Commentary on “Multidetector CT in Vascular Injuries Resulting from Pelvic Fractures”

Raniga et al (1) provide a comprehensive discussion of the utility of CT for a potentially lethal and relatively common traumatic injury. Approximately 10% of patients admitted to level 1 trauma centers for blunt-force injury sustain pelvic fractures (2). The leading cause of death in the first 6 hours after pelvic fracture is abdominal and pelvic hemorrhage (3) with mortality resulting from a vicious cycle of acidosis, hypothermia, hemodilution, and trauma-induced coagulopathy (4). In-hospital mortality increases from 3%–4% to 8%–17% for the subset of patients with mechanically unstable pelvic fractures (5,6). These are more common in younger adults who have experienced high-energy trauma such as motorcycle injuries, motor vehicle collisions involving pedestrians, and falls from heights. Because of frailty, chronic comorbidities, and preexisting anticoagulant therapy, pelvic fractures in elderly patients, which typically result from low-energy mechanisms (eg, domestic falls from standing), can be equally lethal (7). After adjusting for covariates, there is a nearly eightfold increase in the mortality rate for patients with pelvic fractures who are 64 years old and older compared with those aged 15–34 years (6,8). Shock from multisystem trauma is common in both age groups (6). Mortality rates increase considerably in patients who present with shock, which occurs in 20%–50% of patients who have experienced high-energy trauma, according to Costantini et al (5).

Mortality rates in patients have declined in part because of improved resuscitation, diagnostic, and hemorrhage control interventions and damage control techniques (9). New nonimaging diagnostic modalities such as viscoelastic testing (eg, thromboelastography [TEG] and thromboelastometry [ROTEM]) provide quantitative information that allows individualized data-driven tailoring of the transfusion strategy (4). Pelvic circumferential compression devices (binders) and resuscitative endovascular balloon occlusion of the aorta (REBOA) have been adopted at high-volume trauma centers (2,10). The availability of hybrid operating rooms with angiographic and cone-beam CT capabilities is also increasing (11). Meanwhile, 64–detector row scanners, which have been commercially available since 2004, remain the CT workhorse of most trauma centers. Detector rows and dual-source technology can increase the speed of a single sweep, but the time that it takes to perform a multiphasic study is primarily dependent on contrast material kinetics and has therefore remained essentially unchanged. Improvement in image quality has been relatively incremental. Meanwhile, dual-energy CT remains uncommon or not fully utilized in the trauma setting. With the growing armamentarium of diagnostic and therapeutic options available to the trauma team, one might suspect that the role of CT has diminished, but the opposite has happened.

A recent American Association for the Surgery of Trauma (AAST) multicenter study (5) to assess the current management of pelvic fractures at high-volume level I trauma centers reported use of CT at admission in 85% of patients who presented with shock. Improvement in resuscitation techniques and the emergence of new temporizing measures that elevate systolic blood pressure (binders and REBOA) may facilitate the more liberal use of CT. This represents a major shift away from the traditional use of CT as a screening tool reserved for trauma patients who are hemodynamically stable (12).

In patients with pelvic fractures and refractory shock who are not responsive to initial volume resuscitation, focused abdominal sonography in trauma (FAST) is used in lieu of CT as a rapid bedside method to determine whether damage control laparotomy is needed for concurrent abdominal bleeding sources (13). The AAST study (5) showed that in the remaining patients with mechanically unstable pelvic fractures, approximately 60% have arterial sources of bleeding in the pelvis, a majority have polytrauma (mean injury severity score, 28.0), and a substantial number of patients have major injuries and potential sources of bleeding in other regions of the torso. For these reasons, contrast material–enhanced whole-body CT remains a critical and routine element of patient triage (whether patients undergo surgery, are sent to the angiography suite, or are treated conservatively, with transfusion). A review article on this topic is timely and apropos.

The authors primarily emphasize leveraging three major imaging features in the evaluation of pelvic fracture and related bleeding: (a) the pelvic fracture pattern and degree of pelvic instability, (b) the direct signs of vascular injury evaluated with multiphasic imaging, and (c) the location and size of pelvic hematomas. Taken as a whole, these features are used to predict the need for hemostatic intervention. Raniga et al (1) suggest that both angioembolization and retroperitoneal pelvic packing can be used as the first-line method to address high-pressure arterial bleeding. Most institutions favor angiography; however, some employ pelvic packing as an initial temporizing measure to avoid treatment delays associated with activation of the angiography suite. Practice patterns are largely influenced by the preferences of the surgeon and the resources available (5,14). Isolated venous bleeding is addressed with provisional stabilization with binders and external fixator devices.

In principle, the information gleaned at CT should improve patient triage by allowing rapid and objective determination of the source of bleeding. CT provides exquisite anatomic detail, and it is paradoxical that almost all other information available to the surgical team is quantitative, including vital signs, lactate levels and base deficits, and results of viscoelastic testing. However, CT assessment remains coarse and subjective. Management decisions are influenced by an overall gestalt impression because no individual CT feature is uniformly deterministic of a pelvic bleeding source. Overall, the authors provide a measured and realistic discussion of the merits and pitfalls of each CT feature. Outcome prediction in patients with bleeding pelvic fractures is inherently multifactorial and probabilistic. A familiarity with the evidence presented in this article should help the reader to render realistic recommendations.

