Introduction

Non-occlusive mesenteric ischemia (NOMI) is an acute mesenteric circulatory disorder that does not involve the organic occlusion of blood vessels. Its common cause is a decrease in cardiac output resulting in splanchnic hypoperfusion. NOMI is seen in patients with myocardial infarction, congestive heart failure, aortic insufficiency, and renal or hepatic disease [1–4], and it accounts for 20–30 % of acute bowel ischemia [4, 5].

Because its symptoms are nonspecific and it frequently occurs in patients with postoperative consciousness disturbance [6], early diagnosis of NOMI is difficult. Within 3–6 h, intestinal ischemia becomes critical [7], and the morbidity rate is high. According to Han et al. [8], the mortality rate of patients with NOMI is 70–90 %.

Siegelman et al. [9] suggested angiographic criteria for the diagnosis of mesenteric vasospasm. Although angiographic study of the SMA is essential for the diagnosis of NOMI [10], and a useful scoring system for the assessment of angiographic findings in NOMI has also been reported [11], such studies cannot be performed in patients in poor or unstable condition [6]. The utility of multidetector computed tomography (MDCT) for the diagnosis of NOMI has been reported [6, 12–14]. Woodhams [12], who studied 4 patients with NOMI, found that the diameter of the superior mesenteric artery (D SMA) on CT images was smaller in NOMI patients than in their normal controls. Although the diameter of the superior mesenteric vein (D SMV) is narrowed in patients with bowel ischemia [15], there are no reports of changes in D SMV in patients with NOMI. The purpose of the study described in the present paper was to investigate the diagnostic capability of MDCT for detecting NOMI, especially by focusing on the quantitative analysis of D SMA and D SMV.

Materials and methods

Our retrospective study was approved by our institutional review board; prior informed patient consent was waived.

Study population

We retrospectively studied 11 NOMI patients (8 men and 3 women aged 43–87 years, mean 74.9 years) who underwent MDCT between July 2008 and October 2011. All had undergone abdominal surgery to address bowel ischemia, and their diagnosis was based on histopathologic evidence [16]. To determine the diagnostic capability of MDCT for NOMI, we implemented a matched (1:4) case–control design.

The control subjects were chosen regardless of whether they could be matched to patients in our hospital’s database. The subjects we recruited had undergone contrast-enhanced CT studies in our hospital between March 2012 and May 2012 and did not have abdominal diseases. We paired our 11 NOMI patients with age- and gender-matched consecutive control subjects (same sex, same age ±5 years). Our control group comprised 32 men and 12 women aged 42–88 years (mean 74.5 years).

Risk factors for NOMI

NOMI is commonly caused by a decrease in cardiac output that results in splanchnic hypoperfusion. Reported causes of mesenteric vasospasm are myocardial infarction, congestive heart failure, renal or hepatic disease, digitalis, various forms of shock, septicemia, dehydration and hypotension following dialysis, and heart and major abdominal surgery [1–4, 17]. As the risk factors for NOMI are known, one board-certified radiologist reviewed our patient records.

CT scans

All patients underwent MDCT scanning on a 16- (Light Speed Ultra 16, GE, Milwaukee, WI, USA), a 64- (Light Speed VCT, GE; or Aquillion64, Toshiba Medical Systems, Otawara, Japan), a 128- (Definition AS+, Siemens, Forchheim, Germany), or an 8-detector CT scanner (Brightspeed Edge, GE). The settings for each instrument were as follows. Light Speed Ultra 16: rotation time, 0.7 s; beam collimation, 16 × 1.25 mm; section thickness and intervals, 5.0 mm; helical pitch (beam pitch), 1.375; table movement, 39.3 mm/s, scanning field-of-view (FOV), 40 cm; voltage, 120 kV; and auto mA (noise index 10). Light Speed VCT scanner: rotation time, 0.4 s; beam collimation, 64 × 0.625 mm; section thickness and intervals, 5.0 mm; helical pitch (beam pitch), 1.375; table movement, 137.5 mm/s, scanning FOV, 40 cm; voltage, 120 kV; and auto mA (noise index 10). Aquillion64: rotation time, 0.5 s; beam collimation, 64 × 0.5 mm; section thickness and intervals, 5.0 mm; helical pitch (beam pitch), 0.703; table movement, 45 mm/s; scanning FOV, 35 cm; voltage, 120 kV; and auto mA (noise index 8). Definition AS+: rotation time, 0.6 s; beam collimation, 64 × 0.5 mm; section thickness and intervals, 5.0 mm; helical pitch (beam pitch), 0.703; table movement, 45 mm/s, scanning FOV, 35 cm; voltage, 120 kV; and auto mA (noise index 8). Brightspeed Edge: rotation time, 0.5 s; beam collimation, 8 × 1.25 mm; section thickness and intervals, 5.0 mm; helical pitch (beam pitch), 1.675; table movement, 16.75 mm/s; scanning FOV, 35 cm; voltage, 120 kV; and auto mA (noise index 10). We obtained unenhanced and contrast-enhanced images (equilibrium phase), which were taken about 120–150 s after the start of contrast injection. The contrast dose for all patients was 600 mgI/kg patient body weight. The contrast medium was injected into the cubital vein at a rate of 2.0 ml/s through a 22-G intravenous catheter using an automatic injector. Arterial-phase images and reconstructed images could not be evaluated because these were not performed at all participating institutes. Therefore, we only performed unenhanced and equilibrium-phase imaging, taking axial images with a slice thickness of 5 mm.

