To investigate the correlation of iodine concentration (IC) generated by spectral computed tomography (CT) with micro-vessel density (MVD) and vascular endothelial growth factor (VEGF) expression in patients with advanced gastric carcinoma (GC).

Thirty-four advanced GC patients underwent abdominal enhanced CT in the gemstone spectral imaging mode. The IC of the primary lesion in the arterial phase (AP) and venous phase (VP) were measured, and were then normalized against that in the aorta to provide the normalized IC (nIC). MVD and VEGF were detected by immunohistochemical assays, using CD34 and VEGF-A antibodies, respectively. Correlations of nIC with MVD, VEGF, and clinical-pathological features were analyzed.

Both nICs correlated linearly with MVD and were higher in the primary lesion site than in the normal control site, but were not correlated with VEGF expression. After stratification by clinical-pathological subtypes, nIC-AP showed a statistically significant correlation with MVD, particularly in the group with tumors at stage T4, without nodular involvement, of a mixed Lauren type, where the tumor was located at the antrum site, and occurred in female individuals. nIC-VP showed a positive correlation with MVD in the group with the tumor at stage T4 and above, had nodular involvement, was poorly differentiated, was located at the pylorus site, of a mixed and diffused Lauren subtype, and occurred in male individuals. nIC-AP and nIC-VP showed significant differences in terms of histological differentiation and Lauren subtype.

The IC detected by spectral CT correlated with the MVD. nIC-AP and nIC-VP can reflect angiogenesis in different pathological subgroups of advanced GC.

Despite a recent decrease, gastric cancer (GC) remains the most common cancer and is the third leading cause of cancer-related death globally[1]. In China, the incidence and mortality rates of GC remain high, and the vast majority of cases are in the advanced stage[2], which requires more attention.

Angiogenesis is fundamental to the growth, invasion, and metastasis of GC, and greatly influences the response to anti-tumor therapies[3]. To date, the standard method for studying angiogenesis has been histopathological counting of micro-vessel density (MVD), which is specimen- and immunostaining-dependent. This is impractical in advanced patients undergoing anti-angiogenesis or chemotherapy. Compared with MVD counting, using preoperative imaging modalities for the noninvasive assessment of tumor angiogenesis is more acceptable and feasible. A previous study has revealed that MVD, vascular endothelial growth factor (VEGF), and the absolute enhanced value show some positive correlations with conventional contrast-enhanced computed tomography (CT)[4]. Additionally, CT perfusion has shown the potential for evaluating tumor angiogenesis[5,6], but the complicated measurement and high radiation dosage have limited the extensive application of CT in this regard.

The recently developed spectral CT yields material-decomposition (MD) images that can quantitatively map the iodine concentration (IC) of the tissue in enhanced images. This IC value has been proven to show a strong correlation with the actual iodine concentration in the phantom[7]. Recently, preliminary studies have reported the use of the IC value to differentiate benign and malignant lesions, to find embolisms, and to evaluate the efficacy of anticancer therapy[8-11]. Particularly in the evaluation of neoadjuvant chemotherapy of GC, the IC value was proven to be a more robust imaging biomarker than the morphology index[11]. However, it was not clear how this correlated with radiological-pathological features. On the other hand, the IC value has shown a correlation with vascularization in solid tumors, such as pancreatic carcinoma, hepatocarcinoma, and non-small-cell lung cancer[12-14], while there has been no such report for GC, to provide a basis for using image indicators, other than morphology, for assessing chemo-efficacy. Therefore, our purpose was to investigate the correlation between IC value and angiogenesis in GC cases with clinical-pathological data.

A total of 34 advanced GC patients (23 males, 11 females; mean age: 56 ± 3.7 years; age range: 30-73 years) were included (Table 1). Tumor size ranged from 1.5 cm to 13.0 cm (median: 4.0 cm). Fourteen tumors located in the gastric cardia-fundus, nine in the gastricum, and 11 in antrum. All patients were treated surgically by radical gastrectomy and D2 lymph node dissection. Based on pathologic results, the adenocarcinoma was well differentiated in one patient, moderately differentiated in 16 patients, and poorly differentiated in 17 patients. According to the 7^th American Joint Committee on staging classification (AJCC), five patients were classified as T2, two as T3, twenty-two as T4a and five as T4b; eleven as N0, eight as N1, seven as N2, and eight as N3. No patient had distant metastasis. The Lauren classification was as follows: 15 were intestinal type, seven were mixed type, and 12 were diffuse type. Additionally, the immunostaining analysis revealed that 25 patients (73.53%) stained positive for VEGF with a score of 4 (n = 11), 5 (n = 9), or 6 (n = 5). The mean MVD count for these 34 tumors was 26.94 ± 1.35.

