Contrast-enhanced MDCT was performed with a 4-MDCT scanner (MX 8000, Philips Medical Systems) before April 2003 and one of two 16-MDCT scanners (Brilliance 16 Power and Brilliance Big Bore, Philips Medical Systems) after April 2003, according to the techniques described in Table 2. IV and oral contrast media were administered. A biphasic IV injection of contrast material (iohexol, Omnipaque 300, GE Healthcare) was administered routinely with a power injector and a 16- to 20-gauge IV cannula. Unenhanced CT images were not obtained through the abdomen or pelvis. A 2% solution of 600 mL of meglumine diatrizoate (2% Hypaque, GE Healthcare) in water was administered orally or through a nasogastric tube if the patient was unable to drink; 300 mL was given 30 minutes before scanning and 300 mL in the scanning suite. Images of the abdomen and pelvis were acquired as part of the scan from the thoracic inlet to the symphysis pubis on the 4-MDCT scanner and as part of the whole-body scan from the circle of Willis to the symphysis pubis with the arms extended superiorly and placed next to the neck on the 16-MDCT scanner. Images of the abdomen and pelvis were acquired during the portal venous phase of contrast enhancement, and delayed images were acquired from the diaphragm to the iliac crest during the excretion phase, approximately 2–3 minutes after injection. Evaluating the spleen during these two phases was helpful in more accurately visualizing splenic vascular injuries and the renal collecting system.

The new grading system (Table 3) was based on experience from multiple trauma centers, including ours, indicating that CT evidence of active splenic hemorrhage and vascular injuries is predictive of the need for splenic arteriography and transcatheter embolization or splenic surgery [6–10]. Because of the risk of failure of nonoperative management [7–9], it is important to identify patients with these CT findings even when AAST classifications of injuries are low-grade. In the new system, such previously low-grade splenic injuries are upgraded to grade 4a or 4b. In the new system, patients with grade 4 injuries are candidates for splenic arteriography or splenic surgery. The quantity of blood in the peritoneal cavity was not taken into consideration in assigning grades with this modified system.

Initial reconstructed axial MDCT images (3- or 5-mm slice thickness) in each case were prospectively and independently reviewed by a fifth- or sixth-year resident with 1–2 years of experience or by four board-certified emergency radiologists with 3–20 years of experience. Splenic injury was graded according to the CT injury grade modeled on the AAST splenic injury scale (Table 1) [2]. Admission MDCT images in all cases were then reviewed by two radiologists who did not have access to the original interpretations or outcomes. These radiologists regraded injuries according to the modified injury grading system that specifically incorporates the findings of active bleeding or vascular lesions (Table 3).

Active bleeding (Figs. 1A and 1B) was defined as a linear or irregular area of contrast enhancement with an attenuation value similar to or greater than that of the aorta or an adjacent major artery [11]. On CT, the area of active bleeding identified during the portal venous phase may appear to expand on delayed imaging during the renal excretory phase because of continuing bleeding. For this study, vascular injuries (Figs. 2A and 2B) included both splenic pseudoaneurysm and arteriovenous fistula. Pseudoaneurysm and arteriovenous fistula were defined as well-circumscribed areas of contrast enhancement with an attenuation value similar to that of an adjacent contrast-enhanced artery. These vascular injuries may be surrounded by low-attenuation parenchyma or hematoma. On delayed imaging during the excretory phase, vascular injuries typically lose density from washout of IV contrast material, and the attenuation becomes similar to or slightly higher than that of adjacent organ parenchyma.

Splenic intervention was defined as splenic arteriography or surgery performed as a result of the CT-diagnosed splenic injury. Splenic arteriography was performed on all patients in hemodynamically stable condition with high-grade splenic injury (AAST grades 3–5) and on patients with low-grade splenic injury (grades 1 and 2) who had active bleeding or a vascular injury on MDCT. Splenic arteriography was performed on 164 patients. A total of 130 splenic artery embolizations were performed. Embolization was performed on 71 patients with active bleeding or vascular injury. Splenic artery embolization was performed on 59 other patients because of abrupt truncation of a splenic artery branch on arteriography or if, at the discretion of the interventional radiologist and attending trauma surgeon, the patient's clinical condition warranted a more aggressive interventional approach. For example, main splenic artery embolization was performed to control blood pressure in some patients with major head injuries in order to prevent delayed hemorrhage from a high-grade splenic injury.

At our institution, hemodynamic instability is defined with the following criteria: systolic blood pressure less than 100 mm Hg, heart rate greater than 120 beats per minute, and lack of response to a fluid challenge of 2 L of a crystalloid solution. Splenectomy was performed on 45 patients after MDCT, usually because of hemodynamic instability after CT (n = 35). Other reasons (n = 10) included injuries to adjacent organs, need for anticoagulation, high-grade splenic injury in a pregnant patient, and initiation of splenic hemorrhage when the spleen was mobilized at laparotomy.

Medical records were reviewed to determine outcome of management of splenic injuries. Statistical analysis was performed with JMP statistical software, versions 5.1 and 6.0 (SAS Institute) and SAS, version 9.1 (SAS Institute). The Stuart-Maxwell and the McNemar tests of homogeneity were used to compute statistics for assessment of the association between the old and new grading systems. To discern whether the new splenic grading system was better than the AAST injury scale in identifying patients needing splenic arteriography or surgery, receiver operating characteristic (ROC) curves were generated for both grading systems for all splenic interventions. These curves were generated for splenic arteriography and surgery separately and for all splenic interventions combined.

Nonparametric approaches were selected because they provided the foundation on which the ROC curves were based. The nonparametric method of DeLong et al. [12] was used to compare the area under the correlated AAST grade curve with that of the new-grade ROC curve from the same subjects on the basis of concepts proposed by Hanley and McNeil [13, 14]. Hanley and McNeil used Kendall's tau rank-order correlation to calculate the critical statistics for comparing the areas under two correlated ROC curves. DeLong et al. used generalized U statistics to calculate the areas under the ROC curves and to test for differences between areas under correlated ROC curves. The approach of DeLong et al. was selected because the computations yielded more precise estimates of the variance–covariance matrices and CIs of the ROC curve areas than did Hanley and McNeil's approximated estimates.

The JMP nominal logistic platform was used to generate logistic regression (logit) models for the old and new grading systems, compute the areas under the ROC curves, and produce the plots shown in Figures 3A, 3B, 4A, 4B, 5A, and 5B. The JMP logit model was reproduced with the SAS LOGISTIC procedure. The SAS %ROC macro was used to compute the 95% CIs for areas under the ROC curves and to test for significance of differences between the areas under the correlated ROC curves of the old and new grading systems.