHD could be modified by the subsequent introduction of either beta-blocking or non–beta-blocking antianginal medications.
We retrospectively included all patients with CHD admitted to our institution from 1982 to 1992, who underwent, over a six-week period, exercise Tl-201 SPECT and coronary angiography, and who had no myocardial revascularization during the following three months.
Diagnosis of CHD was based on the presence of either significant (≥50%) coronary stenoses or, in patients without significant coronary artery narrowings, of a documented myocardial infarction. The exclusion criteria were as follows: previous history of cardiac surgery or coronary angioplasty, congenital or valvular heart disease and hypertrophic or idiopathic dilated cardiomyopathy. Among the 362 patients who met these criteria, 10 were lost to follow-up, and the 352 remaining (97%) constituted our study population.
Three groups were constituted. The first group, GI, comprised the patients for whom the classes of antianginal medications (1, beta-blockers; 2, calcium antagonists; 3, nitrates or molsidomine) used at the time the exercise SPECT test was performed were the same as those ordered at hospital discharge, after the test. The second group, GII, comprised patients in whom beta-blocking medications were added between exercise test and hospital discharge. The third group, GIII, comprised patients in whom classes of only non–beta-blocking antianginal medications were added between the test and hospital discharge.
As previously described (4,24), coronary angiograms were analyzed visually on end-diastolic frames by experienced observers. Single-plane X-ray left ventricular angiography in the 30° right anterior oblique projection was performed in 342 patients (97%) and used to measure ejection fraction (area–length method ).
Exercise Tl-201 SPECT
The exercise test was performed in the upright position on a bicycle ergometer. The protocol was started at 40 W and increased by 30 W every 3 min. Leads V1, V5and a VFwere continuously monitored, 12-lead electrocardiogram and arterial blood pressure being recorded at each step.
Exercise end points were physical exhaustion, development of angina pectoris, >2 mm ST-segment depression, sustained ventricular tachyarrhythmia, ≥10 mm Hg fall in systolic pressure or achieving maximal predicted heart rate (220 − age). ST-segment depression was considered to be significant when a ≥1-mm horizontal or down-sloping depression occurred 0.08 s after the J point, when compared with baseline.
At peak of exercise, 37 MBq of Tl-201 per 25 kg body weight (without exceeding 111 MBq) was injected intravenously; Exercise was then prolonged for 1 min at the same workload. Stress imaging was initiated 10 to 15 min later, and redistribution imaging 3 to 5 h later. For the 177 patients (50%) who underwent the test since 1988, a Tl-201 activity, corresponding to one third of that given at exercise, was reinjected 10 min before redistribution imaging (26).
The technique to acquire, reconstruct and analyze the SPECT images has been detailed elsewhere (4,27). Briefly, the reconstructed slices were analyzed visually by a blinded experienced observer using both a 20-segment division of the left ventricle and a four-point grading system: 0, normal uptake; 1, equivocal; 2, moderate, and 3, severe reduction of uptake. Total extent of exercise defects was determined by the percentage of segments showing an uptake score ≥2 on exercise Tl-201 tomograms, and extent of reversible defects was determined by the percentage of segments with exercise defects which had a ≥1-point decrease of the uptake score on the rest acquisition.
Follow-up data were obtained by mailed questionnaires or telephone interviews, from the patients, their families or attending physicians. Cardiac death and myocardial infarction were defined as major ischemic events.
Cardiac death was defined as a death of demonstrated cardiac origin or unknown origin, and diagnosis of myocardial infarction was based on the presence of at least two of the following: prolonged (≥20 min) chest pain, electrocardiographic abnormalities (new Q-wave or ST-segment elevation) or enzyme changes (more than twice the upper limit of normal).
Continuous variables were expressed as mean ± SD and compared with Mann-Whitney tests. Discrete variables were expressed as percentages and compared with chi-square or Fisher exact tests when appropriate.
The relations between baseline data (listed in Table 1) and the subsequent occurrence of events were analyzed by means of Cox proportional hazards regression models (28)(BMDP Statistical Software Inc., Los Angeles, California ). The events taken into account were the major cardiac events (cardiac death and myocardial infarction) but also, in an additional analysis, the overall cardiac events including both major events and cardiac interventions (cardiac surgery and coronary angioplasty). Patients who had subsequent cardiac surgery or coronary angioplasty were always censored at the date of their intervention. The chi-square value was calculated from the log of the ratio of maximal partial likelihood functions, and variables for which p values of the univariate chi-square test were less than 0.05 were considered to be significant predictors of prognosis.
