No one likes getting a pulmonary embolism. Or a deep vein thrombosis. And because about 60% of all blood clots are associated with hospitalisation for acute illness1, we take the idea of prophylaxis very seriously. Emerging data suggests that this seriousness has reaped reward in the UK2. We give information about mobilisation and hydration. We sometimes use stockings. We risk assess patients early on and give pharmacological thromboprophylaxis if they have a low risck of bleeding. And we keep our eye on people. They still get clots sometimes, but they seem to be getting less clots overall, which is good news for everyone.
But sometimes, patients can be very complicated. This is particularly true of polytrauma patients, who often have a complex pelvic injury (putting them at high risk of VTE) alongside a traumatic brain injury (putting them at high risk of a complication from thromboprophylaxis) as well as a critical care admission (putting them at high risk of VTE) in addition to a splenic injury (putting them at…… you get the idea). These cases are challenging. And when things are challenging, people often start to champion a solution with face validity but not much hard evidence. “We’ll never get the evidence!”. “We must take action!”. “Lives are at stake!” I am sure you have heard them cry.
Arise the Inferior vena caval (IVC) filter. What’s not to like about the idea of an umbrella in a big proximal vein that will stop any clot travelling to the cardiorespiratory system, where it can potentially kill. You can even put it in yourself these days…. However, the absence of data supporting this intervention is concerning3. As are the reports of fragmentation, thrombosis and complication. You will perhaps have noted that NICE have recently removed all reference to IVC filters4 from their updated prophylaxis guideline in 2018. You may also have seen this comparative effectiveness study reporting increased harm with filter use5.
But what this issue needs is a randomised controlled trial, yes? In the right patients, with appropriate methodology and reported transparently? Of course it does. And hats off to the author group for tackling this challenging issue head on. The abstract is below, but as we always say, please go readit for yourself6.
Available at https://www.nejm.org/doi/full/10.1056/NEJMoa1806515
The paper quotes initially that pulmonary embolism accounts for 12% of all deaths after major trauma. A quick point on that figure; firstly, that came from a single centre study of an intensive care population (by the same authors), automatically excluding pre-hospital deaths and deaths from haemorrhage. Secondly, that 12% was equal to 16 patients of the 143 that died in total. Thirdly, it used an autopsy diagnosis of PE to ascertain it as the cause of death. This may be perhaps unnecessarily pedantic; just an opportunity to remind us to check references and quoted figures.
Nonetheless, there is no debate that the research topic is important and relevant.
What was the question?
In trauma patients with a contraindication to pharmacological thromboprophylaxis, does the insertion of a retrievable IVC filter in the first 72 hours result in a lower incidence of pulmonary embolism (PE), when compared to no filter.
How did they set out to answer it?
In an interesting way. This was a multicentre open label randomised controlled trial across 4 tertiary hospitals in Australia. Great. Patients were included if they were aged over 18, had an estimated injury severity score (ISS) >15 (denoting ‘major’ trauma in most modern systems) and a contraindication to pharmacological thromboprophylaxis within the first 72hrs. Patients were excluded if there was suspicion of imminent death, a confirmed PE on admission, pregnancy, or if no interventional radiologist was available to insert the IVC filter. Interesting already I am sure you will agree – they don’t specify in the manuscript what they considered to be a contraindication to anticoagulation, so it’s tricky to tell if this was subjective or objective. Clearly, there are contraindications and contraindications, if you know what I mean. Also, the lack of an available interventional radiologist as an exclusion criteria introduces a degree of selection bias, or convenience sampling. Providing allocation concealment was maintained up to the point of randomisation, there should be limited impact from this. But still something to bear in mind.
