Boring Question | Do patients with liver disease need FFP before procedures?

I was recently asked to give fresh frozen plasma to a patient with liver disease and an elevated INR before a therapeutic thoracentesis. He was otherwise healthy and had no history of bleeding diathesis. While I was asking for consent and explaining the risks and benefits of blood product transfusion, I began to wonder about the true benefits in his case. Knowing that liver failure leads to both pro and anti-coagulant deficiencies, I wondered: Is an elevated INR in an individual with hepatic disease really a good indicator of their coagulability?

Do patients with liver disease need FFP before procedures?

The coagulation cascade can be simplified into two competing forces: clot formation and clot breakdown. Within the micro-environment of a vessel wall tear, whichever force is largest will govern whether, and how quickly, a thrombus is generated. So, how does liver disease modify these forces? During my literature search I found a very good review article entitled “The prothrombin time test as a measure of bleeding risk and prognosis in liver disease” [1]. The following discussion owes much to this article.

The “INR” test is a measure, in vitro, of the time to coagulate (formation of a fibrin clot as detected optically) in a patient’s isolated sera that has been seeded with tissue factor. This mimics the process that happens in vivo when a vessel wall is damaged, exposing this factor and setting off the coagulation cascade. However, time to coagulation calculated by an INR only reflects how long it takes for the clot to reach a certain size, and not the speed of its subsequent growth (which is also physiologically important for coagulation). This is where anticoagulant factors (which may also be absent in a patient with liver failure), such as protein C, play a role. They put the ‘breaks’ on clot enlargement after it is formed so their deficiency will not necessarily be reflected with an elevated aPTT or INR.

This problem does not influence the measurement of coagulability in otherwise healthy patients on warfarin (e.g. those with atrial fibrillation). In this setting, coagulability is dependent specifically on the vitamin-K dependant proteins and INR does a good job of indicating the extent of anticoagulation. While vitamin-K factors may also be low in a patient with liver failure, so are other proteins (e.g. protein C) which makes INR a poor reflection of in vivo coagulation. Effectively, both coagulation and anticoagulation can be impaired in liver failure.

Effect of thrombomodulin

Effect of thrombomodulin on clotting in liver cirrhosis [2]

An article entitled “Evidence of Normal Thrombin Generation in Cirrhosis Despite Abnormal Conventional Coagulation Tests” [2] addresses the question of how to measure coagulability in these individuals. It found that thrombin generation was not statistically different between those with cirrhosis and healthy controls once thrombomodulin, a factor essential for protein C function, was added to the assay. Thrombomodulin is found in vivo attached to the endothelium so it is not present in vitro unless added. As you can see from the figure, without thrombomoduliun, cirrhotics (the circles) had much less thrombin generation (ie. clotting activity), whereas when it was added, there was no difference between those with liver disease and healthy controls.

That being the case, coagulability (as measured by thrombin generation) in an in vitro setting that more resembles the in vivo environment (by taking into account protein C) is no different in those with cirrhosis than those without, despite elevated INRs in the former. In addition, numerous observation studies exist that show little relation between INR and bleeding risk in those with liver disease. For example, one study [3] found no relation between INR and severity of bleeding post liver biopsy, a review article [4] of the surgical management of cirrhotic patients found that fresh frozen plasma has not yet been shown to be effective, and another study [5] showed that the addition of fresh frozen plasma to the sera of cirrhotic patients decreased the INR but did not change thrombin generation.


In a patient with cirrhosis, INR is a poor predictor of bleeding risk and the common practice of transfusing fresh frozen plasma before surgery or procedures in otherwise stable patients with liver disease and an elevated INR is questionable.


  1. Tripodi, A., et al. Review article: the prothrombin time test as a measure of bleeding risk and prognosis in liver disease” Alimentary pharmacology & therapeutics. 2007: 2, 141-148.
  2. Tripodi, A, et al. Evidence of normal thrombin generation in cirrhosis despite abnormal conventional coagulation tests. Hepatology, 2005: 3. 553-558.
  3. Ewe, K. Bleeding after liver biopsy does not correlate with indices of peripheral coagulation. Digestive diseases and sciences, 1981: 5. 388-393.
  4. Westerkamp AC, Lisman T, & Porte RJ. How to minimize blood loss during liver surgery in patients with cirrhosis. HPB, 2009: 6, 453-458.
  5. Tripodi, A, et al. Thrombin generation in plasma from patients with cirrhosis supplemented with normal plasma: considerations on the efficacy of treatment with fresh-frozen plasma.” Internal and emergency medicine, 2012: 2, 139-144.