Image analysis

Two board-certified radiologists evaluated the quality of the CT images consensually for mesenteric vessel occlusion and bowel ischemia based on the following signs: (1) bowel wall thickening (>2 mm) or thinning (paper-thin wall), (2) high attenuation of the bowel wall on unenhanced scans, (3) localized mesenteric fluid or edema, (4) asymmetric intravenous contrast enhancement (decreased or increased) of segments of the bowel wall, (5) pneumatosis intestinalis, and (6) portal venous gas [18–20].

For the quantitative analysis, two radiologists used CT images with a slice thickness of 5 mm to measure D SMA and D SMV immediately inferior to the inferior pancreaticoduodenal artery bifurcation at a window level and width of 40 and 350 Hounsfield units. This was done after the images had been magnified but before placing cursors [12], and the D SMA and D SMV values of the two radiologists were averaged.

Statistical analysis

The statistical difference in D SMA and D SMV between NOMI patients and the controls was tested with the two-sided Mann–Whitney U test. Bland–Altman plots were used to evaluate the agreement and the existence of systematic differences between the D SMA and D SMV values of the two radiologists, with 95 % limits of agreement (LoA) being calculated (mean difference ± 1.96 × standard deviation of the difference) [21].

We performed linear discriminant analysis (LDA) to find a linear transformation value (“discriminant function”) from the values of D SMA and D SMV that permits more accurate discrimination than either parameter alone.

We used receiver operating characteristic (ROC) analysis to compare the diagnostic abilities of D SMA, D SMV, and the Z value in our LDA. We calculated the area under the best-fit ROC curve (A z ) fitted to each parameter. We also calculated the optimal cut-off value for the D SMA, D SMV, and Z values by ROC analysis. Using the optimal cut-off values, we then calculated the sensitivity and specificity of each parameter for diagnosing NOMI.

We performed the Mann–Whitney U test and LDA with free statistical software (R version 2.15.0). We performed all Bland–Altman tests using the MedCalc software package (version 11.3.7.0, MedCalc Software, Mariakerke, Belgium) and ROC analysis was with the JMP10 software package (SAS Institute, Cary, NC, USA). Differences with p < 0.05 were considered significant.

Results

All 11 NOMI patients had undergone abdominal surgery. Macroscopic examinations and palpation of intestines and mesenteries showed segmental necrotic lesions in the jejunum, and bowel resection was performed for the part of the jejunum with necrosis. Sporadic ischemic and necrotic lesions of the intestine were observed in the resected jejunum of every patient. There were no clear findings of arteriovenous thrombosis in the histopathological specimens.

All NOMI patients presented with risk factors for NOMI. Of the 11 patients, 4 had chronic renal failure, 3 septicemia, 2 congestive heart failure, 1 dehydration following heart surgery, and 1 had a severe burn injury.

At qualitative analysis of the CT images, all 11 NOMI patients manifested asymmetric intravenous contrast enhancement (decreased or increased) of segments of the bowel wall, 9 had localized mesenteric fluid or edema, 8 had pneumatosis intestinalis, 6 had portal venous gas, 6 had high attenuation of the bowel wall on unenhanced CT scans, and 3 had bowel wall thickening (>2 mm) or thinning (paper-thin wall) (Fig. 1). While all NOMI patients presented with more than 2 CT findings of bowel ischemia, none manifested signs of mesenteric vessel occlusion. Segmental necrotic lesions that were approximately consistent with the area of bowel ischemia found on CT images were confirmed during operation.

