Why is hypoxia not part of the Wells Criteria?

Whether you have done medical or surgical runs you will have spent plenty of time trying to figure out whether a patient has a pulmonary embolism. Many clinicians will hang their hat on the presence or absence of hypoxia. You may then wonder why hypoxia is not actually part of the esteemed Wells criteria.

Well, it turns out the presence of hypoxia in PE is quite variable. In fact, not uncommonly patients with massive PE may have normal oxygen saturations, a phenomenon confirmed both by the literature (1) and my own observations. To understand why, we have to understand why hypoxia might occur in the first place.

It has actually taken a while for people to figure out why hypoxia occurs in PE, although it is still not 100% transparent. Many people assume that the problem is V (ventilation)/Q (perfusion) mismatching, where Q is decreased due to the obstruction. This is not quite correct. V > Q results in hypercarbia, but not hypoxia. The problem is that Q is redistributed to other lung units, without a matching rise in V. This results in regions of lung with low V/Q, away from the embolism (2,3). This seems to be the most likely cause of hypoxia.

This explains why massive pulmonary embolism may not cause hypoxia. Remember that massive pulmonary embolism is defined by the presence of RV strain and cardiogenic shock. If most of the pulmonary arterial tree is obstructed there is nowhere for the blood to be redistributed, minimizing the ability for areas of low V/Q to form. Additionally, if the patient has cardiogenic shock, low cardiac output reduces Q, reducing the inequality. Therefore, paradoxically, improving oxygen saturations may be a sign of worsening shock (4).

Hypoxia is therefore not correlated with the severity of pulmonary embolism. Patients with severe PE may not be hypoxic. If your patient appears shocked, or just looks terrible, you cannot use the absence of hypoxia to rule out PE.

On the other hand, small PEs may also not cause hypoxia, if they are too small for significant redistribution of pulmonary blood flow to occur.

All of this leads onto the next point, which is the utility of ABGs when you suspect PE. No doubt at some point you will have been asked to obtain an ABG in a patient where PE is suspected. The origins of this were some small, flawed studies suggesting that a normal A-a gradient on an ABG could rule out PE in combination with other features. This has been proven false in a more rigorous study (5) – a normal A-a gradient is equally likely in patients with or without PE initially suspected of having PE. This study concluded that ABGs had limited diagnostic value in the investigation of PE. Hopefully now you understand why.

Additionally most of these ABGs are requested on patients in whom it is clearly obvious from the end of the bed that there is an elevated A-a gradient. If you are on 5L of oxygen to maintain normal saturations and there is no clinical reason for hypoventilation, then you will have an elevated A-a gradient.

Till next time…


  1. Intensive Care Medicine. June 1977, Volume 3, Issue 2, pp 77–80| Cite as Massive pulmonary embolism without arterial hypoxaemia Pathophysiology in two cases. F. Jardin, J. Bardet, A. Sanchez, F. Blanchet, J. P. Bourdarias, A. Margairaz.
  2. Pulmonary Circulation. Gas Exchange and Pulmonary Hypertension following Acute Pulmonary Thromboembolism: Has the Emperor Got Some New Clothes Yet? John Y. C. Tsang, James C. Hogg First Published June 1, 2014 Review Article.
  3. Journal of Applied Physiology. Pulmonary embolization causes hypoxemia by redistributing regional blood flow without changing ventilation. William A. Altemeier, H. Thomas Robertson, Steve McKinney, and Robb W. Glenny. 01 DEC 1998https://doi.org/10.1152/jappl.1998.85.6.2337
  4. Hemodynamic Factors Influencing Arterial Hypoxemia in Massive Pulmonary Embolism with Circulatory Failure FRAN(OIS JARDIN, M.D., FRANCIS GURDJIAN, M.D., PIERRE DESFONDS, M.D., JEAN-LUC FOUILLADIEU, M.D., AND ANDRI MARGAIRAZ, M.D. Circulation 59, No. 5, 1979.
  5. Diagnostic Value of Arterial Blood Gas Measurement in Suspected Pulmonary Embolism. MARC A. RODGER , MARC CARRIER , GWYNNE N. JONES , PASTEUR RASULI , FRANÇOIS RAYMOND , HELENE DJUNAEDI , and PHILIP S. WELLS. https://doi.org/10.1164/ajrccm.162.6.2004204   PubMed: 1111212 Received: April 24, 2000

Why shock is not all about the blood pressure


When calling the surgical or gastro registrar on the phone, one of the surefire questions you will be asked is ‘is the patient hemodynamically stable?’ This is no doubt an important question but unfortunately hemodynamic stability means different things to different people. There is generally a lack of willingness to look beyond a ‘normal’ vs ‘low’ blood pressure.