Perhaps the most important pitfall highlighted by Raniga et al (1) relates to the diagnostic performance of contrast material extravasation. The authors note that contrast material extravasation is absent at CT when compared with catheter angiography in up to 20%–40% of instances, and that more than half of patients with contrast material extravasation at CT do not require embolization when digital subtraction angiography is performed. The statement is consistent with the highly variable accuracy metrics of contrast material extravasation reported in the literature (15). Newer-generation scanners appear to allow detection of a greater number of clinically insignificant microbleeds (15,16) and may have reduced specificity. Diameter-based threshold values that are used to differentiate between clinically inconsequential foci of contrast material extravasation from foci that require hemostatic intervention are associated with inherent information loss, leaving room for improvement. False-negative or false-positive results that lead to delayed treatment or unnecessary mobilization of limited resources and invasive procedures are bound to occur.

Despite more than a decade of experience with multiphasic CT protocols for pelvic trauma, only a handful of studies (17–20) have been performed to evaluate their accuracy in the characterization of pelvic contrast material extravasation and their efficacy for guiding management. There is a trend toward improved performance over single-phase protocols, but larger multicenter comparative effectiveness studies are necessary before we can be dogmatic on this issue. The clinical value of detecting sources of venous hemorrhage with the use of delayed phase imaging is unclear, because the prevalence of venous injury with arterial injury approaches 100% (2). If arterial bleeding can be occult at CT in 20%–40% of patients, identifying a single source of venous bleeding with dynamic information may lead to a false sense of security.

Similar ambiguity exists in considering extraperitoneal pelvic hematomas. When contrast material extravasation is present, it virtually always will coexist with a perivascular hematoma (intramuscular contrast material extravasation may be an important exception, although it is more likely to be self-limiting). In the absence of contrast material extravasation, a small hematoma is associated with a very small probability of high-pressure arterial bleeding that would benefit from hemostatic intervention, and its location is not relevant to management. Conversely, hematomas of increasingly size (even if they arise predominantly from venous sources) have increasingly high probabilities of concurrent management-altering arterial bleeding that is independent of the presence or absence of contrast material extravasation (16).

Large hematomas are typically multicompartmental, spreading along the space of Retzius, the pelvic sidewalls, and the presacral space, precluding the identification of culprit arterial territories with confidence. A sentinel clot (in the 40–70-HU range) may not be present in a patient with hemodilution and some degree of trauma-induced coagulopathy (21). So, the value of pelvic hematoma relates much more to volume than to location. But how should hematoma be measured? Diameter-based measurements are a poor estimate of volume (22). Semiautomated methods are relatively rapid but require access to and familiarity with advanced postprocessing tools (16,22). Both hematoma volume and contrast material extravasation should be quantified volumetrically (16,23), but the time effort required is prohibitive unless segmentation and measurement are automated. Recent work (24,25) has shown that pelvic hematomas and contrast material extravasation can be measured reliably with the use of deep learning computer vision methods.

Grading the severity of pelvic fractures at the point of care is also difficult. The authors describe a variety of limitations, including the modest interobserver agreement of popular grading systems and the potential masking effect of pelvic binders (26,27). “Binder-off” plain radiography is advocated by some but must be weighed against the risk of releasing tamponade and rendering sharp bone ends temporarily mobile (28). The authors point to a range of secondary signs of stabilizing ligament avulsion that can improve the accuracy and agreement of grading. These include L5 transverse process fractures (iliolumbar ligament), inferior sacral fractures (sacrospinous and sacro-ischial ligaments), and rectus abdominis insertional tears (26).

Both the Young-Burgess and Tile grading systems are widely used, and the system used in the radiology report should mirror institutional practice. Radiologists should be cognizant that the Young-Burgess system reflects several parallel grading systems based on the presumed force vector (eg, APC1–APC3 and LC1–LC3). Studies in which the utility of the Young-Burgess grading system for outcome prediction is evaluated tend to group injury types according to instability in a manner similar to the original intent of the Tile classification. For example, LC3, APC3, and vertical shear injuries may all be grouped in a vertically unstable category analogous to a Tile C injury (29). This issue is nontrivial in the clinical setting.

The authors offer additional valuable but often overlooked insights with regard to the spectrum of pelvic vascular injuries. These include (a) vessels with transection, thrombosis, or pseudoaneurysms under transient tamponade to result in delayed life-threatening bleeding and (b) the potential for nontarget embolization in patients with arteriovenous fistulae. Most scientific articles related to direct evidence of vascular injury have been focused primarily on contrast material extravasation as a predictor of angiographic evidence of bleeding (30). The discussion is clinically valuable and should stimulate further scientific inquiry. The authors provide helpful practical guidance regarding the search for vascular injuries, including using symmetry to advantage when examining vessel size, contour, and opacification. Emphasis is placed on the importance of complete knowledge of third- and fourth-order pelvic vascular anatomy. The detailed discussion of this topic is augmented with excellent images and online videos.

CT provides a large amount of information that is useful for decision making and outcome prediction in patients with bleeding pelvic fractures. However, the often complementary imaging features and patient-specific factors such as age and frailty are difficult to integrate mentally into an objective personalized estimate of risk. This is especially challenging when there are several outcomes of immediate interest—not just the need for hemostatic intervention but also the relative contribution of pelvic bleeding sources to the transfusion requirement and the risk of death.

Ideally, CT should provide quantitative information for personalized prediction of outcomes and decision-making support in the same way that continuous parameters from viscoelastic testing can be used to guide the transfusion strategy. Ultimately, computer vision tools will offer quantitative information that augment the radiologist’s ability to render personalized and objective treatment recommendations (24,25). Until such time, readers should be mindful of the strengths and limitations presented in this measured and comprehensive discussion. Radiologists should have a comfort level with ambiguity and avoid rigid dogmatic thinking when encountering this complex problem.