Материал: Advanced Imaging of the Abdomen - Jovitas Skucas

and more adverse features than older patients (28). In Japan, the prevalence of gastric cancer continues to decrease in young patients.

Epstein-Barr virus infection is associated with some gastric cancers, but reported association varies, being greatest in East Asia. These virus-associated cancers tend to be superficial depressed or ulcerated in appearance.

The Netherlands Cohort Study on diet and cancer revealed a strong inverse association between onion consumption and risk of stomach carcinoma (29).

A number of gastric cancer families exist. Transmission tends to be autosomal dominant.

The Leser-Trélat sign consists of eruption of multiple seborrheic keratoses or similar lesions in association with an internal malignancy and is considered to be a paraneoplastic autoimmune phenomenon. Occasionally such seborrheic keratosis regresses after tumor resection but returns with cancer recurrence. Paraneoplastic hypercalcemia occasionally develops. Rare associations exist between gastric cancer and such conditions as acromegaly and polyarteritis nodosa.

An occasional gastric carcinoma perforates; perforation into the peritoneal cavity generally results in an acute abdomen. Among patients with a perforated gastric carcinoma collected from the Japanese literature, perforation occurred with all tumor stages, being 19% with stage I tumors, 12% with stage II, 30% with stage III, and 39% with stage IV tumors (30).

A rare gastric carcinoma is first detected as a metastasis (Fig. 2.11).

Pathology

An adenoma-to-carcinoma transformation sequence, as found with colorectal tumors, is not evident in most gastric tumors. Carcinomas do occur, however, in adenomatous polyposis syndromes. Gastric dysplasia is a precancerous condition. A gastric cancer is often already present when dysplasia is detected.

Epstein-Barr virus–positive early gastric carcinomas have a mostly CD8+ T-lymphocytic surrounding infiltrate, with adjacent gastric mucosa being atrophic and depleted of parietal cells (31). An occasional undifferentiated carcinoma has Epstein-Barr virus limited to tumor epithelial cells, pointing toward a possible pathogenesis for these tumors.

Gastric cancer classifications in use are those of the WHO and the Lauren and Borrmann systems. Lauren subdivides gastric cancer into two types: intestinal and diffuse. The intestinal type tends to be more differentiated than the diffuse type. These classifications are only roughly comparable; the WHO classification of tubular, mucinous, and papillary carcinomas are mostly equivalent to the Lauren classification of intestinal type, and the WHO signet ring and many undifferentiated carcinomas fall into the Lauren diffuse tumor type. The Borrmann classification (Table 2.3) is based on

gross morphology and is in common use by Japanese and some European investigators.

Lauren intestinal-type carcinomas appear to be related to environmental factors and diet, and in countries with decreasing prevalence of gastric cancer it is the intestinal type that is declining, with little change in diffuse type (32). These two subtypes appear to have different pathogenetic processes. Predisposing atrophic gastritis and intestinal metaplasia and dysplasia more often lead to intestinal type of cancer.

The Goseki classification of gastric cancer is based on intracellular mucus production and degree of tubular differentiation, but this classification does not provide additional prognostic value beyond what was available with the tumor, node, metastasis (TNM) staging and the Lauren classification.

An infiltrating gastric scirrhous carcinoma incites a mostly desmoplastic reaction that leads

Table 2.3. Borrmann classification of gastric carcinomas

Type Description

IPolypoid tumor with no ulceration

to a fixed and rigid-appearing gastric wall. Most scirrhous carcinomas result in a linitis plastica appearance, although an occasional one is focal in extent. Lymphoma and some metastases have a similar imaging appearance, but with these tumors the infiltration is primarily by tumor cells.

A signet ring cell carcinoma is a poorly differentiated adenocarcinoma, with tumor cells invading surrounding tissues singly or in small groups. Some of these are associated with an intense lymphocytic and plasma cell infiltrate.

Extrahepatic hepatoid adenocarcinomas exhibit true morphologic and immunohistochemical hepatocellular differentiation. The most common site for a primary extrahepatic hepatoid carcinoma is in the stomach. These tumors are aggressive and metastasize readily to the liver. In fact, some of the older studies of a double malignant primary in the stomach and liver probably represent gastric hepatoid carcinomas with liver metastases. Nevertheless, this topic is complex, and not only patients with liver metastasis but also patients with a synchronous primary liver hepatocellular carcinoma have been reported.

