Level Pegging? JC and the PEGeD study @StEmlyns

It’s not often you see a mate as a lead author in the New England Journal of Medicine (NEJM). When you do, it needs celebrating. Hats off to Associate Professor Kerstin de Wit (nee Hogg), who with a group of world leading thrombosis experts, published an impressive piece of work this week on the use of clinical probability adjusted d-dimer thresholds for exclusion of PE in the emergency department/outpatient setting​1​. Kerstin is a Virchester alumnus and produced her doctoral thesis in the UK, publishing the MIOPED study​2​ and pioneering the evidence base for ambulatory management of VTE. She has since emigrated to Canada, working first at Ottawa with Phil Wells and co, then onto Hamilton where she now practices as an emergency physician and thrombosis/haemostasis clinician. She has done loads in the last decade, but this recent project gives us a clear reason to officially celebrate her contribution to the literature. Congratulations Kerstin et al.

HOWEVER. This does not mean the PEGeD is exempt from critical appraisal. I am sure Kerstin would think the same and want to encourage scientific discourse. No stone is left unturned in Virchester, irrespective of friendship status. As such, I sat down to have a good look at this article. We suggest you do the same, and the full text (unfortunately not FOAMed) can be found here:

What’s it all about then? 

The premise behind this paper is that we are still overtesting in our attempt to exclude PE. The authors suggest only 30% of the patients we evaluate clinically have both a low pretest probability by Wells score and a negative d-dimer by normal conventional cutpoint. As such, the implication is that 70% of patients go on to have definitive investigation through chest imaging. I am not sure this is entirely true in clinical practice (as a number of patients get better/don’t show up for imaging/are discharged after consultant review with no scan), but I do agree in general with the concern on overtesting. 

If the patient does get imaged, this can be useful in confirming diagnosis or identifying alternate pathology, but comes with multiple risks; radiation exposure, costs, delays in care, contrast reactions and most importantly, the diagnosis of small PE’s unlikely to progress or recur. If anticoagulation treatment is started in this latter group, say hello to 3 months of tablet taking, a 2% major haemorrhage rate​3​ and a 9.1% case fatality rate – even in the best of circumstances. If your patient has clear bleeding risks, renal failure, poor compliance or other complex medical issues, you have just made life very difficult for everyone.

Thus, the PEGeD authors propose a hypothesis to reduce imaging rates in two ways: firstly by adjusting the cutpoint for a ‘positive’ d-dimer test (which you will remember is a continuous, not a binary variable from our previous discussions​4​) in those patients with a low clinical pretest probability (C-PTP) and secondly by applying a d-dimer test to more than just ‘low’ C-PTP patients. Both good and interesting ideas. 

I was surprised though, to see little mention in the background of other recent approaches towards this same goal. The concept of age adjusted d-dimer​5​ has been around for a long time, which follows a similar argument. In addition, other groups have already experimented with differing risk scores and graded thresholds for d-dimer​6​. We’ll come back to this in the results, but keep it in mind.

This seems like a very good time to mention that the new NICE VTE 2020 guidelines are out for stakeholder consultation in the UK at present​7​, and have several specific recommendations supporting the use of age adjusted d-dimer testing for exclusion of PE.

You have until Christmas Eve to comment. Don’t say we never tell you anything…

How was this trial designed?

PEGeD was a prospective management study. Patients with symptoms or signs suggestive of PE were managed according to a set PEGeD protocol; if you had a low C-PTP (as per the three tier Wells score) and your d-dimer was <1000ng/ml, you were discharged with no imaging. If you had a moderate C-PTP and your d-dimer was <500ng/ml, you were discharged with no imaging. If you were high risk, you got imaging. All participants were followed up for 90 days to determine relevant outcomes. 

Interestingly, patients were approached to participate either during diagnostic work up or afterwards, and provided verbal or written consent depending on the ethics committee viewpoint. I say this is interesting, as verbal consent after discharge is really a type of waiver/ deferred consent, such as the type you would use for emergency care research in unconscious patients. This particular group should have been perfectly capable of considering whether or not they wanted to be involved, and then agreeing at the time. 

