RIETE Score for bleeding risk during treatment of pulmonary embolism

Clinical background

In acute venous thromboembolism (VTE), anticoagulation is always indicated, but the treatment carries a bleeding risk that is highest during the first months of therapy. The decision is rarely whether anticoagulation should be started, but which agent to use, how intensively to monitor, and, in the longer term, how long treatment should continue. The RIETE score was developed to structure this risk assessment using variables available at the time of diagnosis, without requiring advanced biomarkers or imaging. The score therefore addresses the baseline risk of major bleeding during the first three months, the period in which the bleeding risk is most pronounced and clinical decisions on agent and monitoring intensity are most pressing [1].

Calculating the RIETE score

The score rests on six clinical and laboratory variables that were independently associated with an increased bleeding risk on multivariable analysis in the derivation cohort:

$$\text{RIETE} = 2 \times \text{recent major bleeding} + 1 \times \text{active cancer} + 1{.}5 \times (\text{creatinine} > 1{.}2\ \text{mg/dL}) + 1{.}5 \times \text{anaemia} + 1 \times (\text{age} > 75\ \text{years}) + 1 \times \text{clinically overt pulmonary embolism}$$

The variables are binary (0 or the stated point value). Anaemia is defined as a haemoglobin <13 g/dL in men and <12 g/dL in women. "Recent major bleeding" refers to bleeding within one month before the VTE diagnosis. "Clinically overt pulmonary embolism" scores points, in contrast to isolated deep vein thrombosis.

The derivation cohort comprised 19,274 consecutive patients with objectively confirmed acute VTE in the international RIETE registry, enrolled between 2001 and 2008. The cohort was randomly divided into a derivation group (13,057 patients) and a validation group (6,572 patients). The figures in the original publication do not add up: the subgroups total 19,629, not 19,274, and the publication does not reveal which figure is the misprint. The 13,057 patients of the derivation group are however verifiable, since its risk groups (2,654 + 9,645 + 758) sum exactly to that number [1]. The outcome measure was major bleeding within three months of starting anticoagulation. In the derivation group, 314 patients bled (2.4%), of whom 105 fatally [1].

The risk categories according to the score are:

The likelihood ratio was 0.14 (95% CI 0.07–0.27) for low risk and 2.96 (95% CI 2.18–4.02) for high risk [1].

Interpretation in practice

The RIETE score should be interpreted as an aid to clinical judgement, not as the sole basis for a decision. No bleeding risk score has so far been tested in a clinical trial aimed at withholding anticoagulation on the basis of its result [3].

Low risk (0 points): The bleeding risk is very low, below 0.5% over three months. This supports standard dosing without any need for particularly intensified monitoring. The score should not be used to justify withholding anticoagulation, but it may support the choice of an agent with a favourable safety profile and help avoid unnecessary hospital stay.

Intermediate risk (1–4 points): The bleeding risk is around 2.5–3%, corresponding to the average risk in the VTE population. Clinical assessment of the individual patient determines whether further measures are needed. Reversible risk factors, above all renal function and anaemia, should be addressed. In patients with cancer, the choice of agent (low molecular weight heparin versus a direct oral anticoagulant) may need to be considered.

High risk (>4 points): The bleeding risk is raised, around 7%. This justifies careful monitoring, correction of reversible factors and consideration of the agent with the lowest bleeding risk. In patients with recent major bleeding, which alone scores 2 points, the benefit of anticoagulation must be weighed against the risk in discussion with the patient. In extreme cases a temporary inferior vena cava filter may be considered, but this is a clinical decision that cannot be supported by the score alone.

Validation and performance

In the internal validation cohort of the derivation study, the bleeding incidences were reproduced well: 0.1%, 2.8% and 6.2% for low, intermediate and high risk respectively [1].

A large comparative study within the RIETE registry, comprising 82,239 patients enrolled between 2001 and 2019, evaluated the RIETE score against the newer VTE-BLEED score over different time intervals. During the first 30 days after the VTE diagnosis, the AUC for RIETE was 0.71 (95% CI 0.70–0.73), compared with 0.69 (95% CI 0.67–0.70) for VTE-BLEED. The negative predictive value for low risk was 98.7%, while the positive predictive value for high risk was only 3.9%. Performance was similar during days 31–90 (AUC 0.70) and remained stable up to one year. The RIETE score performed somewhat better for extracranial bleeding, particularly early in treatment, while VTE-BLEED tended to be better for late intracranial bleeding [2].

An external validation in a Latin American cohort (470 patients, median age 65 years, 59% women, 57% deep vein thrombosis, 30% active cancer, treated mainly with direct oral anticoagulants or low molecular weight heparin) showed poorer performance. The AUC for RIETE was 0.64 (95% CI 0.51–0.76), below the conventional threshold for clinical usefulness. Calibration was poor: the observed bleeding rate was higher than predicted in every risk category, with an observed-to-expected ratio of 2.8, suggesting that the score underestimates the risk in this population. Sensitivity at the threshold of ≥4 points was only 33%, with a specificity of 94% [4].

In summary, discrimination is moderate in large registry cohorts (AUC approximately 0.70–0.71) but poorer in smaller external cohorts with a different case mix. The low positive predictive value means that the majority of patients classified as high risk do not bleed, which limits the score's ability to guide specific treatment decisions.

Limitations

The RIETE score captures only the baseline risk at the time treatment is started. The variables are static, and the score takes no account of clinical changes during treatment, such as improved renal function, a new cancer diagnosis or an intercurrent bleeding tendency. A patient initially judged to be at low risk may develop new risk factors, and the score should therefore not be used for long-term decisions without being reassessed.

The derivation cohort is dominated by patients treated with low molecular weight heparin and vitamin K antagonists, since direct oral anticoagulants were not established when the registry began in 2001. In the Colombian validation cohort, in which a larger proportion of patients were treated with direct oral anticoagulants, performance was poorer [4]. It is unclear whether the score's variables have the same predictive value in patients treated with modern anticoagulants that carry a lower bleeding risk overall.

The score classifies about 75% of patients as intermediate risk, the category in which clinical guidance is least directive. It is in this large middle group that the score has least decision value. The low positive predictive value for high risk (below 4% in large cohorts) means that the score cannot be used to identify patients in whom anticoagulation should be avoided [2].

The creatinine threshold of 1.2 mg/dL is the same for men and women, even though the reference range differs between the sexes. This may lead to systematic underestimation of renal impairment in women with lower muscle mass. The threshold corresponds roughly to an estimated eGFR of 50–60 mL/min depending on age and sex, but the conversion is not exact and the score uses raw creatinine values, not eGFR.

References

1. Ruíz-Giménez N, et al. Predictive variables for major bleeding events in patients presenting with documented acute venous thromboembolism. Findings from the RIETE Registry. Thromb Haemost 2008;100(1):26–31. PMID: 18612534
2. Lecumberri R, et al. Prediction of Major Bleeding in Anticoagulated Patients for Venous Thromboembolism: Comparison of the RIETE and the VTE-BLEED Scores. TH Open 2021;5(3):e319–e328. PMID: 34568742
3. Nopp S, Ay C. Bleeding Risk Assessment in Patients with Venous Thromboembolism. Hamostaseologie 2021;41(4):267–274. PMID: 33626580
4. Ortiz Gómez S, et al. Validation of the RIETE, Kuijer, and HAS-BLED Models to Assess 3-Month Bleeding Risk in Anticoagulated Patients Diagnosed with Venous Thromboembolic Disease. Clin Appl Thromb Hemost 2024;30. PMID: 39106353