Enter the data as you have it — vitals, scores, D-dimer, imaging, biomarkers. The pathway lights up progressively and carries you from first assessment to risk category, treatment and disposition.
The pathway tells you what to do. This chapter is about why — why we start with a probability and not a test, why the D-dimer cutoff moves, why a positive scan is not always a diagnosis, and where the pathway changes shape in pregnancy. The flagship section is the one question this tool gets asked most: should I send a D-dimer in a pregnant woman? It is meant to be read before the shift, not during it.
A 29-year-old woman, 24 weeks pregnant, comes in breathless with left-sided pleuritic chest pain. Heart rate 104, saturations 97% on air, chest clear, calves soft. She is frightened. The FY2 has already sent a D-dimer “to be safe”; it comes back at 1,900 µg/L FEU. A colleague says that means she needs a CTPA now. Another says a D-dimer is meaningless in pregnancy and should never have been sent. The registrar wants to start treatment-dose LMWH while everyone argues.
By the end of this chapter you should be able to say who is right — and why the safest next step was decided before that D-dimer was ever taken. It is answered at the bottom.
Every step in the tool, and the reason it is there.
| What we do | Why | Basis |
|---|---|---|
| Assess probability before any test | The same D-dimer means different things at different pretest probabilities. A test read without a prior is a test misread. | NG158 Christopher |
| PERC to avoid testing at all, only when suspicion is low | In a patient whose overall clinical suspicion is low, eight negative criteria make PE rare enough that even a D-dimer does more harm (false positives → scans) than good. | NG158 1.1.16 PROPER |
| Otherwise: Wells + D-dimer (fixed or age-adjusted) | The NICE pathway. A fixed 500 µg/L FEU cutoff over-scans older patients, so NICE suggests considering age adjustment over 50. | NG158 1.1.14, 1.1.17–21 ADJUST-PE |
| Or, as a separate strategy: YEARS | A self-contained alternative in which the item count sets the D-dimer threshold. It is not in NG158; use it only as a local protocol, and never mix it with a Wells score or age adjustment. | YEARS |
| CTPA (or V/Q) only after the prior is set | CTPA's predictive value depends on the clinical probability it is read against; a positive scan against a low prior can be a false positive. | PIOPED II |
| Risk-stratify before disposition | Confirmed PE is not one disease. Haemodynamics and RV strain, not the clot's size on CT, decide the pathway. | NG158 |
| Pregnancy: no D-dimer, image directly | UK guidance does not use D-dimer in pregnancy; the pathway starts with a leg ultrasound if there are DVT signs, then V/Q or CTPA. | GTG-37b |
| Pregnancy: LMWH while you investigate | The imaging takes time; the risk of untreated PE does not wait. Anticoagulate on suspicion unless contraindicated. | GTG-37b |
Why the pathway asks for a Wells or Geneva score before it will look at a D-dimer.
A D-dimer does not diagnose or exclude PE on its own. It modifies a probability you already hold. The Christopher study built the modern pathway on exactly this: dichotomise Wells into “PE unlikely” or “PE likely”, and only the unlikely group with a normal D-dimer can be sent home without imaging. In 3,306 patients, that combination — unlikely plus a normal D-dimer — carried a 3-month VTE risk of just 0.5% (95% CI 0.2–1.1). Christopher 2006
Wells and the revised Geneva score perform similarly; Geneva uses only objective items, which is why the tool offers it when recall of “is PE the most likely diagnosis?” feels too subjective. Teaching The single most gameable item in Wells is exactly that one — “PE is the most likely diagnosis” — and it is where anchoring creeps in.
Three-tier scores (low/intermediate/high) leave an ambiguous middle. Collapsing to “unlikely / likely” makes the D-dimer decision binary and reproducible, and it is the version validated in the large outcome studies. The tool keeps the raw score visible so you can see how close to the threshold a patient sits.
When the safest D-dimer is the one you never send.
In a patient whose gestalt probability is already low, a D-dimer's false positives generate CTPAs that find nothing, cost radiation and contrast, and occasionally find a subsegmental clot of doubtful significance. The PERC rule identifies the patients in whom that harm outweighs the benefit: if all eight criteria are negative and your gestalt is low, the risk of PE is under about 2% and no further testing is warranted.
