Prevalence and Associations of Abdominal Aortic Aneurysm Detected through Screening

Methods

Participants

The Aneurysm Detection and Management (ADAM) study is an ongoing randomized clinical trial comparing two strategies for the management of AAA (immediate surgery and surgery reserved for aneurysms that enlarge to 5.5 cm, enlarge rapidly, or cause symptoms) in patients 50 to 79 years of age with asymptomatic AAAs 4.0 to 5.4 cm in diameter [13]. Ultrasonography screening clinics were established at the 15 participating Department of Veterans Affairs medical centers to support recruitment into the trial. Active patients at these centers (that is, patients treated during the current or previous fiscal year and those who had future appointments) who met the age criteria for the study were invited by mail to attend the clinic.

Throughout the screening period, the letters were mailed in batches according to patient social security number using the continuously updated administrative patient data at each participating center. A second mailing that was done at some centers excluded persons who had previously been seen in the screening clinic. Active patients at the participating centers were also accepted for screening on a walk-in basis; patients 50 to 79 years of age were included in the analysis because they were members of the target population and eventually would have been invited to participate. Inadvertent repeated screenings were identified by patient social security number, and results of second screenings were excluded. Patients who reported previously having been told that they had an AAA were also excluded.

We include data from patients who were screened from the beginning of the program in October 1992 through March 1995. The study was approved by the human rights committee at the Veterans Affairs Cooperative Studies Program coordinating center and by the institutional review boards at the 15 participating centers.

Assessment of Associated Factors

Before ultrasonographic examination, all patients completed a brief questionnaire that asked about demographic information and possible risk factors for AAA. Patients were asked whether they had ever been told by a physician that they had the condition in question (for example, high blood pressure). The questionnaire was developed for the study, tested at the participating centers, and revised accordingly before the study began.

The reliability of the data from the self-reported questionnaire was assessed by comparing this data with data collected in the subset of screened patients who had an AAA and who were later included in the clinical trial. These data were recorded by a study nurse after discussion with the patient and review of the patient's medical records.

Ascertainment of Abdominal Aortic Aneurysm

The abdominal aorta was measured above and below the renal arteries using a 3.5-MHz real-time sector scanner. The ultrasonographers were instructed to 1) scan the aorta in the anteroposterior and lateral planes and 2) report the maximum external diameter at the widest point of any dilatation for the suprarenal and infrarenal segments. The ultrasonographers met twice during the study period to review techniques and compare measurement distributions. Patients found to have an AAA or another suspected abnormality were referred for evaluation, and discrepancies between the results of subsequent testing and the results of the initial screening were reported back to the study ultrasonographer. Screening appointments for patients whose examinations were inadequate because of insufficient fasting were rescheduled.

An infrarenal aortic diameter of 4.0 cm or larger is generally agreed to constitute AAA, but no method for defining a smaller AAA has gained wide acceptance [14]. Most investigators have used unadjusted aortic diameter (with such cutoff points as ≥ 3.0 cm, which is known to be associated with risk for rupture [15]) but this practice may exaggerate the prevalence of AAA in larger people. To avoid this problem, use of a ratio of infrarenal aortic diameter to suprarenal aortic diameter of 1.5 or greater has been proposed [14]. However, this method labels some small aortas as having an AAA and fails to account for aneurysmal dilatation of the suprarenal aorta. The latter problem has, in turn, been addressed by a proposal to define AAA as an aortic diameter more than 1.5 times the diameter that would be expected on the basis of age, sex, body size, and other factors. However, values for the expected diameter have not been well defined [14]. Therefore, we considered several definitions of AAA, including an infrarenal aortic diameter of at least 3.0 cm, a diameter of at least 4.0 cm, and a ratio of infrarenal to suprarenal aortic diameter of 1.5 or greater.