In general, the bi-phasic nIC values and the MVD counts were positively correlated (P = 0.013, P < 0.001, respectively). VEGF did not correlate with either nIC value or with MVD, as shown in Table 2. When stratified by different clinical features, the correlation coefficient value increased. nIC-AP positively correlated with MVD in patients with tumor stage less than T4 (r = 0.851, P = 0.015), tumor of N0 stage (r = 0.620, P = 0.042), with a tumor located in the pylorus region (r = 0.616, P = 0.044), and who were female (r = 0.696, P = 0.017). On the other hand, nIC-VP correlated with MVD in the more advanced group of patients, with tumors above T4 stage (r = 0.656, P < 0.001), with nodular involvement (r = 0.644, P = 0.001), that were poorly differentiated (r = 0.799, P < 0.001), were of mixed and diffused Lauren subtypes (r = 0.827, P = 0.022; r = 0.765, P = 0.004, respectively), and male gender (r = 0.606, P = 0.002), as shown in Table 3.

The nIC values in the primary GC and normal gastric wall were 0.116 ± 0.033 and 0.101 ± 0.023 in arterial phase (P = 0.033), 0.386 ± 0.061, and 0.286 ± 0.066 in the venous phase (P < 0.001) (Figure 4A). For the VEGF-positive and -negative group, neither nIC-AP nor nIC-VP showed statistically significant differences (Figure 4B).

Open in New Tab   Full Size Figure   Download Figure

Figure 4 Comparison of normalized iodine concentration in arterial phase; and normalized iodine concentration in venous phase between the normal gastric wall and tumor site (A); comparison of normalized iodine concentration in arterial phase and normalized iodine concentration in venous phase between the vascular endothelial growth factor-positive and -negative group (B). ^aP < 0.05, ^bP < 0.001. VEGF: Vascular endothelial growth factor; nIC: Normalized iodine concentration; AP: Arterial phase; VP: Venous phase.

When stratified by clinical subgroups, both nIC-AP and nIC-VP were higher in patients with serosal involvement, lymph node metastasis, poor differentiation, and diffused Lauren type than in those with depth invasion under T4, nodal status N0, high and moderate differentiation, intestinal and mixed Lauren type, but these differences did not reach statistical significance. nIC-AP and nIC-VP showed statistically significant differences between differentiation categories (P = 0.003, P = 0.001, respectively) and between Lauren subtypes (P = 0.016, P = 0.006, respectively). The LSD test revealed that nIC-AP and nIC-VP were significantly different between intestinal and diffuse Lauren subtypes (P = 0.005, P = 0.004, respectively). nIC-VP was also significantly different between intestinal and mixed Lauren subtypes (P = 0.013), as shown in Table 4.

There were four major findings from this study. First, the bi-phasic nICs showed a significantly linear positive relationship with MVD in primary GC. The nIC-AP and nIC-VP correlated with MVD in different subgroups; the former correlated significantly with MVD in the relative earlier stage of advanced GC, while the latter correlated with MVD in the more advanced stage. Second, no significant correlation was found between nICs and VEGF. Third, nIC values in the normal gastric wall were observed to be significant lower than that in the tumor. Fourth, nICs were observed to differ between histological grades and Lauren subtypes of advanced GC; the greater the malignancy, the higher the nICs. Taken together, these observations demonstrate that quantification of iodine in spectral CT imaging have the potential to reflect angiogenesis of advanced GC.

Iodine, a commonly used CT contrast material, is generally known to produce higher attenuation at low tube voltage settings[18]. Based on this effect, spectral CT using high and low voltage switching settings could differentiate materials of the same density[19] and the IC value could be extracted[20]. With water-iodine based material decomposition images, Lv et al[9] concluded that the IC in images of hepatic lesions acquired in a quantitative parameter was highly precise. Thieme et al[21] found a very good correlation between vessel occlusion depicted at CTA and IC defects in the dual-energy image, indicating that iodine distribution in the parenchyma is closely related to pulmonary perfusion. Therefore, the IC may be considered as an indirect marker of perfusion and tumor vascularity.

