Opening case: An unexpected outcome
Once upon a time there was a patient with cardiogenic shock due to a myocardial infarction. Although normotensive, he wasn't perfusing his kidneys. His echocardiogram showed all of the expected features of cardiogenic shock, including a severely reduced ejection fraction and dilated inferior vena cava. He had a Swan-Ganz catheter which showed very low cardiac output, high filling pressures, and high systemic vascular resistance.
The patient's Swan numbers and echocardiogram suggested that he should respond well to milrinone, a drug that ought to increase his ejection fraction while simultaneously reducing his afterload. This would shift his physiology closer to normal.
However, a while after starting the milrinone the patient's blood pressure plummeted to a systolic blood pressure in the 60s. A high-dose epinephrine infusion was needed to raise his blood pressure until the milrinone wore off. What went wrong?
With milrinone, his systemic vascular resistance decreased but his cardiac output remained unchanged, causing his blood pressure to fall. The patient's cardiac output was not milrinone responsive. His ventricle was already working at maximal capacity, in the setting of a massive infarction. The echo and swan data provided the illusion that we understood his hemodynamics, but failed to predict how he would respond to milrinone. They caused us to be wrong with confidence, the most dangerous combination in medicine (1).
Introduction
Use of Swan-Ganz catheterization in heart failure is increasing, despite evidence-based recommendations to the contrary (Pandey 2016). This trend may reflect increased utilization of advanced therapies in heart failure (e.g., LVADs). With a resurgence of the Swan, a resident recently asked me: why don't we use the Swan to guide sepsis resuscitation? Answering this question forced me to recognize that many problems with the Swan continue to haunt us today when using our new darling, bedside echocardiography.
First reason we fail: The flaw of averages
Let's consider a simple question: Does dobutamine increase cardiac index in septic shock? There have been about a dozen prospective RCTs investigating this, with a meta-analysis showing that dobutamine increases cardiac index (Nadeem 2015). This is also what every textbook and review article says. Seems simple, right?
Closer reading of the hemodynamic literature yields a more complex answer. For example, Enrico 2012 published data from individual patients with septic shock treated with dobutamine (below). In some patients dobutamine increased the cardiac output, but in many patients it had little effect:
Unfortunately, most research on hemodynamics focuses on average patient data, ignoring this heterogeneity. Average data can be misleading. For example, the average data from this same study by Enrico 2012 shows that dobutamine increases cardiac index (red box below). Without seeing the individual patient data, there is a natural tendency to assume that dobutamine increases the cardiac output in all patients.
Many other studies have also found striking heterogeneity in the responses of septic patients to dobutamine. For example, Kumar 2007 found that patients who survived were more likely to experience an increased stroke volume in response to dobutamine (adjacent figure).
The flaw of averages refers to inappropriate assumptions that the average applies to all cases. This gets us into trouble clinically, because it deludes us into believing that we can accurately predict the effect of a drug. For example, in the opening case, we were sure that milrinone would increase the patient's cardiac output because this is the average response to milrinone.
Second reason we fail: The Swan's Curse
The swan provides only static variables
The swan attempts to give us a snapshot of the patient's hemodynamics. Although this contains a lot of information, it isn't the information that we need. The Swan provides only static hemodynamic measurements, which don't actually help us determine how to treat the patient. The following two examples illustrate this problem.
Example #1: The Swan doesn't help guide fluid management
Historically, the first variable of interest after a Swan was placed was the pulmonary capillary wedge pressure. Ideally, the wedge pressure should reflect the preload on the left ventricle (2). Thus, the wedge pressure was used as a tool to guide fluid resuscitation.
Unfortunately, this doesn't work. Technical problems aside, filling pressures (e.g. wedge pressure, central venous pressure) cannot guide fluid resuscitation because they are static variables which don't measure fluid responsiveness.
Whether the patient responds to fluid depends on interactions between the preload and cardiac function. For example, a patient with severe heart failure might reach a maximal cardiac output rapidly, even with a low pulmonary capillary wedge pressure (point B). Alternatively, a patient with good cardiac function might be fluid-responsive at exactly the same wedge pressure (point A). Thus, the wedge pressure cannot predict fluid responsiveness.
Example #2: The Swan doesn't help guide inotropic therapy
The other proposed use of a Swan in septic shock is to identify patients with low cardiac output who would benefit from inotropic therapy (e.g. dobutamine)(3). Unfortunately, this doesn't work either.