Unusual features of gastric hepatoid adenocarcinomas include high levels of serum a- fetoprotein and an ability to secrete bile. An occasional gastric adenocarcinoma has no hepatoid features, but perigastric lymph node

metastases contain hepatoid characteristics, suggesting hepatoid differentiation only in metastatic foci. The histology of these tumors shows an adenocarcinomatous component intermingled with hepatoid regions.

Some degree of neuroendocrine differentiation is common both in conventional and hepatoid gastric cancers. In most tumors endocrine cells are either scattered or grouped in small clusters and are independent of tumor stage or histologic type. One unusual gastric carcinoma variant is the presence of an osteoclast giant cell stromal component. An occasional odd histologic mix is encountered, such as a primary coexistent adenocarcinoma and choriocarcinoma.

Detection

Imaging

Currently most initial detection of gastric cancer is by endoscopy and biopsy of a suspicious tumor. Barium studies are generally performed for other indications, and gastric cancers are detected incidentally (Fig. 2.12). Nevertheless, an occasional patient still presents with gastric outlet obstruction secondary to an extensive cancer (Fig. 2.13). Currently, many physicians believe that imaging has a role primarily in staging rather than initial cancer detection, yet some studies point to barium having a role in these patients. Thus in patients with pathologically proved scirrhous gastric carcinomas, tumor location and extent were correctly identified by endoscopy in only 33%, while barium studies were accurate in 68% of

Figure 2.12. Fundal adenocarcinoma (arrows) in a patient evaluated for reflux.

patients (33); upper gastrointestinal studies reveal thick and irregular folds, ulcerations and nodularity.

Calcifications are rare in primary gastric cancers, with most calcifications developing in well-differentiated mucinous adenocarcinomas, although an occasional scirrhous carcinoma contains calcifications.

Giant gastric ulcers can be either benign or malignant. Statistically, however, a giant ulcer is more likely to be malignant. The presence of fistulas generally signifies an advanced gastric cancer.

Many radiologists believe that at initial presentation an antral adenocarcinoma does not spread across the pylorus into the duodenal bulb, the exceptions being if a patient is immunocompromised or during late presentation with extensive tumor spread when bulbar involvement is common. Radiologists use this observation to differentiate among an antral cancer, lymphoma, and peptic ulcer disease, with the latter two readily extending across the pylorus. In a minority of patients, however, transpyloric spread occurs with an antral carcinoma. Duodenal invasion is either direct, through deep muscle layers or lymphatics, or through venules. Barium studies reveal a rigid, open, and eccentric pylorus and a deformed bulb. Computed tomography identifies infiltration extending across the pylorus.

A retrospective review found perceptual error to be the reason for not detecting 41% of 27 carcinomas by upper gastrointestinal radiography performed within 3 years prior to tumor diagnosis (34); the most common overlooked finding was an intramural depression.

Computed tomography tumor detection is aided by gastric distention with water (called hydro-CT). A hypotonic agent helps maintain gastric distention.

In a collection of advanced gastric cancers, a triphasic CT technique achieved a 98% detection accuracy (35); of these, 28% were best seen on parenchymal phase during gradual enhancement from the mucosal side. Computed tomography of a typical gastric carcinoma reveals focal gastric wall thickening and mild-to- moderate arterial phase enhancement. Extensive gastric wall thickening develops with tumor growth, with some tumors exhibiting marked enhancement. Computed tomography can miss early linitis plastica, although when it is well established, CT reveals diffuse gastric wall thickening. When extensive, scirrhous carcinomas result in extensive circumferential infiltration of the stomach, although some of these tumors are plaque-like in appearance. A less common linitis plastica cancer appearance is diffusely thickened gastric rugal folds with no apparent overlying ulcerations; the appearance mimics diffuse lymphoma, less often Ménétrier’s disease. Dynamic CT shows most nonscirrhous carcinomas to have either homogeneous contrast enhancement or at least a thin outer layer, whereas scirrhous ones tend toward

ADVANCED IMAGING OF THE ABDOMEN

a gastric wall consisting of two layers: a lower attenuation thick outer layer and a higher attenuation thick inner layer.

