Making Sense of Apples-to-Oranges Comparisons
K. Craig Kent, MD, describes the advantages and disadvantages of each of the available datasets on endovascular repair in high-risk patients, including what has been learned from the extensive Medicare database.
Endovascular Today: How would you characterize the current debate over how best to define and manage high-risk aneurysm patients?
Dr. Kent: It is a complicated issue that we have not adequately answered despite multiple attempts. During the very robust, interactive session about this issue at the Society for Vascular Surgery Annual Meeting, a variety of different types of data were presented. It was a wonderful session because rather than all of the different groups working in isolation, we had an opportunity (on the podium mind you) for people to really interact and share their results and experiences. I felt very certain that by the end of the session, we all actually came to the same conclusions. We were able to interpret everyone's data in the same way.
Two of the conclusions that all of the participants came to at the end of the SVS session were that there is a high-risk group, and that perhaps some patients who are currently receiving endovascular repair should not be. This group is likely small, and the task still not accomplished is to be able to identify who these patients are preoperatively.
Endovascular Today: What are some of the concerns regarding the available data?
Dr. Kent: Before EVAR-2, many clinicians seemed to believe that endovascular repair was so minimally invasive that it could be performed with very little risk in any patient. Consequently, it may have been liberally used in patients who did not have tremendous longevity. Since its publication, EVAR-2 has shown us that there are some patients who die with this procedure, and there are some who do not benefit from it even if they survive the initial procedure because their overall longevity related to their comorbidities is diminished. The work done by the EVAR-2 investigators is compelling and groundbreaking, but because of certain issues with the trial's design, it is also difficult to compare it to many of the practices and clinical study trends we see in the US.
Endovascular Today: In what ways does EVAR-2 differ from AAA repair in the US?
Dr. Kent: When EVAR-2 was first presented and we got word of its high mortality rate, many people began to think that there was something different about how its high-risk patients were defined versus how this distinction is made in the US. The EVAR-2 study had a mortality rate of 9%, and clearly this is much higher than we have seen in US studies. After looking at the EVAR-2 data more closely, it becomes evident that some of these patients were extraordinarily ill, and were patients who, in our standard of practice, would not have been considered candidates for aneurysm repair in the first place.
The method of randomization used in EVAR-2 also adds to the difficulty in comparing its data to those observed and/or expected in the US. In EVAR-2, one of the randomization adjudicators was the surgeon who determined that the patient was at high risk-in essence, the "eyeball" test. The eyeball test can be a good test; those of us who have been practicing for years know that sometimes just looking at a patient can tell you if they will or will not fare well during or after a procedure. But when the intent is to provide tangible criteria that can be translated to other physicians that define which patients should or should not undergo endovascular repair, the eyeball test is not particularly helpful.
Another concern is that some of the very ill patients included in the trial died awaiting the procedure to which they were randomized, impacting dramatically upon EVAR-2's mortality rates. Some died of their underlying disease, which indicates that they should probably never have had a procedure in the first place. Others died of a ruptured aneurysm, which means that once the decision was made to randomize them, they should have undergone a procedure immediately. Using an intention-to-treat approach, which is appropriate in a study such as this, these patients are included in the deaths in the endovascular repair group, significantly contributing to the 9% mortality rate.
Endovascular Today: What has been learned from the Medicare database, which you presented during the SVS panel?
Dr. Kent: The Medicare database is, essentially, a logistic regression analysis of 48,000 patients who had an overall 30 day mortality rate of less than 2%. A few factors were determined as strong predictors of mortality, one of which was renal failure. Patients who have renal dysfunction are at much higher risk of dying than patients who do not. In fact, in our analysis, the 30-day mortality rate for patients with renal dysfunction is 7%. However, in the Medicare database, the specific level of renal failure cannot be determined; it could be anything from a creatinine of 1.5 to dialysis.
