2462

Primary Gastrointestinal Lymphoma in Japan

A Clinicopathologic Analysis of 455 Patients with Special Reference to its Time Trends

Shotaro Nakamura, M.D.1 Takayuki Matsumoto, M.D.1 Mitsuo Iida, M.D.1 Takashi Yao, M.D.2 Masazumi Tsuneyoshi, M.D.2
1

BACKGROUND. An optimal treatment modality for patients with primary gastrointestinal lymphoma has not yet been established. This study aimed to elucidate the clinicopathologic features of this disease and the inuence of therapeutic modalities on the prognosis in Japanese patients METHODS. The clinicopathologic features of 455 patients with primary gastrointestinal lymphoma were investigated retrospectively regarding treatment modalities and time trends.

Department of Medicine and Clinical Science, Graduate School of Medical Sciences, Kyushu University, Fukuoka, Japan. Department of Anatomic Pathology, Pathological Sciences, Graduate School of Medical Sciences, Kyushu University, Fukuoka, Japan.

RESULTS. This study comprised 342 patients (75%) with gastric lymphoma, 96 patients (22%) with intestinal lymphoma, and 17 patients (4%) with both gastric and intestinal lymphoma. Two hundred thirty-one (51%) patients were classied as having low-grade B-cell lymphoma including 200 marginal zone lymphoma of mucosa-associated lymphoid tissue (MALT) type, 185 (41%) patients were classied as having high-grade B-cell lymphoma including 76 diffuse large cell lymphoma plus MALT lymphoma, and 39 (9%) patients were classied as having T-cell lymphoma. The frequency of nonsurgical treatment, including Helicobacter pylori eradication, chemotherapy, and radiation, increased during the latest decade. Patients who received nonsurgical treatment showed a better overall survival than those treated by surgery, but event-free survival did not differ between two groups. Cox multivariate analysis revealed that early stage, younger age, gastric localization, B-cell phenotype, and absence of B symptoms were independent prognostic factors for better overall and event-free survivals. Mucosa-associated lymphoid tissue-derived lymphoma was also an independent prognostic factor for event-free survival, but not for overall survival. CONCLUSIONS. Nonsurgical treatment may be an optimal therapeutic modality for patients with primary gastrointestinal lymphoma. Cancer 2003;97:246273. 2003 American Cancer Society. DOI 10.1002/cncr.11415 KEYWORDS: gastrointestinal tract, malignant lymphoma, mucosa-associated lymphoid tissue, Helicobacter pylori, nonsurgical treatment, time trends.

Address for reprints: Shotaro Nakamura, M.D., Department of Medicine and Clinical Science, Graduate School of Medical Sciences, Kyushu University, Maidashi 3-1-1, Higashi-ku, Fukuoka 8128582, Japan; Fax: (011) 81-92-642-5273; E-mail: shonaka@intmed2.med.kyushu-u.ac.jp Received December 3, 2002; revision received February 10, 2003; accepted February 12, 2003. 2003 American Cancer Society

rimary gastrointestinal lymphoma is the most frequent type of extranodal malignant lymphoma, accounting for 30 45% of all extranodal lymphomas and for 4 20% of all non-Hodgkin lymphomas.1 4 Although many published articles have reported on various prognostic factors for primary gastrointestinal lymphoma,525 the number of subjects in such series was small. In addition, the majority of patients in these reports were treated by surgery-based modalities. During the past two decades, the diagnosis and treatment of gastrointestinal lymphoma have changed tremendously. Isaacson and Wright26 introduced the concept of mucosa-associated lymphoid

Gastrointestinal Lymphoma in Japan/Nakamura et al.

2463

tissue (MALT) lymphoma in 1983. This new categorization of lymphoma has become widely accepted and has been incorporated into the Revised EuropeanAmerican Lymphoma27 and World Health Organization (WHO) classications.28 To maintain patients quality of life, nonsurgical treatments such as Helicobacter pylori eradication, chemotherapy, and radiation are becoming increasingly popular even for patients with resectable gastrointestinal lymphoma.29 31 However, the optimal treatment for this disease still remains controversial. Previous studies on the prognostic factors of primary gastric32 and intestinal33 lymphomas included mainly patients who underwent surgical resections. However, in recent years, patients with gastrointestinal lymphomas have received nonsurgical treatments at our institution (Kyushu University, Fukuoka, Japan).34 In addition, the prognostic data on patients with lymphoma of multiple involvements remain to be elucidated. To determine the prognostic factors for gastrointestinal lymphoma and to evaluate the inuence of therapeutic modalities on the prognosis, we analyzed a large number of patients with special reference to the MALT concept and time trends.

MATERIALS AND METHODS

Subjects
The series consisted of 455 consecutive adult Japanese patients with primary gastrointestinal lymphoma who were diagnosed at Kyushu University between 1963 and 2002. All cases satised the criteria for primary gastrointestinal lymphoma as dened by Lewin et al.6 The subjects in our previous studies3234 were included in the current study. Pediatric patients younger than 15 years of age were excluded from this study. Based on the time of the initial diagnosis, we divided the times into eight segments of 5 years each. Subsequently, these segments were subgrouped into the following three periods: Period A (19631982, before the introduction of MALT lymphoma26); Period B (19831992, after the introduction of MALT lymphoma); and Period C (19932002, after the introduction of H. pylori eradication for gastric MALT lymphoma29).