Computed tomography arteriography, performed by injecting contrast through a catheter in the celiac trunk, correctly detected seven of eight early gastric cancers and all 13 advanced gastric cancers (36). Computed tomography arteriography is not used as a primary tool for gastric cancer detection, but it is of potential use for cancer staging.

Endoscopic US shows promise in detecting gastric malignancies, achieving sensitivities over 95%. It delineates gastric wall and adjacent structure infiltration. Mucinous carcinomas result in a hyperechoic endoscopic US appearance due to mucinous content and surrounding fibrosis. Endoscopic US of scirrhous carcinomas reveals irregular hypoechoic thickening of submucosa and muscularis propria.

Some scirrhous carcinomas are hypointense on both T1and T2-weighted images, presumably due to the often associated fibrosis. Such a pattern is seen with linitis plastica. Little enhancement is evident with linitis plastica. Magnetic resonance reveals most gastric cancers as early enhancing thickened gastric wall tumors.

Both intestinal adenocarcinomas and various endocrine neoplasms contain receptors for vasoactive intestinal peptide (VIP) and scintigraphy with I-123 VIP appears advantageous. Binding of labeled VIP by primary tumors and metastases is visible for up to 24 hours in primary and recurrent gastric adenocarcinomas.

Endoscopy

Accuracy of gastric cancer detection with endoscopy varies considerably, especially with diffusely infiltrating cancers that are covered by intact mucosa. Even some biopsies are noncontributory, with cancer cells being dispersed in a fibrous matrix. Tumor location is better appreciated on an imaging study than with endoscopy.

In general, endoscopy of a polyp revealing normal overlying gastric mucosa suggests a mesenchymal tumor. Nevertheless, on rare occasion an adenocarcinoma will be mostly intramural and even invade to the serosa and not be detected by endoscopy or even biopsies.

STOMACH

Early Gastric Cancer

Clinical

Controversy surrounds the definition of an early gastric cancer. In Japan, a diagnosis of an early gastric cancer is made considerably more often than in the West. Some observers believe that an early gastric cancer is simply an early stage of a conventional gastric cancer rather than a different disease entity, an opinion not universally shared. Early gastric cancers tend to be intestinal type rather than diffuse type. In Japan about 10% of early gastric cancers are multiple.

Metastasis to lymph nodes is rare if a gastric cancer is limited to the mucosa; nodal involvement is more common with submucosal tumor spread. In the occasional patient with a recurrence after curative resection for early gastric cancer, recurrence is believed to be secondary to occult metastases in perigastric lymph nodes.

Endoscopic resection of early gastric cancer is performed in some countries for elevated mucosal cancers <2cm in size and depressed mucosal cancers without ulceration <1cm in size. Although endoscopic resection is possible for some tumors with submucosal extension, those with deep submucosal invasion are not cured with this approach.

Imaging

Detection and staging cannot be discussed separately for early gastric cancers because the definition already implies that a tumor is limited to the mucosa or at most submucosa. Computed tomography has a limited role in detecting early gastric cancer. Among patients with an early gastric cancer, axial CT identified 64% of tumors, whereas 3D CT images revealed 94% (37). On the other hand, using a triphasic CT technique and a water-filled stomach, another study detected only 23% and achieved a staging accuracy of 15% (35); those early cancers detected were best seen during the arterial or parenchymal phases.

Early gastric cancers detected by CT are polypoid, elevated, or invade submucosa. Wall thickening with early gastric cancers is limited to the inner layer only. These cancers show CT contrast enhancement. Advanced cancers also have con-

trast enhancement, but gastric wall thickening generally is more diffuse.

Postcontrast CT (mucosal phase, 38 to 45 seconds after the start of contrast injection) in patients with early gastric cancer revealed three patterns (38): (1) localized thickening of an inner hyperenhancing layer; (2) focal interruption of an inner hyperenhancing mucosal layer; and (3) focal protrusion of the inner hyperenhancing layer. The lesions became less distinct on a delayed phase. Overall, CT early gastric cancer detection rate was 57%. Of note is that some advanced gastric cancers had an appearance similar to early gastric cancer. Considerable CT difficulty exists differentiating between T1 cancers with submucosal invasion and more advanced cancers.