However, it is without doubt a simpler version of emergency care research to include all patients and manage them to a particular protocol within the ED, when they may be tired/ in pain/ anxious and reluctant to think about research. It is essentially opt-out consent – you call them after the fact and ask if it is OK to still use their data. But you have already essentially done the research intervention.

Of course, this should be fine for all observational research. The reason I find this particularly interesting, is that the study authors were actually changing clinical care (assuming that prior to the study, most sites were using conventional cutpoints). 

Who was involved?

Over 3 years, the investigators enrolled participants through ED or outpatient clinics in Canada (apart from one random inpatient) into PEGeD, but it is not entirely clear across how many centres.  The aim was to exclude children (<18), pregnancy, palliative cases, those on anticoagulation or who had received at least 24h therapeutic dose anticoagulation, those patients who had undergone major surgery (undefined) in the last 21 days or those who had violated the protocol (recent chest imaging or d-dimer result was known prior to inclusion). They also excluded those patients geographically inaccessible for follow up. These last 2 exclusion criteria are always hard to police within a pragmatic study like this one – how do you stop someone taking a cheeky look at the d-dimer (taken at triage) before assessment of C-PTP? Tricky. It’s always really tough to get people to do a C-PTP score in practice without looking at anything else, especially in the context of a pragmatic trial like this. If they do see the d-dimer result, this would undoubtedly creep into their C-PTP score either through conscious or subconscious bias.

What was the primary outcome?

Although with studies like this it would be nice to know who actually had a PE and who didn’t, you simply cannot image every patient for ethical reasons. As such, similar to previous diagnostic accuracy studies, the primary outcome here was one of symptomatic, objectively confirmed VTE over the next 90 days, evaluated by a central adjudication committee with predefined criteria. 

This is good, because it addresses the core question at the mind of every clinician. If I discharge this patient without imaging, using this diagnostic strategy, what is the chance I have missed a clinically relevant VTE event, or someone will come back dead? 

However, this strategy for outcomes is not without issue. The inclusion of DVT within the composite means that false positive outcomes are likely – if I discharge someone using this strategy, who does not have a PE, but then gets a DVT in 4 weeks time for another reason, is this truly a failure of the original diagnostic strategy? Likewise, we talk about symptomatic disease, but remember these patients have been counselled to return in the event of any symptoms that may be related to PE. They will be better informed, and more anxious, than your average patient. And you as a clinician will be more likely to image them if they return with symptoms, having not had imaging in the first place. Thus we are not being entirely objective, some would argue.

The choice to use this outcome drives the sample size calculation. If you want to be sure that discharging patients with a low C-PTP and d-dimer <1000ng/ml has a low subsequent VTE event rate, then you need to estimate this rate and further estimate how many patients you require, to ensure a high level of precision (narrow confidence intervals around this estimate). The authors propose an event rate of 0.8%, but do not provide a reference for this. The next sentence is a little confusing, but I think they are suggesting that in a cohort of patients with a VTE prevalence of 2%, a sample of 1036 patients would be able to identify a miss rate of 0.8% with an upper boundary of the 95% confidence interval at 1.5%. This suggests that in order for their sample to be adequately powered, and for the authors to deem their diagnostic strategy a success, they could still report a VTE event rate of around 1 in 66 patients discharged. They double the 1036 (assuming that low/moderate C-PTP makes up only 50% of those evaluated) and add a further 1.5% potential drop-out rate, ending with a sample size required of >2000 patients needed, who have a low/moderate C-PTP and d-dimer below their proposed cutpoint. Make of all that what you will.  

A word on d-dimer testing

It is always worth highlighting that d-dimer assays are not universal and reported in variable units. The two common assays report in either d-dimer units (DDU) or in fibrinogen equivalent units (FEU). FEU are essentially double the quantitative value of DDU. Therefore 250 DDU is equal to 500 FEU. The kicker is that most assays will report both these units in ng/ml. You therefore often need to ask your lab what they are using, to be clear how this affects your care. 

