Diagnosis of venous thromboembolism in pregnancy: a study in extrapolation or a science in evolution?

Venous thromboembolism (VTE) remains a major preventable cause of maternal morbidity and mortality in developed nations [1–3]. There are no prospective studies investigating appropriate diagnostic algorithms for the diagnosis of deep venous thrombosis (DVT) and pulmonary embolism (PE) in pregnant women. There are several reasons for this, including concerns surrounding radiation risks to the fetus with diagnostic procedures and the fact that performing studies in pregnant patients is difficult because ‘experimentation’ in pregnancy is still not widely acceptable to women, their physicians or research ethics committees. Furthermore, the prevalence of VTE is low in pregnant women, making the requisite sample size prohibitive.

Owing to the paucity of definitive diagnostic studies in pregnant patients, diagnostic algorithms for DVT or PE derived for nonpregnant patients [4–6] have been extrapolated to pregnant patients. For reasons detailed later, this might be inappropriate.

In the last 50 years, the diagnosis of DVT has evolved from clinical assessment alone to use of the ‘gold standard’ – leg venography – to use of less-invasive techniques, such as compression ultrasound (CUS). Although CUS is highly sensitive for detecting proximal (popliteal vein or higher) DVTs in the thighs, it may be limited in its ability to detect clots isolated to the pelvis or calf veins [7]. To mitigate the decreased sensitivity of CUS for the latter clots, early studies proposed (and validated) the use of serial CUS testing over several days to detect the calf vein clots that propagate into the proximal veins [8,9]. In the last 10–15 years, structured prediction rules (low scores indicate that a clot is probably absent) were developed to help clinicians estimate the probability of DVT [10].

D-dimer testing has also been investigated for its role in excluding DVT. Owing to a lack of standardization of D-dimer testing among different D-dimer assays, cut-off points for each assay must be validated before they are used to manage patients [11]. By combining the use of structured prediction rules with D-dimer testing, patients can be triaged into those who require no further testing (negative D-dimer and low or moderate pretest probability) and those who should have CUS (everyone else) [12,13].

Can we extrapolate results from these studies to pregnant patients? Normally, D-dimer levels increase as pregnancy progresses [14–17]. D-dimer levels are further increased in pregnancy-associated conditions such as hypertensive disorders and bleeding, as well as during normal labor [18]. This limits the use of D-dimer testing in pregnant women with suspected DVT. Current cut-off points for DVT, established for several ‘sensitive’ D-dimer assays, in nonpregnant patients, are exceeded in normal asymptomatic pregnant women after the first trimester of pregnancy [16,17]. Until cut-off points are established for these assays in pregnant women, the use of D-dimer testing is limited by the high proportion of false-positive results.

The derivation of all structured prediction rules for DVT diagnosis excluded pregnant patients [10]. Pregnant patients are younger and less likely to have concurrent comorbidities, such as malignancies or surgeries, compared with the general population [10]. Pregnant patients frequently develop leg swelling that is not associated with DVT. In addition, certain conditions that are specific to pregnancy, such as ovarian hyperstimulation syndrome, which are risk factors for VTE [19], are omitted in the prediction rule. It is, however, reassuring to note in a recent retrospective study that clinicians’ clinical assessment based on subjective criteria was reasonably accurate in predicting DVT in pregnant patients [20]. In that same study, it was also suggested that a few objective variables might ‘predict’ the presence of DVT [20]. These variables include left leg presentation, asymmetry of the calf circumference (≥2 cm) and symptomatic in the first trimester of pregnancy. The results from this study, however, require prospective validation.

With the lack of studies investigating the utility of D-dimer testing and the need to validate the aforementioned clinical prediction rule, the current approach for the diagnosis of DVT in pregnant women should rely primarily on CUS [4]. When a pregnant patient presents with suspected DVT, CUS should be performed, visualizing the entire proximal venous system. As stated earlier, CUS, which is sensitive for the detection of proximal clots, may be limited in its ability to diagnose clots isolated to the calf or pelvis. With the use of serial CUS testing over 7 days, calf vein clots should be diagnosed when they propagate above the knee, into the proximal (popliteal or more proximal) deep veins; these extending calf DVTs (∼25% of patients with calf DVT) are associated with high potential for PE. The accuracy of CUS to detect isolated pelvic vein thrombi is uncertain. This is important because pelvic vein thrombi have a high risk of embolization, and they have been reported with high frequency in pregnant women who develop DVT [21–24]. Beyond CUS, there are few imaging options for DVT diagnosis in pregnancy. If isolated pelvic vein thrombosis is suspected in a pregnant patient (i.e., groin pain with asymmetric leg swelling and pain), direct visualization with ultrasound could detect some abnormalities within pelvis veins, and valsalva maneuvers could further reveal the presence of these clots [25]. Previously, contrast-enhanced venography, the gold standard, had been performed with lead shielding of the abdomen in pregnant patients [21–23]. Since CUS has largely replaced contrast venography in most centers, many radiologists have lost their skills or never became adept at performing venography. Owing to concerns surrounding fetal radiation, contrast computed tomography (CT) is not an ideal option if pelvic clots are suspected in an undelivered patient, owing to the associated high fetal (and maternal) radiation exposure [4]. An alternative tool to detect the presence of DVT is MRI [26]. This test is likely to be accurate for pelvic and calf DVT in pregnant patients but the long-term effects of exposure, as well as the safety of the contrast agent required are, as yet, undetermined [27]. In addition, the test is very expensive and may not be readily available in many centers.

