Derivation and Validation of a Prediction Model for Venous Thromboembolism in Primary
Care
Venous thromboembolism (VTE) is diagnosed in the outpatient setting in at least 70%
of the cases[1]
[2] and prevention of ambulatory cases might substantially contribute to a reduction
of its socio-economic burden.[3]
[4] Notwithstanding, as we enter the second decade of this millennium, the evidence
that was generated after decades of clinical research remains unable to support decision-making
beyond in-hospital thromboprophylaxis. Clearly, the exposure to major risk factors
for thrombosis is highest during hospitalisation, and this is when the highest absolute
rates of VTE are observed. Prophylactic anticoagulation is, therefore, routinely recommended
based on the usually favourable benefit-to-risk ratio in this setting.[5] On the other hand, if we turn our attention to primary VTE prevention in the non-hospitalised
population, only selected patient groups with active cancer have been targeted by
clinical trials in view of their substantial baseline risk of developing VTE.[6]
[7] Although the VTE risk of individuals without cancer might also suffice for considering
primary thromboprophylaxis in primary care, current evidence falls short of quantifying
this risk and reliably identifying patients who may benefit from pharmacological preventive
strategies.
In this issue of Thrombosis and Haemostasis, Dentali et al make a new attempt to identify predictors of VTE in primary care.[8] Their risk assessment model was derived using data from a large Italian database
of more than one million adults followed by 1,100 general practitioners. After derivation
and internal validation, they performed external validation in an independent cohort
used by local authorities for health care assessment. The analysis was conducted as
a nested case–control study, where VTE diagnoses were defined by a combination of
International Classification of Diseases-9th Edition codes. Control patients who did
not develop VTE during same-length follow-up were randomly matched to VTE cases within
each risk set. The main finding of the study by Dentali et al is that patients who
had recently been hospitalised, admitted to the emergency room, or had suffered fracture,
stroke, acute infection, or prior VTE, had an at least twofold higher risk of suffering
VTE during follow-up. To make their risk assessment model more practical and facilitate
clinical decisions, the authors went further by developing a classifier for patients
into the different risk categories.
In a world where new clinical scores are constantly developed, published, and then
frequently discarded as clinically irrelevant, the authors must be commended for scrutinising
their risk assessment model by determining its discrimination, calibration and potential
clinical benefit if it were to be used for thromboprophylaxis.[8] For the readers who are not familiar with these parameters, discrimination corresponds
to the probability of correctly classifying patients into those who will and those
who will not have the outcome, in this case VTE. Discrimination alone, however, has
no clinical utility and is a poor method for comparing risk assessment models.[9] Moreover, the minimum threshold for defining the adequate level of discrimination,
as reflected by the concordance statistics (or c-statistics), may largely vary across
different clinical settings. In contrast, calibration is a measure of “absolute accuracy”
and possibly more important for making individual-level decisions, as it refers to
how closely the predicted VTE risk matches the observed VTE risk. In an additional
decision curve analysis,[10] the authors provided initial proof that using this model in decision-making concerning
thromboprophylaxis might provide a benefit, in terms of both VTE and bleeding risk,
compared with treating all patients or treating none.[8]
So, how should these results be interpreted in the context of the available strategies
for primary and secondary VTE prevention? Concerning primary prevention, the first
important comment on the present study is that the strongest predictors of VTE in
primary care were related to recent hospitalisation or to other conditions that would
have received thromboprophylaxis anyway based on current standards. Therefore, this
model may be more helpful for identifying candidates for extended thromboprophylaxis than for truly primary VTE prevention. Such an improved selection
model may indeed be necessary, particularly since recent major trials ([Fig. 1]) yielded rather equivocal results on who, among the medically ill patients, should
receive extended anticoagulant prophylaxis after discharge from hospital.[11]
[12]
[13]
[14]
[15]
Fig. 1 Trials on thromboprophylaxis in ambulatory (recently hospitalised) medically ill
patients ADOPT, Apixaban Dosing to Optimize Protection from Thrombosis trial; APEX,
Acute Medically Ill VTE Prevention With Extended Duration Betrixaban study; DVT, deep
vein thrombosis; EXCLAIM, Extended Prophylaxis for Venous ThromboEmbolism in Acutely
Ill Medical Patients With Prolonged Immobilization trial; MAGELLAN, Multicenter, Randomized,
Parallel Group Efficacy and Safety Study for the Prevention of Venous Thromboembolism
in Hospitalized Acutely Ill Medical Patients Comparing Rivaroxaban with Enoxaparin
trial; MARINER, Medically Ill Patient Assessment of Rivaroxaban versus Placebo in
Reducing Post-Discharge Venous Thrombo-Embolism Risk trial; PE, pulmonary embolism;
VTE, venous thromboembolism.
The second comment relates to the other major predictor that the authors identified
was a prior diagnosis of acute VTE. In the era of (low-dose) oral anticoagulation
for the long-term secondary prevention of VTE, an increasing number of patients will
be candidates for extended anticoagulation after a first episode of acute VTE.[16]
[17] The results of a meta-analysis of clinical trials showed that the use of non-vitamin
K oral anticoagulants for extended anticoagulation was associated with a reduction
in overall mortality.[18] Therefore, it is possible that the scenario that general practitioners will face
in a few years from now will be much different from that of the present study.
Third, a potential discrepancy between the setting of the current study and evolving
clinical scenarios concerns patients with cancer, a factor not recognised in the present
study as a potential predictor of VTE in primary care. Several clinical and statistical
reasons may explain this phenomenon. The most obvious is that patients with cancer
were likely to have already received anticoagulation based on their perceived higher
thrombotic risk and therefore were spuriously classified as being “at no risk” for
VTE. The same argument may also apply to other established VTE risk factors. A recent
practice-based study confirmed that these factors do influence the physicians' decision
to opt for prolonged post-discharge prophylaxis.[19]
The results of the study by Dentali et al should be seen as hypothesis generating
and necessitate further investigation in the setting of an interventional study. What
they highlight is that current preventive strategies appear insufficient to cover
the entire spectrum of patients at risk for VTE, since this risk clearly extends beyond
the period of hospitalisation. As the burden of VTE remains substantial[20] and global public awareness low,[21] such a tool may serve to attract the attention of general practitioners and stimulate
them to increase the level of VTE suspicion, with implications not only for VTE diagnosis
and management but also for primary VTE prophylaxis in primary care.
