Hypoxia in Pulmonary Embolus

mechanism hypoxia featured

Hypoxia is almost ubiquitous in PE (In the comments Liam suggests otherwise. I honestly can’t find a clear answer. in PEITHO, a trial of submassive PE, 85% were receiving oxygen suggesting hypoxia. Certainly in the tiny PEs, lots of people don’t have hypoxia. If you have a better answer let us know in the comments). Yet it is not immediately clear why. You might think you know but certainly when I start to think about it too much it all becomes very muddy. This is mainly due to my poor understanding of respiratory physiology no doubt. I’ve tried to correct that somewhat with this post.

My basic thinking has always been. A PE is a big clot in the lung, this means part of the lung doesn’t work right, ipso facto then there must be hypoxia. That gets you through day to day existence in emergency medicine but it’s hardly a detailed description of the problem.

The terminology doesn’t really help here as V/Q mismatch technically seems to mean that there is an imbalance ventilation to perfusion. In the context of PE people state that the hypoxia is due to V/Q mismatch but don’t clearly state if it’s a high or low V/Q state.

The terms shunt and venous admixture are also used with some frequency which has a tendency to confuse the anatomist in me where the I can tell where the superior oblique is and what direction it runs in purely from the name…

I started with Rosen’s 8th Edition. The PE chapter is written of course by Jeff Kline.

A lodged clot can redistribute blood flow to areas of the lung with already high perfusion relative to ventilation and therefore cause more blue blood to pass through the lung without being fully oxygenated. This venous admixture is probably the primary cause of hypoxemia with PE and the increased alveolar-arterial oxygen difference

This actually is a pretty decent explanation of what seems to be happening. PE causes redistribution and instead of the left lung getting 50% of the cardiac output it suddenly gets 80% and the ventilation isn’t sufficient to oxygenate the blood.

But  figured I’d do a bit of further reading just to see what else is out there.

Paper #1

Huet Y, Lemaire F, Brun-Buisson C. Hypoxemia in acute pulmonary embolism. Chest. 88(6):829-36. 1985. [pubmed]


These guys studied a whopping 7 people,  all 1-9 days post PE, but 2 hrs post their formal angio that was used to diagnose it. Most had greater than 50% pulmonary vascular occlusion and most had some CXR changes too (something we don’t see very commonly). They all had Swan-Ganz catheters placed hence all the lovely data they got to play with.

They were all on heparin and were given urokinase after their “gas exchange test”. This involved infusing a bunch of inert gases dissolved in dextrose and measuring lots of things.

They suggest that the hypoxia had different causes depending on the patient and interestingly could be related to their CXR changes.

  • if they had atelectasis on CXR then there hypoxia was from shunt.
  • if the CXR was normal it was due to perfusion of lung units with low V/Q ratios (ie overperfusing a lung unit with no increase in ventilation)


In PE there is a degree of shifting of ventilation away from and around the underperfused lung, this presumably, is the lung autoregulating itself. This shifting is not particularly well done and after a while atelectasis occurs and as a result of that you now have shunt as an additional cause of hypoxia.

They conclude from their data that initial hypoxia is due to V/Q mismatch (in particular ,perfusion of lung units with low V/Q ratios) and later in the disease course it is likely shunting.

Paper #2

Burton GH, Seed WA, Vernon P. Observations on the mechanism of hypoxaemia in acute minor pulmonary embolism. British medical journal (Clinical research ed.). 289(6440):276-9. 1984. [pubmed]

These were probable PEs, all diagnosed on V/Q scans. All were tachy and breathless with normal CXRs and patients were identified through chart review. Mostly post op. They describe them as minor PE but I suspect they were all quite impressive and may well be termed “submassive” in these days of right heart strain and trops…

Most were a week post symptoms and the ABG was taken just after the diagnostic V/Q scan.