A multivariate ascending Cox regression analysis was performed stepwise to assess the additional prognostic information provided by Tl-201 SPECT compared with that already provided by the other investigations. The assumed limit for significance to enter a variable was 0.05 and the limit to remove a variable was 0.10. The Cox model was initially only applied to the significant univariate predictors from clinical history, exercise testing and catheterization. The selected independent predictors were then forced to remain in the model, whereas SPECT variables were tried and entered at a p value of 0.05.
The Cox model analyses were applied in the overall population, and in a comparative way, in each of the three groups (GI, GII and GIII). Finally, to provide a more precise evaluation of the effect of the further prescription of each of the three medication classes after the test, univariate Cox analysis was applied comparatively between patients who were without a given class both at exercise test and at follow-up and those who had only this class added for follow-up. The relative risks (±SD) provided by the Cox analyses of the baseline variables in the different groups were compared by means of a one-tail test on a table of critical values of the normal distribution, the null hypothesis being that the relative risks might be lowered in patients for whom antianginal medications were added after the test.
Exercise SPECT was performed on antianginal treatment in 198 patients (56%), and an antianginal treatment was prescribed at hospital discharge after the test in 331 (94%) (Table 2). When comparing exercise test data between patients with (n = 81) and those without (n = 271) beta-blockers at the time of exercise, there was a significant difference only for maximal heart rate (125 ± 22 vs. 144 ± 21 bpm, p < 0.0001) and maximal systolic blood pressure (178 ± 29 vs. 185 ± 28 mm Hg, p = 0.04).
GI comprised 136 patients (39%). In most (89%) of the remaining 216 patients, additional classes of antianginal medications were introduced between the test and hospital discharge. The added medications included beta-blockers in 79 patients (GII), and did not include beta-blockers in 113 patients (GIII).
The antianginal treatment was decreased between exercise SPECT and follow-up in only 24 patients. Most of them (79%) had no evidence of myocardial ischemia at Tl-201 SPECT; the reasons for antianginal medications withdrawal could not be ascertained retrospectively in all patients.
The patients’ baseline characteristics are detailed in Tables 1 and 2. Regarding antianginal treatments, there were obvious differences among the three groups in accordance with the way they were defined (Table 2). For example, GII and GIII patients, in whom antianginal medications had been added after exercise SPECT, also had, compared with GI patients, a higher number of antianginal medications at follow-up and a lower number of antianginal medications at exercise SPECT (Table 2).
In addition, because none of them had beta-blockers at exercise SPECT, GII patients reached higher values of maximal heart rates at exercise than those from GI or GIII (Table 1). All other baseline characteristics (Table 1)were similar in the three groups, except for the frequency of proximal left anterior descending coronary artery (LAD) stenosis, which was lower in GIII than in GI or GII.
During a follow-up of 5.3 ± 1.6 years, 30 patients (9%) had cardiac death, 45 (13%) had major ischemic events (cardiac death or infarction), 57 (16%) had cardiac interventions (coronary angioplasty: 24, bypass grafting: 35 [after a previous angioplasty in five cases], heart transplantation: 4 [after a previous angioplasty in one case]) and 100 (28%) had a cardiac event (major ischemic event and/or cardiac intervention).
There were no significant differences between the three groups in the follow-up durations (GI: 5.2 ± 1.7 years, GII: 5.3 ± 1.3 years, GIII: 5.4 ± 1.8 years) and in the rates of nonfatal myocardial infarction (GI: 5%, GII: 6%, GIII: 3%). Although the differences were not statistically significant, there were trends toward lower rates in GII than in GI or GIII, when considering cardiac death (GI: 10%, GII: 5%, GIII: 12%), major ischemic events (GI: 15%, GII: 11%, GIII: 14%), cardiac interventions (GI: 20%, GII: 11%, GIII: 19%) and cardiac events (GI: 33%, GII: 23%, GIII: 33%).
Univariate prediction of major ischemic events
The risk of major events was not significantly related to the prescription of beta-blockers, calcium antagonists or nitrates, even when the analyses were restricted to either GI, GII or GIII.
Significant variables by univariate analysis are listed in Table 3. Among the clinical and exercise testing data, significant predictors were as follows: increase in heart rate (p = 0.01), age (p = 0.025) and a positive exercise test (p = 0.045). The relative risks associated with these three parameters were not significantly different when determined in either GI, GII or GIII.
Among catheterization data, the main predictors of major ischemic events were: left ventricular ejection fraction (p < 0.0001), number of ≥70% diseased vessels (p = 0.0001) and presence of a stenosis on the proximal LAD (p = 0.0049) (Table 3). As illustrated in Figure 1, the relative risks provided by the Cox models for each of these parameters were similar in the three groups.