Some of the more obvious risk factors, such as pulmonary disease, congestive heart failure, myocardial infarction, etc., had mortality rates of 3% to 5%. One of the tasks that we are trying to accomplish now, and we have not yet published our data, is to create a scoring system. The goal of the scoring system would be to allow a physician to preoperatively evaluate a patient for the likelihood of death with endovascular repair. Endovascular repair is no different than any other surgical procedure-there is an associated perioperative mortality. The real issue is to try to figure out exactly which patients are at high risk for death.
Another part of the equation is how long the patient is expected to live. If a patient has congestive heart failure or renal failure, his longevity is not as robust a patient who does not have these disease processes. When repairing aneurysms prophylactically, the challenge is to figure out whether they will die of their natural disease process first, or of rupture of their aneurysm. Ironically, it is likely that those same patients who have a high perioperative risk for endovascular repair also probably do not have great longevity.
When you put the two together, you realize that there are in fact some patients in the high-risk group who are not worth treating because (1) they may die during the procedure, or (2) even if they make it through the procedure, they could very likely end up dying of their natural disease process before they would have died of their ruptured aneurysm. It is the combination of those two elements that will help us predict who will or will not benefit from EVAR. This is not an easy analysis to do. Whether it be the Cleveland Clinic data, our Medicare analysis, EVAR-2, or the US IDE data, we do not currently have sufficient long-term data about these patients to fully understand who should not have the procedure.
So far in our analysis, the renal failure patients seem to be at highest risk. This does not mean that you would never place an endograft in a renal failure patient, but perhaps it does mean that if you are considering endografting a patient with renal failure, you should look closely at the overall picture and attempt to predict longevity as well as perioperative risk. Patients still need to be individualized, but will a dialysis patient with myocardial and lung disease and a 5.5-cm AAA real benefit from intervention?
Endovascular Today: What are the limiting factors that need to be considered when evaluating the available datasets in order to determine which patients are at high risk?
Dr. Kent: Each of these datasets has its advantages and limitations. I have already mentioned some of the concerns with EVAR-2, which I must emphasize is a very valuable dataset that has caused all of us to look much closer at our treatment of high-risk patients.
The Cleveland Clinic dataset is from a single, high-profile institution that has an extraordinary amount of experience with endovascular repair and a selected referral population. We can certainly gain information from their data, but this is a select group of patients.
With regard to the US IDE dataset, there is little doubt that these patients are also carefully selected. There were fairly stringent criteria for patients being accepted into these trials, and although there is a group of patients who were at high risk but were still entered, the selection criteria kept the extraordinarily ill patients from being included. The sponsoring companies did not want patients to die during these procedures.
One advantage of the Medicare database is that the number of patients studied is extraordinarily large. Another advantage is that this data set is that it is truly reflective of what is happening in the US. A disadvantage of the Medicare database is that many of the variables that we analyzed are not well defined. For example, there are codes for congestive heart failure, but we cant tell if a patient has an ejection fraction of 45% or 22%. There is a code for renal failure, but we don't know if the creatinine is 1.5 or 4.8. The other limitation of the Medicare dataset is there is no anatomic information about these patients.
When comparing these studies, it really comes down to how we define "high risk." If we use traditional definitions, the mortality rate even in "high risk" patients remains low, around 2%. If patients who are very ill are enrolled, the mortality rate will naturally increase. It is important that we consider all of the factors that go into a study's design and its definition of "high-risk" before trying to make definitive comparisons.
Endovascular Today: How will the US IDE, Medicare, Cleveland Clinic, and EVAR-2 data impact the standard of care for high-risk patients?
Dr. Kent: They will absolutely impact the standard of care. These studies have eliminated the myth that endovascular repair is safe in all patients, but we must also keep in mind that there isn't need for alarm; many high-risk patients do well with endovascular repair and very much benefit from this procedure.
K. Craig Kent, MD, is Chief, Combined Columbia and Cornell Division of Vascular Surgery, New York-Presbyterian Hospital, New York, New York. He may be reached at firstname.lastname@example.org.