They garnered a huge 11 pts. The V/Q all showed reduced perfusion in areas well ventilated (which seems to be the definition of PE on a V/Q scan) and lots of other distal areas that were overperfused comared to how well ventilated they were. The more severe the V/Q scan changes the more severe the ABG abnormalities.

One of the big issues here is that they assume the cardiac output was normal or raised i their calculations but they don;t actually measure it. The prior study did measure CO in their patients and that it was reduced in all patients.

This doesn’t really help much in working out what’s going on to be honest

Paper #3

D’Alonzo GE, Bower JS, DeHart P, Dantzker DR. The mechanisms of abnormal gas exchange in acute massive pulmonary embolism. The American review of respiratory disease. 128(1):170-2. 1983. [pubmed]

I had to work off the abstract here as I couldn’t get full access. They studied two patients here, both with “massive PE” (they don’t provide a definition in the abstract) using the inert gas method and found that shunt was the main issue, not V/Q mismatch . They state that 20% and 39% of blood flow was through unventilated lung. They conclude that shunt is the main issue.

Paper #4

D’Angelo E. Lung mechanics and gas exchange in pulmonary embolism. Haematologica. 82(3):371-4. 1997. [pubmed]

This a review article written by one of the authots above. It’s the best I’ve found and is open access too. Bottom line: shunt and V/Q mismatch are the main causes of hypoxia. (are there any other options???)

It does try and explain why V/Q mismatch causes low O2 – apparently due to the sigmoidal shape of the O2 curve increases in ventilation cannot keep pace with either increased or decreased perfusion. Note this does not apply to CO2 as its curve is linear. This did ring a bell from my ICU reading years ago…

It highlights that some reports have noted bronchoconstriction and airway narrowing is part of PE – though perhaps this is clinically silent for most patients as I can’t say I’ve heard a great deal of wheezing in there.

The atelectasis that occurs (with resulting shunt) could be down to pneumoconstriction from low CO2 or it could be down to humoral mediators from the platelets in the clot surrounding it.



Both shunt (perfusing a totally unventialted lung segment) and V/Q mismatch (poor matching of ventilation to perfusion) are important causes of hypoxia in PE. The shunt is probably the important take away point as we commonly see patients with pleuritic pain for a week with what looks like consolidation/atelectasis on a CXR and we don’t entertain the diagnosis of PE as most of us were brought up believing that we should think of PE in patients with SOB and a clear CXR.

In my search I did find some great videos on basic mechanisms of hypoxia in all conditions which I’ve embedded below.

Any questions, comments, corrections are always welcome.

UPDATE: Martin on twitter noted that shunt is really just V/Q mismatch in the extreme. In particular shunt refers to perfusion of a completely unventilated lung segment. If the segment is perfused and partially ventilated then it is just an area of high V/Q (numerator bigger than the denominator). I hope this doesn’t confuse matters.

Image source: https://commons.wikimedia.org/wiki/File:Saddle_thromboembolus.jpg

Magnesium for Migraine

Migraines are fun to treat. Not so fun for patients but at least we can fix most of them.

It’s always nice to have another tool or two in the armamentarium for treating migraine.

While hardly surprising given this noble element’s history, it turns out magnesium has been studied for migraine before with some equivocal results. This new paper was an interesting read and as I’m prepping for the FCEM critical appraisal exam I may have overthought the stats and methods here. Please let me know if I’m barking up the wrong tree, I’m sure Carley will 😉

The paper:

Comparison of Therapeutic Effects of Magnesium Sulfate vs. Dexamethasone/Metoclopramide on Alleviating Acute Migraine Headache. [pubmed]


