Despite pulmonary embolism being the leading cause of pregnancy-related maternal mortality in the developed world,1–3 it remains a diagnostic dilemma. Multiple factors likely contribute, the primary of which is that high-level data informing physicians on diagnostic testing for pulmonary embolus are from trials that excluded pregnant women. Thus physicians, faced with the clinical suspicion of the life-threatening yet rare diagnosis of pulmonary embolism in the pregnant or postpartum woman, are required to extrapolate diagnostic testing strategies from nonpregnant study groups. The Prospective Investigation of PulmonaryEmbolism Diagnosis II trial4 provided Level I evidence that computed-tomographic angiography is the superior diagnostic test for pulmonary embolism compared with ventilation–perfusion scan, but the trial excluded pregnant woman from the study groups. Several leading experts in computed tomography (CT) radiology have suggested that although it is tempting to extrapolate these high-level data to pregnant women, the underlying physiologic changes in pregnancy, such as alterations in cardiac output, changes in plasma volume, and varied distribution of fluid between body fluid compartments, might alter the ability of computed-tomographic angiography to be diagnostic in this population. These alterations could lead to a higher nondiagnostic-study rate of computed-tomographic angiography than the 6%4 reported in the nonpregnant population. Proponents of computed-tomographic angiography in pregnancy cite the high rate of intermediate probability ventilation–perfusion scans in the general population, but these references universally rely on studies of older patients with much higher rates of underlying lung disease. The rate of nondiagnosis for imaging for pulmonary embolism in pregnant women is unknown.
In addition to the rate of diagnosis by radiologic tests that may be altered in pregnancy, the results of these tests are interpreted in the context of clinical suspicion, or “pretest” risk. Consideration must be given to the possibility that the clinical suspicion of pulmonary embolus in a pregnant or immediately postpartum patient may not be directly generalizable from a nonpregnant population.5 Physiologic changes in gestation can produce benign symptoms such as dyspnea, which can mimic or mask pulmonary embolism. Attempts have been made to assign pretest risk for pulmonary embolism based on the clinical presentation,5,6 but they have not been predictive in pregnant patients.7
Given the lack of data to identify the optimal diagnostic test for pulmonary embolism in pregnant women, we performed a retrospective cohort study to test the hypothesis that CT angiography has a higher rate of nondiagnosis in pregnant woman than ventilation–perfusion scan. We also assessed clinical factors and their ability to predict pulmonary embolism to identify factors that could be used clinically to adjust pretest risk.
MATERIALS AND METHODS
We performed a 5-year retrospective cohort study of all consecutive women, pregnant or within 6 weeks postpartum, who underwent radiologic testing for pulmonary embolism caused by clinical suspicion. Women were evaluated between 2001 and 2006 in the Mallinckrodt Institute for Radiology in Washington University in St. Louis Medical Center and underwent at least computed-tomographic angiography or ventilation–perfusion. After institutional review board approval was obtained from Washington University in St. Louis, the cohort was identified using International Classification of Diseases, 9th Revision diagnosis codes and procedure billing codes.
Computerized medical records within the 5 study years were searched using the billing codes for computed-tomographic angiography of the chest or thorax, and ventilation–perfusion scan. Patients were identified as pregnant or postpartum using International Classification of Diseases, 9th Revision codes for pregnancy, pregnant state, child birth, and puerperium. To assure complete identification of the study cohort, the radiology department census for the 5-year study period was used to confirm complete identification of the potential cohort.
The electronic charts for all identified patients were available for review. The medical records were reviewed, and women were excluded if they were not pregnant or not within 6 weeks from delivery. After identifying all women meeting inclusion criteria, two investigators (A.G.C. and M.J.S.) extracted all medical records using a closed-end data extraction tool, blinded to the final diagnosis. Data extracted included maternal demographics, medical and obstetric history, pregnancy, and complications. Detailed information was extracted on chief complaint, presentation, and clinical evaluation, including the signs and symptoms that led to radiologic evaluation for pulmonary embolism. Importantly, although the majority of women who were evaluated for pulmonary embolism during pregnancy went on to deliver at our institution, institutional review board approval was obtained to contact the few women who delivered elsewhere to assure complete ascertainment of follow-up information for the pregnancy and final maternal diagnosis.