Statistical Analysis

Analyses to determine the association between the items on the questionnaire and the presence of AAA were done using univariable and multivariable logistic regression. Because the purpose of the analysis was descriptive, the multivariable models included all variables that were considered in the questionnaire. Composite variables were formed from several questions that were used to identify coronary artery disease; hypertension; hypercholesterolemia; and, in a separate analysis, any atherosclerosis. Responses of “don't know” were treated as responses of “no” for specific diseases under the assumption that if a person had a disease, he or she would be aware of it. This assumption was not made for hypercholesterolemia or hypertension, which are often asymptomatic; responses of “don't know” for these conditions were considered to be “missing.” Because of the resulting large number of missing responses for hypercholesterolemia, an indicator variable was included for this item to prevent deletion of these patients from the logistic models (as was otherwise done for missing responses) [16]. Extreme values for height (< 152 cm and > 198 cm), weight (< 45 kg and > 182 kg), waist circumference (< 56 cm and > 152 cm), and cholesterol levels (< 2.59 mmol/L and > 12.93 mmol/L) and smoking history incompatible with age were also excluded.

To compare the many cases of smaller, borderline AAA with cases of larger, definite AAA, separate regression models were developed 1) for comparing patients who had AAAs of 3.0 to 3.9 cm with those who had infrarenal aortic diameters less than 3.0 cm and 2) for comparing those who had AAAs of 4.0 cm or larger with those who had infrarenal aortic diameters less than 3.0 cm. A third model was developed for comparing patients whose ratio of infrarenal to suprarenal aortic diameter was 1.5 or greater with those whose ratio was less than 1.5.

Comparison of questionnaire data with data collected for the clinical trial was made for categorical variables by using percent-pair agreement and the κ statistic and for continuous variables by using the mean difference (questionnaire value minus clinical trial value) and limits of agreement (the range within which 95% of the differences would be expected to occur, calculated as the mean difference ± 1.96 times the SD of the differences [17]).

Results

During the study period, 320 000 letters were mailed; 5.3% of these were returned because of an invalid address or death. Of the remaining letters, 30% (91 000 letters) were returned by persons willing to be screened. A total of 73 943 persons who were 50 to 79 years of age and did not have a history of AAA were screened; 492 were excluded because the aorta could not be visualized. The remaining 73 451 patients constitute the study group, of whom 3.5% were walk-ins. As Table 1 shows, the study group consisted primarily of men who had a history of smoking (97.2% were male; 75.5% had a history of smoking). This closely reflects the population of U.S. veterans (96% are male; 74% have a history of smoking [18]); however, the study group also included 2078 women and 17 981 persons who had never smoked.

Abdominal aortic aneurysm, defined as an infrarenal aortic diameter of 3.0 cm or more, was detected in 3366 patients (4.6%); the infrarenal aortic diameter was at least 4.0 cm in 1031 patients (1.4%). The frequencies of larger AAAs were as follows: Three hundred sixty-eight (0.50%) patients had an AAA of 5.0 cm or larger, 224 (0.30%) patients had an AAA of 5.5 cm or larger, 137 (0.19%) patients had an AAA of 6.0 cm or larger, 48 (0.07%) patients had an AAA of 7.0 cm or larger, and 22 (0.03%) patients had an AAA of 8.0 cm or larger. It can be seen that with each 1-cm decrease in AAA diameter, the number of AAAs of that diameter or larger more than doubles. Table 2 shows the prevalence of AAA by sex, race, and smoking status. The values in white men who smoked were substantially higher than those in other groups. The prevalence of AAAs of 4.0 cm or larger was 1.6% in white men and is shown by age and smoking status for all men in Table 3, in which the marked effect of these two variables can be seen. When the ratio of infrarenal to suprarenal aortic diameter (shown as an alternative definition of AAA in Table 2 and Table 4) was used, 118 aortas of less than 3.0 cm in diameter were classified as having AAAs and 201 aortas of 4.0 cm or larger were classified as not having AAAs.

The results of the multivariable analyses are shown in Table 4. The factors listed in Table 4 were significant univariable predictors except for 1) deep venous thrombosis and cancer at a site other than the skin [in all models], 2) weight [in the model for the ratio of infrarenal to suprarenal aortic diameter], and 3) abdominal imaging in the past 5 years (in the model for AAAs ≥ 4.0 cm and in the model for the ratio of infrarenal to suprarenal aortic diameter).