Patients were randomised to receive either an IVC filter or no filter, through a permuted block central system across 4 sites. This should maintain allocation concealment and minimise conscious or subconscious bias. All patients were allowed intermittent pneumatic compression, although we couldn’t clearly find out results on who got what. The protocol recommended commencement of pharmacological thromboprophylaxis at clinical discretion as soon as it was considered safe to do so. Very subjective…
To minimise detection bias, the trial used a predefined criteria for identifying PE. The intention here is good – you could easily envisage that in an open label study, all patients randomised to control would worry clinicians – as such they may scan anyone with any symptoms, detect a load of asymptomatic PEs and quickly announce that filters were the best thing since sliced bread. However, that doesn’t mean that the issue is foolproof. If you read the supplementary appendix, you’ll see that the predefined trigger for PE investigation was oxygen requirement of 5L to maintain target sats, hypotension for more than 30 minutes or unexplained chest pain. If we are talking about the same polytrauma patients that we see all the time in Virchester and London, then a lot of them have pulmonary contusions, rib fractures and all kinds of injuries that can cause those symptoms. As such, you can make the argument that these patients are being screened for the presence of PE and not necessarily investigated for symptomatic disease. But it’s tricky to know without the granular data.
All patients had Doppler ultrasound of the legs at 2 weeks after enrollment, as a safety measure to try and pick up big DVTs early in the control arm. Again, this is screening, not diagnostics resulting from a clinical suspicion or assessment.
The study used a composite primary end point of symptomatic PE, or all-cause mortality at 90 days. This has raised eyebrows in the social media sphere. Composite outcomes are frowned on, although I am sure we all agree that it is vital to know when patients in a trial like this die of PE. Could the authors have been more circumspect here, perhaps using an independent adjudication committee to confirm death from, or potentially related to PE? Perhaps. Many other thrombosis trials use this technique and there are even recently published recommendations on it7, so it was quite interesting not to see it here.
A key secondary end point was the incidence of symptomatic PE in patients who survived to 7 days and did not receive pharmacological thromboprophylaxis (due to ongoing contraindications). The authors also looked at all cause mortality, major and non-major bleeding at 90 days
The study was powered to detect a 9% incidence of PE in the control group. They assumed a negligible PE rate in those randomised to IVC filter insertion, worked out as a 8.5% lower rate of PE with filters (0.5 vs 9%).
Data were analysed with intention to treat, which seems sensible. The study predefined stopping criteria for the study which meant that If 4 fatal PEs occurred in the control group then the trial would be stopped.
What did they find?
The trial was conducted over 2 years and screened 1740 patients, enrolling 240. This is immediately notable and demonstrates one of 2 things – either these patients are rare, or that when pushed most of us will be comfortable with the idea of using pharmacological thromboprophylaxis within 72h. 2 patients from each group crossed over treatment arms; It is understandable that someone randomised to a filter would not get it for logistical reasons or complications, but it is not as clear why someone randomised to no filter would then get one. This is not really explained other than vague hints at ‘clinician discretion’ and ‘clear indication arising’.
The patient demographics were as you might expect from a trauma population; predominantly male, around the ages of 20-40, with an ISS around 25. The median GCS was lower in the IVC filter arm, but this did not equate to a larger number receiving intracranial pressure monitoring. 57.5% of the patients had a form of intracranial haemorrhage (ICH), but you have to go to the supplemental appendix to see that there were slightly more ICH patients in the filter arm. The vast majority (95%) of patients were also recruited through a single-centre; this raises alarm bells about allocation concealment after all, and makes you wonder if this could (?should) really be considered a single centre study. In the IVC filter group, 89% had the filter inserted within the first 24 hours.
There were 27 deaths in total in the study, of which one died from a fatal saddle embolism (in the control arm). It is worth highlighting that this patient died on day 16 after being on pharmacological thromboprophylaxis for 8 days preceding. The study found no statistically significant difference between groups for the composite outcome of symptomatic PE or death (13.9 vs 14.4%) in the IVC filter group vs control respectively. In the pre-specified subgroup that survived 7 days and had no chemoprophylaxis in that time due to ongoing contraindication (a total of 80 patients only; 46 filter vs 34 control), 5 of the patients in the control group developed a symptomatic PE, whilst there were none in the IVC filter group. The authors present this as a hazard ratio for PE of 0 (0 to 0.55), which looks potentially significant. There was no difference in the occurrence of DVT at 2 week screening, and no difference in transfusion requirements.
