The Case of the Thirteen Watches

Door-to-balloon time is probably the most overstated, but least understood surrogate outcome measured in Emergency Medicine today. No sooner does the patient with midsternal chest pressure and ST elevations set foot in the emergency department, when they are packaged up and swept away to the cath lab by a gang of cardiology fellows and attendings. Now with the addition of prehospital EKGs we barely have time to get a set of vitals and toss and aspirin and prasugrel (if your cardiologists are anything like ours) in their mouth as they pass by. On September 4th 2013, the NEJM published an article showing just how rapid our response has become(1).

In this study, Menees et al examined door-to-balloon times from 2005-2009. Over a 4 year period door-to-balloon-times decreased from an average of 82 minutes to 67 minutes. In a disease as critical as ST elevation myocardial infarction, this decrease in time was sure to make a dramatic difference in mortality. The assumption was that this study would prove that all the resources invested had proved their worth, including training an army of cardiologists to be on 24 hour call and mobilizing a pre-hospital system to quickly and effectively deliver our chest pain patients to certified STEMI centers which were equipped with state of the art cath labs designed to be ready for action at a moments notice. What did we get from this massive investment? What is the prize for being faster, prettier and more efficient than we were 5 years ago? The answer appears to be nothing…

The overall mortality was unchanged from 2005 to 2009 despite this drastic reduction in door-to-balloon time. The opinion piece published in conjunction with this article by Dr. Eric Bates and Dr. Alice Jacobs (2), postulates that we have reached the nadir of door-to-balloon time and to further improve STEMI outcomes we must now focus on prehospital time. “Symptom-to-door time” will be this decade’s proclamation. After investing an extraordinarily large quantity of money and resources into this new arena we will surely see the drop in mortality which we all hoped would be seen with our speedy dash to the cath lab. This conclusion and proposal is fundamentally and mathematically wrong, as we will examine.

The fundamental belief that door-to-balloon time is a universally important measure was based on the hypothesis that “time is muscle”. The door-to-balloon time version of the time is muscle theory assumes a linear relationship between time to reperfusion and amount of myocardium salvaged. Any fan of the cardiac literature knows that this is not the case. As far back as ISIS 2 (3), it was shown that the time is muscle theory is only true in the hyper-acute phase of a STEMI. There was a 4.1% ARR if patients were treated with streptokinase within the first 4 hours. This benefit dropped significantly to an ARR of 2% in those that were treated over 4 hours, but stayed relatively consistent all the way out to 24 hours.

In GUSTO 2b, a study comparing PCI to tPA, a similar effect occurred(4). Overall PCI outperformed its competitor. There was a 1.3% ARR in patients who received PCI when compared to those receiving tPA. Contrary to the overall findings, in the patients who were treated within 2 hours of symptom onset, tPA performed better than PCI. A pooled analysis of 25 randomized clinical trials comparing primary percutaneous coronary intervention and in-hospital fibrinolysis in acute myocardial infarction patients published in The European Heart Journal in 2006 (5) found the following:

Overall there was a 2.6% ARR in death with PCI when compared to tPA. Early in the disease process (within the first 2 hours) the difference between the interventions is far less, because in this stage of a STEMI time actually does matter and the small delay to get to the cath lab has an effect. After 3 hours the difference in treatments remains stable all the way out to 12 hours.

In the CAPTIM trial published in the Lancet in 2002, patients were randomized to prehospital fibrinolysis (PHF) or PCI (13). There was an average difference in reperfusion time of 60 minutes between the two treatment arms.In patients whose symptoms began less than 2 hours earlier, those who received PHF fared better than those who received PCI. Inversely, in patients whose symptoms began greater than 2 hours earlier, those who received PCI did better. This is not a discussion of the treatment options for STEMI but merely a comparison of PCI and thrombolytics to demonstrate the non-linear effect of time on myocardial necrosis. Time was so valuable in the first few hours that the delay to intervention involved in performing PCI is important enough to make tPA (overall an inferior choice for reperfusion) superior or equivalent to PCI.

Even with modern treatment of STEMIs this pattern holds true. The STREAM trial published in the NEJM in March of 2013 (6) compared immediate thrombolytic therapy to PCI with a delay of 60 minutes (around the average door-to-balloon time at present (1)). The caveat was the symptoms had to have started within 3 hours of presentation. In this trial the median difference in symptom onset to start of reperfusion was 78 minutes. 30-day mortality was similar between the two groups, 3.3 vs. 3.4 respectively, but there was more heart failure and episodes of cardiogenic shock in the primary PCI group. So even in the modern era of STEMI treatment, time is critical and even small delays make a difference in this hyper-acute phase of the infarction.

This temporal relationship is not linear. Just because a delay of 1 hour is critical early in the course of the disease, does not mean that same delay matters later on. In both ISIS 2 and GUSTO 2b patients who were seen over 4 hours into their infarction, a delay of 2 hours made no difference in outcomes(3,4). So where did our universal application of door-to-balloon time come from and how did we stray so far from the practical?