This is particularly important with any graduation of d-dimer cutpoint. The authors report 500ng/ml as the routine cutpoint in Canada, which is assumed to be FEU. They are changing this in the study to 1000ng/ml in the low C-PTP group. All well and good. In my institution however, our lab reports in DDU – as such, our routine cutpoint is 250ng/ml. As such, I would transfer this study design to mean that in my patients with a low C-PTP, I might be able to discharge patients with a d-dimer below 500 – not 1000ng/ml. Other centres will have a routine cutpoint of 230ng/ml (DDU), therefore the novel cutpoint would become 460.

Interesting stuff.  Speak to your lab, understand what you use, and be clear how it relates to the established literature figure of 500ng/ml FEU as a routine cutpoint. The authors pragmatically allowed different institutions to use their routine assays in the study, which is helpful and makes the research more generalisable. They report five different assays used and touch on this issue, including different local outpoints in the legend for table 1. Have a read.  

Ok, tell me about the patients recruited?

Centres assessed 3133 patients as meeting the inclusion criteria, although there is no denominator for this to cover which patients were missed. As such, and given we don’t know the number of recruiting centres, we can’t really be entirely clear on whether this was a consecutive or convenience sample. 

941 met one or more exclusion criteria, 136 did not provide consent and 39 were recruited in error, leaving us with 2017 patients for analysis. Only 1325 of these had a low/mod C-PTP and a negative d-dimer by study cutpoint, but this is indeed overall more than the 2000 they set out to get.

Of interest, only 2.3% of these 2017 patients had a high C-PTP, and only 10.8% had a moderate C-PTP (irrespective of d-dimer result). This, I think, gives us a good insight into the study population; >85% were low risk. Remember, the inclusion criteria were allcomers with signs or symptoms of PE. This worries me a little that recruiting centres made a decision to consciously or subconsciously include/recruit only those who they deemed to be at very low risk for PE. This compounds the concerns above, about convenience versus consecutive recruitment. I’d be interested to know if these numbers are generalisable to other countries and clinical settings. 

In the low C-PTP group (Wells 4 or below) with a positive d-dimer result (>1000ng/ml), 18.6% had PE on initial clinical imaging. Of the group discharged after initial imaging was negative (but remember, who had a positive d-dimer), 1 (0.5%) had a VTE confirmed at 90 day follow up. 

In the moderate C-PTP (Wells 4.5 to 6) group with a positive d-dimer result (>500ng/ml), 24.2% had PE on initial clinical imaging. Of the group discharged after initial imaging was negative, 0 had a VTE confirmed at 90 day follow up. 

In the high C-PTP group (Wells >6), 40.4% had PE on initial clinical imaging. Of the group discharged after initial imaging was negative, 0 had a VTE confirmed at 90 day follow up. 

Other points of interest in the baseline characteristics? Patients were a mean age of 52 (18), and median duration of symptoms was 5 days. The authors also break the Wells score down by individual factors contributing to documented C-PTP category, which is a fascinating insight into the medical mind. Tachycardia was the most frequent positive variable in the low risk group, for example. PE was the most likely diagnosis in 98% of patients deemed high risk (remember less than half of this group actually had PE on imaging) and 83% in the moderate risk group (<25% had PE on imaging). Just goes to show. A total of 13 patients were lost to follow up. 

And what are the headline results?

Well, in 1325 patients with a low or moderate C-PTP and a negative d-dimer (below the clinical probability adjusted threshold), 0 had subsequent VTE within the next 90 days. This is good news. The knock on from this is that 315 chest imaging scans were avoided in patients with a low C-PTP (who had a d-dimer between 500-999) and 40 avoided in the moderate group (assuming they would routinely be imaged irrespective of d-dimer result). The authors suggest a relative reduction in the need for imaging of a third (from 51.9% with a conventional strategy, down to 34.3% with the PEGeD protocol).