NICE NG158 supports this. Recommendation 1.1.16 reads, verbatim: “If clinical suspicion of PE is low based on the overall clinical impression (from general medical history, physical examination and any initial investigations such as electrocardiography or chest X-ray), and other diagnoses are feasible, consider using the pulmonary embolism rule-out criteria (PERC) to help determine whether any further investigations for PE are needed.” It adds that PERC has not been validated in people with COVID-19. NG158 1.1.16 · [2020, amended 2023] The committee deliberately worded “low” as clinical gestalt rather than a percentage, although the largest trial (PROPER) defined it as an estimated risk below 15%.
PROPER, a cluster-randomised trial across 14 French EDs, tested this prospectively: a PERC-based strategy was non-inferior to conventional work-up, with one missed PE (0.1%) in the PERC arm, while cutting CT use (13% vs 23%) and shortening ED stay by 36 minutes. Freund 2018
PERC only applies when your unstructured suspicion is already low. Applying it to a moderate- or high-probability patient to justify not testing is a misuse of the rule, not a shortcut through it.
PERC is decided on gestalt, not on a Wells score: an earlier version of this tool opened PERC only below a Wells score of 2, and that has been removed. If any PERC criterion is positive, PERC has done its job and is finished. You then start one formal strategy from scratch, either Wells + D-dimer (NICE) or YEARS (local protocol). Nothing from PERC carries into that score.
Age adjustment and YEARS both raise the bar safely, but they belong to different strategies and must never be stacked.
D-dimer is sensitive but not specific. Its specificity collapses with age, malignancy, pregnancy, sepsis, surgery and inflammation — the very patients who present breathless. A fixed 500 µg/L FEU cutoff therefore rules out fewer and fewer people as they get older, sending them to CT instead.
ADJUST-PE, which used a Wells or Geneva score first, raised the cutoff to age × 10 µg/L (FEU) for patients aged 50 or over. NICE NG158 1.1.14 says to consider an age-adjusted threshold for people “aged over 50”; at exactly 50 both give 500, so the wording difference changes no number. Among those 75 and older with a non-high probability, this lifted the proportion in whom PE could be excluded from 6.4% to 29.7% — without loss of safety: the failure rate in the newly-excluded band was 1 of 331 (0.3%). Righini 2014
YEARS does something subtler: it makes the cutoff depend on how worried you are. Three items — clinical signs of DVT, haemoptysis, and “PE the most likely diagnosis” — set the threshold. With no items, PE is excluded below a D-dimer of 1,000 ng/mL (= µg/L) FEU; with one or more, below 500 ng/mL FEU. YEARS is used instead of a Wells or Geneva score, not after one, and its thresholds are never age-adjusted. It is not part of NICE NG158. In 3,465 patients this avoided CTPA in 48% versus 34% with the Wells pathway, at a 3-month failure rate of 0.61% (0.36–0.96). van der Hulle 2017
| Strategy | Who | Rule-out threshold | Status | Never |
|---|---|---|---|---|
| PERC | Low clinical suspicion on gestalt, with another diagnosis feasible | No test: all 8 criteria negative | NG158 1.1.16 consider | Use it on a Wells score, or in moderate/high suspicion |
| Wells (or Geneva) + D-dimer | Everyone else who is stable. NICE uses two-level Wells: ≤4 → D-dimer, >4 → CTPA | 500 µg/L FEU, or age × 10 µg/L FEU over 50 | NG158 1.1.14, 1.1.17–21 (Geneva is not in NG158) | Apply YEARS thresholds to it |
| YEARS | Instead of Wells, where it is local protocol | 1000 µg/L FEU (0 items) · 500 µg/L FEU (≥1 item) | Not in NG158 | Age-adjust it, or add a Wells score |
The failure rates of age-adjusted and YEARS pathways sit alongside the fixed-cutoff pathway, not above it. The higher threshold is not “accepting more risk” — it is correcting a cutoff that was set for a younger, healthier population than the one in front of you.
Why a positive CTPA is not always a diagnosis, and where V/Q still belongs.