As recently as 1999 the ACC/AHA “Guidelines for the Management of Patients With Acute Myocardial Infarction” (7) the 1A recommendation for “early PTCA” is stated as follows:

As an alternative to thrombolytic therapy in patients with AMI and ST-segment elevation or new or presumed new left BBB (LBBB) who can undergo angioplasty of the infarct-related artery within 12 hours of onset of symptoms or beyond 12 hours if ischemic symptoms persist, if performed in a timely fashion by persons skilled in the procedure and supported by experienced personnel in an appropriate laboratory environment.

There is no mention of door-to-balloon time anywhere in the document. It is not until 2004 when the AHA published its “Guidelines for the Management of Patients With ST-Elevation Myocardial Infarction”(8), that the recommendation of a door-to-balloon time of 90 minutes appeared. The citation used to justify this recommendation is a 2000 JAMA article entitled “Relationship of Symptom-Onset-to-Balloon Time and Door-to-Balloon Time With Mortality in Patients Undergoing Angioplasty for Acute Myocardial Infarction”(9).

This article included data from a prospectively gathered registry of 27,080 patients with ST elevation myocardial infarctions undergoing primary PCI. Using multivariable logistic regression they calculated mortality rates at 30- minute intervals. Using these methods they found a “statistically” significant increase in mortality when delay to PCI was longer than 120 minutes. Interestingly this same benefit was not found when examining symptom onset-to-balloon time. Instead of questioning this seemingly physiological impossibility the authors concluded that door-to-balloon time was the metric of value and we as a system bought in. Why would door-to-balloon time matter while symptom-onset-to-balloon time holds no clinical significance? It is important to point out that this was not a randomized trial. At best we can say there seems to be an association with door-to-balloon time and decreased mortality. Patients who had longer door-to-balloon times were older, had more prior and severe MIs, more CABGs and higher rates of DM.In a word, they were sicker. Maybe the reason these patients had longer delays to reperfusion were that they were not stable enough to go directly to the cath lab. There are countless reasons why people with longer delays fare poorer which may be unrelated to time to reperfusion.

Multivariable logistic regression itself is not always reliable and will not always provide us with the answers we seek. Most importantly when building a logistical regression model is the assumptions you make about the relationship between your dependent and independent variables. All logistic and linear regression models are an attempt to define the slope of the line that explains the relationship between two variables. If variable A increases by 1 unit how does this affect variable B. When conducting multivariable regression you are attempting to control for other variables that may confound your initial measurement. The difference between logistic and linear regression is simply the variables you wish to measure. Linear regression is done on continuous variables whereas logistic regression is performed when you have binary or ordinal variables. In either case it is important to have some idea of the slope of the line you are attempting to define. It is critical to examine your dataset and estimate the type of line that will best fit your data. This is what is called testing for goodness of fit (10).

The authors of the 2000 JAMA trial used stepwise logistic regression to build a model, which included all the “statistically” significant variables that were gathered from the registry data. This involves performing regression analysis on each variable collected and including the variables that best predict the mortality of the cohort. The assumption is if you now control for these variables you will be able to eliminate their effect on the cohort and examine exclusively the effect of the independent variable of your choosing, in this case time. This methodology has many flaws. The first and most obvious being you can only account for variables that have been thought of and recorded. If certain variables were not thought of or the data was missing then they would not be included in the dataset and could not be controlled for. Even the variables that you do control for can lead you astray. In a dataset this large, in which many variables are measured it is easy to have multiple variables appear statistically significant merely due to the large n alone. Step-wise logistic regression assumes each of these variables independently affects our outcomes measure setting the threshold of statistical significance at a p value of 0.05. The more variables you measure in this fashion the more likely you will find statistical significance by chance alone. Secondly, the authors of this JAMA article assumed a linear relationship between time and mortality and failed to check for goodness of fit.

Ironically even with these methodological flaws in the gathering and presenting of their data they may have obtained the right answers. They just drew the wrong conclusions. In a similar article published in the BMJ in 2009 entitled “Association of door-to-balloon time and mortality in patients admitted to hospital with ST elevation myocardial infarction: national cohort study”(11), the authors examined this very same question. Using a national database they analyzed the effect of door-to-balloon time on overall mortality. They too used logistic regression to control for confounders. Only this time proper methodology was utilized. The authors used predefined variables derived and internally validated in this cohort to control for severity of illness. Before running their regression model they tested for goodness of fit and to what should at this point be no one’s surprise, discovered a linear model was not ideal for defining door-to-balloon time’s effect on mortality.

Their results were almost identical to the JAMA study. They found a statistical association between increasing door-to-balloon times and increasing mortality. Like the previous study the authors drew the conclusion that because an increased door-to-balloon time was associated with an increased mortality, we can improve patient outcomes by shortening that time. This is not necessarily true. Longer door-to-balloon times are markers of severity of illness and difficulty of the procedure. The sicker patients were more likely to have longer door-to-balloon times and for this reason were more likely to have poorer outcomes.