There were no attributable deaths within the cohort. There were 2 cases of VTE on follow up; 1 in a patient with a low C-PTP, positive d-dimer (1200) and negative CTPA, who then went on to be diagnosed with a DVT during follow up, and second in a patient with low C-PTP, positive d-dimer and positive CTPA who went on to be diagnosed with recurrence despite treatment. Hardly a failure of the diagnostic strategy in question, I am sure you’ll agree. 

The authors also then went on to assess how their strategy would compare to other recently published options, such as the use of an age adjusted cut point in the over 50’s, or the YEARS algorithm​8​. This is where it gets very interesting. In the study cohort, both the age-adjusted and YEARS strategies also reduced chest imaging rates compared to a conventional strategy. In the low C-PTP group, PEGeD was the best by a pretty comfortable margin. However, the PEGed strategy actually resulted in more chest imaging compared to the others, in the moderate C-PTP group. Lastly, Age adjusted d-dimer and PEGeD faired equally in high C-PTP chest imaging rates, but YEARS avoided imaging in a further 10 patients here.

The difficulty with interpreting this last paragraph, is that we aren’t presented with any data on specificity – would the YEARS algorthim have missed some disease in these high C-PTP patients? It’s hard to tell. 

All sounds very impressive. Any concerns?

The authors rightly point out early on in the limitations that without capturing a screening log, there are worries about this cohort being a convenience sample. The concern here is that investigators would selectively recruit uncomplicated patients or people at fairly low risk of VTE, such that they could prove the success of their new strategy. The authors defend their results with a post-hoc analysis of excluded patients, showing a low VTE event rate. These patients of course got into the study (and were later excluded) therefore I am not sure they say much about those patients who the investigators did not approach/screen. A more convincing argument follows that the prevalence of PE by C-PTP category in this study, were similar to those found within other recent North American publications​9​. This supports the idea that the authors have not misjudged baseline risk through selective enrollment.   

I think it is worth also highlighting that this is not a randomised study. This raises 2 issues. First, although you may conclude from a piece of work like this that the PEGeD strategy is safe, you have not compared it contemporaneously to standard care. As such, you don’t really know whether patients have got better care or not, and we can’t really be sure that they have. If the additional chest imaging investigations in a standard care group picked up a few aortic dissections for example, that would be worth noting. Outcomes like that are not found in prospective management research such as this, and all benefits are theoretical compared to what the authors perceive would have been done, and what would have resulted. It is also difficult to use studies like this when evaluating the totality of the evidence; bodies like NICE will often exclude observational management studies, or downgrade them significantly due to lack of randomisation. 

The primary outcome in PEGeD is not mortality. And it is not singular, it is composite​10​, with does carry some baggage. Is it a patient related outcome measure? I would say so – patients would presumably be very clear that what they want from a diagnostic work up is to have a potentially life threatening pathology excluded, with the minimum fuss and investigation. Would they really be happy without imaging however, even if the risk of an event was deemed very low. Not for me to say. And this endpoint is industry standard and mirrors that of all other trials in this field, so just a discussion point really, rather than a criticism.

A bigger concern is the lack of patients in the moderate C-PTP with negative d-dimer group. The PEGeD authors combine 40 patients from this cohort alongside >1200 in the low risk C-PTP group. I am not certain this is a representative sample of those patients with moderate clinical risk, and would want to see more data here. It also leaves things quite confusing for those of us who use a dichotomised Wells score, such as that recommended by NICE. Which of these 40 patients had a Wells score of 5 or 6? They are the patients of interest for me in particular and the numbers I suspect are unfortunately too low to draw any serious conclusions..