What else? Well, there was a higher rate of all cause death (13.1% vs 9.3%) major bleeding (70.5% vs 66.1%), and non-major bleeding (23.8% vs 17.8%), in the IVC filter group. None of these results would be deemed significant as secondary outcomes and also with the 95% confidence intervals for the hazard ratio spanning 1.
What does this mean?
Like always, we think the meaning is probably in the eye of the beholder.
There is and will always be a degree of confirmation bias when reading papers. When studies are underpowered it can lead to a wide number of opinions. Those already opposed to IVC filters will read this as confirming their beliefs given the lack of significant difference in the primary outcome between groups. Those who favour the use of IVC filters will note the absence of any symptomatic PEs in the IVC filter group compared to the 5 the presented in the control arm, along with the hazard ratio. Were these PEs important? Not sure we know. Does prevention of these PEs justify widespread use of an intervention that can potentially cause harm? Who’s to say.
On that topic however, it is worth highlighting that multiple safety endpoints were seen with filter use. These events included entrapped thrombus within the filter at first removal (4.9% of cases) and adherence to the caval wall requiring surgical removal (0.8% of cases). There is also devil in the detail – if you look at the cox proportional hazards regression analysis in the supplementary appendix, you will see that the risk of death or PE appears to be worse with a filter as your ISS and age increase. Odd, in that these are the sickest and most vulnerable patients who we would expect to potentially benefit. They don’t appear to. They appear to do worse.
Following on from this, it is very much worth looking at the deaths in this study, which are the main contributors to the primary outcome. Again, it’s the supplementary appendix for these, but good of the authors to provide. There is only 1 cause of death attributed to VTE directly; all others talk about TBI, complications of multisystem trauma and multiorgan failure. This reassures us that in >100 complex trauma patients with an initial contraindication to pharmacological thromboprophylaxis who did not get an IVC filter, <1% died as a result of VTE. Again, this one case of saddle PE had also had 8 days of pharmacological thromboprophylaxis prior to dying of PE.
A quick note on power and outcomes
Quite a few clinicians have commented on the power of this study and use of a composite outcome. Some have even begun tinkering with the stats. This can be dangerous.
Say you thought that the primary outcome in this trial should have been symptomatic PE (as defined by the authors) and/or death attributed to PE, in patients with an ongoing contraindication to thromboprophylaxis. It is fairly straightforward to look at the event rate of 5 PEs in the control group and compare it to 0 in the intervention group, then proudly announce an absolute risk reduction of 14.7% and a NNT of 7. Well done everyone. However, the variation around this estimate is unclear and any sample size calculator will tell you to be cautious around interpreting data from an n=80 population, even if your risk reduction is as large as this one.
More importantly though, looking at the data this way ignores the risks and potential harm caused by the intervention, through focusing solely on VTE outcomes. What if putting filters in causes significant bleeding, or other adverse events? These would not be captured by your filtered primary outcome. How do you address this? By presenting all cause mortality, as the authors have done here, and ensuring that major bleeding and complications are weighted against the benefits of reducing VTE risk. Many other recent thrombosis trials have8 taken the same approach, or presented efficacy and safety data side by side. It is all well and good that an intervention may reduce morbidity or mortality from PE; but if it increases these outcomes through another mechanism, that is essential to know.
What’s the take home?
Trauma is a heterogenous set of diseases; patients surviving haemorrhagic shock are very different to those with significant intra-cranial haemorrhage. To say that there is no role for IVC filters based on this paper would be quite a punchy conclusion to make. This isn’t the first study to identify that IVC filters are associated with lower rates of PE9 (and potentially mortality) and there remains a suggestion that in the right patients it may reduce the development of symptomatic PEs.