Smoking was the risk factor most strongly associated with AAA in the multivariable analyses; the odds ratio (OR) for AAAs of 4.0 cm or larger compared with normal aortas (infrarenal aortic diameter < 3.0 cm) was 5.57 (95% CI, 4.24 to 7.31). The association of smoking with AAA increased significantly with the number of years of smoking (ORs per 10-year increments: 1.18 [CI, 1.13 to 1.24] for AAA of 3.0 to 3.9 cm, 1.09 [CI, 1.02 to 1.16] for AAA ≥ 4.0 cm, and 1.13 [CI, 1.07 to 1.19] for ratio of infrarenal to suprarenal aortic diameter ≥ 1.5). This association decreased significantly with the number of years after quitting smoking (ORs per 10-year increments: 0.81 [CI, 0.76 to 0.86] for AAA of 3.0 to 3.9 cm, 0.72 [CI, 0.65 to 0.79] for AAA ≥ 4.0 cm, and 0.79 [CI, 0.73 to 0.85] for ratio of infrarenal to suprarenal aortic diameter ≥ 1.5) when these variables were added to the models. In a separate age-adjusted linear regression analysis that was limited to smokers who had an AAA, the number of years of smoking was not significantly related to the diameter of the AAA. After adjustment for the number of years of smoking (and all other variables), current smokers were still more likely than former smokers to have an AAA (OR, 1.63 [CI, 1.37 to 1.94] for AAA of ≥ 4.0 cm). The excess prevalence associated with smoking (the etiologic fraction [19]) accounted for 78% of all AAAs of 4.0 cm or larger in the population, which suggests that smoking may be responsible for most clinically important, previously undiagnosed AAAs.

Chronic obstructive pulmonary disease, which was only associated with an AAA of 4.0 cm or larger in the general multivariable models, was no longer significantly associated with AAA in that model after adjustment for the number of years of smoking (OR, 1.06 [CI, 0.90 to 1.24]).

In all three models, age and height were positively associated with AAA; female sex and black race were negatively associated with AAA. Sex and height remained significantly associated with AAA when AAA was defined by the ratio of infrarenal to suprarenal aortic diameter. This finding indicates that these factors were not artifacts of the effect of body size on the definition of AAA. Waist circumference was associated with an AAA of 4.0 cm or larger but was not significantly associated with AAA as defined by the ratio of infrarenal to suprarenal aortic diameter. A family history of AAA was associated with AAA but was reported by only 5.1% of the population.

Patients who reported having had abdominal imaging (ultrasonography, computed tomography, or magnetic resonance imaging) in the past 5 years were only moderately less likely to have previously undiagnosed AAA, a potentially useful finding for the design of future screening programs. The prevalence of AAA was not significantly different for walk-in patients (those who came in for screening before being invited) compared with patients who responded to the letter of invitation.

Coronary artery disease, high cholesterol levels, and hypertension were positively associated with AAA in all models. Weight and a history of cancer were not associated with AAA in any of the multivariable models. Other diseases, including cerebral vascular disease, claudication, and deep venous thrombosis, were independently associated with some but not all definitions of AAA. When the regression models in Table 4 were reanalyzed after coronary artery disease, cerebral vascular disease, and claudication were combined into a single variable of “any atherosclerosis,” the ORs for this variable were 1.57 (CI, 1.44 to 1.72) for AAA of 3.0 to 3.9 cm, 1.68 (CI, 1.47 to 1.92) for AAA of 4.0 cm or larger, and 1.67 (CI, 1.50 to 1.85) for a ratio of infrarenal to suprarenal aortic diameter of 1.5 or larger. These values are all slightly higher than those for coronary artery disease alone.

Diabetes showed an unexpected negative association with AAA in all three regression models. To assess the possibility that this finding could be an artifact caused by limiting our study to cases of AAA that were diagnosed through screening, we reviewed data on all patients who were in the same age range as those included in our study, had AAAs of 4.0 cm or larger that had not been diagnosed through screening, were referred to the trial from the same medical centers during the period of our study, and had a diabetes status recorded. Of these 849 referred patients, 12% had diabetes; this was identical to the 12% rate seen in screened patients who had an AAA of 4.0 cm or larger and significantly less than the 19% rate seen in screened patients who did not have an AAA (P < 0.001). This finding suggests that the negative association between AAA and diabetes in screened patients did not result from diabetic patients being more likely to have previously received a diagnosis of AAA.