  • Study type
    • RCT double blind
  • Population
    • ED patients who someone thought had migraine by ICHD criteria
  • randomisation was computer generated
  • Intervention
    • the drugs appear well blinded
    • 10mg metoclopramide/8mg dex v 1g mag
    • allowed rescue meds but don’t say what they were
  • Power calculation
    • This struck me as a bit funny. They don’t state it clearly but it looks like power was based on a primary outcome of a 2cm difference on pain scale at 2 hrs.
    • It’s also not clear if they’ve done the power calculation to compare one drug versus the other or just looking to see if there was a 2cm pain reduction from baseline at 2 hrs. If it was powered simply to show that either drug is effective at 2 hrs then it’s not really a comparative study.  Here’s the quote to see what you think:
      • With power set at 0.9 (b = 0.01) [Note this must be a simple typo: beta should = 0.1 not 0.01] and error level at 0.05 (a = 0.05), we estimated the minimum sample size for the study to be 31 subjects on each arm to detect a 2-cm difference in the pain intensity score (NRS at baseline vs. NRS at 2 h).


  • 70 patients
  • both interventions worked but mag worked quicker, there’s a nice graph to show the effect
  • as for primary outcome pain score at 2 hrs was 0.66 cm v 2.5 cm with a p value of <0.0001. This also smells a bit funny as if the trial was powered to find a difference of 2 cm between the two drugs and the actual difference they found was only 1.84 cm it’s hard to see how that gets them a p value with so many zeroes. However if the trial was powered to show that either drug can reduce headache from baseline then the p value makes sense.


  • the big issue here is what they were actually testing. from reading the conclusions the authors make it looks like a comparative trial but if it is then the stats look and power calculation don’t smell right.
  • setting aside the complicated machinations of the stats that I may be misinterpreting, it seems from simply looking at the numbers that this seems to have an effect. Whether or not it’s better is up for debate and it may well be the natural course for migraines to get better over time no matter what we do.
  • as always would be nice to see a bigger study in a setting more similar to ours.


Some other magnesium headache studies (from the reference list)

  • Cete Y, Dora B, Ertan C, Ozdemir C, Oktay C. A randomized pro- spective placebo-controlled study of intravenous magnesium sulphate vs. metoclopramide in the management of acute migraine attacks in the emergency department. Cephalalgia 2005;25: 199–204.
  • Corbo J, Esses D, Bijur PE, Iannaccone R, Gallagher EJ. Randomized clinical trial of intravenous magnesium sulfate as an adjunctive medication for emergency department treatment of migraine head- ache. Ann Emerg Med 2001;38:621–7.

Some #FOAMed

RCT of ED Renal Ultrasound for renal colic

The Trial

Smith-Bindman R, Aubin C, Bailitz J, Bengiamin RN, Camargo CA Jr., Corbo J, et al. Ultrasonography versus Computed Tomography for Suspected Nephrolithiasis. N Engl J Med. 2014 Sep 18;371(12):1100–10. 

They managed to come up with the STONE trial as acronym for this one. [Study of Tomography Of Nephrolithiasis Evaluation]

This is big news for USS, as it’s an RCT of the use of ED US use. Ultrasound, of course, makes sense to lots of us who see the probe as some sort of prehensile extension of the human body able to go forth and grasp the diagnosis. Still it’s nice to have some data to help us better understand how it helps our practice.

Here’s the details as I read them.


  • multi centre randomised trial. All good so far
  • Randomisation could have been better described I thought
  • 3 Groups
    • ED performed US by an EP credentialed in US – unclear if these were US super users or just regular punters with basic level US skills
    • Radiology perfromed US
    • CT
  • The first major concern is in the lack of blinding. Though it’s hard to see how you could blind this.
  • the ultimate decisions on imaging and disposal after the randomised imaging were down to the EP looking after the pt. So people afer US could go on to have CT if this was felt to be needed
  • Unclear if the EP looking after the doc was also looking after the patient
  • 3 Primary outcomes (which is a tad naughty. The prmary outcome should of course be, primary I would have thought)
    • “high-risk diagnoses with complications that could be re- lated to missed or delayed diagnoses”. The obvious one here might be a missed AAA for example. the missed pathology was pre defined and categorised by a number of the authors all independent of each other.
    • cumulative radiation exposure (does imaging beget imaging)
    • total costs (which I presume is for a different paper, as it’s not reported here)
  • follow up was by repeated phone calls and a structured interview
  • diagnostic accuracy was a secondary outcome but the gold standard here was the patient reporting stone passage or surgical removal. This is important as most people consider CT as the gold standard but as this is one of the modalities being assessed it would be “incorporation bias”  to include CT in the gold standard