Vallet 1993 evaluated the response of 50 patients with septic shock to an infusion of 10 mcg/min/kg dobutamine. Only about half of the patients responded well, with improved oxygen consumption and cardiac output (5). The Swan hemodynamic profiles of responders and nonresponders were identical (table). Thus, a Swan cannot predict which patients will benefit from dobutamine. This finding was confirmed by Rhodes 1999.
Inability of cardiac output to predict dobutamine responsiveness may be due to many factors. However, a simple illustration of how this could occur is as follows. Lets again consider two septic patients, one of whom has a cardiomyopathy. To compensate for the cardiomyopathy, Patient #2's endogenous sympathetic activity is increased (so Patient #2 starts out at point B, prior to administering any dobutamine). These two patients have the same baseline cardiac output, but only Patient #1 will respond well to exogenous dobutamine.
Could the Swan be saved by serial measurements?
Theoretically, by frequently repeating Swan numbers, we could get a better sense of the responsiveness of the hemodynamic system to various interventions. Unfortunately, in usual clinical practice the Swan isn't precise enough to reliably detect changes in serial hemodynamics (Marik 2013). Meanwhilele, when scrutinizing differences between serial sets of Swan data, it's easy to be misled by random variations.
The Swan's curse
The Swan's curse is that even if we know intellectually that static variables don't predict responsiveness, it is exceedingly difficult to ignore them. We see a wedge pressure of 5 mm in a patient with sepsis, and even though we know that this doesn't predict fluid responsiveness, we still want to give fluids. Although wrong, this response is intuitive and nearly irresistible. This explains why it is probably wisest not to measure a wedge pressure (or a CVP) in the first place.
Bedside echocardiography: The curse continues
Diagnosis vs. hemodynamic monitoring
Currently, the Swan has largely been replaced by echocardiography for hemodynamic evaluation. The diagnostic capacity of echocardiography is vastly superior to the swan. For example, echocardiography can immediately and definitively identify a range of hemodynamic diagnoses (e.g. RV failure, tamponade, hypovolemic shock, valvular regurgitation).
Echocardiography is also being used to guide resuscitation, as a hemodynamic monitor. Although this shows considerable promise, it also suffers from many of the same pitfalls that the Swan did. Most findings on echocardiography are, once again, static variables.
IVC diameter is the new CVP
Perhaps the most notable example is the IVC diameter. Some clinicians who scoff at using CVP to guide sepsis resuscitation will nonetheless infuse crystalloid until the IVC is distended. However, IVC diameter is largely a reflection of CVP:
A common error is assuming that a small IVC indicates volume depletion. IVC size and CVP are not measurements of volume status, but rather they are determined by the interaction of cardiac function and volume status (e.g. a small IVC can be caused by vasodilation). Thus, assuming that a patient with a small IVC needs fluid is wrong for two reasons:
- Small IVC doesn't necessarily mean the patient is volume depleted
- Like CVP, IVC size doesn't predict volume responsiveness.
Unfortunately, an underfilled IVC is visually striking, it's sitting right there, and it looks empty. Regardless of our intellectual understanding of hemodynamics, we feel compelled to fill the IVC. The curse continues (4).
Ejection fraction is the new cardiac index
Ejection fraction may have similar problems compared to the cardiac index. For example, remember the patient above with cardiogenic shock who was nearly killed by milrinone? That patient had a severely reduced ejection fraction, which could suggest a benefit from milrinone. Unfortunately, baseline echocardiography fails to reveal the responsiveness of the ejection fraction to milrinone. Thus, newer technology may generate the same errors that were previously encountered with the Swan (5).
Responsiveness: The cure for the Swan's curse
The cure for static variables is understanding dynamic variables. Considerable progress has been made in the past few years in understanding fluid responsiveness (explored in an epic podcast by Scott Weingart).
In addition to fluid responsiveness, we may need to pay more attention to vasopressor responsiveness and inotrope responsiveness. The response of individual patients is more heterogeneous and unpredictable than suggested by textbooks. Therefore, rather than assuming that the patient will respond in an average way, we may need to carefully administer a drug and monitor the individual patient's response.
To be continued… this is part three of a four-part series on hemodynamics.