In fact these are the same results that were found in the recent NEJM article by Menees et al (1). Patient with door-to-balloon times greater than 90 minutes had a significantly higher mortality than those who were under 90 minutes. Again this was a measure of how sick the patients were, and not a variable that if manipulated affects outcomes, proven by unchanged mortalities despite the decreasing annual door-to-balloon times.

The most extraordinary part of this misconceived urgency is that this is not a case where we have very little data and have to draw conclusions from observational cohorts. On the contrary, we have thousands and thousands of patients in true randomized control trials demonstrating that except for the hyper-acute phase of a STEMI where time is vitally important, even a 2 hour delay to the PCI did not make a clinical difference(3,4,5).

What is concerning about the Menees study is not the lack of benefit decreasing door-to-balloon times had on national mortality, this should have been obvious, but rather we are at risk of making the same flaw in logic that we made 13 years ago. Instead of accepting that we have reached a natural plateau of the disease process and further attempts to reduce reperfusion times will be met with diminishing return, we have decided we must redouble our efforts and mobilize the prehospital resources to improve our symptom-onset-to-door times.

In 2009, Dr. Jerry Hoffman and Dr. Richelle Cooper published an article in the CJEM addressing this exact topic(12). Discussing the efforts of Canada to restructure their prehospital resources to divert chest pain patients to STEMI capable facilities, Dr. Hoffman and Cooper make an strong argument that though this plan will most certainly benefit a few patients mathematically it will surely cost more than it is worth. They estimate that out of 100,000 chest pain patients who call for an ambulance if 50% are critical enough to have ACS (a very high estimate by American standards) and 5% of these patients have a myocardial infarction, then 2,500 out of 100,00 are having an AMI. If even 20% of these patients (again very high for American standards) are having a true STEMI then we are down to 500 patients out of the original 100,000 who could possibly benefit from rapid mobilization of prehospital system. Now if the benefit of PCI over thrombolytics has an absolute risk reduction of death of 1-2%, then only 5-10 of these 500 STEMI patients would benefit from this diversion. How many of these patients are in the hyper-acute stage where every minute counts? Even if half of these patients presented in this hyper-acute period 5 patients would benefit out of 100,000, is it worth the resources required in restructuring an entire prehospital system?

Though not directly related these thoughts are highly relevant and a excellent example of how inefficient and costly it would be to mobilize a prehospital system to at best benefit so few. It is actions like these, investing exorbitant amounts of money and resources into questionable interventions that has gotten our healthcare system into the crisis state that we are in today. The results from the Menees article is a clear example of an overreaction to assumptive data on a national level. It is a warning rather than a cry to redouble our efforts in this futile attempt at perfection.

Sources Cited:

Menees DS et al. Door-to-balloon time and mortality among patients undergoing primary PCI. N Engl J Med. 2013 Sep 5;369(10):901-9
Bates E, Jacobs A. Time to Treatment in Patients with STEMI. N Engl J Med 2013; 369:889-890
ISIS-2 Collaborative Group. Randomized Trial of Intravenous Streptokinase, Oral Aspirin, Both, or Neither Among 17187 Cases of Suspected Acute Myocardial Infarction: ISIS-2 The Lancet – 13 August 1988 ( Vol. 332, Issue 8607, Pages 349-360 )
The GUSTO IIb Angioplasty Substudy Investigators. A Clinical Trial Comparing Primary Coronary Angioplasty with Tissue Plasminogen Activator for Acute Myocardial Infarction. N Engl J Med 1997; 336:1621-1628 June 5, 1997
Boersma E et al. Does Time Matter? A Pooled Analysis of Randomized Clinical Trials Comparing Percutaneous Coronary Interventions and In-Hospital Fibrinolysis in Acute Myocardial Infarction Patients. The European Heart Journal in 2006 (5) (2006) 27, 779–788)
The STREAM Investigative Team. Fibrinolysis or Primary PCI in ST-Segment Elevation Myocardial Infarction. N Engl J Med 2013; 368:1379-1387 April 11, 2013
1999 Update: ACC/AHA Guidelines for the Management of Patients With Acute Myocardial Infarction: Executive Summary and Recommendations. Circulation. 1999; 100: 1016-1030
ACC/AHA Guidelines for the Management of Patients With ST-Elevation Myocardial Infarction ??Executive Summary: A Report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines (Writing Committee to Revise the 1999 Guidelines for the Management of Patients With Acute Myocardial Infarction. Circulation. 2004;110:588-636
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.
Barrett T, Brywczynski J, Schriger D. Is the Golden Hour Tarnished? Registries and Multivariable Regression: Answers to the March 2010 Journal Club Questions. Annals of Emergency Medicine – August 2010 (Vol. 56, Issue 2, Pages 188-200
Association of door-to-balloon time and mortality in patients admitted to hospital with ST elevation myocardial infarction: national cohort study BMJ 2009
Hoffman JR, Cooper RJ. To divert or not to divert prehospital ST-elevation myocardial infarction: What is the question? CJEM 2009 ;11(5):423-425
The CAPTIM study group. Primary angioplasty versus prehospital fibrinolysis in acute myocardial infarction: a randomised study. Lancet 2002; 360: 825-829