CTPA is excellent but not perfect: in PIOPED II its sensitivity was 83% and specificity 96%, rising to 90%/95% when combined with venous-phase imaging. Stein 2006 The crucial finding was that its predictive value depended on the clinical probability it was read against: a positive scan carried a 96% positive predictive value when the clinical assessment agreed, but was non-diagnostic when they were discordant. A positive CTPA in a genuinely low-probability patient may be a false positive — this is the origin of over-called subsegmental PE.
V/Q (or perfusion-only) scanning retains a role when CTPA is contraindicated or the chest radiograph is normal — most importantly in younger patients and in pregnancy, where the radiation trade-off differs. Teaching
Haemodynamics and RV strain decide disposition, not the clot on the picture.
Once PE is confirmed, the question changes from “is it there?” to “how sick is this patient?” A saddle embolus in a well, normotensive patient with a normal right ventricle may be lower-risk than a small clot in a shocked one. The tool follows the standard tiers: high-risk (haemodynamic instability → consider thrombolysis), intermediate (RV dysfunction and/or raised troponin, normotensive → admit, monitor), and low-risk (sPESI 0, no RV strain → consider early discharge with anticoagulation). The decision is physiological, not radiological.
What UK guidance actually says, why it says it, and what the international evidence adds.
Pregnancy is where this pathway changes shape. The pretest tools were not derived in pregnant women, the physiology raises D-dimer in almost everyone, and the imaging carries two people's radiation. It is also where the most common single question arises: should I send a D-dimer?
The RCOG is unambiguous. Green-top Guideline 37b states, verbatim: “D-dimer testing should not be performed in the investigation of acute VTE in pregnancy.” GTG-37b · Grade D RCOG 2015 The same guideline adds that “there is no evidence to support the use of pretest probability assessment in the management of acute VTE in pregnancy” Grade C. In other words, UK practice removes both the score and the D-dimer and goes straight to imaging.
1. ECG and chest X-ray for every suspected PE. 2. If there are also signs or symptoms of DVT, do a compression duplex ultrasound of the legs first — if it confirms DVT, treat, and no chest imaging is needed. 3. If there are no DVT signs, do a V/Q scan or CTPA. 4. If the CXR is abnormal, choose CTPA over V/Q. 5. Start treatment-dose LMWH immediately on clinical suspicion and continue until PE is objectively excluded, unless strongly contraindicated. GTG-37b RCOG 2015
The V/Q-versus-CTPA choice is a radiation trade-off, and the RCOG says patients should be counselled on it: compared with CTPA, V/Q “may carry a slightly increased risk of childhood cancer but is associated with a lower risk of maternal breast cancer; in both situations, the absolute risk is very small.” Grade D
Two reasons. First, D-dimer rises physiologically through normal pregnancy, so a “positive” result loses almost all of its specificity — it sends nearly everyone to CT and rules out almost no one. Second, and more decisively, the UK's own diagnostic-accuracy study, DiPEP, looked for a usable threshold in 310 pregnant and postpartum women and found none: D-dimer's area under the curve was 0.668 by laboratory ELISA and 0.651 near-patient, and the authors concluded that conventional and candidate biomarkers had “no utility either for their negative or positive predictive value” and that “no diagnostically useful threshold” existed. Hunt 2018 That is the evidence behind the RCOG's Grade D line.
Here is the honest counterpoint, and the reason this question keeps coming up. Two large prospective management studies have shown that a structured pregnancy-adapted pathway that includes D-dimer — with a raised or conditional cutoff, always coupled to clinical criteria and leg ultrasound — can safely avoid some CT scans.
The European Society of Cardiology's 2019 guideline reflects this and considers a validated pregnancy-adapted algorithm — including D-dimer — a reasonable option. ESC 2019 This is the “international evidence” that sits behind the question.
The disagreement is narrower than it looks. Everyone agrees a raw, unstructured D-dimer in pregnancy is unhelpful — DiPEP proves it and Artemis never used one that way. The international studies do not endorse “send a D-dimer and act on it”; they validate a whole pathway in which D-dimer is one gated step alongside clinical criteria and leg ultrasound. Until UK guidance changes, the safe position on a live shift is: do not rely on a D-dimer to rule PE in or out in a pregnant woman, and never let a raised one push you toward a scan a structured assessment did not justify — or a normal one reassure you out of imaging UK practice requires. If your unit runs a validated pregnancy-adapted protocol, follow it as a protocol, not as an isolated test.