Both imaging studies are performed in a standard fashion at our institution. Pulmonary embolism–protocol computed-tomographic angiography is performed with a 100–125-mL injection (adjusted for weight) of intravenous iodinated contrast (350 mg I/mL) at 4 mL per second by 20-gauge antecubital vein, using 4- and 16-row multidetector scanners. Automated bolus timing software is used, with a region of interest placed over the main pulmonary artery and triggering at 100 HU. Images are reconstructed at 1×1 mm and 5×5 mm and viewed on Picture Archiving and Communication System. Final diagnostic reads are established by an attending radiologist who specializes in chest CT and is blind to patient anticoagulation status. Nondiagnosis is defined as a final interpretation by a radiologist who cannot definitively determine if the patient did or did not have a pulmonary embolism caused by the technical limitations of the study. For quality assurance purposes, these studies were reevaluated by blinded radiologists who confirmed the interpretation of nondiagnosis. Ventilation–perfusion scans are performed using 10–30 mCi of Xe-133 gas for the ventilation portion. Technetium-99m macroaggregated albumin at a dose of 2.0–4.0 mCi (containing 200,000–600,000 particles) is used for the perfusion portion. Final diagnostic interpretation of the ventilation–perfusion scan is made by an attending radiologist specializing in nuclear imaging. Last, chest X-rays were obtained with an upright posteroanterior (PA) radiograph, and a second view from the lateral position was only obtained if the frontal view did not adequately assess all lung fields.
The two primary study groups were defined by the initial study obtained to diagnose pulmonary embolism: computed-tomographic angiography compared with ventilation–perfusion scan. The primary outcome of the study was the rate of nondiagnosis from the initial imaging modality. For computed-tomographic angiography, this was defined as a final attending radiologist's interpretation that the study was too technically limited to determine whether a patient had a pulmonary embolism. For ventilation–perfusion scan, this was defined as a final attending radiologist's interpretation as “moderate” or “low” probability. Descriptive statistics were used to estimate the proportion of nondiagnosis in the cohort and in each of the two study groups as well as final diagnosis of pulmonary embolism in the cohort and each group. Baseline characteristics of the two groups, those initially evaluated by computed-tomographic angiography, were compared with those evaluated by ventilation–perfusion and compared using Student t test for continuous variables and χ2 or Fisher exact test as appropriate for categorical variables. Stratified analysis was performed to identify potential confounding effects and interactions. Rate of nondiagnosis was also compared in a subgroup analysis, based on whether the initial chest X-ray was normal. Next, multivariable logistic regression models were fit to estimate the effect of initial diagnostic imaging modality on the rate of nondiagnosis in the main groups and subgroups, while adjusting for potentially confounding effects identified historically and by the stratified analysis. Regression models were created in a backward, step-wise fashion, removing potentially confounding variables that were not significant.
To assess whether specific clinical factors could be used to adjust pretest risk for pulmonary embolism in pregnancy, the clinical findings were also compared between women diagnosed with pulmonary embolism and those who were not. The frequencies of each of the factors was compared between the groups, using Student t test for continuous variables and χ2 or Fisher exact test as appropriate for categorical variables. Then the sensitivity, specificity, and positive and negative likelihood ratios were calculated for each factor. Finally, receiver operating characteristic curves were used to estimate whether any single factor, or several in combination, could efficiently predict pulmonary embolism and thus be used clinically to adjust pretest risk. All statistical analyses were performed using Stata 10, Special Edition (StataCorp LP, College Station, TX).