We did a similar analysis for coronary artery disease (the only other condition for which data from referred patients were available) that yielded similar results: Fifty-seven percent of referred patients who had AAAs of 4.0 cm or larger had a history of coronary artery disease. This rate was similar to the rate of 53% seen in screened patients who had AAAs of 4.0 cm or larger (P > 0.09) but was significantly higher than the rate of 38% seen in screened patients who did not have AAAs (P < 0.001).

Three hundred sixty screened patients who had AAAs were later included in the clinical trial. These patients did not differ significantly from the other screened patients who had AAAs of similar size in any of the characteristics listed in Table 1 except mean age (67.5 years compared with 68.9 years; P < 0.001). When questionnaire data were compared with trial data in these 360 patients to assess the reliability of the questionnaire, pair agreement was more than 85% for all categorical variables except high cholesterol levels (77%). In addition, the κ value was 0.6 or more for all variables except chronic obstructive pulmonary disease (κ = 0.59), high cholesterol levels (κ = 0.47), and claudication (κ = 0.36). The lesser degree of agreement seen with high cholesterol levels and claudication was presumably caused by erroneous self-reporting, thus creating “noise” and probably reducing the strength of the associations for these variables as reported in Table 4. For continuous variables, the mean differences and limits of agreement were −0.1 years ( −3.6 to 3.3 years) for age; 1.1 cm ( −6.7 to 8.9 cm) for height; −0.4 kg ( −8.2 to 7.3 kg) for weight; −2.3 cm ( −14.7 to 10.1 cm) for waist circumference; and 1.0 years ( −13.3 to 15.3 years) for number of years of smoking.

Discussion

We report the largest study of AAA screening to date and the first multivariable analysis of potential risk factors for all previously unrecognized cases of AAA occurring in a population. Our findings may help resolve several disagreements that arose from previous studies. Our finding of a 1.6% prevalence of previously unrecognized AAAs of 4.0 cm or larger in older white men agrees with the rates found by O'Kelly and Heather [4] and Pleumeekers and colleagues [10]; however, our rate is higher than the 0.9% rate found by Lindholm and colleagues [2] and lower than the 2.0% to 3.3% rate reported in other screening studies [5-9]. Our findings suggest that differences of this magnitude could be explained by differences in smoking history and age, although only two of the earlier studies included patients older than 80 years of age [8, 10]. Female sex and black race were associated with lower rates of AAA in our study, as previously reported [10, 20, 21].

Small AAAs were much more common than large AAAs in our population and in other studies. The size distribution that we observed helps explain the importance of determining a size threshold for elective surgery [1] because lowering this threshold by 1 cm more than doubles the number of patients considered for surgery.

Smoking was the factor most closely associated with AAA in our study. Although an association was reported in multivariable analyses in the two previous case–control studies [11, 12] and in univariable analyses in several large prospective studies [22-24], smoking has not previously been considered the principal cause of AAA. If the association we observed between smoking and AAA is assumed to be causal, then most AAAs 4.0 cm or larger that were discovered in our screened population could be attributed to smoking. Our findings provide further evidence that AAA is a smoking-related illness.

The ways in which smoking might promote the formation of aneurysms remain unknown [3, 25]. The lack of an association between duration of smoking and AAA diameter in our patients could imply that smoking is more important in genesis than in growth of AAAs, but other researchers have found continued smoking to be associated with an increased growth rate [26].

Coronary artery disease was significantly associated with AAA, as was “any atherosclerosis,” but the associations were considerably weaker than the association with smoking. High serum cholesterol levels were also independently associated with AAA; this is a point of disagreement between the two previous multivariable analyses [11, 12]. This association presumably reflects the connection with atherosclerosis [12], although other explanations, including a direct inflammatory effect of some oils, have been suggested [3].

A family history of AAA was an independent predictor of AAA, but only a small proportion of patients reported it. Family history has been shown by others to be a univariable predictor [27, 28] but has not previously been included in multivariable models. Our findings support the hypothesis that at least some AAAs have a genetic basis [3].