It should be made clear that hypotension is a sign of ‘decompensated’ shock. The idea that hypotension is a late sign of shock has been long recognized in the trauma literature where 18% of penetrating abdominal trauma can have over 750ml of blood in the abdomen despite normal vital signs (including 7% who had over 1500ml). In the majority of patients evidence of tissue hypoperfusion precedes the development of hypotension (1).

Why is this? Well, it is all about the concept that pressure does not equal flow. The body is quite adept at trying to maintain a normal blood pressure through various compensatory mechanism, chief amongst which is the sympathetic response. But this does not tell us whether perfusion at the tissue level is adequate- inadequacy of this is after all the definition of shock. This is especially true in young patients where the compensatory mechanisms are quite strong and blood pressure may be preserved till late in the piece, as illustrated by the graph at the top, of my creation (credit to MS paint).

This phenomenon is true in all situations- not just trauma- so it is relevant to the patients you will be seeing on ward calls. For example ‘normotensive shock’ is recognized in sepsis (2), cardiogenic shock (3), and just generally (4). What then are the signs of ‘normotensive shock’ you should look out for? Well, these are simply the signs of inadequate tissue perfusion- cool and clammy skin, oliguria, mental state changes and elevated lactate. Other signs which may accompany this that are not necessarily indicative of hypoperfusion but that do indicate a compensatory response are tachycardia and tachypnea.

It is also important to note two slightly related things. Firstly, a systolic blood pressure of 110 may be normal for a 20 year old but grossly hypotensive for a 70 year old with chronic hypertension. Secondly, it is sometimes difficult to tell whether a ‘soft’ blood pressure in a young person is just normal for them or whether they are actually hypotensive- a normal heart rate cannot be used to reassure you in this instance because not uncommonly shocked patients may have paradoxically increased parasympathetic tone (1) the exact mechanism for which is unclear. Older patients may also be on Beta Blockers.

The conclusion in all of this- next time you call the gastro reg with a patient who has vomited blood and is clammy with a lactate of 4 but has a normal blood pressure, the answer to the question ‘is the patient hemodynamically stable?’ is a ‘HELL NAH”.

Today’s post may sound basic but it is all about fundamentals- a fundamental which is often simply not done. Rather than spending 10 minutes documenting dual heart sounds, look and feel for the signs of shock!

Till next time.



  1. Identification and Resuscitation of the Trauma Patient in Shock Michael N. Cocchi, MDa , Edward Kimlin, MDa , Mark Walsh, MDb , Michael W. Donnino, MD. Emerg Med Clin N Am 25 (2007) 623–642
  2. Septic Shock. Advances in Diagnosis and Treatment. Christopher W. Seymour, MD, MSc and Matthew R. Rosengart, MD, MPH. JAMA. 2015 Aug 18; 314(7): 708–717.
  3. Menon V et al. Acute myocardial infarction complicated by systemic hypoperfusion without hypotension: Report of the SHOCK trial registry. Am J Med 2000 Apr 1 108 374380
  4. Approach to Hemodynamic Shock and Vasopressors. Stefan Herget-Rosenthal, Fuat Saner and Lakhmir S. Chawla. CJASN March 2008, 3 (2) 546-553; DOI:https://doi.org/10.2215/CJN.01820407

Measuring Blood Pressure

Since hypotension is probably the most common “serious” ward call you will be asked to see, it is somewhat dispiriting that zero time is spent in medical education going over how the tools we use to measure this important parameter actually work.

If you’ve sat in on any tutorial on managing hypotension you will likely have been exposed to the old chestnut that if a patient’s radial pulse is palpable their systolic is at least 80, and if their femoral is palpable then it is 70-80, and if the carotid is palpable it is 60-70, or some other similar variation on these parameters.

How accurate is this seemingly simple bedside guide? Unfortunately not very! This (1) elegant study compared arterial line blood pressure measurement with assessment of pulses (done by blinded assessors). The graph below (reproduced without any permission) shows the expected blood pressures in green shading based on which pulses are palpable (groups 1-4), while the scatter points shows the actual systolic blood pressure recorded. As you can see, the degree of hypotension was severely underestimated by the above rules.


However, while the presence of a radial pulse is not reassuring, it would still appear that the absence of one is probably correlated with severe hypotension. Therefore if your BP cuff is not able to record the blood pressure, and you cannot feel radial pulses, it is not a machine fault! For some reason people stop trusting machines when they give extreme readings.