Of 305 women with a clinical suspicion of pulmonary embolism, one did not deliver at our institution and could not be reached for follow-up, leaving 304 for analysis. In the cohort, 18 (5.9%) women were ultimately diagnosed with a pulmonary embolism by completion of 6 weeks postpartum; five in first trimester, one in second trimester, five in third trimester, and seven postpartum. Evaluation was prompted with increasing frequency throughout pregnancy and postpartum: 10.9% first trimester, 16.1% second trimester, 38.5% third trimester, and 34.5% postpartum. The initial diagnosis of pulmonary embolism was sought by computed-tomographic angiography in 108 (35.1%) and by ventilation–perfusion in 196 (64.9%) women. Women initially evaluated by computed-tomographic angiography were more likely postpartum and tended to be less likely to use tobacco (Table 1).
Women who underwent computed-tomographic angiography tended to have a slightly higher rate of nondiagnostic study (17.0% compared with 13.2%, P=.38) (Table 2). Next we considered rate of nondiagnosis stratified by chest X-ray results in the 281 (92.4%) women who had a chest X-ray obtained before imaging, 262 (93.2%) of whom had a single-view chest X-ray. For women with a normal chest X-ray, there was a fivefold higher rate of nondiagnosis from a CT scan as compared with a ventilation–perfusion scan (relative risk [RR] 5.3, 95% confidence interval [CI] 2.1–13.8), and increased risk remained even after adjusting for gestational trimester or postpartum state, hypoxia, and chest pain (adjusted odds ratio [aOR] 5.4, 95% CI 1.4–20.1, P<.01). In contrast, when we examined the subgroup of women who had an abnormal chest X-ray, women were much less likely to have a nondiagnostic result from a CT scan when compared with a ventilation–perfusion scan, even after adjusting for the same significant confounding effects (aOR 0.4, 95% CI 0.2–0.8, P<.01).
The most common reason a patient underwent evaluation for pulmonary embolism was shortness of breath (60%), followed by tachycardia (54%) and desaturation less than 95% (40%) (Table 3). There was no significant risk association between specific clinical symptoms and findings and the risk of pulmonary embolism. The factor with the greatest risk association was a Pao2 less than 65 mm Hg (RR 2.8, 95% CI 1.4–5.8), but was not sufficiently predictive of a diagnosis of pulmonary embolism (positive likelihood ratio 2.8). A multivariable model with the three factors of strongest association (Pao2 less than 65 mm Hg, chest pain, and desaturation) was not predictive of pulmonary embolism in receiver operating characteristic analysis (area under the curve 0.68). We also examined specific subgroups based on presentation. Comparing the 199 patients who were evaluated during pregnancy to the 105 evaluated postpartum, there was no difference in the rate of pulmonary embolism diagnosis (11 [5.5%] compared with 7 [6.7%], P=.69). None of the 33 patients evaluated for tachycardia as the only clinical finding were diagnosed with a pulmonary embolism.
In this study of consecutive women radiologically evaluated for a clinical suspicion of pulmonary embolism, pregnant or within 6 weeks postpartum, we found that the rate of nondiagnostic initial study was dependent on chest X-ray results. Women with a normal chest X-ray were significantly more likely to achieve diagnosis when studied by ventilation–perfusion as compared with CT scan. Conversely, for women with abnormal chest X-ray findings, a ventilation–perfusion scan was significantly more likely to result in nondiagnosis. These results suggest that patients who are pregnant or immediately postpartum with clinical suspicion for pulmonary embolism may be triaged to the test most likely to give a useful diagnostic answer based on chest X-ray findings.