A negative association was seen between AAA and diabetes. This finding may be an artifact caused by limiting our study to prevalent, previously undiagnosed AAA if diabetic patients are more likely than those who do not have diabetes to have died with an AAA or to have received a previous diagnosis of AAA. However, the two previous case–control studies did not find a significant positive association between diabetes and dying with or of AAA [11, 12], and we saw equally low rates of diabetes associated with AAAs found through screening and through other means.

The literature offers some support for a negative association between AAA and diabetes. Patients in Massachusetts who had resection of an AAA had a significantly lower prevalence of diabetes than did controls who had appendectomy (OR, 0.78) [29]. In addition, nonsignificant reductions were seen in the frequency of diabetes in patients who had AAA in the two largest previous screening studies to date that included information on diabetes (relative risks, 0.83 for each [9, 10]) and in a recent update of one of the nested case–control studies (relative risk, 0.73) [30]. None of these findings was addressed by the researchers.

Diabetes appears to have an effect on large arteries that is distinct from atherosclerosis and is characterized by increased aortic stiffness [31, 32] and, at least in peripheral arteries, by medial calcification [33]. These changes could stabilize the aorta and resist aneurysmal dilatation [34]. However, aortic stiffness has been found to be increased in patients who have an AAA [35, 36], and no association has been seen between calcification and the expansion rate of AAA [37]. Therefore, the reason for the negative association between AAA and diabetes remains unclear.

The association of AAA with hypertension was marginally significant, a finding that reflects the results of previous reports. Most screening studies have found no univariable association between AAA and hypertension [5-7, 9], whereas both case–control studies found hypertension to be an independent risk factor for AAA [11, 12].

We did not find an independent association between AAA and any of several other factors, including cancer at sites other than the skin, body weight, and chronic obstructive pulmonary disease (after quantity of smoking was considered). Several other researchers found a univariable association between AAA and chronic obstructive pulmonary disease and suggested an abnormality of elastin as a possible link [6, 9]. Our multivariable models suggest that the univariable association of chronic obstructive pulmonary disease with AAA is mediated through heavy smoking rather than through a preexisting abnormality, but our findings do not preclude a similar effect of smoking on tissue integrity in the lung and aorta.

Our study had several limitations. First, it was limited to U.S. veterans who were 50 to 79 years of age; generalization of the results to other populations must be made with caution. Second, as other screening studies have been, our report is limited to previously unrecognized prevalent cases of AAA. The true prevalence of AAA is therefore higher than the values we (and others) report. Also, as noted above in relation to diabetes, previously unrecognized prevalent cases of AAA may differ from other cases of AAA with respect to associated factors. The similar prevalences of diabetes and coronary artery disease that we saw in patients in whom AAA was detected through screening compared with those in whom AAA was detected by other means provide some reassurance in this regard. A study of all prevalent or incident cases of AAA in a large population, requiring universal screening plus universal review of charts and death records, is probably not feasible. Third, the response rate to our letters of invitation was 30%, raising the possibility of selection bias if AAA prevalence or associations differed between responders and nonresponders. Fourth, the assessment of potential risk factors was by self-report and may therefore contain inaccuracies. Comparison with carefully collected data in patients who were randomly assigned in the clinical trial showed reasonably high agreement on most topics, although only moderate agreement was shown for high cholesterol levels and claudication. Erroneous self-reported data would be expected to create “noise” and reduce associations. Fifth, results of ultrasonography may be inaccurate. However, we previously reported [38] that results produced by our ultrasonographers for AAA in this population differed from results of computed tomography by less than 0.5 cm in two thirds of cases.

Our study provides information that should be useful for planning future screening programs for AAA, but it does not directly address whether such programs are justified. This issue, which depends on the quality of the evidence that screening reduces morbidity or mortality rates and is cost-effective, remains unresolved [39]. The U.S. Preventive Services Task Force [40] has given AAA screening a “C” recommendation (“insufficient evidence to recommend for or against screening”). In a recently reported randomized trial, screening for AAA resulted in a 55% reduction in AAA ruptures in men, but the associated 41% reduction in AAA-related deaths did not reach statistical significance despite a sample size of more than 15 000 [41]. Obtaining more definitive information on the effectiveness of screening will be difficult, but the ADAM trial [13] and a similar study in the United Kingdom [42] will provide related evidence about the effectiveness of surgery for small AAAs.