Reviewing with the Staff | Dr. Kerstin de Wit

Dr. de Wit (@kerstinhogg) is an Assistant Professor in the Department of Medicine at McMaster University.  She is cross-appointed between the divisions of Emergency Medicine and Thrombosis. She has reviewed the above piece, and has also written the following.

Bleeding can occur in patients with liver disease for a variety of reasons. The liver synthesis of Factors factors II, V, VI, IX, X, XI, XIII, Protein S, Protein C and fibrinogen can be reduced. Malnutrition is common and vitamin K deficiency results in a further reduction of factors II, VII, IX, X, Protein C and Protein S. Cirrhosis is commonly associated with thrombocytopenia. Furthermore, cirrhotic patients are at risk of bleeding secondary to portal hypertension, in places such as the esophagus where tamponade is challenging to achieve. Falls are common in alcoholic patients and the relative fragility of their tissues is seen in the high prevalence of subdural hematoma in those with minor head injury.

Clotting is not infrequent in cirrhotic patients, with many patients diagnosed with portal vein thrombosis. This venous thrombosis is likely to result from venous stasis in the portal vein, accompanied by reduced presence of our natural anticoagulants Protein C and Protein S.

It is difficult to know whether the bleeding propensity is ‘measureable’, for the reasons stated in this article. However, we should be aware that regardless of coagulation test results, all cirrhotic patients have a high risk of bleeding.

What can we do to mitigate risk of bleeding and improve outcomes in those who present with bleeding?

  • Limiting exposure to anticoagulants and antiplatelet agents (particularly aspirin). Weight appropriate dosing of prophylactic heparin is important. (Patients weighing <50kg should have reduce dose low molecular weight heparin).
  • For bleeding cirrhotic patients we should give IV vitamin K in the emergency department.
  • It is unlikely that fresh frozen plasma is harmful and should be administered in major bleeds.
  • Treatment will also include transfusion of packed red blood cells to a targeted hemoglobin and platelet transfusion to maintain a platelet count >50.

Of course, the ultimate treatment is hemostasis.

This post was peer-reviewed by Dr. Brent Thoma (@Brent_Thoma)

Author information

Michael Garfinkle
Internal Medicine Resident at the University of Saskatchewan and Creator of the LR Database website and Dx Logic app.

The post Boring Question | Do patients with liver disease need FFP before procedures? appeared first on BoringEM and was written by Michael Garfinkle.

Life Beyond Medicine | Humanities and Visual Arts

Do What You Love

Obviously, med school is stressful. It’s especially stressful in third year, when students are starting to reflect on their extra-curricular involvements in the past two years for CaRMs applications and residency. For those who have managed to push out 5 publications, sit in student council, and still managed to go out on a Tuesday night and stumble into a physiology class at 830 on a Wed morning: kudos to them because I don’t know how they did it. For me, one of the most important things I’ve learned in medical school has been this: do what you love, and everything else will fall into place. I hope in sharing my experience, it helps junior medical students and pre-medical students pursue and get involved in their passions, rather than do activities simply to boost their resumes.

Getting involved with med school was a bit different for me. I was never interested in running for student reps, or taking on administrative roles as the class’s treasurer, or even joining intramurals (mostly because I have two left feet, zero hand-eye coordination and my body is built like a brick).

image 1 However, I found my niche in my school’s humanities program. I was lucky enough to attend a school at which humanities were actively encouraged and woven into our curriculum. We had the opportunity to get involved in varied interests ranging from visual arts to theatre to creative writing, and the program hosted frequent events such as art shows and history of medicine lectures to engage students in the humanities.

