That’s a leading question. Of course you would.
Who wouldn’t want to know more about a disease that has killed >1 million people to date, worldwide. Who wouldn’t want to know more about the thromboembolic risk associated with COVID-19 when we have been told that the VTE prevalence approaches 50% and that some centres have been using systemic thrombolysis based purely on clinical evidence of respiratory deterioration.
We always want to know more, right? But there is an important caveat to knowledge in the modern age. Do you want to hear more from your own echo chamber? Or do you want to know more from the broader literature, including findings that may disagree with what you initially believed. I watched the social dilemma the other night (which I highly recommend) and it occurred to me that we have similar issues with modern medicine and the receipt of new information on a contentious topic. It is very easy through social networking, pubmed feeds, grand rounds and anecdote to only hear information that supports our initial stance on an issue. If you were worried about VTE risk early and searched for information/confirmation to support a proposed change in practice, this information feed will continue. We all saw how the bar dropped for publication during the early stages of the pandemic, due to a clear thirst for knowledge and experience on this topic. It’s time now for us to raise this bar back up – through scepticism, critical appraisal of the literature and perhaps broadening our reading lists.
What do we know about VTE prevalence?
If you cast your mind back to our previous webinar, we talked about early case series, cohort studies and reported rates of VTE approaching 50%. We also discussed the challenges with interpretation of these early findings, which may have arisen through selection bias, improved access to imaging and a ‘screening’ type approach, particularly in the critically ill.
Has this high prevalence borne out in further studies? Not quite to the original degree. Several systematic reviews have now been conducted with pooled analysis to try and capture the overall prevalence. In May, Porfidia et al1 reported an overall VTE prevalence of 26% with a fairly broad confidence interval (20 to 32), within 3487 patients participating in 30 published studies. VTE rates ranged from 0 to 79% and were significantly lower in those studies pursuing a diagnostic work up in patients with suspected VTE, rather than for other reasons (suggestive of a screening approach). In the cohort of patients admitted to intensive care, VTE incidence was similar overall but the proportional burden of PE was far higher.
Nopp et al2 followed this in September 2020 with a further systematic review identifying >28,000 patients participating in 66 studies. I think that probably highlights the challenges we have in dealing with the volume of information and the speed of research on this topic; four months after the first, we have a second systematic review with >10* the patients and double the studies. Interesting. The authors report a lower VTE prevalence, at 14.1% and with tighter confidence intervals (11.6 to 16.9). Again, a screening type approach to the diagnosis of VTE was associated with quadruple the rate of diagnosis in clinically suspected VTE. If you look for it, you WILL find it, as we said before.
Prevalence of VTE and PE in patients admitted to critical care was similar to the previous review, as was proportional PE burden in these patients. Lastly, the authors report a very interesting decrease in reported prevalence over time with publication. I think this is fascinating – is this academic proof of the social dilemma? Is this representative of the dust settling? Is this proof of efficacy in regards to immunosuppressive therapy lowering the risk of VTE? Not sure.
What do we know about the anatomy of VTE?
This is hugely important. There is a big difference between a saddle pulmonary embolus following proximal deep vein thrombosis, and in-situ pulmonary thrombosis reported as ‘possible’ subsegmental PE in isolation. Unfortunately, neither systematic review addresses this. All PEs are equal. However, as we know, some are more equal than others.
In their study from Kings, Martin Whyte et al3 reported PE in 5.4% of hospitalised patients, including those admitted to ICU. More than 50% of cases were unilateral and isolated segmental or subsegmental. Desborough et al4 reported a case series of 66 patients with critical illness admitted to Intensive Care @ GSTT and reported 40% of imaging confirmed PE to be segmental/subsegmental. Middledorp et al5 describe an even higher proportion of likely immunothrombosis, with over 90% of their PE burden described as isolated segmental/subsegmental PE.
We all need to look at this in context. Our local data from Salford is being presented on the 14th October in the Rod Little prize session at the RCEM VSC. We followed up 496 patients with confirmed COVID 19 requiring hospitalisation for 90 days using our Hospital Acquired Thrombosis (HAT) database, case note review and regional imaging software. PE occurred in only 2.4% of our patients, with >80% reported as isolated segmental/subsegmental disease.
As such, it seems like this preliminary finding of local immunothrombosis in-situ as a common complication of COVID-19 is being replicated within the literature and within local practice. This is good science and should impact how we deliver care.
What do we know about VTE prophylaxis?