In conclusion, previously undiagnosed AAA in older U.S. veterans was associated with several factors. Smoking showed the strongest association and appears to be responsible for most clinically important cases of previously undiagnosed AAA. Age, coronary artery disease, any atherosclerosis, high cholesterol levels, and family history of AAA also showed important associations; female sex, black race, and diabetes were negatively associated. These data may help guide future research on the causes of AAA and future programs for the detection of this condition.

Appendix

The following are members of the ADAM Veterans Affairs Cooperative Study Group. Asterisks indicate past members.

Frank A. Lederle, MD (Co-Chairman), Veterans Affairs Medical Center, Minneapolis, Minnesota; Samuel E. Wilson, MD (Co-Chairman), University of California, Irvine, Medical Center, Orange, California; Jean M. Larson (Project Coordinator), Veterans Affairs Medical Center, Minneapolis, Minnesota; and Donovan B. Reinke, MD, and Howard J. Ansel, MD, Central Computed Tomographic Scan Laboratory, Veterans Affairs Medical Center, Minneapolis, Minnesota.

Investigators at Veterans Affairs Medical Centers: Louis M. Messina, MD, Mary Lingg, RN, and Latonya Trohallis (Ann Arbor, Michigan); Linda Graham, MD, Charles L. Mesh, MD, Evelynn Wingard, RN, MS, John Lindesmith, RN*, and Julie Francosky, RT (Cleveland, Ohio); Anselmo A. Nunez, MD, Christine Estep, RN, and James Taylor, RT (Miami, Florida); William C. Krupski, MD, Pamela Strecker, RN, and Linda Schoening, CCVT (Denver, Colorado); Fred N. Littooy, MD, John Maggio, PhD, and Wendy Cote, RVT (Hines, Illinois); Gary W. Barone, MD, Bernard W. Thompson, MD*, Brenda Kackley, RN, Rob Little, RT, and Terry Wood, BSRT* (Little Rock, Arkansas); Ian Gordon, MD, Cheryl Kohl, RN, Rebecca Complin, RVT, and Manju Akkinepali* (Long Beach, California); Charles W. Acher, MD, Judy Archibald, RN, Traci Jo McGuire, RTR, and Jennifer Fitzsimons, BS* (Madison, Wisconsin); Edmund P. Chute, MD, William D. Payne, MD, Michael L. Schwartz, MD, Catherine Proebstle, RN, Ronald Hedblad, RT, and Michael Richardson, RT* (Minneapolis, Minnesota); Michel S. Makaroun, MD, Tricia DeBoo, RN, and Jean Fuhs, RT (Pittsburgh, Pennsylvania); Gregory L. Moneta, MD, Richard A. Yeager, MD, James M. Edwards, MD, Roland Jemerson, RN, and Robert McCartney, RVT (Portland, Oregon); Raymond G. Makhoul, MD, Laura Bartnicki, RN, and Linda Tyree Joyner, RT (Richmond, Virginia); Robert J. Hye, MD, Edward J. Plecha, MD, Yehuda G. Wolf, MD*, Gerry Cali, RN, Terri Scala, RN*, and Roger Hull, RCVT (San Diego, California); Dennis Bandyk, MD, Dolores Bou-Eid, RN, and Maureen Jackson, RT (Tampa, Florida); and Julie Freischlag, MD, Christa Kallio, RN, Susan Framberg, RN, Sandra Towne, RN*, and Cleveland Weller, RT (Milwaukee, Wisconsin).

Planning Committee: Frank A. Lederle, MD (Chairman), Minneapolis, Minnesota; Samuel E. Wilson, MD, Long Beach, California; Gary R. Johnson, MS, West Haven, Connecticut; Louis Messina, MD, Ann Arbor, Michigan; Charles W. Acher, MD, Madison, Wisconsin; Fred N. Littooy, MD, Hines, Illinois; Jack L. Cronenwett, MD, Hanover, New Hampshire; David J. Ballard, MD, PhD, Decatur, Georgia; David B. Matchar, MD, Durham, North Carolina; and Christopher R.B. Meritt, MD, New Orleans, Louisiana.