My partner has a degree in the arts. With my history of being a science nerd, he jokingly mocks the idea of a medical humanities program and specifically my involvement in it. Much to his surprise, I loved the humanities program. My involvement in art shows, presenting at conferences on various humanities projects, putting up my photography for display, and painting pieces representing the medical school experience did not feel like work. I did not feel as though I was just going through the motions of “padding up” my CaRMS application. I enjoyed photography, painting, and writing in humanities and I would’ve enjoyed them in my spare time as hobbies even if it wasn’t for credit. It made me happier. My art pieces became conversation starters for my peers. We bonded over the work I created. I took pride in sharing my passion with others and was encouraged when they decided to get involved too.
I’ve always felt in the past that the visual arts were rarely showcased in school but in medicine (of all places!) I was lucky to be able to incorporate what I love- photography, painting and writing- into my academic experience. For me, combining my hobbies with medicine was a break from endless lectures and classroom work. It was a significant outlet for my stress. I met some wonderful, like-minded people who appreciated the more “artsy” side of medicine as I did, and they, like me, understood how the arts could be used as a medium to portray the unique experiences we have. More importantly, the peers that I worked with understood how important the arts are in my efforts to stay well in medical school.


My passion for the humanities and the arts created new opportunities and opened new doors. I had the chance to present at confernces and visited new cities. My artwork was picked to be on the cover of a humanities magazine and was shared nationally. It was also selected to be on the cover of Academic Medicine, as well as the Dalhousie Medical Journal. I then went on to be a humanities editor for our school journal, and combined my loimage 2ve for the humanities and writing. I had incorporated my love of photography into a mandatory classroom project. As a part of this project, I had the privilege to photograph palliative care patients and share their stories with a visual narrative. The project, titled “When Words Are Not Enough: The Role of Visual Narratives in Palliative Care” was presented at a national humanities conference in Calgary. The aim of the project was to use visual narratives to portray the difficulty of the experiences and complex emotions of a patient nearing the end of his/her life. This work was eventually presented to the Provincial Health Minister on behalf of the palliative care team at the hospital, which helped shape funding and policies in palliative care. More importantly, the photographs were gifted to the families of the palliative care patients after they had passed. I had the chance to develop a special relationship with those patients and their families, and learned another significant lesson in medicine: that often, death is not the end. Their stories will carry on through their loved ones and the lives they have touched.

All that time dedicated to humanities- spending hours on a painting or editing photos, preparing presentations and writing reflective pieces- none of it felt like work. It gave me energy, and it was something I looked forward to doing at the end of the day. Sure, I was a little bit bleary eyed getting up early on a Sunday morning post ER-shift, but once I was photographing my classmates swimming with developmentally challenged children at a local swimming pool as part of our Making Waves Program- I did not mind in the least having to get only 3 hours of sleep the night prior. I wanted to do the things I enjoyed, and I was happy to do them. How simple is that?
One of my friends is in first year medical school, and she asked me for advice on surviving medical school at the start of the year. The only advice I could give her with absolute certainty was to get involved with something she was passionate about. I was lucky enough to find my niche in the humanities, and I don’t think my medical school experience would have been the same without it. Do what you love, and everything will fall into place.

image 3

Author information

Jessie Kang
Jessie Kang
Jessie Kang is a medical student at Dalhousie University.

The post Life Beyond Medicine | Humanities and Visual Arts appeared first on BoringEM and was written by Jessie Kang.

Life Beyond Medicine | Consulting for medical television

The “Life Beyond Medicine” section is supposed to bring a new perspective to our readers about what it means to be a doctor and still have a life outside of medicine.  That said, one of the things that I have had a chance to do recently is work as a consultant on a TV show, and that would be something that I would likely not have been able to do were it NOT for my MD degree. 😀

What I’ve been up to… Consulting for medical television

2014-12-11 20.48.12-1

A selfie on set with my friend Keri!

For the past year I’ve been a consultant for the Canadian drama Remedy, which airs on Global TV in Canada tonight.  If you haven’t tuned in to it in the past, I would highly recommend it (and not merely because of my obvious conflict of interest)! I was a fan of the show well before my consulting gig. I particularly love how it doesn’t JUST focus on doctors and nurses.  Remedy presents a multitude of perspectives by tapping into the lives of the healthcare aides, porters, administrators and other people who make the hospital run.

I landed the job through pure serendipity. One day last year I noticed my friend from high school, Keri Ferencz (featured in the picture with me), posting on Facebook that she was working as a researcher for the show. A quick joke post about how I’d love to help her out anytime resulted in my recruitment. Fast forward a few months and I was sitting in the Season 2 writers room being asked a pile of questions by Greg Spottiswood (the showrunner) and his talented writers.  Since then I have begun suggesting medical content to help craft scenes that enhance the show’s inter-personal drama.

Me. On set!

Me. On set!