D1, bcl-2, bcl-6, c-myc, p53, Ki-67, immunoglobulin light chains ( , ), or F1. All H & E and immunohistochemical slides were reviewed separately by two observers (S.N. and T.Y.) and a common consensus was reached in all cases. Histologic classication was done according to WHO criteria28 with the following modications:33 low-grade B-cell lymphomas comprised marginal zone B-cell lymphoma of MALT type (MALT lymphoma), follicular lymphoma, mantle cell lymphoma, and plasmacytoma; high-grade B-cell lymphomas comprised diffuse large B-cell lymphoma (DLBL) plus MALT lymphoma (formerly referred to as high-grade MALT lymphoma35), DLBL without MALT lymphoma, Burkitt lymphoma, or lymphoblastic lymphoma; and T-cell lymphoma. Marginal zone B-cell lymphoma (MALT lymphoma) was dened as a diffuse proliferation of centrocytelike cells with lymphoepithelial lesions.26,35 Diffuse large B-cell lymphomas were divided into two entities according to the presence or absence of MALT lymphoma areas.33 Immunoproliferative small intestinal disease (IPSID) was classied into either marginal zone B-cell lymphoma or DLBL plus MALT lymphoma.35 Therefore, each lymphoma was divided histologically into either MALTderived lymphoma (marginal zone B-cell lymphoma and DLBL plus MALT lymphoma) or nonMALT-derived lymphoma (others).33

Staging and Diagnostic Procedures

The stage of lymphoma was determined according to the Lugano International Conference classication (I, II1, II2, IIE, or IV), which was proposed as a modied version of the Ann Arbor criteria for gastrointestinal lymphoma.36 The staging workup included physical examination with inspection of the Waldeyer tonsillar ring, blood cell count and serum chemistry, chest radiographs (all patients), abdominal ultrasound (416 patients), computed tomography scan of the chest and abdomen (334 patients), gallium scintigraphy (248 patients), and bone marrow aspiration or biopsy (357 patients). Endoscopy with biopsy and/or barium studies of the upper and lower gastrointestinal tracts were used to assess the extent of lymphoma. The macroscopic type of lymphoma was either supercial, polypoid, ulcerative, polyposis, diffuse, or mixed type.32,33 This typing was determined based on the resected specimens or on the double-contrast radiographs with endoscopic ndings. The depth of tumor invasion was determined by endoscopic ultrasonography in 150 patients.34 Otherwise, the depth

Histologic Classication
All histologic materials were obtained by endoscopic biopsy, surgery, and/or autopsy. These tissue specimens were stained routinely with hematoxylin and eosin (H & E). Immunohistochemical staining for CD3 and CD20 was performed on all 455 specimens. In some selected cases, additional staining was done using antibodies CD5, CD10, CD23, CD30, CD56, cyclin

2464

CANCER May 15, 2003 / Volume 97 / Number 10

TABLE 1 Histologic Classication and Sites of Origin

Histologic type Low-grade B-cell lymphoma Marginal zone lymphoma (MALT) Follicular lymphoma Mantle cell lymphoma Plasmacytoma High-grade B-cell lymphoma DLBL plus MALT lymphoma DLBL without MALT lymphoma Burkitt lymphoma Lymphoblastic lymphoma T-cell lymphoma Gastric group (n 342) (%) Intestinal group (n 96) (%) Combined group (n 17) (%) Total (n 455) (%)

170 (50) 20 (6) 0 2 (0.6) 52 (15) 77 (23) 1 (0.3) 1 (0.3) 19 (6)

27 (28) 7 (7) 2 (2) 0 20 (21) 20 (21) 5 (5) 3 (3) 12 (13)

3 (18) 0 0 0 4 (23) 2 (12) 0 0 8 (47)

200 (44) 27 (6) 2 (0.4) 2 (0.4) 76 (17) 99 (22) 6 (1.3) 4 (0.8) 39 (9)

MALT: mucosa-associated lymphoid tissue; DLBI: diffuse large B-cell lymphoma.

was determined by histology of the resected specimens.

RESULTS
Clinical and Histologic Features
The 455 patients had a mean age of 58 years (range, 16 90 years) and comprised 247 men and 208 women. The most frequent primary site was the stomach (gastric group; 342 patients [75%]), followed by the intestine (intestinal group; 96 patients, [21%]). The remaining 17 (3.7%) patients had both gastric and intestinal involvement (combined group). Among patients in the gastric group, 93 had tumors mainly in the upper third portion, 132 in the middle third, 76 in the lower third, and 41 in two or all portions. Among patients in the intestinal group, 6 had tumors in the duodenum, 10 in the jejunum, 43 in the ileum, 6 in the duodenum, jejunum, and ileum, 11 in the ileocecal region, 13 in the colorectum, 1 in the appendix, and 6 in both small and large bowel. The histologic classication and primary site are shown in Table 1. The study sample comprised 200 (44%) patients who were classied as marginal zone B-cell lymphoma, 27 (5.9%) as follicular lymphoma, 2 (0.4%) as mantle cell lymphoma, 2 (0.4%) as plasmacytoma, 76 (17%) as DLBL plus MALT lymphoma, 99 (22%) as DLBL without MALT lymphoma, 6 (1.3%) as Burkitt lymphoma, 4 (0.8%) as lymphoblastic lymphoma, and 39 (9%) as T-cell lymphoma. In the gastric group, marginal zone B-cell lymphoma was the most frequent histologic type (50%), whereas DLBL and Tcell lymphoma were the most frequent types in the intestinal (43%) and combined groups (47%), respectively. Two hundred forty-eight patients (55%) had Stage I disease, 95 (21%) had Stage II1 disease, 56 (12%) had Stage II2 disease, 10 (2.2%) had Stage IIE disease, and