Executive Committee: Frank A. Lederle, MD (Chairman), Minneapolis, Minnesota; Samuel E. Wilson, MD, Long Beach, California; Gary R. Johnson, MS, West Haven, Connecticut; Louis M. Messina, MD, Ann Arbor, Michigan; Charles W. Acher, MD, Madison, Wisconsin; Fred N. Littooy, MD, Hines, Illinois; David J. Ballard, MD, PhD, Decatur, Georgia; and C. William Cole, MD*, Vancouver, British Columbia, Canada.

Data Monitoring Board: G. Patrick Clagett, MD (Co-Chairman), University of Texas Southwestern Medical Center, Dallas, Texas; John D. Corson, MB, ChB (Co-Chairman), University of Iowa, Iowa City, Iowa; William C. Cushman, MD, Veterans Affairs Medical Center, Memphis, Tennesee; C. Seth Landefeld, MD, University Hospital of Cleveland, Cleveland, Ohio; Theodore G. Karrison, PhD, University of Chicago, Chicago, Illinois; John P. Matts, PhD, University of Minnesota, Minneapolis, Minnesota; and Dorothea Collins, ScD (ex officio), CSPCC, Veterans Affairs Medical Center, West Haven, Connecticut.

Cooperative Studies Program Coordinating Center, West Haven, Connecticut: Gary R. Johnson, MS (Study Biostatistician); Rae Bartozzi, Ray Kilstrom, MBA*, and Kathy Riester, BA* (Statistical Assistants); Kathy Newvine, BS, and Robert Goodwin, MS (Computer Programmers); Dorothea Collins, ScD (Chief); Peggy Antonelli (Administrative Officer); Mary Smith (forms design); Darrell Burns (Travel Clerk); and Lillie Franklin, Pattie Collins, Stella Marcinauskis, and Bonita Hunter (data entry).

Cooperative Studies Program Administration at the Department of Veterans Affairs Central Office: John R. Feussner, MD (Chief); Daniel Deykin, MD (past Chief); Janet Gold (Administrative Officer); and Ping C. Huang, PhD (Staff Assistant).

Presented at the 19th Annual National Society for General Internal Medicine Meeting, 3 May 1996, Washington, D.C.

From the Veterans Affairs Medical Center, Minneapolis, Minnesota; the Veterans Affairs Medical Center, West Haven, Connecticut; the University of California, Irvine, Orange, California; the Veterans Affairs Medical Center, Hines, Illinois; the Veterans Affairs Medical Center, Tampa, Florida; the Veterans Affairs Medical Center, Denver, Colorado; the Veterans Affairs Medical Center, Little Rock, Arkansas; the Veterans Affairs Medical Center, Madison, Wisconsin; and Emory University Center for Clinical Evaluation Sciences, Decatur, Georgia.

Mr. Johnson: Cooperative Studies Program Coordinating Center (151A), Veterans Affairs Medical Center, 950 Campbell Avenue, West Haven, CT 06516.

Dr. Wilson: Department of Surgery, University of California, Irvine, Medical Center, 101 The City Drive, Orange, CA 92668.

Dr. Littooy: Veterans Affairs Medical Center (112), Fifth Avenue and Roosevelt Road, Hines, IL 60141.

Dr. Bandyk: Veterans Affairs Medical Center (112), 13000 Bruce B. Downs Boulevard, Tampa, FL 33612.

Dr. Krupski: Veterans Affairs Medical Center (112), 1055 Clermont Street, Denver, CO 80220.

Dr. Barone: Veterans Affairs Medical Center (112), 4300 West Seventh Street, Little Rock, AR 72205.

Dr. Acher: Veterans Affairs Medical Center (112), 2500 Over-look Terrace, Madison, WI 53705.

Dr. Ballard: Emory University Center for Clinical Evaluation Sciences, Decatur Plaza Suite 620, 101 West Ponce de Leon Avenue, Decatur, GA 30030.