Single photon emission computed tomography predictors were: total extent of exercise defects (p < 0.0001), extent of reversible defects (p = 0.0028) and extent of irreversible defects (p = 0.0026). The relative risk related to extent of irreversible defect was similar when determined in either GI, GII, or GIII (Table 3, Fig. 1). In contrast, the relative risk related to the extent of reversible defects was significantly lower when determined in GII than in either GI (p = 0.045) or GIII (p = 0.015) (Table 3, Fig. 1). Although the extent of reversible defect was a potent predictor in both GI (p = 0.0045) and GIII (p = 0.0037), no such relationship was observed in GII (p = 0.54), where the only significant SPECT predictor was the extent of irreversible defects (Table 3). The extent of reversible SPECT defect remained a predictor (p = 0.045) when the analysis in GI was restricted to patients who had beta-blockers (n = 42; four major events). A similar subgroup analysis could not be done in GIII; 19 were on beta-blockers and only one had a major event.
These results are illustrated in Figure 2by the major-events–free survival curves, stratified as a function of presence or absence of a reversible exercise defect at baseline, significant differences being reached in both GI and GIII, but not in GII.
Influence of the type of events analyzed or of the duration of follow-up
When all cardiac events, including cardiac interventions, were taken into account, extent of reversible Tl-201 SPECT defect remained a potent predictor in GI (p = 0.017; relative risk [RR] [95% confidence interval (CI)]: 1.03 [1.01 to 1.06]) and in GIII (p = 0.034; RR [95% CI]: 1.03 [1.00 to 1.06], but was still totally unrelated to the occurrence of events in GII (p = 0.34; RR [95% CI]: 0.98 [0.93 to 1.03]). In all groups taken together, extent of irreversible defect lost its prognostic significance for all cardiac events.
When the prediction of major events (cardiac death and myocardial infarction) was restricted to the initial two-year follow-up period (22 major events), the extent of reversible defect was still a significant predictor in GI (p = 0.0093; RR [95% CI]: 1.06 [1.02 to 1.11]) and not in GII (p = 0.38; RR [95% CI]: 1.04 [0.96 to 1.13]), whereas the result obtained in GIII was of borderline significance (p = 0.062; RR [95% CI]: 1.04 [1.00 to 1.10]). In all groups taken together, extent of irreversible defect was not a significant predictor of major events over the first two years of follow-up.
Multivariate prediction of major ischemic events
Multivariate analysis was first applied to the significant clinical, exercise test and catheterization data listed in Table 3. In the overall population, the main independent predictors of major events were: 1) left ventricular ejection fraction (p = 0.0024; RR [95% CI]: 0.97 [0.95 to 0.99]); 2) number of ≥70% diseased vessels (p = 0.012; RR [95% CI]: 1.73 [1.13 to 2.64]); 3) proximal LAD stenosis (p = 0.021; RR [95% CI]: 2.23 [1.15 to 4.32]), and 4) age: (p = 0.023; RR [95% CI]: 1.04 [1.00 to 1.08]).
Compared with the prediction provided by this model, the extent of reversible defects was the sole SPECT variable providing significant additional information in the analysis performed in the overall population (p = 0.017) and also in both subgroup analyses performed in GI (p = 0.020) and GIII (p = 0.026). However, no SPECT variable provided significant additional information in the analysis performed in GII (p = 0.70 for extent of reversible defects) (Fig. 3).
Analysis of the specific impact of each antianginal class
To provide a more direct analysis of the impact related to the prescription of beta-blockers, only patients without beta-blockers on the day of the test were kept in the analysis (n = 271; 39 major events), and patients for whom beta-blockers were added after the test (n = 79) were compared with the others (n = 192). The extent of reversible defect was still the sole baseline variable for which the relative risk showed a significant difference between the groups (RR [± 95% CI] per percent of left ventricle, addition of beta-blockers: 0.98 [0.92 to 1.05] vs. no addition of beta-blockers: 1.05 [1.02 to 1.07], p < 0.05).
In contrast, when considering the patients who were without calcium antagonists on the day of the test (n = 246; 30 major events), there was no difference in the risk related to extent of reversible defect between that determined in patients for whom calcium antagonists were added after the test (n = 89; RR [95% CI]: 1.04 [1.00 to 1.07]) and that determined in the others (n = 157; RR [95% CI]: 1.03 [1.00 to 1.07]).
In a similar way, in patients who were not on nitrates or molsidomine on the day of the test (n = 228; 22 major events), there was no difference in the relative risk related to the extent of reversible defect between that determined in patients for whom nitrates or molsidomine were added after the test (n = 104; RR [95% CI]: 1.02 [0.99 to 1.06]) and that determined in the others (n = 124; RR [95% CI]: 1.02 [0.96 to 1.09]).