Lastly – some of the numbers don’t quite add up. If you look at the patient flow diagram in figure 1 for example, you will see that 467 patients in the low C-PTP group and 179 in the moderate C-PTP group had chest imaging for investigation of PE. In table 1 where the same data is presented, these numbers are 465 and 179 respectively. Small differences, but they can make a big difference in a study like this. I think also, the PEGeD minus Age adjusted chest imaging sum for low C-PTP patients is wrong in Table 3. This should be a negative, surely? Little things, but a bit worrying to see in the NEJM.

Global heath perspective from Stevan Buijns

It is worth noting that clinical probability testing is not as straight forward in settings with high HIV and tuberculosis prevalence – such as seen in many low- and middle-income settings. Both diseases are associated with an increased risk in VTE. What makes this complex, is that these also substantially increase the differential diagnosis: HIV increases the risk for opportunistic infections, and tuberculosis causes haemoptysis, is strongly associated with COPD (which also increases VTE risk) and often results in a chronically abnormal chest x-ray which is difficult to interpret. Given that lower respiratory tract infection is the top cause of mortality on the African continent, dyspnoea is considered a very common acute presentation.

There is very little research on the topic of C-PTP for PE in low- and middle-income settings. A small, retrospective study reported poor correlation between the revised Geneva Score and finding PE in such a population.​11​Although C-PTP for PE is often applied in these settings, it should be used with due caution.

It goes without saying that resource limitations often dictate the diagnostic work-up – the absence of testing for d-dimer or imaging for PE poses significant challenges in low- and middle-income settings.

Bottom line – Do you think we should adopt this?

I think the PEGeD supports what most of us already think. That in patients with a low C-PTP for PE, the current d-dimer cutpoint is probably set too low. This data provides compelling evidence that in this low risk group, the bar can probably be raised higher than it can with age adjustment alone, can be raised higher in younger patients (not just the over 50’s) and can be raised up to where YEARS will allow. However, due to some minor concerns around selection/inclusion and the low numbers of patients with a moderate C-PTP, I am not sure I am willing to use the rule on this group. In the UK I am also fairly hamstrung by NICE guidelines​12​, which dichotomise the Wells score into unlikely (4 and below) and likely (>4) risk; if I went against this by using the suggested d-dimer cutpoint in someone with a Wells score of 5 (moderate risk by three tier / likely by dichotomised Wells) and there was any sort of adverse outcome, I would feel very very vulnerable indeed. Despite my protestations on the evidence base.  

As such – if my patients have a Wells score of 4 or below and I don’t really think they have PE, I will make sure to PEGeD or age adjust their d-dimer if over 50 and now consider PEGeDing their d-dimer threshold if under 50. I would hope this may reduce my rate of V/Q and CTPA imaging, but I will retain my gestalt and clinical skill – if I can’t find any other possible reason for the symptoms, I’ll pursue whatever imaging I think necessary.

Any last messages?

The low incidence of PE in this ambulatory cohort reinforces how often we look to exclude this disease, and how we still probably overconsider this issue. Remember to use other diagnostic strategies when working up these patients, including the PERC rule​13,14​ where appropriate (also featuring in the new NICE guidelines) and just because someone mentions the letters P and E, that doesn’t commit you to excluding the pathology. This is all about risk, gestalt and your clinical decision making. As we have said many times​15​. Some of this can be shared with the patient of course. Tell them what you are thinking. Then, they can tell you what they think. That matters.

Congrats again to Kerstin on a really impressive piece of work and we look forward to hearing what you all think. 