Speaking of which, it is worth understanding how automatic BP cuffs actually work. The machine is actually an oscillometer, which measures oscillations in the brachial artery that are transmitted through the air filled tubing. The mean arterial pressure (MAP) is determined when the amplitude of the oscillations is maximal. Above and below the MAP, the amplitude of oscillations will decline, and the systolic and diastolic are determined when the amplitude reaches a certain percentage of the maximal amplitude (see the figure below for an illustration). The reason this is important is that the cutoff for systolic/diastolic is determined based on a mathematical formula, so it is estimated in a way (2).



Where this becomes relevant is thinking about how reliable your NIBP is in hypotension. This study (3) looked at correlation between NIBP and invasive arterial line measurement in 150 ICU patients, about half of whom had circulatory failure with hypotension or needing vasopressor drugs. They found that non-invasive measurement of MAP was pretty accurate, especially when the patient was hypotensive. However, non-invasive systolic and diastolic measurements were not accurate, with the systolic pressure often being overestimated.

What about manual blood pressure? The literature provides very divergent evidence, but it seems that in the population that we are interested in (the hypotensive) that automated cuffs overestimate the systolic blood pressure, sometimes to quite alarming degrees, at least if this study of trauma patients is to be believed (4). Manual blood pressure is better correlated to injury severity and markers of shock in this group.

The interesting corollary to this is that the often employed delay tactic by house officers when paged about hypotension, of asking the blood pressure to be repeated manually in the hope that it will give a more encouraging reading, is a waste of time. If the automatic cuff is sized properly, any hypotension should likely be believed.

The conclusion of the second part of this article is basically

  • Either learn to work with MAP rather than SBP, or…
  • If you can’t be stuffed doing the above or find it too difficult, then use a manual to assess the SBP; automatic cuffs are unreliable for this parameter and will overestimate it in the setting of hypotension.

Till next time…


  1. Charles D Deakin and J Lorraine Low. Accuracy of the advanced trauma life support guidelines for predicting systolic blood pressure using carotid, femoral, and radial pulses: observational study. BMJ. 2000 Sep 16; 321(7262): 673–674.
  2. https://lifeinthefastlane.com/ccc/non-invasive-blood-pressure
  3. Lakhal K, Macq C, Ehrmann S, Boulain T, Capdevila X. Noninvasive monitoring of blood pressure in the critically ill: reliability according to the cuff site (arm, thigh, or ankle). Crit Care Med. 2012 Apr;40(4):1207-13. doi: 10.1097/CCM.0b013e31823dae42.
  4. Are automated blood pressure measurements accurate in trauma patients? J Trauma. 2003 Nov;55(5):860-3. Davis JW1, Davis IC, Bennink LD, Bilello JF, Kaups KL, Parks SN.


The Hb in acute bleeding


Imagine you are called to review a patient who has started vomiting blood an hour ago. You arrive to find them tachycardic with a heart rate of 110. Amongst all the other stuff you would do for this patient, you check an Hb. Its 130, unchanged from baseline. What does this tell you about the severity of the bleed?

The answer is nothing at all. Zilch. De nada.

The reason for this is explained brilliantly in the chapter on acute blood loss anemia in the book Clinical Haematology: Theory and Procedures (1) and supported by more recent articles (2).

When blood is lost from vessels, it is both plasma and red cells that are lost in equal numbers. Therefore acutely the haemoglobin concentration will not change. What is responsible for the haemoglobin dropping is shift of fluid from the extravascular space to the intravascular space in response to reduced intravascular volume, thus diluting haemoglobin.  It can take 48 hours for the full effect of this to be seen.

The earliest haematological findings that are seen are actually an increase in the platelet count which can take place in as little as an hour. Soon after a neutrophilia with left shift of white cells develops. The latter can take 2-4 days to resolve.

So in the above example, don’t be reassured by the fact that the Hb is normal, even though many times even your seniors will try and tell you “the patient can’t be bleeding significantly because there is no haemoglobin drop”.  In fact, the presence of a resting tachycardia, as in the above example, indicates fairly significant blood loss, potentially consistent with greater than 750ml if the classification of shock systems are to be believed.

The other corollary to this is that you don’t base the decision to transfuse blood on the Hb level, but rather on the patient’s clinical progress/your assessment. The full ins and outs of this are something that is difficult to discuss in a blog post. Additionally, the presence of thrombocytosis or neutrophilia/left shift can give you clues to the presence of bleeding when you are unsure, as they develop quite early.


  1. Clinical Hematology: Theory and Procedures, Volume 936. Mary Louise Turgeon. Lippincott Williams & Wilkins, 1999. Pages 116-119.
  2. Clinical review: Hemorrhagic shock. Guillermo Gutierrez, David Reines and Marian E Wulf-Gutierrez. Critical Care2004(8):373