But who is the right patient? That’s where we suspect there will be ongoing debate. For some of us, this paper suggests that restricting IVC filter use does not bring about death in this population, may avoid harm, will save cost and resource and allow opportunity to focus on other things. For others, it may look like it prevents symptomatic PE without causing additional mortality. A quick point on the latter – if we are talking about preventing symptomatic PE in the subgroup analysis of patients who did not recieve pharmacological prophylaxis for >7 days, then this data suggests there is no great rush to put the filter in. I can’t find any evidence that any patient got a PE in the first 7 days from admission. That in itself is useful, as the logistics of arranging this intervention quickly can be challenging outside of centres with IR on site.
Are you a naysayer? I think I am (Dan). I have never been overly convinced by the data on IVC filters. On our major trauma/neurosciences ICU I don’t pursue them unless I asbolutely have to, in particularly complex cases with ongoing super high VTE risk and continuing lengthy contraindication to thromboprophylaxis. So I am starting from a position of scepticism. This work does not change my mind. In fact it supports my opinion that there is harm with filters, that despite complexity very few of these patients seem to die of VTE, and that the more complex the patient actually the higher the hazard of a filter. This latter point and the additional supplementary data suggest to me that it could be the younger pelvic fractures, who are not super sick, that have an ongoing contraindication to anticoagulation who may stand to benefit the most from filter placement. But that’s just me.
What do you think? Are you a fan of the filter? If so does this work support your beliefs?
What do you believe? And when do you change? Sounds like a good presentation that10, if only someone could cobble something together…..
Happy Summers everyone.
Dan and Rich
- 1.Movement WT. Know Hospital-Associated VTE. World Thrombosis Day. http://www.worldthrombosisday.org/issue/hospital-associated-vte/. Published 2019. Accessed 2019.
- 2.Hunt BJ. Preventing hospital associated venous thromboembolism. BMJ. June 2019:l4239. doi:10.1136/bmj.l4239
- 3.Bikdeli B, Chatterjee S, Desai NR, et al. Inferior Vena Cava Filters to Prevent Pulmonary Embolism. Journal of the American College of Cardiology. September 2017:1587-1597. doi:10.1016/j.jacc.2017.07.775
- 4.NICE N. Venous thromboembolism in over 16s: reducing the risk of hospital-acquired deep vein thrombosis or pulmonary embolism. NICE. https://www.nice.org.uk/guidance/ng89. Published 2018. Accessed 2019.
- 5.Turner TE, Saeed MJ, Novak E, Brown DL. Association of Inferior Vena Cava Filter Placement for Venous Thromboembolic Disease and a Contraindication to Anticoagulation With 30-Day Mortality. JAMA Netw Open. July 2018:e180452. doi:10.1001/jamanetworkopen.2018.0452
- 6.Ho KM, Rao S, Honeybul S, et al. A Multicenter Trial of Vena Cava Filters in Severely Injured Patients. N Engl J Med. July 2019. doi:10.1056/nejmoa1806515
- 7.Kraaijpoel N, Tritschler T, Guillo E, Girard P, Le Gal G. Definitions, adjudication, and reporting of pulmonary embolism‐related death in clinical studies: a systematic review. J Thromb Haemost. July 2019. doi:10.1111/jth.14570
- 8.Meyer G, Vicaut E, Danays T, et al. Fibrinolysis for Patients with Intermediate-Risk Pulmonary Embolism. N Engl J Med. April 2014:1402-1411. doi:10.1056/nejmoa1302097
- 9.Haut ER, Garcia LJ, Shihab HM, et al. The Effectiveness of Prophylactic Inferior Vena Cava Filters in Trauma Patients. JAMA Surg. February 2014:194. doi:10.1001/jamasurg.2013.3970
- 10.Carley S. What to Believe and when to Change. SMACC. https://smacc.net.au/2014/08/carley-simon-what-to-believe-when-to-change/. Published 2016. Accessed 2019.