Treatment and Response

Therapeutic modalities were divided into either surgical resection, nonsurgical treatment (chemotherapy, radiation, or antibiotic treatment such as H. pylori eradication), or both. Complete remission (CR) was dened as the disappearance of all clinical evidence of lymphoma. Partial remission was dened as a decrease of 50% or greater in the sum of the tumors. No response was deend as tumor reduction of less than 50%, or disease progression.

Prognosis and Statistical Analysis

Overall survival (OS) was measured from the date of diagnosis to death from any cause. Event-free survival (EFS) was measured from the date of diagnosis to disease progression, disease recurrence, or death from any cause. The probability of OS and EFS was calculated by the KaplanMeier method and the value was compared using the log rank test. All variables that inuenced the prognosis (P 0.1) were put into a multivariate analysis using the Cox proportional hazards model. Other statistical differences were evaluated using either the Fisher exact probability test, the chi-square test, the MannWhitney U test, or the KruskalWallis test. A value of P less than 0.05 for each test was statistically signicant. For multiple comparisons, however, P values were interpreted after the Bonferroni correction. All statistical analyses were performed using the Statistical Software Package for the Social Sciences version 8.0J (SPSS Japan Inc., Tokyo, Japan).

Gastrointestinal Lymphoma in Japan/Nakamura et al. TABLE 2 Comparison of Clinicopathologic Features and Sites of Origin
Characteristics Mean age (yrs) Gender Male Female Histology Low-grade B-cell High-grade B-cell T-cell Stage I II1/II2 IIE/IV Depth of invasion Mucosa/submucosa Beyond submucosa Treatment Surgery alone Nonsurgical Both No Response to treatment (n Complete remission Partial remission No response Gastric (n 342) (%) 57.8 172 (50) 170 (50) 192 (56) 131 (38) 19 (6) 215 (63) 94 (28) 33 (10) 181 (53) 161 (47) 178 (52) 78 (23) 82 (24) 4 (1) (n 338) 312 (92) 23 (7) 3 (1) Intestinal (n 96) (%) 58.3 64 (67) 32 (33) 36 (38) 48 (50) 12 (13) 30 (31) 53 (55) 13 (14) 20 (21) 76 (79) 45 (47) 15 (16) 34 (35) 2 (2) (n 94) 64 (68) 21 (22) 9 (10) Combined (n 17) (%) 57.6 11 (65) 6 (35) 3 (18) 6 (35) 8 (47) 3 (18) 4 (24) 10 (59) 7 (41) 10 (59) 0 9 (53) 4 (24) 4 (24) (n 13) 2 (15) 6 (46) 5 (39)

2465

P valuea NS 0.0118

0.0001

0.0001

0.0001

0.0001

445)

0.0001

NS: not signicant. a P 0.007 is signicant using the Bonferroni correction.

46 (10%) had Stage IV disease. Macroscopically, 155 tumors (34%) were classied as supercial type, 95 (21%) as polypoid type, 119 (26%) as ulcerative type, 6 (1.3%) as polyposis type, 40 (9%) as diffuse type, and 40 (9%) as mixed type. Two hundred twenty-three patients (49%) underwent surgical resection alone, whereas 102 (22%) received nonsurgical treatment (60 had H. pylori eradication alone, 16 chemotherapy alone, 15 eradication plus chemotherapy, 5 eradication plus radiation, 2 chemotherapy plus radiation, and 4 eradication plus chemotherapy and radiation). Major regimens for H. pylori eradication consisted of a proton pump inhibitor (omeprazole, lansoprazole, or rabeprazole) with a combination of antibiotics (amoxicillin plus clarithromycin with or without metronidazole) for 14 days. A combination of both surgery and nonsurgical treatment was performed in 120 patients (26%); 105 were treated by surgery plus chemotherapy, 4 by surgery after eradication, 4 by surgery after eradication plus chemotherapy, and 7 by surgery plus chemotherapy and radiation. The remaining 10 patients (2%) did not receive any treatment. Among the 153 patients

treated by chemotherapy, 142 received four or six cycles of the cyclophosphamide, doxorubicin, vincristine, and prednisolone regimen and the remaining 11 patients were treated by oral monochemotherapy with cyclophosphamide.34 Of the 445 treated patients, 378 patients (83%) achieved a CR, 50 (11%) achieved a partial remission, and 17 (4%) showed no response.