In patients with CHD, myocardial ischemia is an important prognostic factor which can be corrected by appropriate therapeutic interventions. Exercise SPECT gives the opportunity to assess the ischemia induced by a physiologic stress, and has been demonstrated to provide strong prognostic information (1–9)complementary to that of cardiac catheterization (2,4). Exercise ischemia as assessed by scintigraphy (reversible defects) has been shown to correlate with a poorer prognosis and to be a more accurate predictor than ischemic signs on exercise electrocardiography (1,4,5).
Exercise SPECT under daily life antianginal medications
Because the exercise ischemia, evidenced by SPECT, may be decreased by antianginal medications (13–18), the withdrawal of such medications is generally recommended for patients having exercise SPECT to detect CHD. However, when the exercise SPECT is performed on antianginal treatment, it has been shown that a normal result conferred an excellent long-term outcome (30). Conversely, for the purpose of prognostic determination, it is likely that performing the exercise SPECT on daily life antianginal treatment will provide more potent information, as it will show the perfusion abnormalities likely to occur during daily life.
Exercise SPECT was performed on the subsequently ordered daily life treatment in 39% of our patients (GI). In this subgroup, the SPECT variables, including that reflecting the amount of ischemia (reversible defect extent), were strongly correlated with prognosis, and exercise SPECT provided clearly original and complementary prognostic information compared with cardiac catheterization. This finding gives evidence of the high prognostic value of exercise SPECT when performed in such daily life conditions.
Impact of the further addition of beta-blocking antianginal medications
A definite trend toward lower rates of cardiac death and of cardiac interventions was observed in patients who had a further addition of beta-blockers (GII). Possibly because of the relatively small sizes of the groups, however, the differences with the other groups were not significant. Nevertheless, because of the retrospective and nonrandomized design of our study, it was difficult to directly compare the event rates between the groups, and only the relative risk of events were used for those comparisons. When so determined, only the risk related to stress ischemia (reversible defect extent), showed significant differences between the groups. It was, indeed, significantly lower in patients for whom antianginal medications including beta-blockers were added after exercise SPECT. Therefore, for a given amount of exercise ischemia, the risk of major events was significantly lowered when medications including beta blockers had been subsequently added.
This observation cannot be explained by a less severe ischemic risk at baseline in the GII patients. Indeed, compared with the patients who had no change in antianginal treatment (GI), those who had a further addition of beta-blockers (GII) had equivalent baseline characteristics, even for variables related to the ischemic risk (reversible defect extent, positive exercise testing). Finally, the number of classes of antianginal medications prescribed after the test was higher in GII than in GI, giving evidence that GII patients were not considered at a lower ischemic risk after having performed the baseline investigations.
Contrary to what was observed for exercise SPECT ischemia, all of the other conventional prognostic parameters, such as those reflecting the amount of irreversibly damaged myocardium (irreversible defect extent, left ventricular ejection fraction) or those reflecting the severity of coronary artery disease (proximal LAD stenosis, number of diseased vessels), had a similar prognostic significance in the three groups. Therefore, a comprehensive analysis of our results suggests that there was a significant prognostic effect of beta-blockers principally related to the risk related to stress ischemia.
The heart protection documented for beta-blockers in a certain subset of CHD patients is considered to be mainly related to an anti-ischemic action, but also to their potential antiarrhythmic effects including the prevention of ventricular fibrillation during ischemia (23). The anti-ischemic action of beta-blockers has been documented in a number of exercise scintigraphic studies: propranolol has been reported to decrease perfusion abnormalities in 57% to 100% of cases (13–15)and to erase any abnormalities in 27% (15). In a recent Tl-201 SPECT study performed after myocardial infarction, it was observed that oral therapy including beta-blockers led to a reduction in exercise ischemia equivalent to that provided by coronary angioplasty (19).
The influence of the further addition of non–beta-blocking antianginal medications
The exercise ischemia evidenced by scintigraphy can also be decreased by the oral administration of non–beta-blocking antianginal medications; limited effects were observed for nifedipine (16)or isosorbide-5-mononitrate (17), but a more pronounced effect was documented at short term for isosorbide dinitrate (16,18). In our group comparisons, however, the risk related to stress ischemia (reversible defect extent) was unchanged in patients who had a subsequent administration of only non–beta-blocking antianginal medications (GIII). This suggests a specific role for beta-blockers. This finding is in keeping with the observation that only beta-blocking antianginal medications have been consistently reported to improve prognosis in certain subsets of CHD patients (20–23).