Critical appraisal of earlier publications on this topic revealed a worrying number of hospitalised patients not receiving any thromboprophylaxis. We need to move on from this discussion. The evidence base for thromboprophylaxis in hospitalised patients with any serious medical illness is very strong. Furthermore, implementation of standardised risk assessment and tailored prescribing has shown a clear reduction in mortality over recent years. In addition, unpublished data emerging confirms the risk reduction in patients with COVID-19. Thromboprophylaxis works. Full stop.
Do we need even more of it though? Intermediate dosing of LMWH (double the standard prophylactic dose) gained some traction in the early stages of the pandemic and has been adopted at various individual sites and within national guidance documents in some countries. However, others were less optimistic about clinical benefit. What more do we know, 6 months down the line? Well, the honest answer is not that much. Retrospective observational cohort studies of up to 100 patients such as Mattioli et al6 suggest the practice to be feasible and safe, but also report major bleeding rates of approximately 2%. This is higher than expected, which we will come on to it shortly. Other studies report variable VTE prevalence rates with the strategy, ranging from 5% (Fauvel et al) to 50% (Bompard et al), which is clearly as high as seen with any other strategy. Of note, all studies published to date allow intermediate dosing on clinician preference. As such, we are left with cohort data on patients receiving higher dosing stragies who have likely been selected due to perceived increased baseline risk or illness severity. Such findings are likely to be highly biased, which makes the prevalence data challenging to interpret.
There is also a question around the difference between intermediate dosing and weight adjusted dosing of thromboprophylaxis. Are they really any different? Although guidelines provide limited advice on this issue, recent updates to the Summary Product Characteristics for several LMWH regimens and Q&A sessions within the UK Clinical Pharmacy Association highlight the potential for underdosing at extremes of weight and offer guidance on weight adjusted prescribing. Examples of weight adjusted prescribing can be easily found online. Are some centres therefore already use intermediate dosing, without really realising? Or is intermediate dosing a double dose of weight adjusted LMWH? Unpublished attempts at systematic review and pooled analysis are already lumping all these strategies together, but they can be very different. This is another important point to remember, when thinking about changing practice.
What do we know about therapeutic anticoagulation, in the absence of confirmed VTE?
Last time, we discussed how some sites had moved towards a strategy of empirical therapeutic dose anticoagulation for higher risk patients. An example from Mount Sinai can be found here. Is this clinically effective at reducing VTE rates? Again, that’s tricky to unpick from the observational cohort data. The Mount Sinai team published data on 2773 patients hospitalised with COVID 19, in the May edition of JACC this year.7 They report no difference in mortality with therapeutic anticoagulation compared to those not receiving therapeutic anticoagulation (22.5% versus 22.8%). However, they go on to perform a cox proportional hazards model adjusted for a variety of covariates which demonstrates an improved median duration of survival, concluding that anticoagulation may be associated with improved outcomes. This is a little tricky to swallow, given the limitations of an observational study, unknown indication for anticoagulation, median duration of only 3 days anticoagulation and unobserved confounders.
Other studies looking at the efficacy of treatment dose anticoagulation in COVID19 again show widely variable VTE outcome rates, ranging from 3.6%, to 27.0% and all the way up to 55.6%. There has even been a recently published Cochrane review looking at the relevant evidence and trying to summarise the available data. The result? There is no good quality evidence or reliable signal that using therapeutic anticoagulation outside of suspected/confirmed VTE reduces either prevalence of VTE or all cause mortality. Unpublished work comparing the pooled incidence would suggest it is very similar between intermediate and therapeutic dosing, implying a threshold effect for LMWH prophylaxis.
What do we know about bleeding?
What about bleeding, I hear you cry (ask politely)? Well, we do know a little more here. And observational data on bleeding can perhaps be a bit more informative, given the standardised definitions in the literature, the well-established baseline comparison rates and the clear impact on clinical decision making. We have previously discussed how any change in practice should be predicated on the evidence supporting the benefits and the harms. If we know that escalated anticoagulation strategies carry additional harm, the necessary burden of proof to change practice in this direction should be much higher.
In the Mount Sinai paper referenced above, major bleeding increased with therapeutic anticoagulation compared to any other strategy, but the difference did not reach statistical significance. Hanny Al-Samkari et al recently reported retrospective findings in Blood regarding bleeding events in 400 hospitalised patients with COVID-19, with 90% receiving standard dose thromboprophylaxis.8 They found a major bleeding rate of 2.3%, with an overall bleeding event rate of 4.8%. In the critically ill, major bleeding was seen in 5.6% of patients. This is a lot higher than would be expected within an average cohort of medically unwell patients and raises real concerns about escalated therapy.