Fig. 1
figure 1

Contrast-enhanced MDCT images obtained in a 43-year-old man with a severe burn injury. a The arrows point to bowel wall thickening due to jejunal ischemia. b The diameter of the superior mesenteric vein (D SMV; 5 mm) was smaller than of the superior mesenteric artery (D SMA; 7 mm)

Full size image

At quantitative analysis, Bland–Altman analysis showed that the values measured by two readers were very similar, with low biases (−0.18 ± 0.82 for D SMA and −0.05 ± 0.99 for D SMV), and the limits of agreement at 95 % were in the clinically acceptably range (−0.40 to +0.04 for D SMA and −0.32 to +0.21 for D SMV). Thus, we regarded the measurements of the two readers as being in good agreement. The median D SMA was 6.0 mm (range 3.5–8.0 mm) in NOMI patients and 7.6 mm (range 5.5–9.0 mm) in the control group (Fig. 1). D SMA was significantly smaller in the NOMI patients than the controls (p < 0.01) (Fig. 2). The A z value and the optimal cut-off value for D SMA to diagnose NOMI were 0.83 and 6.5 mm, respectively (Fig. 5). At a cut-off value of 6.5 mm for the diagnosis of NOMI, sensitivity and specificity were 81.8 % and 81.8 %, respectively (Table 1).

Fig. 2
figure 2

D SMA values for the NOMI patients and the controls. The solid line indicates the median D SMA in each group

Full size image
Table 1 Diagnostic capability of NOMI using D SMA and D SMV
Full size table

The median D SMV was 7.5 mm (range 5–9 mm) in NOMI patients and 11.0 mm (range 8.0–14.0 mm) in the controls. It was significantly smaller in the NOMI patients (p < 0.01) (Fig. 3). The A z value and the optimal cut-off value for D SMV to diagnose NOMI were 0.90 and 9.0 mm, respectively (Fig. 5). At a cut-off value of 9.0 mm for the diagnosis of NOMI, sensitivity and specificity were 81.8 and 88.6 %, respectively (Table 1).

Fig. 3
figure 3

D SMV values for the NOMI patients and the controls. The solid line indicates the median D SMV in each group

Full size image

The linear discriminant function was

$$ Z \, = \, - 0. 7 4 { } \times {D}_{\text{SMA}} { - }0. 9 4 { } \times {D}_{\text{SMV}} + { 13}. 7 2. $$

Z > 0 indicates a positive and Z < 0 a negative finding (Fig. 4). The A z value for the transformed value of LDA (the Z value) that was used to diagnose NOMI was 0.94, and the optimal cut-off for the Z value was 0.93 (Fig. 5). The sensitivity for the Z value was 81.8 %, and the specificity was 97.7 % (Table 1). At discriminant analysis, the p values of D SMA and D SMV were 0.11 and <0.01, respectively, which show that D SMV was the only significant factor in linear discriminant analysis.

Fig. 4
figure 4

Scatter plot of D SMA and D SMV. Black circles NOMI patients, open circles controls. The solid line shows the classification boundary obtained by linear discriminant analysis

Full size image
Fig. 5
figure 5

Receiver operating characteristic curves for the three methods (D SMA, D SMV, and Z value)

Full size image

Discussion

Intra-arterial infusion of papaverine hydrochloride into the mesenteric arteries via a catheter is currently the only therapy for NOMI [9, 10]. While it lowered the mortality rate from 70–90 to 10–50 % [10, 22], the procedure is invasive and requires skill. Because NOMI often occurs in patients in poor or unstable condition [6], it should be performed only after a definite diagnosis of NOMI has been made. Therefore, if NOMI is suspected, the first-line examination should rule out NOMI. This requires high specificity. When we investigated the diagnostic abilities of D SMA and D SMV, their specificities at a cut-off value were 81.8 and 88.6 %, respectively. This suggests that D SMV is superior to D SMA for the diagnosis of NOMI. On the other hand, at LDA, among the three diagnostic parameters, the Z value with both D SMA and D SMV returned the highest A z value upon ROC analysis and satisfied the requirements for high sensitivity and specificity. Therefore, we concluded that the Z value obtained by LDA was the best diagnostic parameter for NOMI.

According to our results, the median D SMV was significantly smaller in the NOMI patients than in the controls. Moreover, ischemic and necrotic lesions of the intestine were observed during operation in all NOMI patients. To the best of our knowledge, no previous reports have commented on the smaller D SMV in patients with NOMI. In many patients with acute superior mesenteric arterial occlusion, D SMV is smaller than D SMA [23]; Hayakawa et al. [15], who called this phenomenon “the smaller SMV sign,” posited that it was attributable to a decrease in the venous return due to blockage of the arterial flow. As NOMI is commonly the result of splanchnic hypoperfusion, and flow in the superior mesenteric artery is decreased in NOMI patients, their superior mesenteric venous return may be decreased, making it reasonable to expect that their D SMV is smaller than in normal subjects.