Last winter I had the chance to visit the set of Remedy.  The episode that I watched being filmed and rehearsed aired last week, so now I’m allowed to talk about it. First of all I have to confess: IT WAS SUPER COOL TO BE ON SET. The big stars were just walking around in between scenes and running their lines in the hallways of the set. The set itself was super cool. It was on a sound stage / warehouse next door to where Ontario’s provincial emergency response team houses and runs their operations. Funny coincidence, right? Upon arrival, Keri showed me around the set, and I watched a guy getting his make-up on in preparation for a big trauma scene, a scene being filmed (see picture of me viewing the scene), and a rehearsal of the choreography for a trauma scene.

Personally, I find the process fascinating and interesting. As a noted drama geek (at least behind the scenes) throughout high school and university, this is the sort of thing that I’ve always fancied doing!

Worlds Colliding

2014-12-11 20.51.49

The behind the scenes view. Sets look SUPER realistic… till you go backstage!

While on set watching the actors rehearse the trauma scene, I had an epiphany about medical education and simulation. At times I have been skeptical about simulation’s role in medical education. However, during the rehearsals I realized how important rehearsing can be to help economize motion and minimize chaos. At the beginning of my visit, the scene that I watched was a rough and uncoordinated walk through on the steps. The technical consultant (Lindsay, a paramedic) was very insightful and had carefully planned out the blocking of the scene. However, even with her directing the flow, it was difficult for everyone to figure out where they should stand, where they move, where the equipment should be, etc.. After some discussions with her team (and after she kindly allowed me to provide a few insights from my personal experiences in trauma rooms), we moved on.

2014-12-11 22.04.11

When Keri and I came back to see how the rehearsal was progressing I was astonished! The team in front of me had been transformed.  They were now a well-oiled machine that could easily pass for practitioners in any ED trauma suite. The actors now flowed around the room, anticipating everyone else’s movements and their needs…  Effectively, they worked as a high performance team!

This got me really thinking that, just maybe, our own ED teams could be doing more of this deliberate practice (a là Anders Ericsson’s model) by incorporating more in situ simulation. In Hamilton we have been doing this a lot more, but I think finding opportunities to train together repeatedly as a team is a lesson that I will take from my visit to the set of Remedy.

I would like to thank Keri, Greg, and the rest of the Remedy team for welcoming me to their team this past year. More than just a fun job, they’ve given me a lot to think about, both as a doctor and an educator, and I really have appreciated the chance to reflect about my discipline(s) with their guidance.

The season finale of Remedy Season 2 airs tomorrow, but if you’re Canadian you can watch all of the episodes here and catch up!

Edited by Dr. Brent Thoma (@Brent_Thoma)

Author information

Teresa Chan
Managing Editor at BoringEM
Emergency Physician. Medical Educator. #FOAMed Supporter, Producer and Researcher. Assistant Professor, Division of Emergency Medicine, Department of Medicine, McMaster University. + Teresa Chan

The post Life Beyond Medicine | Consulting for medical television appeared first on BoringEM and was written by Teresa Chan.

KT Evidence Bite: Cardioversion and Thromboembolism

Editor’s note: This is a series based on work done by three physicians (Patrick ArchambaultTim Chaplin, and our BoringEM Managing editor Teresa Chan)  for the Canadian National Review Course (NRC). You can read a description of this course here.

The NRC brings EM residents from across the Canada together in their final year for a crash course on everything emergency medicine. Since we are a specialty with heavy allegiance to the tenets of Evidence-Based Medicine, we thought we would serially release the biggest, baddest papers in EM to help the PGY5s in their studying via a spaced-repetition technique. And, since we’re giving this to them, we figured we might as well share those appraisals with the #FOAMed community! We have kept much of the material as drop downs so that you can quiz yourself on the studies.

Paper: The FinCV (Finnish CardioVersion) study of cardioversion of acute atrial fibrillation


Airaksinen KE, GronbergT, NuotioI, et al. The FinCV (Finnish CardioVersion) study. J Am Coll Cardiol. 2013;62(13):1187-1192. PMID: 23850908

Summarized by: Tim Chaplin
Reviewed by: Teresa Chan & Patrick Archambault

Clinical Question

What is the incidence of and risk factors for thromboembolic events after ED cardioversion of acute atrial fibrillation?