Other studies have tried to evaluate whether data on imaging for pulmonary embolism obtained in nonpregnant populations can be extrapolated to pregnant women. Chan et al8 provided a consecutive retrospective chart review of 120 pregnant patients evaluated for pulmonary embolism with ventilation–perfusion scanning and showed a 25% nondiagnosis rate. However, only 50% of those patients had a chest X-ray before the ventilation–perfusion scan, and due to the single mode of diagnostic evaluation they were unable to compare ventilation–perfusion to CT scan. Radiology literature, in addition to the obstetric literature, has acknowledged that inherent characteristics of pregnancy may alter the interpretation of CT angiography, making its utility in providing a diagnostic answer different from that seen in nonpregnant populations.9–11 In 2008, Bae et al12 evaluated contrast enhancement in cardiovascular CT imaging in 73 patients. These data showed that body mass index (or, perhaps even more importantly, body surface area) is strongly inversely correlated with the magnitude of contrast enhancement, suggesting that individual physiologic differences may alter important aspects in CT imaging. Physiologically altered states such as pregnancy, with known changes in body weight, body fluid distributions, or cardiopulmonary status, might then necessarily alter the radiology protocols or interpretations of these studies.11–13
Finally, we were unable to identify a specific clinical sign or symptom in our cohort that conferred significant increased risk for subsequent diagnosis of pulmonary embolism. We considered likelihood ratios as a type of odds, and because no single clinical finding significantly doubled the risk of pulmonary embolism, we were unable to conclude that any single presenting sign or symptom was clinically meaningful for clinicians with specific regard to the ability to predict risk of pulmonary embolism.
As an additional measure to minimize selection bias, the women for whom delivery and follow-up information was not available in our medical record system were contacted by telephone to obtain complete data. All but one woman was available. In addition, all data were available from a single institution, which some may argue may decrease generalizability. However, we would suggest that the uniform imaging techniques and reading standards for the imaging studies provide fewer confounding variables between the groups of women who had diagnostic compared with nondiagnostic results.
Our study has some limitations. We were unable to obtain accurate data regarding body mass index. However, surrogate markers for the complications of obesity such as diabetes or hypertension were able to be collected. In addition, one could argue that the study results may be in part confounded by indication, in the sense that individual patient-level decisions were made by physician discretion regarding whether the initial mode of imaging was computed-tomographic angiography or ventilation–perfusion. Given the retrospective nature of this study we are unable to adjust for that fact. However, when the groups were compared on characteristics that may have influenced the choice of one test over the other (for example, cardiomyopathy, asthma, tobacco exposure) the groups were statistically similar. In addition, the relationship between test and rate of nondiagnosis remained the same when adjusted for the variables by which the groups differed. Last, chest X-ray data were not available for every subject, which could have increased the risk for selection bias. However, the majority of patients did have a chest X-ray (92%), and a sensitivity analysis on clinical and historical factors for those without one showed them to be statistically similar to those with a chest X-ray, making the potential for bias less likely if it exists at all.
The amount of radiation exposure is small for both ventilation–perfusion (2.5 millisievert [mSv]) and computed-tomographic angiography (7 mSv)14,15 with respect to recommendations for limits put forth by the American College of Obstetricians and Gynecologists,16 and fairly similar. The radiation exposure from a single PA chest X-ray is substantially less (0.02 mSv), but this is tripled when a second, lateral view is added.14 In the setting of ruling out the life-threatening entity of pulmonary embolism in a pregnant or immediately postpartum woman, physicians should feel comfort using radiologic imaging to obtain a firm diagnosis, because the risks of nondiagnosis far outweigh any theoretical exposure risks. However, it is also prudent to minimize unnecessary exposures, which is most efficiently accomplished by ordering the initial test that will most often yield a diagnosis and prevent the requisite additional studies. Further in our experience, when ordering the chest X-ray, a single high-quality view yields sufficient information the majority of the time; reserving an additional lateral view for rare cases when the PA is not adequate.
This study provides clinically useful information regarding choice of radiologic imaging to diagnose or confidently exclude pulmonary embolus in pregnant women. Although recent high-level evidence in the nonpregnant literature has suggested that CT scan is a superior test to ventilation–perfusion imaging, there are many physiologic reasons why this likely may not be universally the case for pregnant women. Although CT and ventilation–perfusion scans showed statistically similar rates of nondiagnosis in pregnant women in this study, the rates of nondiagnosis were significantly different based on chest X-ray results. With consideration of the clinical judgment of the caring physician, the stability of the patient in question, and the diagnostic modalities available, we would suggest using chest X-ray results to help choose the diagnostic modality that will most often yield a definitive answer. At our institution, pregnant or postpartum women with a normal chest X-ray undergo ventilation–perfusion scan, whereas those with an abnormal chest X-ray are evaluated by CT angiography.