Based on our results, the median D SMA was also significantly smaller in the NOMI patients than in the controls. However, we also found that at LDA, D SMV but not D SMA was a significant diagnostic factor, while most other studies focused solely on the SMA to diagnose NOMI [6, 9, 10, 12], Because the intravenous pressure is low, the amount of fibroelastic tissue or smooth muscle is lower in venous than in arterial walls, suggesting that changes in D SMV due to a decrease in the blood flow may be larger than those in D SMA.

All of our NOMI patients had risk factors for NOMI [1–4, 17]. Reported risk factors for acute mesenteric arterial occlusion are atrial fibrillation, myocardial infarction with subsequent impairment of the cardiac wall, and structural heart defects [24], while the risk factors for mesenteric venous thrombosis are direct injury, local venous congestion or stasis, and thrombophilia including a deficiency in antithrombin III or protein S or C [5, 25]. There is a tendency not to consider risk factors in the diagnosis of acute mesenteric arterial occlusion or mesenteric venous thrombosis because occlusion of these vessels can be identified on contrast-enhanced CT images [20, 26, 27]. On the other hand, as the CT imaging findings in NOMI can be complex, it has been reported that it is rather difficult to diagnose NOMI from among the various acute mesenteric ischemic diseases known [4, 20]. Consequently, recognition of the risk factors for NOMI may be aid in the diagnosis of NOMI.

MDCT has been reported to represent an excellent diagnostic tool for the early and accurate detection and localization of acute mesenteric ischemic changes [28, 29]. All 11 of our NOMI patients manifested more than 2 CT signs of bowel ischemia [18–20]. While qualitative assessment of D SMA and D SMV narrowing may be a difficult task, the identification of ischemic changes in the bowel on enhanced CT scans may provide a clue to the presence of NOMI. In addition, unlike catheter angiography, MDCT findings make it possible to rule out other disorders in attempts to reach a differential diagnosis in patients with mesenteric ischemia [29, 30].

Our study has some limitations. First, our study population was relatively small. For an accurate evaluation of the ability of MDCT to diagnose NOMI, we selected patients who were diagnosed based on histopathologic evidence, so our findings should be considered to be preliminary. Second, the CT scanners and scanning parameters used varied because we included patients treated at several hospitals. Nonetheless, we think that differences in the imaging techniques had only a relatively small impact on our results because we only evaluated bowel ischemia and D SMA and D SMV. However, it was difficult to evaluate morphological changes in the SMA in detail in our study because some images were not scanned with a protocol for evaluating arteries, such as CT angiography, although morphological changes in the SMA were reported to be important for diagnosing NOMI not only in angiographic studies but also in MDCT. Further investigation is needed [9–12]. Third, while there were several types of mesenteric ischemia, including acute mesenteric arterial occlusion and mesenteric venous thrombosis, we only evaluated NOMI patients and control subjects. Mazzei et al. [31] reported the CT appearance of the mesentery with acute mesenteric ischemic disease due to various causes. However, the diagnosis of other types of acute mesenteric ischemia is not difficult with MDCT, because the occlusion of these vessels can be identified on contrast-enhanced CT images. On the other hand, the diagnosis of NOMI at MDCT is difficult. NOMI patients may sometimes be mistakenly diagnosed as not having severe mesenteric ischemia because the mesenteric vessels are not occluded. Therefore, it is more important to clarify the differences between CT images of NOMI patients and control subjects and between those of NOMI patients and patients with other types of mesenteric ischemia; our study is the first to clarify the differences between NOMI patients and control subjects. Fourth, it is uncertain from our results whether decreased D SMA and D SMV values without any CT signs of bowel ischemia implies ischemia or necrosis of the bowel, because all of the NOMI patients in our study showed not only decreased D SMA and D SMV values but also more than 2 CT signs of bowel ischemia. Again, further investigation is needed [18–20]. Finally, in the paper reported by Mazzei et al. [31] from Italy, the median D SMA and D SMV values in patients with NOMI (6.95 and 6.55 mm, respectively) are different from those obtained in the present work. This difference may be due to race. Therefore, cut-off values of D SMA and D SMV for the diagnosis of NOMI derived from our results are only thought to be indicative for Japanese or Asian patients.

In conclusion, D SMA and D SMV were significantly smaller in NOMI patients than in the controls, and D SMV may be a more diagnostically significant parameter than D SMA for the diagnosis of NOMI.