  • screened 3700, took 2700
  • 3-5% lost to follow up which may be a problem – is the reason they didn’t answer the phone at follow up due to the fact they were dead? The reassuring thing is that lost to follow up was similar between groups
  • 40% in this trial had a prior history of stone
  • most were youngish and the most of the time the doc had a >50% pre-test probability of stone. Which is common – stones are usually obvious and most CTs we do for stone are positive, at least in my experience anyhow
  • only 8% were admitted from the ED. This is amazing to me as we admit almost all our query stones. Either because we can’t get a CT at 3am for a stone (let’s face it the priority is pain control not diagnosis here) or because we need to get them a urologist (who only have a weekday service in our place). Our admissions are short but still, it’s nice to see that it is more than possible to manage these as out patients in a less dysfunctional system than ours in Ireland.
  • Primary outcomes
    • the missed high risk diagnoses were tiny (you have to look in the supplementary appendix for this
      • ED US – 6. One bowel obstruction but mainly people returning with infections
      • Rad US – 3  – one missed ovarian torsion and the others infective
      • CT – 2 – infective complicatons
      • All of these were less than 1% and of course when the numbers are this tiny, there’s no statistical difference between them. Worth noting that the infective complications are going to be there no matter what you do, no imaging is going to be definitive for pyelo most of the time.
  • There was significantly less radiation in both US groups. Which is hardly surprising. The excess radiaiton in the CT group was all due to the index CT and not lots of follow up CTs thankfully.
  • One of the most interesting things to me was the accuracy of all 3 tests. Remember that the gold standard here was stone passage of surgery. All 3 tests had identical sens/spec. Sensitivity of 85% and spec of 50%.
  • 40% of those in the ED US group went on to have a CT anyhow at the docs discretion. 27% of the rad group went on to have CT. Again, this is hardly surprising. People simply don’t trust US and most urologists want a CT, I know ours do. Despite the fact that even their guidelines suggest US as the investigation of choice if available.


  • This is a great effort and a substantial trial. As we probably already knew, EP performed US appears safe and accurate when we pose a focused question. There will always be misses but the numbers here are tiny and are more clinical judgement related than imaging related.
  • The issue will be dealing with the specialists who may not be able, or willing to deal with just the US. I don’t say that to be critical, there may be lots of good reasons to pursue further imaging, but there doesn’t seem to be much need for the young, uncomplicated, clinically typical stone.

Finally, the entire study protocol is available as a PDF of supplementary material on the NEJM site and is a fascinating insight into the background of putting together an RCT and the sheer volume of work required for it.

Want more renal ultrasound:

The LINC trial

I wouldn’t regard myself as a total nihilist when it comes to cardiac arrest but I’m usually not on the optimistic side. I work in a typical inner city setting where outcomes from OOHCA are on the poor side. I suspect are numbers are polluted by the fact that our local EMS system only occasionally stops CPR on scene. We get a lot of patients transported with (manual) CPR in progress. Most of these patients have not responded to a substantial period of ACLS and are simply dead and typically I stop resus as soon as they arrive.

There are of course some very important exceptions to this, the people who we know have a more favourable outcomes: bystander CPR, short down times, reversible causes, VF/VT, younger age etc… These are the people where the mantra of maximally aggressive care goes into action. Thus there’s a stark contrast in my practice between the patients who appear to me to be obviously dead and don’t need ACLS – they need a death certificate – and the patients for whom ACLS is simply not enough, or at the very least it doesn’t account for the nuances needed in advanced critical care.