- Although echocardiography is safer than Swan-Ganz catheterization, it is susceptible to some of the same cognitive errors.
- Our understanding of how medications affect hemodynamics is based on the average response of groups of patients. Unfortunately, it may be incorrect to assume that an individual patient will respond similarly to the average (“Flaw of averages”).
- Most variables obtained from either Swan-Ganz catheterization or bedside echocardiography are static variables. These describe the current state of the patient's hemodynamics, but don't predict how the patient will respond to various interventions. Unfortunately, there is a natural tendency to base therapeutic decisions on static variables which are easily observable (“Swan's Curse”).
Must-read article:
- Best article ever written about hemodynamic monitoring: Alice in Intensiveland.
Related posts
- Fluid responsiveness (EMCrit)
- Empty IVC + hyperkinetic heart doesn't equal volume depletion (PulmCrit)
- IVC for decisions on fluid status (EMCrit)
- Current series on hemodynamics
Notes
- One key factor that made this dangerous is that milrinone is renally cleared, so that it may accumulate over time. Thus, it may initially seem that the patient is doing fine, but over time the drug level may build up and cause problems.
- There are about a dozen potentially confounding variables, which can cause the wedge pressure to be inaccurate. However, for now let's ignore these and suppose that the wedge pressure can actually measure the preload on the LV.
- Epinephrine is my inotrope of choice in sepsis for reasons which have been explored previously. Dobutamine is used here because there is more data about its effect on hemodynamics. More on epinephrine next week.
- Please note that this discussion pertains to IVC size, not IVC variability. IVC variability is an entirely different topic which has been explored previously here.
- This study defined “responders” vs. “non-responders” based on whether there was an increase in oxygen consumption (VO2). However, VO2 was closely correlated with DO2 and cardiac output, such that patients who were “dobutamine responders” were predominantly patients who experienced significant increases in cardiac output.
Image credit: Black Swan from Wikipedia.
- PulmCrit Wee – Loading infusion auto-titration (LIAT) for infused medications with intermediate half-lives - March 23, 2025
- PulmCrit Wee: Michelin Chest Syndrome - March 15, 2025
- PulmCrit: ADAPT and SCREEN trials are full of sound and fury, signifying little - December 13, 2024
Once again a very interesting and thought provoking post. Was mechanical support considered/used in the treatment of your patient with cardiogenic shock? Milrinone often works too well as an inodilator and a little goes a long way. I find that most times a vasopressor either norepinephrine or vasopressin is required to maintain SVR/MAP after initiating Milrinone in cardiac surgery patients who start out with high SVR and are normotensive. Some centers bolus milrinone which can contribute to profound vasodilatation. As long as cardiac output is maintained, reasonably dosed vasopressors to maintain MAP shouldn’t harm end organ perfusion. Trending downstream markers… Read more »
Must admit that I’m not a big milrinone fan, largely due to its long half-life which makes it difficult to shut if off it things start going sideways. In renal failure this is a greater problem. You could combine milrinone with an agent to maintain SVR (i.e. norepinephrine), and some of these patient may be on norepinephrine allready. My preference is often to use low-dose epinephrine as a inotrope, because this is simpler. Of course, as you point out, in heart failure patients mechanical support may theoretically be better (with correlative data suggesting that inotropes track with poor outcomes). I… Read more »
Wonderful post, as always. Seems the take away message (as always) is to err on the side of caution with anything. “Give a little, stop, watch, repeat or change as necessary”
Yep. Sometimes a careful trial-and-error approach is better than going nuts with hemodynamic monitoring and trying to predict the future.