Anticoagulation in pregnancy is LMWH, dosed on booking or early-pregnancy weight. DOACs and warfarin are contraindicated (warfarin is teratogenic; DOACs cross the placenta and are not recommended). Start LMWH on suspicion and continue until PE is excluded. Massive PE with cardiovascular compromise is a multidisciplinary emergency: IV unfractionated heparin is the preferred initial agent, with thrombolysis considered for confirmed (or, in extremis, strongly suspected) massive PE. GTG-37b RCOG 2015
Normal oximetry is common in PE, especially in the young and in pregnancy. It does not exclude it.
UK guidance does not use it to exclude PE in pregnancy — a “normal” one is not a rule-out.
Physiologically expected. On its own it neither confirms PE nor mandates a scan.
Against a low prior it may be a false positive; against a high one it can still matter. The prior decides.
Only rules out with a normal D-dimer. Score alone is not an exclusion.
Burden on CT does not equal instability. A normotensive patient with a normal RV is not high-risk.
Pregnancy raises VTE risk severalfold; youth does not offset it.
The 29-year-old at 24 weeks. Here is what the chapter should let you say:
How much of this pathway rests on trials, and how far can each study carry the weight put on it? Every figure below was taken from the primary paper. The pregnancy section is set out at length because that is where guidance and evidence appear to diverge — and where the divergence is narrower than it looks.
The strength column is this module's appraisal, not a published grade.
| Question | Best evidence | Design | Strength |
|---|---|---|---|
| Probability + D-dimer to rule out | Christopher 2006 | Prospective management, n=3306 | High — large, outcome-based |
| PERC to avoid testing (NICE NG158 1.1.16: consider when suspicion is low) | PROPER 2018 | Cluster RCT, n=1916 | Moderate–high — randomised, non-inferiority |
| Conditional D-dimer cutoff (a separate strategy; not in NG158) | YEARS 2017 | Prospective management, n=3465 | High — large, outcome-based |
| Age-adjusted cutoff (NG158 1.1.14: consider over 50) | ADJUST-PE 2014 | Prospective management, n=3346 | High |
| CTPA accuracy | PIOPED II 2006 | Prospective accuracy, n=824 | Moderate — sensitivity 83%, prior-dependent |
| D-dimer in pregnancy | DiPEP 2018 | Observational biomarker cohort, n=310 | Moderate — shows no usable threshold |
| Adapted pathway in pregnancy | Artemis 2019; CT-PE-pregnancy 2018 | Prospective management, n=498 / 395 | Moderate — safe, but no RCT vs direct imaging |
Consecutive patients with suspected PE, managed by dichotomised Wells, D-dimer and CT. van Belle 2006
| “PE unlikely” + normal D-dimer | 1057 patients (32%); 3-month VTE 0.5% (95% CI 0.2–1.1) |
| PE excluded by negative CT | 3-month VTE 1.3% (0.7–2.0) |
| Missed PE at 3 months | 1 (0.1%) PERC arm vs 0 control — non-inferior |
| CT imaging | 13% vs 23% |
| ED length of stay | 36 minutes shorter |
Patients with a low gestalt clinical probability only; mean age 44. Freund 2018 This is the evidence behind NICE NG158 1.1.16, which recommends considering PERC when clinical suspicion is low.