Relationship between Clinicopathologic Features and Primary Site of Lymphoma

Table 2 indicates the clinicopathologic features of the three groups classied by primary site of lymphoma. Male patients predominated in both the intestinal and combined groups. The most frequent histologic subgroup was low-grade B-cell lymphoma in the gastric group, high-grade B-cell lymphoma in the intestinal group, and T-cell lymphoma in the combined group (P 0.0001). The clinical stage also differed among the three groups: 63% of the patients in the gastric group had Stage I disease, 55% of patients in the intestinal group had intermediate-stage disease (i.e., Stage II1 or II2), and 59% of patients in the combined

2466

CANCER May 15, 2003 / Volume 97 / Number 10

number of patients in the gastric and intestinal groups, increased over time. An analysis of three chronologic periods revealed that 144 patients (32%) belonged to Period A, 117 (26%) to Period B, and 194 (43%) to Period C. Table 3 comparatively shows the clinicopathologic ndings of patients in Periods AC. Over these three periods, the average age of patients increased from 52 years to 61 years (P 0.0001), whereas the male-to-female ratio did not change. The histology and depth of tumor invasion did not differ between Periods A and B. However, the frequency of low-grade B-cell lymphomas and tumors restricted to the submucosa increased in Period C. In addition, the frequency of the combined group, as well as that of advanced-stage disease (i.e., Stage IIE or IV), increased in Period C. Nonsurgical treatment was applied more frequently to the patients in Period C than in the earlier two periods (P 0.0001). Antibiotic treatment such as H. pylori eradication was performed in 91 of the 194 (47%) patients in Period C, but only in two patients with IPSID in the former two periods. The response to treatment did not change over the three periods. Greater than 80% of the patients achieved CR in each period.

Survival and Prognostic Factors

Ten patients who did not receive any treatment were excluded from the follow-up analysis. The follow-up after the diagnosis in 445 treated patients ranged from 1 to 366 months (mean, 56 months). The OS and EFS rates after 5 years were 72% and 68%, respectively. Among the three groups classied by site (Table 4), the gastric group showed the highest OS and EFS rates, whereas the intestinal group showed intermediate values and the combined group showed the lowest values (P 0.0001; Fig. 2). Signicant differences were also observed between the gastric and intestinal groups (OS, P 0.0001), but not between the intestinal and combined groups (OS, P 0.2269). Patients with low-grade B-cell lymphoma had better OS and EFS rates than patients with high-grade B-cell lymphoma (OS, P 0.0001). Patients with T-cell lymphoma had poorer OS and EFS rates compared with patients with high-grade B-cell lymphoma (OS, P 0.0001; Fig. 3). Patients with Stage I disease had better OS and EFS rates compared with patients with Stage II1 or II2 (OS, P 0.0001) or patients with Stage IIE or IV disease (OS, P 0.0001). A signicant difference was also found between the latter two groups (OS, P 0.0002). Patients with MALT-derived lymphoma demonstrated signicantly better OS and EFS

FIGURE 1.

Trends in the number of patients with primary gastrointestinal lymphoma among eight chronologic segments. (A) All patients (n 455). (B) Patients with gastric lymphoma (n 342) and intestinal lymphoma (n 96).

group had advanced-stage disease (i.e., Stage IIE or IV). Tumor invasion was deeper among patients in the intestinal group compared with the other two groups. The therapeutic modalities did not differ between the gastric group and intestinal groups. Nonsurgical treatment was applied more frequently in the combined group than in the other two groups. The response to treatment differed among the three groups: CR was achieved in 92% of the patients in the gastric group compared with 68% of patients in the intestinal group and only 15% of patients in the combined group.

Time Trends of Gastrointestinal Lymphoma

There were 14, 30, 48, 52, 58, 59, 96, and 98 patients, respectively, in eight chronologic segments. As shown in Figure 1, the number of all patients, as well as the

Gastrointestinal Lymphoma in Japan/Nakamura et al. TABLE 3 Comparison of the Clinicopathologic Features and Chronologic Periods
Period A (n 144) (%) Mean age (yrs) Gender Male Female Site of origin Gastric Intestinal Combined Histology Low-grade B-cell High-grade B-cell T-cell Stage I II1/II2 IIE/IV Depth of invasion Mucosa/submucosa Beyond submucosa Treatment Surgery alone Nonsurgical Both No Response to treatment (n Complete remission Partial remission No response 52.0 79 (55) 65 (45) 119 (83) 24 (17) 1 (0.7) 62 (43) 70 (49) 12 (8) 90 (63) 44 (31) 10 (7) 59 (41) 85 (59) 114 (79) 2 (1) 26 (18) 2 (1) (n 142) 123 (87) 18 (13) 1 (0.7) Period B (n 117) (%) 59.3 67 (57) 50 (43) 86 (74) 29 (25) 2 (2) 52 (44) 54 (46) 11 (9) 51 (44) 54 (46) 12 (10) 44 (38) 73 (62) 70 (60) 4 (3) 39 (33) 4 (3) (n 113) 98 (87) 12 (11) 3 (3) Period C (n 194) (%) 61.4 101 (52) 93 (48) 137 (71) 43 (22) 14 (7) 117 (60) 61 (31) 16 (8) 107 (55) 53 (27) 34 (18) 105 (54) 89 (46) 39 (20) 96 (49) 55 (28) 4 (2) (n 190) 157 (83) 20 (11) 13 (7)

2467

P valuea 0.0001 NS

0.0047

0.0091

0.0004

0.0069

0.0001

445)

NS

NS: not signicant. a P 0.006 is signicant using the Bonferroni correction.

rates than non-MALT lymphoma patients (P 0.0001; Fig. 4). Patients with the following characteristcs had signicantly better OS and EFS rates: younger age, female gender, tumors restricted to the submucosa, supercial tumors, small tumors ( 8 cm), tumors without perforation, absence of B symptoms, and radical resection. Patients who underwent nonsurgical treatment showed a better OS than those who had surgerybased modalities (P 0.0118). However, such a difference was not found in EFS (P 0.3056; Fig. 5). Chronologically, patients in Period C (19932002) demonstrated a better OS than those in the earlier two periods (P 0.0108), but EFS did not differ among the groups (P 0.6823). Results of multivariate analysis are shown in Table 5. Signicant independent prognostic factors for better OS and EFS were early stage, younger age, gastric localization, B-cell phenotype, and absence of B symptoms. The presence of MALT-derived lymphoma

was not a signicant factor for OS, but it was a significant independent prognostic factor for EFS. A time trend was a signicant prognostic factor for OS, but not for EFS.