First principles of patient care in the CVICU: what is the gradient of my patient’s Starling curve, and where is my patient in relation to that gradient? Perfusion is the goal. Numbers say what numbers say, but there’s no substitute for twelve hours of ongoing assessment, making small changes and assessing the patient’s response, all in real time, listening always to the small voice in the back of my mind. I wonder what dose the milrinone was initiated at, what titration orders the patient had, what therapies were concurrent with that medication administration, etc. I, too, find that primacor requires… Read more »
Well said. There is no substitute for meticulous ongoing assessment. This entire series on hemodynamics was motivated by observations made by ICU nurses that a patient would respond well to one drug but poorly to another drug. For a long time I tried to figure out how to predict whether a given patient would respond to various medications. Ultimately I came to the conclusion that this was impossible (or, at least, too difficult for me to figure out). Agree about the problems with chemically flogging a failing heart in cardiogenic shock. Although we have done this for years, I’m not… Read more »
Oh, I’m unaware of any negative data regarding the use of IABP in the setting of cardiogenic shock. I read where it was not shown to have significant benefit when used with certain patient populations, but negative? As in significantly increasing mortality or morbidity? SHOCK 2 didn’t show significant negative data. What new findings have I missed, or perhaps have misinterpreted? Stepping away from the posted patient, I certainly don’t see data supporting the use of inodilators alone in the setting of cardiogenic shock. Ahem. Regarding drip selection, I’ve used low-dose epi (0.02-0.05 mcg/kg/min) with nitro to get the primacor… Read more »
Regarding the IABP, none of the studies about it have been positive. Some meta-analyses have suggested that IABP use correlated with increased bleeding. There is a risk of various procedural complications including impaired renal perfusion if it is incorrectly located and also infection, ischemia, etc. From a purely “evidence-based” standpoint this is a hard intervention to justify. Maybe there are some individual, highly selected patients who might benefit. Regarding your last question, I think the concept that doctors have 100% control is unrealistic and potentially a bad goal. In reality, the entire ICU team has some control over the patient,… Read more »
I don’t really know what else we could do with our severe takotsubo cardiomyopathy patients if not IABP. Maybe Impella? But I’ve seen more coagulopathies recently, maybe related to the anticoagulant requirements of the PVAD? I don’t know. We don’t leave the balloons in like we once did, that’s for sure. Now we only leave as a bit of insurance for the proximal LAD (mostly superstition, they tell me) or tight LM on the way to the OR. It’s still indicated for severe left ventricular obstruction but the bread-and-butter of using IABP for two and three days on our STEMI… Read more »
New age hemodynamics- first try to get DO2 to a low-normal value which includes targeting hb and CO, but then the real question is how to augment CO: dynamic measures trump static ones so PP variation, passive leg raise etc reign high, though they have requirements (ie A-line, good windows, sinus rhythm, passive etc.). FALLS protocol by Daniel Lichtenstein says A profile means pt can tolerate more fluids! IVC change with breathing helpful esp if passive with vent, and at extremes. Keep in mind that venous return is just as important and this requires an understanding of what determines mean… Read more »
As much as I love Daniel Lichtenstein, I disagree with the FALLS protocol. For those not familiar with this, the concept is essentially giving a patient fluid until they start to develop early pulmonary edema (as measured by B-lines on lung ultrasonography). More about it here: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3848672/ Classically, the greatest fear about giving a patient too much fluid has been the development of pulmonary edema. Although this remains a major concern, volume overload can hurt patients in a variety of other ways (e.g. renal intra-capsular edema impairing renal perfusion, abdominal compartment syndrome, increased IVC causing reduced perfusion pressure, gut edema,… Read more »
Static numbers were never the way the Swan was to be used. A static wedge pressure is relatively meaningless; but from the wedge response to a fluid challenge, something can be learned. Cardiac output increase after fluid challenge is monitoring the same thing as the Cheetah Nicom, and I believe how the technology was verified and tested. It is interesting that the Swan never died in cardiac surgery. In the hands of experts, both physicians and nursing staff, it still provides very useful data. Also, there is some value in static numbers… for example, the much despised CVP is useful… Read more »
Many patients are coming back from the OR without swans these days even after surgery to repair diseased valves, though I can’t disagree with the intent of your statement. Sure, an optimally-calibrated system in the hands of an expert only errs sometimes, and can provide useful data. But how often do the planets align in such a manner, and when they are, how often are decisions made that positively affect the patient’s course/recovery in a significant way, decreasing length of stay, decreasing mortality and morbidity, justifying the risks of its presence? Not often enough, all things considered, although–I do like… Read more »
I would argue that the Swan was never really tested for mortality in a study where the doctors and nurses were expert in its use. No mortality benefit was shown probably because of too wide of spread of use and lack of knowledge. Anyway, has anyone seen the article that showed gauging fluid responsiveness by ETCO2 respsonse? This is preliminary, but very interesting!
The PA cath doesn’t only provide you with static numbers. You can measure the change in stroke volume before and after a passive leg raise and that is likely more accurate than all the non-invasive measurements or surrogates of cardiac output that we do.