| Items | DVT signs · haemoptysis · PE most likely |
| Cutoff | <1000 ng/mL (0 items) · <500 ng/mL (≥1 item). The paper gives no unit basis; labelled FEU here (assays: VIDAS, Tina-quant, STA-LIA, Innovance) |
| Design | YEARS was assessed instead of a Wells score, never on top of one, with no age adjustment |
| CTPA avoided | 48% vs 34% with Wells |
| 3-month failure | 0.61% (0.36–0.96); fatal PE 0.20% |
| Cutoff | age × 10 µg/L for ≥50 years, after a Wells or simplified Geneva score. The paper gives no unit basis; labelled FEU here (six assays incl. VIDAS, Tina-quant, STA-Liatest, Innovance). NICE NG158 1.1.14: “aged over 50” |
| Exclusion in ≥75 yr, non-high prob | 6.4% → 29.7% |
| Failure in the 500–cutoff band | 1/331 (0.3%) |
| CTA alone | sensitivity 83% · specificity 96% |
| CTA + CT venography | sensitivity 90% · specificity 95% |
| PPV, concordant clinical probability | 96% (high or low); 92% intermediate |
| PPV, discordant probability | non-diagnostic |
Blood samples from pregnant and postpartum women with suspected PE, testing conventional and candidate biomarkers against conclusive imaging. Hunt 2018
| D-dimer AUC (ELISA) | 0.668 |
| D-dimer AUC (near-patient) | 0.651 |
| Conclusion | “no diagnostically useful threshold”; biomarkers had “no utility” for negative or positive predictive value |
Three YEARS items, leg ultrasound if DVT symptoms, D-dimer <1000 ng/mL FEU (0 items) or <500 ng/mL FEU (≥1 item); CTPA only if not excluded. van der Pol 2019
| CTPA avoided | 39% (95% CI 35–44); 65% first trimester, 32% third |
| 3-month failure | 1 popliteal DVT (0.21%, 0.04–1.2); no PE missed |
Revised Geneva score, D-dimer, bilateral leg ultrasound, CTPA (V/Q if inconclusive). PE prevalence 7.1%. Righini 2018
| Excluded without CTPA | 46 women (low/intermediate prob + negative D-dimer) |
| 3-month failure, untreated | 0.0% (0.0–1.0) |
| YEARS applied to the cohort | 77/371 (21%) spared imaging, 0 missed Langlois 2019 |
The suspected-PE pathway is one of the better-evidenced in emergency medicine: probability-first testing, conditional D-dimer cutoffs and prior-dependent imaging all rest on large prospective data. Pregnancy is the exception where guidance and evidence appear to part company — but the gap is narrow. UK practice omits D-dimer because a raw one is worthless (DiPEP) and goes straight to imaging with immediate LMWH; the international evidence shows a structured pregnancy-adapted pathway can safely include D-dimer and spare some scans. On a UK shift, follow RCOG GTG-37b unless your unit runs a validated adapted protocol — and either way, never let an off-protocol D-dimer, high or low, override the imaging decision.
This teaching layer sits alongside the PE decision pathway. The pathway is a clinical decision-support tool; this section explains the reasoning behind it and appraises the evidence, with a flagship chapter on the D-dimer-in-pregnancy question.
Every figure in the Evidence tab was taken from the primary publication (PubMed abstracts via NCBI eutils, or the source guideline). The RCOG Green-top 37b recommendations are quoted verbatim from the guideline PDF. Nothing here was written from memory of a study.
UK practice (RCOG GTG-37b, 2015) does not use D-dimer in pregnancy and goes straight to imaging with immediate LMWH. International prospective studies (Artemis 2019, CT-PE-pregnancy 2018) show a structured pregnancy-adapted pathway that includes a gated D-dimer can safely spare some CT scans. This module teaches the UK standard first and explains the international evidence honestly, without endorsing a departure from local guidance on a live tool.
Following review by the ResusDoc Clinical Governance Lead, three targeted changes were made: (1) PERC is now presented as NICE NG158 1.1.16 describes it, applied only when clinical suspicion is low on gestalt, and kept separate from Wells + D-dimer and YEARS, which are now alternative strategies that are never chained; (2) every D-dimer threshold is labelled FEU or DDU, with a prompt to match your laboratory's units and cut-off; (3) NICE NG158 and local anticoagulation protocols are shown first, and the 2026 AHA/ACC guidance is labelled supplementary.
The whole subdomain — the decision tool, Teaching and Evidence — was clinically reviewed in October 2026 against current UK guidance by Dr Asim Ijaz, Consultant in Emergency Medicine and ResusDoc's clinical governance lead. The amendments he requested have been applied; his final sign-off of the amended version is pending. This is decision support, not a substitute for clinical judgement or local protocol.