[Skip to Content]
[Skip to Content Landing]
Editorial
June 14, 2000

Primary Angioplasty—Time Is of the Essence

Author Affiliations

Author Affiliation: Department of Cardiology, Cleveland Clinic Foundation, Cleveland, Ohio.

JAMA. 2000;283(22):2988-2989. doi:10.1001/jama.283.22.2988

Twelve years ago, Ellwood1 described "a technology of experience" by which observational databases would be used to learn about how best to improve implementation of medical interventions. Ellwood envisioned a unified national database containing information on all facets of patient care. Researchers and decision makers would have the ability to analyze relevant data to answer questions of interest. Although a single national database of this kind has not been realized, a number of very large registries have been successfully assembled and used to learn about important associations between treatments and outcomes.2 In this issue of THE JOURNAL, Cannon et al3 report on the relation between speed of performing primary angioplasty and in-hospital mortality among patients with acute myocardial infarction. Taking advantage of a multicenter database containing information about more than 750,000 patients, the authors focused on 27,080 patients who presented with ST-segment elevation or left bundle-branch block and who were treated with primary angioplasty.

Only 2230 patients (8%) underwent primary angioplasty within 60 minutes of hospital presentation; that is, they had a "door-to-balloon time" of less than 1 hour. These patients had an in-hospital mortality rate of 4.2%. As the door-to-balloon time increased to 2 hours or longer than 3 hours, mortality steadily increased to 8.5%. The increased mortality persisted after using logistic regression4 to adjust for baseline differences: patients who had a door-to-balloon time longer than 2 hours still had a 41% to 62% increased risk. The authors concluded that door-to-balloon time is an independent and potentially modifiable predictor of mortality and that this measure could be used as a quality-of-care indicator. Of concern, nearly half the patients studied had a door-to-balloon time longer than 2 hours.

The association noted between door-to-balloon time and mortality meets a number of accepted criteria for a valid epidemiological relationship.5 The relation was strong, because death rates nearly doubled as door-to-balloon time increased from less than 1 to longer than 3 hours. There is certainly biological plausibility, given the known rapid progression of myocardial necrosis after coronary artery occlusion6 and the results of thrombolysis trials demonstrating lower mortality with early restoration of patency in the infarct-related artery.7 A dose-response pattern was present, since mortality rates steadily increased with longer door-to-balloon times. Also, at least according to standard statistical techniques,4 the association was independent of confounding factors.

The issue of confounding is a serious concern in the study by Cannon et al. Patients who had shorter door-to-balloon times were very different from patients with longer times. Those with shorter times were younger, more likely to be male, and less likely to have diabetes, prior myocardial infarction, prior coronary artery bypass graft surgery, and contraindications to thrombolytic therapy. Although standard statistical techniques may be used to calculate adjusted risk or odds ratios, the results of these analyses are inherently untrustworthy when many baseline differences exist.8 Furthermore, a number of unobserved confounding factors may have played a role in producing an apparent association between shorter door-to-balloon times and lower mortality, including patient-specific factors such as collateral vessels; physician-specific factors like previous primary angioplasty experience9; or hospital-specific factors such as use of highly coordinated critical care protocols.10

Because of marked differences in their study cohort's baseline characteristics, the authors appropriately chose to use a technique known as propensity analysis,8,11 which effectively addresses the problem of multiple confounding variables. For each patient a propensity score was generated that estimated the probability of a longer door-to-balloon time given his/her baseline characteristics.11,12 When the authors repeated their mortality analyses after adjusting for the propensity score, a door-to-balloon time longer than 2 hours remained predictive of an increased risk of death, although the increase in risk was 28%, less than the 41% to 62% found using standard techniques.

Propensity analysis has many important attributes, particularly the ability to balance many observed covariates simultaneously.8,11,12 Previous investigators have used propensity analyses successfully to study a number of questions, including use of revascularization procedures for treatment of coronary artery disease.13 Important limitations of propensity analysis must be recognized, however, including inability to balance unobserved confounders and the need for very large databases.8

The inability to account adequately for confounding factors in observational databases has led some researchers to question the use of nonrandomized cohorts to assess effects of treatments.14 The accepted criterion standard for assessing treatments is the randomized controlled trial, which can successfully eliminate observed and unobserved confounders.15 Indeed, the medical literature is replete with examples by which biologically plausible treatments have been found to be ineffective16 or even harmful17 once subjected to the rigors of randomized trials. Nonetheless, there are inherent problems with randomized trials, including expense, time requirements, and enrollment of patients who tend not to be representative of "real" practice.18

Furthermore, some important treatment-related questions simply cannot be answered by randomized trials. It is inconceivable that a randomized trial would be done comparing immediate to "delayed" primary angioplasty. Yet, the issue of whether door-to-balloon time matters is a very important one. If Cannon et al are correct, hospitals that are unable to consistently perform primary angioplasty within 2 hours of patient arrival should make concerted efforts to improve their processes. Given the large number of potentially eligible patients, even a 28% difference in mortality may translate into many lives lost.

Two other important findings in the study by Cannon et al deserve comment. First, while there was a strong association between door-to-balloon times and in-hospital mortality, no such association existed between the time from symptom onset to the performance of angioplasty. This observation was surprising. The authors correctly suggest that this lack of association may have been due to the inability of patients to precisely determine when their coronary arteries occluded. Cannon et al also argue that among late-presenting patients, complete reperfusion may be achieved more often with primary angioplasty19 than with thrombolysis.20 In any case, the greater importance of door-to-balloon time as a predictor of outcome suggests that, at least for determining optimal timing of primary angioplasty, hospital-specific factors are at least as important as patient-specific ones.