The FOAMed sphere is full of advanced resuscitation resources from some of the best resuscitationists in the world. That’s where I’ve learned most of my critical care.

All this is just an introduction to the LINC trial. It’s been out for a while and I’ve linked to some of the other FOAMed reviews out there but I figured I’d throw in my own two cent.


  • industry sponsored RCT
  • Sweden, Netherlands and UK
  • excluded those with an inappropriate body size so we don’t know how many this will be applicable to
  • the LUCAS protocol had a slightly different protocol with 3 mins v 2 mins. It has been suggested that was due to rescuer fatigue in the manual group. But either way it shows that the two groups were not treated the same. The mechanical group also got an extra DCC 90 secs after the LUCAS was put on.
  • outcome was survival at 4 hrs which is a bit soft
  • CPC outcome was done from the notes which is open to some biases
  • those with ROSC got hypothermia for all rhythms though I don’t think this was pre hospital
  • they assumed a 25% 4 hr survival which seems a touch high but then they weren’t measuring good neuro survival here as a primary outcomes


  • 2500 pts 30% VF/VT
  • 23% survival for both groups
  • in those that survived there was about a 35% PCI rate.
  • 8% surviving to hospital discharge with good outcome (CPC 1&2)
  • it is interesting to note that you either died or you did well. There weren’t many survivors with bad outcomes. Which is perhaps the most important finding of the whole study. They were virtually all CPC 1. It may be that when a poor outcome was anticipated (say at day 3 post arrest) then aggressive treatment was withdrawn and the patient succumbed.


This is a remarkably negative study from a strictly EBM point of view. So any routine (ie for every single cardiac arrest) use is hard to justify given the cost of implementing this device in all EDs and all EMS crews.

Good, coordinated, team CPR is hard to do well and like most things you only get really good at it when you practice. A lot. Like most EDs I’ve worked in we don’t practice, we do our ACLS courses but we don’t practice as a group to deal with all the small logistic factors that turn a straightforward resus into a chicken bomb. We use the LUCAS in certain cases  – the exceptions where I think continuing CPR will allow an intervention or produce a positive outcome – an every time I do I’ve noticed a better atmosphere in the room. As the team leader (as well as the interventionalist – it’s usually just me and my nurses and some very junior docs who can’t even put an IO in) it allows me a bit more cognitive space as I don’t need to keep an eye to monitor the rate and quality of the CPR.

If ECLS becomes a mainstream possibility (which in this country seems unlikely) then mechanical CPR is likely to have even more of a role. The key is of course patient selection. The heart too good to die is fairly rare. For most patients I look after, cardiac arrest is the natural course of age and massive comorbidity.

The logistic factors that give you an advantage (freeing up hands and cognitive space) for those long and infrequent resus cases that makes me glad that we still have one in the resus room.


Hallstrom Al, Rea TD, Sayre MR, Christenson J, Anton AR, Mosesso VN, et al. Manual Chest Compression vs Use of an Automated Chest Compression Device During Resuscitation Following Out-of-Hospital Cardiac Arrest: A Randomized Trial. JAMA. American Medical Association; 2006 Jun 14;295(22):2620–8.

Rubertsson S, Lindgren E, Smekal D, Östlund O, Silfverstolpe J, Lichtveld RA, et al. Mechanical Chest Compressions and Simultaneous Defibrillation vs Conventional Cardiopulmonary Resuscitation in Out-of-Hospital Cardiac Arrest. JAMA. 2013 Nov 17.

FOAMed Reviews:

Rory Spiegel at EMLit of Note




Image Credit Wikimedia Commons

Outcomes in Subsegmental PE

This is a big and important topic but still a little bit confusing to say the least.

The way I want to view the world is like this.

PE is not a single disease but instead something we should call by different names.

1) Pulmonary Fluff or Lung Lint

– these are the tiny sub segmental PEs that we diagnose because diagnostic protocols bend our arms and force us to become bad doctors and it’s not our fault… These PEs get 3 or 6 months anticoagulation and we cross our fingers and hope they don’t die from the anticoagulation.