Population Adult patients with primary diagnosis of atrial fibrillation who were successfully cardioverted in the ED within 48hrs of atrial fibrillation onset. Atrial fibrillation of cardiac cause.
Comparison factors (observational study)Large number of clinical characteristics including: age, gender, hypertension, heart failure, diabetes, other comorbidities, medications, time to cardioversion, method of cardioversion etc.
OutcomeThromboembolic events (clinically stroke or systemic embolism confirmed by CT or MRI, surgery, or autopsy) within 30 days after cardioversion.


This was a retrospective database analysis of adult (>18yo) patients who presented to 3 large EDs in Finland with a primary diagnosis of acute (<48hrs) atrial fibrillation and who were cardioverted successfully in the ED. These patients did not receive peri-procedural anticoagulation. Baseline characteristics were recorded. A univariate analysis followed by multivariable logistic regression was performed to identify risk factors. 


n= 2481 patients with a total of 5116 successful cardioversions (patients were included multiple times if multiple successful cardioversion occurred within study period) of which 88% were electrical cardioversions.

  • Overall incidence for thromboembolism was 0.7% with events occurring at an average of 2 days post cardioversion

Risk factors for thromboembolism

  • cardioversion > 12 hrs after onset of atrial fibrillation (1.1%) vs <12 hrs after onset( 0.3%)
  • Multivariable regression identified independent factors:
    • >12 hrs from symptom onset
    • advancing age
    • female sex
    • heart failure
    • diabetes


Early cardioversion (<12hrs) is associated with lower thromboembolic events. High risk patients should be considered for peri-procedural and long-term anticoagulation. This is in accordance with the 2010 Europena guidelines.

Take Home Point

1. Earlier cardioversion (<12hrs) may be safer than later (12hrs) cardioversion for patients who present with acute onset atrial fibrillation.

2. Peri-procedural anticoagulation (i.e. with IV heparin) may reduce the risk of thromboembolic events in high risk patients undergoing ED cardioversion of acute atrial fibrillation

EBM Considerations

  • Retrospective study:  This large database review was retrospective which makes the jump from association to causation difficult to make and introduces opportunity for bias including the completeness of collected information.
  • Multiple comparisons: The large number of variables evaluated in the initial univariate analysis puts the study at risk for Type II error. The authors did use a conservative estimate to include variables in the multivariable logistic regression but even with such correction the risk of error related to multiple comparisons is quite high.
  • Identifying onset of atrial fibrillation: The ability of patients to accurately identify the onset of atrial fibrillation is debatable and as such the 12 hour cutoff point is difficult to accurately identify.

For a pdf version of this summary click NRC – BoringEM – FinCV

Author information

Eve Purdy
Medical Student Editor at BoringEM
Fourth year medical student at Queen's University-happily consuming, sharing, creating and researching #FOAMed

The post KT Evidence Bite: Cardioversion and Thromboembolism appeared first on BoringEM and was written by Eve Purdy.

How to Use Likelihood Ratios in Every Day Practice

A large part of an emergency physician’s job is finding and ruling out emergencies. To do that we ask questions, examine them, look at their blood, and take radiographic images. The goal of all of these tests is to rule-in or rule-out disease. However, not all tests are created equal…

As a student, I became frustrated when I was told that such-and-such test was “good” and I should use it. What did good mean? Is a good test always right? I like statistics and percentages, but these tests were presented with a hazy diagnostic foundation that I found unsatisfying. It jarred with the evidence based medicine that I was, in the same breath, being told I should practice. When I discovered the power of likelihood ratios (LRs) things started to make much more sense.

Any discussion of a test’s diagnostic power requires a working understanding of LRs and basic statistics. Ten years ago, LRs required calculations or a Fagan nomogram to be used effectively. Fortunately, there are now tools out there that do the math for you. However, these tools still require some statistical know-how.

Following a brief statistical interlude, this post will teach you how to 1) determine the diagnostic strength of common tests and 2) how to use likelihood ratios in real world situations.

Statistical interlude

To use likelihood ratios effectively you need to understand these 5 points:

1. Every patient that walks through the ED doors with a complaint has a probability of having any number of diseases. This is called the pre-test probability.

2. We perform tests on patients to increase or decrease the probability of them having each possible disease. A test’s ability to increase or decrease the probability of a certain disease is given by the likelihood ratio. Likelihood ratios can be calculated for positive and negative test results using the sensitivity and specificity.