Second, Cannon et al noted that an association existed between hospital volume of primary angioplasty and mortality that was independent of the relationship between door-to-balloon time and outcome. Hospitals performing more than 1 primary angioplasty per month had lower mortality rates than those performing fewer; those institutions performing 3 or more per month had even lower mortality rates. The reasons for this are not clear, considering the very small number of procedures per institution and likely even smaller number per operator in a significant fraction of the institutions studied. Nonetheless, this pattern is consistent with studies reporting an inverse association between procedural volumes and adverse outcomes.21

The report by Cannon and colleagues demonstrates how a large, carefully assembled observational database can be used to gain important insights about treatment efficacy and quality of care. Their observations support the recommendation that hospitals caring for patients with acute myocardial infarction should offer primary angioplasty only if they can consistently keep door-to-balloon times to less than 2 hours, and if they can maintain an adequate volume of procedures. Patients should not be excluded from rapidly performed primary angioplasty solely because of duration from symptom onset. As widespread implementation of lessons from observational databases may save thousands of lives, Ellwood's envisioned "technology of experience" will become reality.1

References
1.
Ellwood PM. Shattuck lecture—outcomes management: a technology of patient experience.  N Engl J Med.1988;318:1549-1556.Google Scholar
2.
Lauer MS, Fortin DF. Databases in cardiology. In: Topol EJ, ed. Textbook of Cardiovascular Medicine. Philadelphia, Pa: Lippincott-Raven Publishers; 1998:1083-1106.
3.
Cannon CP, Gibson CM, Lambrew CT.  et al.  Relationship of symptom-onset-to-balloon time and door-to-balloon time with mortality in patients undergoing angioplasty for acute myocardial infarction.  JAMA.2000;283:2941-2947.Google Scholar
4.
Hosmer D, Lemeshow S. Applied Logistic RegressionNew York, NY: Wiley; 1989.
5.
Hill AB. The environment and disease: association or causation?  Proc R Soc Med.1965;58:295-300.Google Scholar
6.
Reimer KA, Lowe JE, Rasmussen MM, Jennings RB. The wavefront phenomenon of ischemic cell death, 1: myocardial infarct size vs duration of coronary occlusion in dogs.  Circulation.1977;56:786-794.Google Scholar
7.
The GUSTO Angiographic Investigators.  The effects of tissue plasminogen activator, streptokinase, or both on coronary-artery patency, ventricular function, and survival after acute myocardial infarction.  N Engl J Med.1993;329:1615-1622.Google Scholar
8.
Rubin DB. Estimating causal effects from large data sets using propensity scores.  Ann Intern Med.1997;127:757-763.Google Scholar
9.
McGrath PD, Wennberg DE, Malenka DJ.  et al. for the Northern New England Cardiovascular Disease Study Group.  Operator volume and outcomes in 12,998 percutaneous coronary interventions.  J Am Coll Cardiol.1998;31:570-576.Google Scholar
10.
Knaus WA, Draper EA, Wagner DP, Zimmerman JE. An evaluation of outcome from intensive care in major medical centers.  Ann Intern Med.1986;104:410-418.Google Scholar
11.
Joffe MM, Rosenbaum PR. Invited commentary: propensity scores.  Am J Epidemiol.1999;150:327-333.Google Scholar
12.
D'Agostino Jr RB. Propensity score methods for bias reduction in the comparison of a treatment to a non-randomized control group.  Stat Med.1998;17:2265-2281.Google Scholar
13.
Mark DB, Nelson CL, Califf RM.  et al.  Continuing evolution of therapy for coronary artery disease.  Circulation.1994;89:2015-2025.Google Scholar
14.
Yusuf S, Wittes J, Bailey K, Furberg C. Digitalis—a new controversy regarding an old drug: the pitfalls of inappropriate methods.  Circulation.1986;73:14-18.Google Scholar
15.
Moses LE. Measuring effects without randomized trials? options, problems, challenges.  Med Care.1995;33(4 suppl):AS8-AS14.Google Scholar
16.
Schatzkin A, Lanza E, Corle D.  et al.  Lack of effect of a low-fat, high-fiber diet on the recurrence of colorectal adenomas.  N Engl J Med.2000;342:1149-1155.Google Scholar
17.
The ALLHAT Collaborative Research Group.  Major cardiovascular events in hypertensive patients randomized to doxazosin vs chlorthalidone.  JAMA.2000;283:1967-1975.Google Scholar
18.
Califf RM, Pryor DB, Greenfield Jr JC. Beyond randomized clinical trials: applying clinical experience in the treatment of patients with coronary artery disease.  Circulation.1986;74:1191-1194.Google Scholar
19.
Brodie BR, Stuckey TD, Wall TC.  et al.  Importance of time to reperfusion for 30-day and late survival and recovery of left ventricular function after primary angioplasty for acute myocardial infarction.  J Am Coll Cardiol.1998;32:1312-1319.Google Scholar
20.
TIMI Study Group.  The Thrombolysis in Myocardial Infarction (TIMI) trial: phase I findings.  N Engl J Med.1985;312:932-936.Google Scholar
21.
Grassman ED, Johnson SA, Krone RJ. Predictors of success and major complications for primary percutaneous transluminal coronary angioplasty in acute myocardial infarction.  J Am Coll Cardiol.1997;30:201-208.Google Scholar
×