Other cohort studies report findings that corroborate this picture; Pesavento et al9 followed 324 hospitalised patients in Padua with COVID-19, of whom 80 received escalated dosing prophylaxis strategies. The Hazard ratio for a composite endpoint of major bleeding or clinically relevant non-major bleeding was 3.89 (95% CI 1.90 to 7.97) with escalated dosing strategies. Mortality was also higher in this group, although the observational nature of the study again introduces confounding and selection bias regarding this result.
There are limitations to these results similar to those seen when assessing VTE incidence in observational cohort data. But there is a clear signal that increasing the dose of prophylactic anticoagulation is not without risk. As such, this must factor into decision making on when to change practice and how to individually risk assess.
What do we know about risk prediction?
Lots of proposed algorithms discuss specific criteria that should trigger consideration of higher dosing strategies. Again, we lack the evidence to be confident here.
Great effort was put into evaluation of the d-dimer at the beginning of the pandemic. Unfortunately, the literature does not seem to support this initial enthusiasm. Studies such as those already referenced highlight that a higher D-dimer does unfortunately not predict increased VTE rates in isolation; higher numbers predict increased rates of bleeding, critical illness and death. It is also notable that this variable has not made it onto clinical decision rules such as the 4AC score10, derived through rigorous study in >30,000 patients and designed to provide evidence based mortality estimates. As such it cannot guide individual treatment strategies.
Clearly though, we are still trying on this topic. Hot off the press last month was a new prediction model, derived through a retrospective cohort of 127 patients with confirmed COVID-19 admitted to Intensive Care.11 The authors report 41.7% patients experiencing any VTE and 16.5% developing PE during their ICU stay, although we are not told what type of PE. Interestingly, CRP had the best area under the curve of any blood or respiratory variable, including d-dimer, platelet count and P/F ratio.
Can we use this data? I don’t think so. As already discussed, the selection bias and confounding within all these studies makes it very difficult to identify risk characteristics that are reliably associated with increased VTE rates. And even if we ignored these problems, we would still have no good evidence to suggest that increasing the dose of LMWH in these patients at potentially increased risk would have any net benefit.
What do we know about other treatments?
This is where we can be more confident. We know a lot more about treating this disease than we did earlier in the year.
We know moderate potency steroids started early in hypoxaemic patients can save lives. We know that a lot of things probably don’t work, like Lopinavir/Ritonivir and hydroxychloroquine, no matter what a certain president has to say on the matter. We know more about management of critical illness and COVID-19 pneumonia, including how to provide supportive care (including mechanical ventilation) while minimising iatrogenic harm. We know about the risk of line associated and extracorporeal circuit thrombosis and we know how to manage these issues without systemic anticoagulation. We know more about survival in certain subgroups and can use this information to prognosticate and allocate resources.
We also know a little more (but not enough) about other immunomodulatory therapies that may reduce the degree of inflammation within the lungs, potentially leading to lower rates of immunothrombosis. In an observational cohort study of 764 patients with COVID-19 admitted to intensive care, tocilizumab appeared to be associated with a reduction in hospital related mortality.12 RCT evidence is coming (COVACTA), as it is on convalescent plasma, azithromycin and remdesivir through the platform RECOVERY trial.
Will these drugs attenuate immune response and reduce pulmonary thrombosis? Gergi et al13 report a case series of 6 patients receiving mechanical ventilation for COVID19, who were given compassionate use tocilizumab as a single dose. The authors report a significant and rapid reduction in markers of thromboinflammation. Findings like these push this hypothesis forward, but must be interpreted with caution given the design.
What are we doing in light of all this new information?
All kinds of stuff, would appear to be the shortest answer. With no consensus between international guidelines and limited randomised controlled trial data to advise on practice, many are still practicing in the time of Osler. A well conducted survey study led by Rachel Rosovsky and team proved this point14 well in June this year, highlighting the disparity of practice between 515 physicians from 41 countries. Respondents reported a wide variation in prevalence rates, diagnostic approach and in particular significant practice variation regarding the need for dose escalation of anticoagulation outside the setting of confirmed or suspected VTE.
And what should we be doing?
I think that’s still tricky. But we are in a better position than last time and I think we can use this new evidence to agree on some issues
First, it is clear (and has been for some time) that all patients admitted to hospital who are unwell with COVID-19 should undergo risk assessment for VTE and should receive pharmacological thromboprophylaxis if safe. If community hubs and ‘hospital at home’ services continue to expand, this assessment may need to be extended outside the front door.