As with the diagnosis of DVT, the approach to diagnosis of PE in pregnant women relies on studies performed in nonpregnant patients [4]. As with the diagnosis of DVT, the use of a structured clinical prediction rule and D-dimer assays plays a central role in stratifying patients into various risk groups prior to the use of diagnostic imaging [28,29]. Here, again, the use of the structured prediction helps exclude PE [30], and D-dimer testing has not been studied in pregnant patients [28,29].

For the diagnosis of PE in nonpregnant patients, the use of ventilation–perfusion (V–Q) as an imaging modality, has gradually been supplanted by spiral CT [31]. Unlike V–Q scan, spiral CT scan is usually more accessible after hours, is as accurate as V–Q scanning and can also detect nonthrombotic abnormalities in the chest [31].

The choice between V–Q scan versus CT scan for PE diagnosis in pregnant patients is less clear. The ability of V–Q scan for diagnosing PE in pregnant patients has been described in several retrospective studies [32,33]. In one such study, the distribution of the results of V–Q scanning among 124 pregnant women with suspected PE was 74% normal, 24% nondiagnostic and 2% high probability [32]. The low prevalence of high probability and the high prevalence of normal scans was supported in another retrospective study, in which 92% of the scans were normal and only 1% of the patients had PE [33]. The incidence of subsequent VTE events following normal scans and nondiagnostic scans in pregnant women who were not treated with anticoagulation was low [32]. These studies suggest that the prevalence of PE is low among pregnant patients who present with suspicious symptoms, unlike those seen for the studies in the general population where the incidence of PE is 6–23% [28–30].

The accuracy of spiral CT scans for the diagnosis of PE in pregnant patients is not known, but is likely to be similar to nonpregnant patients. The estimated fetal radiation exposure in the first two trimesters of pregnancy associated with spiral CT when compared with the V–Q scan is marginally lower (0.003–0.077 vs 0.07–0.170 mGy) [34]. The fetal radiation exposure is similar, however, by the third trimester of pregnancy (0.051–0.131 vs 0.07–0.170 mGy ) [34]. Although no ‘safe’ threshold of fetal radiation can be defined, these levels of fetal radiation exposure is considered ‘low’ when compared with an exposure of 50 mGy or more, above which the risk of childhood malignancies is increased [35].

The use of spiral CT scan for PE diagnosis also requires the use of iodinated contrast agent, which can cross the placenta, potentially affecting fetal thyroid tissue [36,37]. The use of spiral CT scan is also associated with significant maternal breast radiation exposure (∼20 mGy) [38]; at this level, an increased risk of breast cancer cannot be excluded [38,39]. With at least three-quarters of all presenting pregnant women having no PE or any intrathoracic abnormalities, the use of the spiral CT scan, exposing maternal breast tissue to high doses of radiation may not be warranted.

Currently, the approach to the diagnosis of pregnant patients with suspected PE should be the use of V–Q scan. In most patients (>70%), the initial finding with the V–Q scan would be normal and no further testing would be required. In the uncommon situation when the V–Q scan is high probability, PE is diagnosed and the patient should be treated. In patients with nondiagnostic scans, with a low clinical suspicion of PE (i.e., in the presence of other concurrent conditions, such as pneumonia or asthma), serial CUS over 7 days may be a safe approach to further imaging [28,30]. In the remaining subset of patients with nondiagnostic scans in which PE could not be excluded based on clinical suspicion, spiral CT can be considered. In this situation, a balanced discussion weighing the risks of undiagnosed PE and both maternal and fetal radiation exposure should be conducted with the patient.

Conclusion

The diagnosis of VTE in pregnant women still poses challenges to clinicians. While VTE is widely recognized as a condition that occurs more frequently during pregnancy and can result in maternal morbidity and mortality, surprisingly few studies are available to guide clinicians in the appropriate diagnosis of PE and DVT in these patients. Currently, CUS and V–Q remain the primary imaging modalities for the DVT and PE diagnosis in pregnant patients.

While much research has occurred within the area of VTE diagnosis in nonpregnant patients, particularly in the use of ancillary tests, such as prediction rules and D-dimer testing, no equivalent studies are available for pregnant patients. The extrapolation of results from these studies in nonpregnant patients to pregnant patients is currently uncertain. Since fetal radiation exposure from imaging tests is always of utmost concern, research into the use of these ancillary tests or any ‘noninvasive’ test should be considered a priority.

Financial & competing interests disclosure

Jeffrey S Ginsberg is recipient of a Career Investigator Award from the Heart and Stroke Foundation of Ontario and the Braley and Gordon Chair for research of Thromboembolic Disease. The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.

No writing assistance was utilized in the production of this manuscript.