DISCUSSION
There are many publications regarding the prognosis of primary gastrointestinal lymphoma. However, most of these studies analyzed a small number of subjects and only ve recruited more than 300 patients.2,17,2325 The current study is the second largest series, which follows an earlier study by Freeman et al.2 Table 6 shows the primary site of gastrointestinal lymphoma in ve studies,2,17,2325 together with a larger population-based report by Gurney et al.37 The most common primary site was the stomach with a frequency ranging from 44% to 75%. In our series, the gastric group comprised 75% of all patients. This percentage was similar to that in studies from Austria17 and Germany25 and the value is higher than that observed in

2468

CANCER May 15, 2003 / Volume 97 / Number 10

TABLE 4 Signicant Prognostic Factors on Univariate Analysis

Overall survival No. of patients Age (yrs) 57 or younger 58 or older Gender Male Female Site of origin Gastric Intestinal Combined Histology Low-grade B High-grade B T-cell MALT-derived Yes No Depth of tumor Within submucosa Beyond submucosa Stage I II1/II2 IIE/IV Macroscopic type Supercial Others Size of tumor (cm) 8 8 or more Perforation Yes No B symptoms Yes No Radical surgery Yes No Treatment No surgery Surgery-based Period A/B (19631992) C (19932002)
MALT: mucosa-associated lymphoid tissue. a Assessed by the log-rank test.

Event-free survival P valuea 5-Yr (%) P valuea

5-Yr (%)

202 243 243 202 338 94 13 228 181 36 272 173 203 242 244 148 53 154 291 248 197 10 435 67 378 158 287 102 343 255 190

81 64 68 77 77 58 16 87 63 27 84 54 86 60 87 62 26 88 64 78 65 0 73 45 77 82 66 86 69 68 79

0.0001

79 59 64 73 74 53 25 81 60 27 79 52 80 59 82 60 23 82 61 74 62 0 69 38 74 80 61 71 67 68 70

0.0001

0.0022

0.0008

0.0001

0.0001

0.0001

0.0001

0.0001

0.0001

0.0001

0.0001

0.0001

0.0001

0.0001

0.0001

0.0149

0.0270

0.0001

0.0001

0.0001

0.0001

0.0015

0.0003

0.0118

0.3056

0.0108

0.6823

four other studies.2,23,24,37 Geographic variations in the prevalence of H. pylori infection, celiac disease, or other environmental factors may be the cause of the difference.35,38,39 In Middle Eastern countries, the frequency of intestinal lymphoma is high (range, 49

81%). This may be partly explained by the high prevalence of IPSID in these areas.22,40,41 Among the three groups classied by primary site, histology and clinical stage were different. The gastric group contained a predominance of early-stage, low-

Gastrointestinal Lymphoma in Japan/Nakamura et al.

2469

FIGURE 2. The survival curves for 445 patients with primary gastrointestinal
lymphoma as stratied by the three groups according to anatomic site of origin. (A) Overall survival among three groups (P 0.0001). Gastric versus intestinal group (P 0.0001) and intestinal versus combined group (P 0.2269). (B) Event-free survival among three groups (P 0.0001). Gastric versus intestinal group (P 0.0001) and intestinal versus combined group (P 0.4832).

FIGURE 3. The survival curves for 445 patients with primary gastrointestinal
lymphoma as stratied by the three histologic subgroups. (A) Overall survival among the three groups (P 0.0001). Low-grade versus high-grade B-cell type (P 0.0001) and high-grade B-cell versus T-cell type (P 0.0001). (B) Event-free survival among the three groups (P 0.0001). Low-grade versus high-grade B-cell type (P 0.0001) and high-grade B-cell versus T-cell type (P 0.0001).

grade B-cell lymphomas, most of which were marginal zone lymphomas. Conversely, the intestinal group contained a predominance of intermediate-stage (II1 or II2), high-grade B-cell lymphomas and the combined group comprised a pedominance of advancedstage, T-cell lymphoma. These differences may have been associated with the difference in response to treatment and patient survival. A better prognosis for gastric lymphoma compared with other localizations has been described previously,1,4,8,11,17,18,21,24,25 although several authors showed no difference in the prognosis between gastric lymphoma and intestinal lymphoma.5,9,1315,19,20,22,23 Few publications have shown prognostic data on patients with both gastric and intestinal involvement.