2) The thrombus of death

– these are the sickish looking people whom you are actually worries about and have high clot burdens and strainy looking hearts and ECGs. These PEs are the ones you think about giving the drug of the big pharma devil to – tPA.

In my imaginary rose tinted world, fluff gets no treatment and the thrombus of death gets tPA.

The problem is that the reality is a whole lot more complicated than that. There is good reason that we treat the small PEs just like the bigger ones. We simply don’t know if they’re dangerous or not so we err on the side of “if in doubt treat” (despite the fact that the evidence for anti-coagulating any PEs is a little ropey)

This paper suggests that these small PEs might be more dangerous than I’d like to believe. [Hat tip to EMU for making me aware of this]

Exter den PL, van Es J, Klok FA, Kroft LJ, Kruip MJHA, Kamphuisen PW, et al. Risk profile and clinical outcome of symptomatic subsegmental acute pulmonary embolism. Blood. 2013 Aug 15;122(7):1144–9. [Full Text Link]


  • data from 2 prior observational diagnostic studies, one was the Christopher study, the other was much smaller. Both European
  • as a result the data is only as good as the original studies which in this case should have been pretty good
  • death where PE could not be ruled out as a cause were considered recurrent PE. Not great but probably the best way to do it.
  • logistic regression used to see if there was an association between where the PE was and the outcomes


  • 3800 pts, 2700 of which got CTPA
  • 21% had PE overall.
  • 115 people with isolated Sub Segmental PE (SSPE)
  • they appeared similar to those with more proximal PEs (except for Wells scores, where they tended to be lower)
  • 3 month follow up was better than 99%
  • 4/115 developed VTE in 3 month follow up. the numbers come out at 3.6% v 2.5% suggesting recurrent VTE more likely in the small PEs!
  • of note those in whom PE was ruled out developed VTE 0.8% of the time


  • I’d love to know the details of the 4 people who developed recurrent VTE in the SSPE group but they don’t tell us
  • the mortality numbers are quite high 10% for SSPE, 6.3% for more proximal PE and 5% for those in whom it was ruled out. This seems very high. 1 in 20 who had a negative work up for PE were dead in 3 months?
  • this seems like a very sick population where 1 in 20 of those with a -ve PE work up were dead in 3 months
  • the numbers with SSPE here are very small (115)compared to the overall numbers and therefore the number of people for whom the outcome of interest occurred (recurrent VTE or death) was absolutely tiny.
  • Everyone in this study was treated which we are assuming is an effective treatment. Therefore this study tells us the natural history of treated SSPEs. It would be really nice to know what the natural history of untreated SSPEs is.

Ryan over at EMLitofnote has reviewed the paper too.

UPDATE 17-5-14

Mattia Quarta of EMPills fame has this to say:

Dear Andy thanks for sharing your thoughts on this paper.

One thing I’ve noticed is that the prevalence of SSPE between the two cohorts that were merged for this study seems pretty different. In the Christopher study 110 SSPE were found which means that only 6 come from Klok publication.
So it’s more less 17% vs 5%. Apparently the two cohorts are not so homogenous after all.
In the Christopher study as opposed to Klok PE level determination apparently was not specifically planned. All 41 PEs excluded because localization was not reported belong to the Christopher study.
Klok and colleagues adopted a specific strategy to locate the level of the clots found. So I wonder whether Cristhopher data are completely reliable when it comes to PE level, considering that a certain grade of variability exists between radiologists in the interpretation of subsegmental clots. With these numbers even one patient can make a huge difference in the results.

I’m not really sure this study will change my point of view on SSPE. For the moment I still lean toward the same imaginary rose tinted idea that fluff gets no treatment unless they have DVT.

I’ll wait until we have more data. http://clinicaltrials.gov/show/NCT01455818