Tumor angiogenesis is defined as the formation of new blood vessels from pre-existing vessels[22]. MVD and identification of VEGF in tissues are commonly used biomarkers of angiogenesis[23]. In previous studies, the relationships between dynamic contrast-enhanced perfusion CT parameters and immunohistological markers of angiogenesis have been studied in different tumors, including colorectal cancer[24], advanced GC[6,25], lung cancer[26], etc. However, the studies produced discrepant results on whether the use of blood flow or permeability surface area product were efficacious.

Our results proved a positive linear relationship between IC and MVD in primary GC. The nIC values were significantly elevated in the tumor as compared to the normal gastric wall. Similar results were acquired by Pang et al[27], who considered the nIC value of the infarcted myocardium to be an important indicator of MVD in the 1-min and 3-min CT images. Additionally, the study by Hu et al[12] indicated that the nIC values of three-phase scans had a positive correlation with MVD for detecting the therapeutic response in a pancreatic carcinoma xenograft nude mouse model. Taken together, spectral CT imaging can be used to evaluate angiogenesis in disease.

On the other hand, nIC had a low correlation with VEGF expression, and no significant differences were found in the comparison between IC parameters and VEGF group. VEGF, one of the most prominent biomarkers of angiogenesis studied to date, has been shown to correlate well with CT perfusion in peripheral pulmonary nodules[28]. Moreover, the study of Zhou et al[13] in a rabbit VX2 liver model suggested that nIC and contrast-enhanced ultrasound parameters positively correlated with VEGF and FGF2 expression, while several reports failed to show such a correlation, for example, the GC study performed by Yao et al[6] and the colorectal cancer research by Goh et al[24]. The relationship between VEGF expression and imaging of tumor vascularity is complex. Further investigations of IC and VEGF in a large population with different clinical-pathology are needed.

At the same time, nIC in arterial phase and venous phase displayed different character when stratified. The nIC-AP showed correlations with MVD in relative earlier and differentiated advanced GC, while nIC-VP correlated with MVD in more advanced stage and poorly differentiated advanced GC.

In terms of the acquisition time of our routine abdominal CT scanning, the AP was performed at about 25-30 s after contrast media injection, when the mucosa at the lesion presented as a focal enhanced line. nIC-AP can reflect the blood supply and functional capillary density. In the VP of GC, the markedly increased interstitial fibrous tissue reduced the flow-out speed of the contrast media[29]; thus, more dysfunctional neo-vessels should be considered. Therefore, the nIC-VP may represent the distribution of iodine in interstitial spaces. Under such conditions, the bi-phasic nIC demonstrated the vascular character of GC, and nIC-VP may be better correlated with MVD in advanced GC.

Furthermore, both nICs were different between histological differentiations and Lauren subgroups. The nIC-VP was more effective than nIC-AP in displaying the difference between Lauren subtypes. In general, the nIC value was higher in poorly differentiated and diffused types than that in highly or moderately differentiated and intestinal types. A previous study[30] has found that the degree of tumor angiogenesis was closely related to the pathological grade, that is, the poorer the differentiation of the tumor, the higher the MVD values. Another study[31] demonstrated a correlation between MVD and tumor histological type according to the Lauren classification. Accordingly, it is likely that the nIC value can be used to evaluate histology by mapping the neovascularization of advanced GC. The finding in the present study was also supported by the results of Pan et al[15] and Wang et al[32] in GC research.

The nICs increased in patients with serosal involvement and lymphatic metastasis, but the differences from patients without serosal involvement and lymphatic metastasis did not reach significant difference, for reasons that are not immediately clear, as these clinical-pathological features should reflect the functional status of the vasculature. Both nIC-AP and nIC-VP correlated with MVD of GC located in the antrum and occurring in different genders. However, nICs were not very effective in differentiating between these groups.

There are several potential limitations in our study. First, the tumor vasculature is spatially heterogeneous. Although we selected specimens carefully under the guide of two major specialists, the excision level was barely achieved. It may be questioned whether the part of the histopathological part selected and matched to the imaging measurement represented the angiogenesis of the whole tumor. Second, the sample numbers were limited, especially when subdivided by clinical classification. A larger prospective investigation is needed to confirm the present findings. Third, given the limitations of CT resolution, flat and light ulcerous lesions without enhancement would have been missed in the images, which inducing selection bias.

In conclusion, the bi-phasic nIC values had a positive linear correlation with MVD. nIC-AP reflected the angiogenesis in relatively earlier and well-differentiated advanced GC, while nIC-VP reflected this in further advanced and poorly differentiated GC. Spectral CT with quantitative IC value offers a new choice to evaluate the angiogenesis of gastric cancer noninvasively.