As found in our study, dAmore et al.23 demonstrated that patients with combined gastric and intestinal involvement showed a signicantly poorer survival than patients with the involvement localized in the stomach or in the intestine. Similar results were reported by Ruskone-Fourmestraux et al.18 and Koch et al.25 The poor prognosis of the combined group in our study may have been associated with more frequent T-cell phenotype and more advanced-stage disease. In three studies,18,23,25 however, T-cell lymphomas were extremely rare. This discrepancy may correlate with the difference in the incidence of T-cell lymphoma between Western countries and Japan.28,42

2470

CANCER May 15, 2003 / Volume 97 / Number 10

FIGURE 4. The survival curves for 445 patients with primary gastrointestinal
lymphoma as stratied by mucosa-associated lymphoid tissue (MALT)-derived lymphoma versus non-MALT lymphoma. (A) Overall survival (P 0.0001). (B) Event-free survival (P 0.0001).

FIGURE 5. The survival curves for 445 patients with primary gastrointestinal
lymphoma as stratied by nonsurgical treatment versus surgery-based modalities. (A) Overall survival (P 0.0118). (B) Event-free survival (P 0.3056).

The incidence of gastric and intestinal lymphoma in the U.S.43,44 and Europe37 has increased over the past two decades. Our data conrmed this trend even in Japan (Fig. 1). Advances in diagnostic procedures may have led to an improvement in the accuracy in the diagnosis of lymphoma. In addition, an actual increase in the occurrence of gastrointestinal lymphoma should also be considered. Increasing exposure or susceptibility to risk factors, such as H. pylori infection, ultraviolet light, chemicals, and excessive protein or fat in the diet, also may have contributed to this increase.37,38 In Period C (19932002), both the number and the proportion of low-grade B-cell lymphoma increased. This is attributable to an increase in the diagnosis of gastric MALT lymphoma by endoscopy with biopsy. In earlier periods, a considerable number of patients

with gastric MALT lymphoma were diagnosed with reactive lymphoid hyperplasia (pseudolymphoma) or chronic gastritis.32,35 In addition, the increase in nonsurgical treatment in Period C seems to have owed to the application of H. pylori eradication. Furthermore, the increase in the number of patients in the combined group and advanced-stage patients in Period C may be due to recent improvements in staging modalities such as colonoscopy. Our multivariate analysis revealed an earlier stage, younger age, gastric localization, B-cell phenotype, and absence of B symptoms to be independent prognostic factors for better OS and EFS. However, these prognostic factors may have a limited value because this study was based on a retrospective analysis of patients with different histologic types and heterogeneous treatment modalities over a prolonged period. To our knowledge, 11 English studies performed a mul-

Gastrointestinal Lymphoma in Japan/Nakamura et al. TABLE 5 Independent Prognostic Factors Determined by Multivariate Analysis
Overall survival Variable Stage (I vs. II1/II2 vs. IIE/IV) Age Gastric localization Immunophenotype (B-cell vs. T-cell) B symptoms MALT-derived lymphoma Time trends (19631993 vs. 19932002) Coefcient/SE 5.1198 4.0000 3.2978 2.3900 2.2113 1.5243 3.0536 P value 0.0001 0.0001 0.0010 0.0168 0.0270 0.1275 0.0023 Event-free survival Coefcient/SE 4.7564 3.8143 3.7568 3.2621 2.4404 2.1081 NE

2471

P value 0.0001 0.0001 0.0002 0.0011 0.0147 0.0350 NE

SE: standard error; MALT: mucosa-associated lymphoid tissue; NE: not evaluated.

TABLE 6 Sites of Origin in Primary Gastrointestinal Lymphoma in the Literature

Anatomic sites of origin (%) Author (yr) Freeman et al. (1972)2 Radaszkiewicz et al. (1992)17 dAmore et al. (1994)23 Liang et al. (1995)24 Gurney et al. (1999)37d Koch et al. (2001)25 Current study (2003) Nation United States Austria Denmark Hong Kong United Kingdom Germany Japan No. of patients 538 372 306 425 1047 371 455 Gastric 346 (64) 258 (69) 175 (57) 238 (56) 463 (44) 277 (75) 342 (75) Intestinal 192 (36) 71 (19) 109 (36) 184 (43) 440 (42) 70 (19) 96 (21) Combined ND 8 (2.2)a 22 (7.2) ND ND 24 (6.5)a 17 (3.7) Others 35 (9.4)b 3 (0.8)c 144 (14)b

ND: not described. a Cases involving multiple gastrointestinal tract organs. b Unspecied. c Cases with primary esophageal lymphoma. d Registry data without prognostic analysis.

tivariate analysis for prognostic factors on patients with both gastric and intestinal lymphoma.9,1116,18,19,22,24 All but one15 demonstrated an early stage to be the most signicant prognostic factor for better survival.9,1114,16,18,19,22,24 The next most frequent independent prognostic factor was a younger age.11,15,16,18,24 Other factors reported to have contributed to survival included female gender,11,12,22 surgical resection,13,14,18 low-grade histology,12,13,24 smaller size of tumor,15,24 and a good performance status.15,22 Anatomic site (gastric localization) was an independent prognostic factor in two studies.12,18 No previous studies have shown the prognostic signicance of T/B-cell immunophenotype in primary gastrointestinal lymphoma. In addition, MALT-derived lymphoma including DLBL plus MALT lymphoma was an independent prognostic factor for EFS in the current study (Table 5). Similar results were demonstrated in two previous studies.12,19 Chronologic periods signicantly inuenced OS, but not EFS in our study. Weingrad et al.9 showed a

similar time trend toward a better survival in the latest than the earliest period. However, a better OS may have been biased by differences in the length of follow-up and by the number of deaths unrelated to lymphoma. It is important to note the fact that the type of therapeutic modalities did not inuence EFS and that patients with nonsurgical treatment showed a better OS than those treated by surgery. The role of surgery in the treatment of gastrointestinal lymphoma has been a matter of debate.1318,30,31,45,46 Although past studies showed that surgical resectability was of signicance, recent studies have demonstrated that nonsurgical strategies and surgery-based modalities had an equivalent efcacy.30,31,45 Based on all these observations, and taking patients quality of life into consideration, nonsurgical treatment is an alternative and optimal therapeutic modality for gastrointestinal lymphoma. Further randomized prospective studies with a large number of patients are still needed to establish