Although we have observed that exercise SPECT ischemia was a strong predictor for patients who had the test while on their daily life treatment (GI), few of them were on beta-blockers. Therefore, the prognostic value of exercise SPECT performed on beta-blocker therapy remains to be more extensively analyzed.
In the present study, no control SPECT was performed after the changes in antianginal medications. Therefore, we were not in a position to determine whether the decreased prognostic impact of exercise ischemia in patients who had a further addition of beta-blockers was directly related to an improvement in exercise perfusion, or whether other protective effects of beta-blockers such as their antiarrhythmic effects might have played a role.
In addition, the subsequent sequential changes in antianginal therapy during follow-up were not taken into account, and this was because this information could not be precisely obtained several years later in a number of cases. However, we found essentially similar results, despite a much lower number of analyzed events, when the analysis was restricted to the initial two years of follow-up and thus, when medications were likely to be unchanged for most of patients. Furthermore, the same limitation applies to previous prospective randomized studies, in which initial treatment with beta-blockers has been reported to improve prognosis (20–23), even at long-term follow-up (≥5 years) and despite the fact that a high number of patients had modified their medications (31,32). Our study shows that the prescription of beta-blockers is sufficient to have an influence on the deleterious prognostic significance of the stress (Tl-201 SPECT) ischemia, whereas it has no effect on the other nonischemic prognostic parameters.
Our analyses were mainly based on the prediction of major events (cardiac death, infarction) because other events, such as cardiac interventions, are subjective and are strongly influenced by the results of the baseline investigations (especially coronary angiography for myocardial revascularization). Nevertheless, in an additional analysis that included cardiac interventions, we found equivalent results on the SPECT prognostic parameters.
Our population of stable and medically treated patients constitutes a selected low risk subgroup, and extrapolations to other populations of CHD patients are uncertain. Most of our patients had a history of infarction, possibly because those without such a history and who have significant coronary artery disease are less frequently referred to medical therapy only. Therefore, it may be wondered whether the protective effect of beta-blockers predominantly acts in the setting of residual ischemia after infarction. This hypothesis, which would require a much larger population to be analyzed, is strengthened by the consideration that it is mainly in postinfarction patients that beta-blockers have been demonstrated to improve prognosis (20–23,31,32). Also, it might be questioned whether beta-blockers would have a similar effect in patients having a higher ischemic risk, such as those referred for myocardial revascularization. Although further studies will be required to clarify these points, the present data document that, when including beta-blockers, oral antianginal therapy is able to have a favorable impact on the deleterious prognostic effect of stress ischemia.
This study, performed in patients with medically treated CHD, gives evidence that: 1) exercise ischemia documented by Tl-201 SPECT has a strong prognostic significance when the test is performed on the same antianginal treatment as that subsequently prescribed; and moreover 2) the introduction of antianginal medications, when including beta-blockers, appears to have a favorable influence on the deleterious prognostic effect of exercise ischemia.
☆ This investigation was supported by the Hospital of Nancy (Projet Hospitalier de Recherche Clinique 1994-A).
- coronary heart disease
- confidence interval
- left anterior descending coronary artery
- relative risk
- single photon emission computed tomography
- Received November 21, 1997.
- Revision received February 16, 1999.
- Accepted March 19, 1999.
- American College of Cardiology
- Brown K.A
- Iskandrian A.D.S,
- Chae S.C,
- Heo J,
- Stanberry C.D,
- Wasserleben V,
- Cave V
- Machecourt J,
- Longère P,
- Fagret D,
- et al.
- Marie P.Y,
- Danchin N,
- Durand J.F,
- et al.
- Dakik H.A,
- Mahmarian J.J,
- Kimball K.T,
- Koutclou M.G,
- Medrano R,
- Verani M.S
- Stratmann H.G,
- Williams G.A,
- Wittry M.D,
- Chaitman B.R,
- Miller D
- Berman D.S,
- Hachamovitch R,
- Kiat H,
- et al.
- Tamaki N,
- Ohtani H,
- Yamashita K,
- et al.
- Rainwater J,
- Steele P,
- Kirch D,
- LeFree M,
- Jensen D,
- Vogel R
- Hockings B,
- Saltissi S,
- Croft D.N,
- Webb-Peploe M.M
- Madias J.E,
- Lee V.W,
- Song S.S
- Dakik H.A,
- Farmer J.A,
- Kleiman N.S,
- et al.
- Cruickshank J.M
- Marie P.Y,
- Karcher G,
- Danchin N,
- et al.
- Cox D
- ↵Dixon WJ. BMDP statistical software R7. Los Angeles, California: University Press of California, 1992.
- Brown K.A,
- Rowen M