1. Martin JA, Hamilton BE, Sutton PD, Ventura SJ, Menacker F, Kirmeyer S, et al. Births: final data for 2005. Natl Vital Stat Rep 2007;56:1–103.
2. James AH, Jamison MG, Brancazio LR, Myers ER. Venous thromboembolism during pregnancy and the postpartum period: incidence, risk factors, and mortality. Am J Obstet Gynecol 2006;194:1311–5.
3. Weindling AM. The confidential enquiry into maternal and child health (CEMACH) [published erratum appears in Arch Dis Child 2004;89:293]. Arch Dis Child 2003;88:1034–7.
4. Stein PD, Fowler SE, Goodman LR, Gottschalk A, Hales CA, Hull RD, et al. Multidetector computed tomography for acute pulmonary embolism. N Engl J Med 2006;354:2317–27.
5. Wells PS, Anderson DR, Bormanis J, Guy F, Mitchell M, Gray L, et al. Value of assessment of pretest probability of deep-vein thrombosis in clinical management. Lancet 1997;350:1795–8.
6. Wells PS, Anderson DR, Rodger M, Stiell I, Dreyer JF, Barnes D, et al. Excluding pulmonary embolism at the bedside without diagnostic imaging: management of patients with suspected pulmonary embolism presenting to the emergency department by using a simple clinical model and d-dimer. Ann Intern Med 2001;135:98–107.
7. Balan KK, Critchley M, Vedavathy KK, Smith ML, Vinjamuri S. The value of ventilation–perfusion imaging in pregnancy. Br J Radiol 1997;70:338–40.
8. Chan WS, Ray JG, Murray S, Coady GE, Coates G, Ginsberg JS. Suspected pulmonary embolism in pregnancy: clinical presentation, results of lung scanning, and subsequent maternal and pediatric outcomes. Arch Intern Med 2002;162:1170–5.
9. Bhalla S, Lopez-Costa I. MDCT of acute thrombotic and nonthrombotic pulmonary emboli. Eur J Radiol 2007;64:54–64.
10. Marik PE, Plante LA. Venous thromboembolic disease and pregnancy. N Engl J Med 2008;359:2025–33.
11. Yamashita Y, Komohara Y, Takahashi M, Uchida M, Hayabuchi N, Shimizu T, et al. Abdominal helical CT: evaluation of optimal doses of intravenous contrast material—a prospective randomized study. Radiology 2000;216:718–23.
12. Bae KT, Seeck BA, Hildebolt CF, Tao C, Zhu F, Kanematsu M, et al. Contrast enhancement in cardiovascular MDCT: effect of body weight, height, body surface area, body mass index, and obesity. AJR Am J Roentgenol 2008;190:777–84.
13. Bae KT, Tao C, Gurel S, Hong C, Zhu F, Gebke TA, et al. Effect of patient weight and scanning duration on contrast enhancement during pulmonary multidetector CT angiography. Radiology 2007;242:582–9.
14. Mettler FA Jr, Huda W, Yoshizumi TT, Mahesh M. Effective doses in radiology and diagnostic nuclear medicine: a catalog. Radiology 2008;248:254–63.
15. Hurwitz LM, Reiman RE, Yoshizumi TT, Goodman PC, Toncheva G, Nguyen G, et al. Radiation dose from contemporary cardiothoracic multidetector CT protocols with an anthropomorphic female phantom: implications for cancer induction. Radiology 2007;245:742–50.
16. ACOG Committee on Obstetric Practice. ACOG Committee Opinion. Number 299, September 2004 (replaces No. 158, September 1995). Guidelines for diagnostic imaging during pregnancy. Obstet Gynecol 2004;104:647–51.