2472

CANCER May 15, 2003 / Volume 97 / Number 10

gastrointestinal non-Hodgkins lymphoma: a review of 175 British National Lymphoma Investigation cases. Br J Cancer. 1993;67:776 782. Tondini C, Giardini R, Bozzetti F, et al. Combined modality treatment for primary gastrointestinal non-Hodgkins lymphoma: the Milan Cancer Institute experience. Ann Oncol. 1993;4:831 837. Franssila KO, Jaser N, Sivula A. Gastrointestinal nonHodgkins lymphoma: a population-based clinicopathological study of 111 adult cases with a follow-up of 10-15 years. APMIS. 1993;101:631 641. Amer MH, El-Akkad S. Gastrointestinal lymphoma in adults: clinical features and management of 300 cases. Gastroenterology. 1994;106:846 858. dAmore F, Brincker H, Grnbk K, et al. Non-Hodgkins lymphoma of the gastrointestinal tract: a population-based analysis of incidence, geographic distribution, clinicopathologic presentation features, and prognosis. J Clin Oncol. 1994;12:16731684. Liang R, Todd D, Chan TK, et al. Prognostic factors for primary gastrointestinal lymphoma. Hematol Oncol. 1995; 13:153163. Koch P, del Valle F, Berdel WE, et al. Primary gastrointestinal non-Hodgkins lymphoma: I. Anatomic and histologic distribution, clinical features, and survival data of 371 patients registered in the German Multicenter Study GIT NHL 01/92. J Clin Oncol. 2001;19:38613873. Isaacson P, Wright DH. Malignant lymphoma of mucosaassociated lymphoid tissue: a distinctive type of B-cell lymphoma. Cancer. 1983;52:1410 1416. Harris NL, Jaffe ES, Stein H, et al. A revised EuropeanAmerican classication of lymphoid neoplasms: a proposal from the International Lymphoma Study Group. Blood. 1994;84:13611392. Jaffe ES, Harris NL, Stein H, Vardiman JW, ed. World Health Organization classication of tumours. Pathology and genetics of tumours of haematopoietic and lymphoid tissues. Lion: IARC Press; 2001. Wotherspoon AC, Doglioni C, Diss TC, et al. Regression of primary low-grade B-cell gastric lymphoma of mucosa-associated lymphoid tissue type after eradication of Helicobacter pylori. Lancet. 1993;342:575577. Maor MH, Velasquez WS, Fuller LM, Silvermintz KB. Stomach conservation in stages IE and IIE gastric non-Hodgkins lymphoma. J Clin Oncol. 1990;8:266 271. Koch P, del Valle F, Berdel WE, et al. Primary gastrointestinal non-Hodgkins lymphoma: II. Combined surgical and conservative or conservative management only in localized gastric lymphomaresults of the prospective German Multicenter Study GIT NHL 01/92. J Clin Oncol. 2001;19:3874 3883. Nakamura S, Akazawa K, Yao T, Tsuneyoshi M. Primary gastric lymphoma: a clinicopathologic study of 233 cases with special reference to evaluation with the MIB-1 index. Cancer. 1995;76:13131324. Nakamura S, Matsumoto T, Takeshita M, et al. A clinicopathologic study of primary small intestine lymphoma: prognostic signicance of mucosa-associated lymphoid tissue-derived lymphoma. Cancer. 2000;88:286 294. Nakamura S, Matsumoto T, Suekane H, et al. Predictive value of endoscopic ultrasonography for regression of gastric low grade and high grade MALT lymphomas after eradication of Helicobacter pylori. Gut. 2001;48:454 460.

the optimal management for patients with primary gastrointestinal lymphoma.

20.