Cheers


Dan

References

  1. 1.
    Kearon C, de Wit K, Parpia S, et al. Diagnosis of Pulmonary Embolism with d-Dimer Adjusted to Clinical Probability. N Engl J Med. November 2019:2125-2134. doi:10.1056/nejmoa1909159
  2. 2.
    Hogg K. Outpatient diagnosis of pulmonary embolism: the MIOPED (Manchester Investigation Of Pulmonary Embolism Diagnosis) study. Emergency Medicine Journal. February 2006:123-127. doi:10.1136/emj.2005.027110
  3. 3.
    Linkins L-A, Choi PT, Douketis JD. Clinical Impact of Bleeding in Patients Taking Oral Anticoagulant Therapy for Venous Thromboembolism. Ann Intern Med. December 2003:893. doi:10.7326/0003-4819-139-11-200312020-00007
  4. 4.
    Horner D. VTE Masterclass. St Emlyn’s. https://www.stemlynsblog.org/vte-masterclass-with-dan-horner-at-rcem15/. Published 2015. Accessed December 2019.
  5. 5.
    Righini M, Van Es J, Den Exter PL, et al. Age-Adjusted D-Dimer Cutoff Levels to Rule Out Pulmonary Embolism. JAMA. March 2014:1117. doi:10.1001/jama.2014.2135
  6. 6.
    Rezzaie S. The Years Study. REBEL EM. https://rebelem.com/the-years-study-simplified-diagnostic-approach-to-pe/. Published 2019. Accessed December 2019.
  7. 7.
    NICE U. Venous thromboembolic diseases: diagnosis, management and thrombophilia testing (2019). NICE. https://www.nice.org.uk/guidance/indevelopment/gid-ng10087/consultation/html-content-2. Published 2019. Accessed December 2019.
  8. 8.
    Pol LM, Dronkers CEA, Hulle T, et al. The            YEARS            algorithm for suspected pulmonary embolism: shorter visit time and reduced costs at the emergency department. J Thromb Haemost. March 2018:725-733. doi:10.1111/jth.13972
  9. 9.
    Sharif S, Eventov M, Kearon C, et al. Comparison of the age-adjusted and clinical probability-adjusted D-dimer to exclude pulmonary embolism in the ED. The American Journal of Emergency Medicine. May 2019:845-850. doi:10.1016/j.ajem.2018.07.053
  10. 10.
    Cordoba G, Schwartz L, Woloshin S, Bae H, Gotzsche PC. Definition, reporting, and interpretation of composite outcomes in clinical trials: systematic review. BMJ. August 2010:c3920-c3920. doi:10.1136/bmj.c3920
  11. 11.
    Bulajic B, Welzel T, Vallabh K. Clinical presentation and diagnostic work up of suspected pulmonary embolism in a district hospital emergency centre serving a high HIV/TB burden population. African Journal of Emergency Medicine. September 2019:134-139. doi:10.1016/j.afjem.2019.05.003
  12. 12.
    NICE U. Pulmonary Embolis,. NICE. https://cks.nice.org.uk/pulmonary-embolism. Published January 2019. Accessed December 2019.
  13. 13.
    MD C. PERC Rule. MD CALC. https://www.mdcalc.com/perc-rule-pulmonary-embolism. Published 2019. Accessed December 2019.
  14. 14.
    Kline JA, Mitchell AM, Kabrhel C, Richman PB, Courtney DM. Clinical criteria to prevent unnecessary diagnostic testing in emergency department patients with suspected pulmonary embolism. J Thromb Haemost. August 2004:1247-1255. doi:10.1111/j.1538-7836.2004.00790.x
  15. 15.
    Carley Si. Making good decisions in the ED. St Emlyn’s. https://www.stemlynsblog.org/making-good-decisions-in-the-ed-rcem15/. Published 2015. Accessed 2019.

How you can support St Emlyn’s


Posted by Dan Horner

Dr Daniel Horner BA MBBS MD PgCert MRCP (UK) FRCEM FFICM is an editorial board member on the St Emlyn’s blog and podcast. He is Professor of Emergency Medicine of the Royal College of Emergency Medicine. He is a consultant in Emergency Medicine and Intensive Care at Salford Royal NHS Foundation Trust. He is chair of the national exemplar centre Thrombosis Committee and Regional lead for Injuries and Emergencies on the NIHR Clinical Research Network. He is a Senior clinical lecturer at the University of Manchester and collaborator with the University of Sheffield. You can find him on twitter as @RCEMProf

Thanks so much for following. Viva la #FOAMed

Translate »