Conventional endoscopy and endoscopic US have roughly similar accuracy in determining the depth of invasion with early gastric cancers.

Staging

The TNM staging classification is outlined in Table 2.4. Preoperative staging of gastric cancer is of obvious importance but is fraught with uncertainty. Intraoperative surgical assessment tends to overstage early invasion and understage deep invasion. Microcarcinosis, defined as scattered carcinoma cells within lymph node sinuses or pulp without surrounding stromal reaction, is common in otherwise pN0 tumors; the number of detected tumor cells and involved lymph nodes carries prognostic significance, a significance differing from that found with gross lymph node metastasis.

Primary gastric cancer typically spreads to adjacent lymph nodes and liver. In a prospective study of regional lymph nodes from consecutive patients with primary gastric cancer, the mean diameter of tumor-free lymph nodes was 4.1mm and those infiltrated by metastases was 6.0mm (39); a practical problem is that although 80% of tumor-free lymph nodes were <5mm in diameter, 55% nodes containing metastases were also <5mm in diameter and the authors concluded that lymph node size is not a reliable indicator for lymph node metastasis. In spite of such findings, many authors assume an arbitrary boundary between normal and metastatic nodes. Currently no imaging modality is sufficiently accurate to reliably predict lymph node involvement. Perigastric lymph

Table 2.4. Tumor, node, metastasis (TNM) staging of gastric carcinoma

Primary tumor:

Lymph nodes:

Nx Regional nodes cannot be assessed

N0 No metastases in regional lymph nodes

N1 Metastases in 1 to 6 regional lymph nodes

N2 Metastases in 7 to 15 regional lymph nodes

N3 Metastases in more than 15 regional lymph nodes

Distant metastasis:

Mx Distant metastases cannot be assessed M0 No distant metastases

M1 Distant metastases

Source: From the AJCC Cancer Staging Manual, 6th edition (2002), published by Springer-Verlag, New York, NY, used with permission of the American Joint Committee on Cancer (AJCC), Chicago, IL.

nodes can be infiltrated by tumor but still be normal in size; conversely, some enlarged nodes do not contain tumor. Compounding the issue is that a nonneoplastic reactive infiltrate develops in some lymph nodes close to a cancer. Such sarcoid-like epithelioid cell lesions are detected in regional lymph nodes in about one third of gastric cancers; they are independent of tumor stage, type, and tumor grade, and do not aid in prognosis.

ADVANCED IMAGING OF THE ABDOMEN

Fluoroscopic guidance of nasogastric biopsies is feasible to establish boundaries of a known gastric carcinoma, but this technique has achieved only limited acceptance.

Comparing published CT, US, and MR gastric cancer staging results is fraught with difficulty. Sensitivities vary markedly, depending on the basic assumptions used. As a gross approximation, helical CT and endoscopic US appear to provide similar T and N staging accuracies. Most problems arise in differentiating T2 and T3 tumors.

Computed Tomography

Computed tomography staging of gastric cancer has been rather disappointing and continues to be controversial. Some patients have obliteration of adjacent fat planes due to cachexia or inflammation. Some CT studies have achieved a tumor (T) and node (N) staging accuracy of only about 50% (40), although multislice triphasic CT results are more accurate. In general, CT differentiation between T3 and T4 cancers is not sufficiently accurate to predict resectability. It is accurate, however, in detecting distant metastases, and some surgeons rely on CT primarily to detect these metastases. Computed tomography more often suggests correctly unresectability rather than resectability.

Depth of tumor invasion and serosal invasion are believed to be difficult to determine even with helical CT. Yet in establishing serosal invasion, use of hydro-CT, induced hypotonia, and prone patient position has achieved a sensitivity of 100% and specificity of 80% to 87% (41).

The ability of multislice CT to detect lymph nodes depends on node size; roughly half of nodes <10mm and about 75% of those >10mm are detected. Detecting tumor in a lymph node is another matter; intrinsically, nothing in the CT appearance suggests that a particular lymph node is involved by tumor.

Staging accuracy of CT arteriography in patients with gastric cancer was 77% for serosal invasion and 76% for regional lymph node metastasis (36); of note is that accuracy for T staging of early gastric cancers was only 50%.

Whether multislice 3D images aid in staging gastric cancers remains to be determined.