Positive likelihood ratio = Sensitivity / (1 – Specificity)

Negative likelihood ratio = (1 – Sensitivity) / Specificity

Whereas sensitivity and specificity tell us how good a test is when the patient already has (sensitivity) or does not have (specificity) the disease in question, likelihood ratios tell us how much a positive or negative test result affects the likelihood of a disease when we do not know if they have it or not! This is why they are superior to sensitivity and specificity for clinical reasoning (read more about this on a previous BoringEM post).

3. A positive test result usually increases a disease’s post-test probability. The greater the positive likelihood ratio, the greater the post-test likelihood of disease. Positive likelihood ratios range from zero to infinity. A negative test result usually decreases a disease’s post-test probability. The smaller the negative likelihood ratio, the less likely the post-test probability of disease is. Negative likelihood ratios range from zero to one.

4. No test is perfect and whether a test influences our clinical decision making in a particular situation depends on the severity of the disease and the pre-test probability, not necessarily how big or small its likelihood ratios are.

5. Tests can only be used one after one another in specific situations. Whether they can be “added” together depends on whether they are dependent upon one another.

For example, hearing rales on exam increases a persons chance of having a pneumonia, as does egophony. If an individual has both does that mean that we can use the first test’s post-test probability as the second’s pre-test probability? Not necessarily: these variables could be dependent on one another. Most patients with egophony might also have rales (meaning they are basically the same test) which would make it erroneous to add them.

When can we add tests? Often we must rely on our intuition. An x-ray has little to do with a stethoscope finding so we would probably be able to use the post-test probability of hearing rales for pneumonia as the pre-test probability for the test of an x-ray finding a consolidation. In other cases we may have a decision rule that has already correctly amalgamated the tests and accounted for any interdependence with the magic of math. The Heckerling score for pneumonia is such a decision rule.

The problem – and the solution?

This sounds great in principle, but in practice we rarely speak in post-test probabilities and LRs. I think this is because they are hard to remember, there are so many of them, and there was, until recently, no easy way to access them. That is why I created my iOS app, DxLogic, and website, The following examples will demonstrate how you to can use this resource to approach some common medical emergencies in a probabilistic and rational manner. The screenshots found below are from the iOS application.

Example 1

A 86 year old gentleman walks into the ED complaining of recent onset chest pain. Because he’s an ED patient in his late 80’s with a presenting complaint of chest pain he has an approximately 12% chance of having an MI. Thus, his pre-test probability of MI is 12%.

How do I know this?

My app and website contain a database of likelihood ratios and disease prevalences. If you click on ‘prevalences’, select the presenting complaint “chest pain, non-traumatic” and then the > 85 age group you will get the appropriate statistic and its reference in the literature.


12% is a substantial risk for a potentially deadly condition so we will need to do some investigations to rule it out. In this situation we commonly ask about cardiac risk factors (diabetes, hypertension, current smoker, hypercholesterolemia, and family history of premature coronary artery disease in a first-degree relative) while other testing (ECG’s, cardiac enzymes, etc) occurs.

It turns out that he does not have any cardiac risk factors. The resource can now calculate the post-test probability.  Setting the pre-test probability to 12% and selecting that he has no cardiac risk factors in the age > 65 subgroup gives us a final post-test probability of 11%. In other words, the absence of cardiac risk factors in a patient aged > 65 does not change his probability of having an MI! In this context,  asking about cardiac risk factors was not a very good test.


Interestingly you can also see that even the presence of 4 or 5 cardiac risk factors doesn’t increase his probability. In an elderly patient cardiac risk factors really don’t matter.


Example 2

But what if the patient was fifty-six and has no cardiac risk factors? If we were to repeat our steps but for the age 55-64 group (~5% pre-test probability, and -LR any cardiac risk factor ~0.5) we would find that the post-test probability of an MI is now 3%. This is not a complete rule-out but a borderline low probability. I would still feel the need to order additional testing (a normal ECG and troponin would lower his risk even more!), but using this app/website can provide real numbers for a discussion with our patients about their risks. Already we can inform him that their chance of having an MI is about 3% – a solid statistic that will allow him to participate more fully in his ongoing care.


Example 3

Let’s shift gears to demonstrate point #4 from the statistical interlude. A fifty-five year old woman presents with a new cough, runny nose and sore throat. Her grand-daughter has been sick all week with similar symptoms. A family friend recently died from a pulmonary embolism and she is quite concerned that she might be having one. A d-dimer was ordered to alleviate her concerns and came back positive. A CT-PE study was then done which showed a single segmental PE. Case closed, right?