Second, in those patients who are prescribed thromboprophylaxis, weight-based dosing algorithms should be strongly considered. There appears to be no clinical evidence of superiority for intermediate dosing strategies over a standard, licensed weight-based approach. I would extend this logic to patients admitted to intensive care.
Third, we should ensure these patients are managed in line with the emerging evidence base outside of anticoagulation. Immunomodulatory agents clearly have face validity in this cohort. Before we weigh up the merits of unproven dose escalations of anticoagulants that we know will cause predictable harm, we must ensure that hypoxaemic patients have received dexamethasone. In the future this will also apply to any other therapies that have been studied and proven effective within the context of large, well conducted randomised controlled trials.
Fourth we must continue to offer research opportunity. All our patients have the right to participate in ongoing studies and potentially benefit from novel therapies.
A final word
Just a short plea to reason, Brandolini’s law and the newly defined concept of evidence-based agility15. In the modern age, it takes an order of magnitude higher to refute a change in practice than it does to produce it in the first place. With the pace and scale of change around this condition, this law is more important than ever.
Given the body of evidence we have now, I am not sure it is right to continue to practice in the time of Osler. We can’t say we ‘just don’t know’ about therapeutic anticoagulation strategies for these patients. We know it does harm. In the world of dexamethasone, risk prediction and immunothrombosis, we don’t really know that it helps, although we thought it might. Until we do know, we should stop doing it, on balance of risk. The same could be said of intermediate dosing strategies.
We need more data. But until then, we should probably start to think about primum non nocere, rather than looking for a magic bullet. The former will never go out of date. The latter is often a futile search, as any intensivist will tell you.
Thanks for reading.
1 – Porfidia et al. Venous Thromboembolism in patients with COVID-19: Systematic Review and Meta-analysis. Thrombosis Research 2020; 196: 67-74
2 – Nopp et al. Risk of venous thromboembolism in patients with COVID‐19: A systematic review and meta‐analysis. Research and Practice in Thrombosis and Haemostasis 2020; Online first
3 – Whyte et al. Pulmonary embolism in hospitalised patients with COVID-19. Thrombosis Research 2020; 195: 95-99
4 – Desborough et al. Image-proven thromboembolism in patients with severe COVID-19 in a tertiary critical care unit in
the United Kingdom. Thrombosis Research 2020; 193: 1-4
5 – Middledorp et al. Incidence of venous thromboembolism in hospitalized patients with COVID‐19. Journal of Thrombosis and Haemostasis 2020; 18(8): 1995-2002
6 – Mattioli et al. Safety of intermediate dose of low molecular weight heparin in COVID-19 patients. J Thromb Thrombolysis. 2020; 13: 1–7.
7 – Paranjpe et al. Association of Treatment Dose Anticoagulation With In-Hospital Survival Among Hospitalized Patients With COVID-19. Journal of the American College of Cardiology 2020; 76(1): 122-124
8 – Al-Samkari et al. COVID-19 and coagulation: bleeding and thrombotic manifestations of SARS-CoV-2 infection. Blood 2020; 136(4): 489-500
9 – Pesavento et al. The hazard of (sub)therapeutic doses of anticoagulants in non‐critically ill patients with Covid‐19: The Padua province experience. Journal of Thrombosis and Haemostasis 2020; 18(10): 2629-2635
10. Knight et al. Risk stratification of patients admitted to hospital with covid-19 using the ISARIC WHO Clinical Characterisation Protocol: development and validation of the 4C Mortality Score. BMJ 2020;370:m3339
11 – Dujardin et al. Biomarkers for the prediction of venous thromboembolism in critically ill COVID-19 patients. Thrombosis Research 2020; Online first
12 – Biran et al. Tocilizumab among patients with COVID-19 in the intensive care unit: a multicentre observational study. Lancet Rheumatology 2020; 2(10): E603-E612
13 – Gergi et al. Thrombo-inflammation response to Tocilizumab in COVID-19. Research and Practice in Thrombosis and Haemostasis 2020; online first
14 – Rosovsky et al. Anticoagulation practice patterns in COVID‐19: A global survey. Research and Practice in Thrombosis and Haemostasis 2020; 4(6): 969-983
15 – Carley et al. Evidence-based medicine and COVID-19: what to believe and when to change. EMJ 2020; 37: 572-575