REFERENCES
1. Loehr WJ, Mujahed Z, Zahn FD, Gray GF, Thorbjarnarson B. Primary lymphoma of the gastrointestinal tract: a review of 100 cases. Ann Surg. 1969;170:232238. Freeman C, Berg JW, Cutler SJ. Occurrence and prognosis of extranodal lymphomas. Cancer. 1972;29:252260. Aozasa K, Tsujimoto M, Sakurai M, et al. Non-Hodgkins lymphomas in Osaka, Japan. Eur J Cancer Clin Oncol. 1985; 21:487 492. Otter R, Bieger R, Kouin PM, Hermans J, Willemze R. Primary gastrointestinal non-Hodgkins lymphoma in a population-based registry. Br J Cancer. 1989;60:745750. Naqvi MS, Burrows L, Kark AE. Lymphoma of the gastrointestinal tract: prognostic guides based on 162 cases. Ann Surg. 1969;170:221231. Lewin KJ, Ranchod M, Dorfman RF. Lymphomas of the gastrointestinal tract: a study of 117 cases presenting with gastrointestinal disease. Cancer. 1978;42:693707. Contreary K, Nance FC, Becker WF. Primary lymphoma of the gastrointestinal tract. Ann Surg. 1980;191:593598. Herrmann R, Panahon AM, Barcos MP, Walsh D, Stutzman L. Gastrointestinal involvement in non-Hodgkins lymphoma. Cancer. 1980;46:215222. Weingrad DN, Decosse JJ, Sherlock P, Straus D, Lieberman PH, Filippa DA. Primary gastrointestinal lymphoma: a 30year review. Cancer. 1982;49:1258 1265. Dragosics B, Bauer P, Radaszkiewicz T. Primary gastrointestinal non-Hodgkins lymphoma: a retrospective clinicopathologic study of 150 cases. Cancer. 1985;55:1060 1073. Aozasa K, Tsujimoto M, Inoue A, et al. Primary gastrointestinal lymphoma: a clinicopathologic study of 102 patients. Oncology. 1985;42:97103. Aozasa K, Ueda T, Kurata A, et al. Prognostic value of histologic and clinical factors in 56 patients with gastrointestinal lymphomas. Cancer. 1988;61:309 315. Azab MB, Henry-Amar M, Rougier P, et al. Prognostic factors in primary gastrointestinal non-Hodgkins lymphoma: a multivariate analysis, report of 106 cases, and review of the literature. Cancer. 1989;64:1208 1217. Bellesi G, Alterini R, Messori A, et al. Combined surgery and chemotherapy for the treatment of primary gastrointestinal intermediate- or high-grade non-Hodgkins lymphomas. Br J Cancer. 1989;60:244 248. Salles G, Herbrecht R, Tilly H, et al. Aggressive primary gastrointestinal lymphomas: review of 91 patients treated with the LNH-84 regimen. A study of the Groupe dEtude des Lymphomes Agressifs. Am J Med. 1991;90:77 84. Rackner VL, Thirlby RC, Ryan JA. Role of surgery in multimodality therapy for gastrointestinal lymphoma. Am J Surg. 1991;161:570 575. Radaszkiewicz T, Dragosics B, Bauer P. Gastrointestinal malignant lymphomas of the mucosa-associated lymphoid tissue: factors relevant to prognosis. Gastroenterology. 1992; 102:1628 1638. Ruskone-Fourmestraux A, Aegerter P, Delmer A, Brousse N, Galian A, Rambaud JC. Primary digestive tract lymphoma: a prospective multicentric study of 91 patients. Gastroenterology. 1993;105:16621671. Morton JE, Leyland MJ, Vaughan Hudson G, et al. Primary 21.

2. 3.

22.

4.

23.

5.

6.

24.

7. 8.

25.

9.

26.

10.

27.

11.

28.

12.

29.

13.

30.

14.

31.

15.

16.

32.

17.

33.

18.

34.

19.

Gastrointestinal Lymphoma in Japan/Nakamura et al.

35. Isaacson PG, Norton AJ. Extranodal lymphomas. Edinburgh: Churchill Livingstone; 1994. 36. Rohatiner A, dAmore F, Coifer B, et al. Report on a workshop convened to discuss the pathological and staging classications of gastrointestinal tract lymphoma. Ann Oncol. 1994;5:397 400. 37. Gurney KA, Cartwright RA, Gilman EA. Descriptive epidemiology of gastrointestinal non-Hodgkins lymphoma in a population-based registry. Br J Cancer. 1999;79:1929 1934. 38. Doglioni C, Wotherspoon AC, Moschini A, Boni M, Isaacson PG. High incidence of primary gastric lymphoma in northeastern Italy. Lancet. 1992;339:834 835. 39. Wright DH, Jones DB, Clark H, Mead GM, Hodges E, Howell WM. Is adult-onset coeliac disease due to a low-grade lymphoma of intraepithelial T lymphocytes? Lancet. 1991;337: 13731374. 40. Tarawneh MS. Non-Hodgkins lymphomas in Jordanians: a histopathologic study of 231 cases. Hematol Oncol. 1986;4: 9199.

2473

41. Salem P, Anaissie E, Allam C, et al. Non-Hodgkins lymphomas in the Middle East: a study of 417 patients with emphasis on special features. Cancer. 1986;58:11621166. 42. Kadin ME, Berard CW, Nanba K, Wakasa H. Lymphoproliferative diseases in Japan and Western countries: proceedings of the United States-Japan Seminar, September 6 7, 1982, Seattle, Washington. Hum Pathol. 1983;14:745 772. 43. Severson RK, Davis S. Increasing incidence of primary gastric lymphoma. Cancer. 1990;66:12831287. 44. Devesa SS, Fears T. Non-Hodgkins lymphoma time trends: United States and international data. Cancer Res. 1992;52 (suppl):5432s5440s. 45. Ferreri AJ, Cordio S, Paro S, et al. Therapeutic management of stage I-II high-grade primary gastric lymphomas. Oncology. 1999;56:274 282. 46. Fischbach W, Dragosics B, Kolve-Goebeler ME, et al. Primary gastric B-cell lymphoma: results of a prospective multicenter study. Gastroenterology. 2000;119:11911202.