You’re gut is probably questioning this conclusion, and this is why. First, the presentation is atypical for a pulmonary embolus. She’s over 50, so the PERC rule is not negative, but her Well’s score is low, giving her a 6% pre-test probability of pulmonary embolus. Her pre-test probability would probably be even lower because the Well’s score was validated in an ER population suspected of having PE and most physicians would not suspect a PE in this case. However, we will allow this conservative overestimation and use it as our pre-test probability.

CT-PE is a very good rule-in test, but no test is perfect (unless it’s the gold standard) and its average likelihood ratio is 24 (high, but still a far way from infinity). This gives a post-test probability for PE of about 60%. This also means that the chance that this woman has a diagnosis other than PE is 40%! That’s almost 50-50!!

So, before we start her on lifelong anticoagulation, we may consider further testing (e.g. the gold standard, angiography) or getting her close follow up with a specialist to consider her case further. In this case, despite how ‘good’ of a test a CT-PE is, it still leaves us with a lot of uncertainty due to the low pre-test probability in this case.



There are thousands of tests out there. Sometimes it feels like the hospital is a test factory with poor quality control. However, selecting the most useful tests for a given situation can both save time time and avoid unnecessary and potentially harmful tests. I hope this article has made you aware of the important role that likelihood ratios can play in everyday practice. By quantitatively understanding what a test does to a patient’s diagnostic probabilities we can relay information to patients to help them confidently participate in the decision making process.

While you are investigating the wonderful world of likelihood ratios, please consider giving my app and website a try (and tell your friends!). I am always open to feedback so if you have any on the app or this article, please contact me!

Reviewing with the Staff | Dr. Heather Murray

Dr. Murray is an Associate Professor of Emergency Medicine at Queen’s, and holds a cross-appointment in the Department of Community Health and Epidemiology.

Likelihood ratios (LRs) are not intrinsically easy to understand, but they are critical for correctly interpreting test results. Most clinicians understand the concept of sensitivity and specificity, at least well enough to know that highly sensitive tests rarely miss disease (and effectively rule out disease) while highly specific tests are rarely falsely positive (and effectively rule in disease). However the problem with sensitivity and specificity is that they are properties of the test. So, while knowing these test properties can help us select the right test to perform, they do not tell us what to do once we have a result.

Many studies and clinicians talk about positive and negative predictive values, or the chance that the test is correct. There are problems with this. First, PPV and NPV change as the prevalence of disease changes. For example, as fewer people in your study population have the disease (decreasing prevalence), the negative predictive value of the test will go up. If we hear that a test has a 99% NPV, that sounds good, but this is only helpful if you know the pretest probability of the disease in your patient. If your patient’s pretest probability of disease is higher than the population studied, their NPV (or risk of a false negative test) will be lower. This is why I discourage students and residents from talking about the PPV and NPV – they are a moving target with an unclear application.

The likelihood ratio is the best option for interpreting test results. When you are faced with a positive test result, the real question is whether your test is a true positive or a false positive. If the positive LR of the test is 9, that means that there are 9 true positive test results for every 1 false positive – pretty good! But a positive LR of 2 means that there are only 2 true positives for every 1 false positive – not as reassuring. The positive likelihood ratio is literally the ratio of the true positive results to the false positive results in a given test, and it is stable (unaffected by pretest probability).

The negative likelihood ratio is a bit trickier since you are starting with a negative premise, and end up with a number less than 1 to interpret. The negative LR asks about the ratio of false negative tests to true negative tests (or “is this negative test messing with me?”). Ideally this should be very small. So a negative LR of 0.1 has 1 false negative for every 10 true negatives (or 1/10) while a negative LR of 0.5 has 1 false negative for every 2 true negatives (1/2).

The beauty of these is not only the stability and more intuitive application (once you understand them), but the mathematical ability to combine them with pre-test probability estimates for a concrete post-test probability, as Michael Garfinkle has shown here with his post, website, and app. Try them out and embrace the likelihood ratio as the powerful practical tool that it is.

This piece was edited Dr. Brent Thoma

Author information

Michael Garfinkle
Michael Garfinkle
Internal Medicine Resident at the University of Saskatchewan

The post How to Use Likelihood Ratios in Every Day Practice appeared first on BoringEM and was written by Michael Garfinkle.