ORIGINAL ARTICLE

Incidence and Long-term Outcome of Young Patients With Gastric Carcinoma According to Sex
Does Hormonal Status Affect Prognosis?
Jong-Han Kim, MD, PhD; Yoon-Jung Boo, MD, PhD; Joong-Min Park, MD; Sung-Soo Park, MD, PhD; Seung-Joo Kim, MD, PhD; Chong-Suk Kim, MD, PhD; Young-Jae Mok, MD, PhD

Background: We investigated the clinicopathological

features and evaluated the prognostic impact of age and sex on patients with gastric cancer.
Design: Retrospective study from 1993 to 2000. Setting: Korea University Medical Center. Patients: A total of 1299 patients with gastric cancer

were divided into young (n = 175 [13.5%]) and older (n=1124 [86.5%]) groups with an age cutoff of 40 years.
Main Outcome Measures: Clinicopathological characteristics were investigated and survival analysis was performed according to sex for each age group. Results: Tumor differentiation was significantly differ-

were significantly different between the young and older groups. In male patients, the 10-year survival rate of the young group was higher (62.5%) than that of the older group (44.6%) (P = .03). Although it was not statistically significant, the survival rate of the older female group was higher than that of the young group (56.2% vs 51.9%). On multivariate analysis, tumor stage (P .001) and sex (P =.042) were proved to be independent prognostic factors.
Conclusions: Only tumor differentiation was an important difference between the 2 age groups, and prognosis was not affected by age. However, when sex was added to age as a factor, the older male and young female groups had an unfavorable prognosis. Therefore, we propose that sex hormones such as estrogens contribute to the survival differences, and further studies are needed to confirm this possibility.

ent between the 2 age groups. Among male patients, the young group had more undifferentiated tumors than the older group (P .001) but, in female patients, both differentiation (P .001) and operative methods (P =.008)

Arch Surg. 2008;143(11):1062-1067 Recently, the prognostic effect of sex hormones in the development of gastric carcinoma has been studied,9 and the presence of estrogen receptors has been shown to be associated with poor survival in young patients with gastric carcinoma.10,11 In other reports, the presence of androgen receptors was found to be an independent unfavorable prognostic factor in gastric cancer.12 Therefore, the relationship between sex hormones and the prognosis of gastric cancer remains unclear. Although there is a difference according to sex, the status of sex hormones is dependent on age. Thus, both age and sex should simultaneously be considered to precisely evaluate the impact of sex hormones on the prognosis of gastric cancer. Therefore, in this study, we retrospectively analyzed the clinical and pathological

usually a disease of the aged, with a mean age of approximately 50 to 60 years1; patients younger than 40 years represent 2% to 8% of all patients with gastric carcinoma.2 The question of whether gastric carcinoma in young patients is different from that in older patient has been raised but remains unresolved. Some authors have suggested that young patients have a worse prognosis, and

ASTRIC CARCINOMA IS

See Invited Critique at end of article

Author Affiliations: Department of Surgery, Korea University College of Medicine, Seoul.

this is thought to be due to delayed diagnosis and more aggressive tumor behavior. Others have shown that age is not an independent factor for survival.3-8

(REPRINTED) ARCH SURG/ VOL 143 (NO. 11), NOV 2008 1062

WWW.ARCHSURG.COM

2008 American Medical Association. All rights reserved. Downloaded From: http://archsurg.jamanetwork.com/ on 04/03/2013

factors for young patients with gastric cancer and compared the findings with those of their counterparts according to sex.
METHODS From January 1, 1993, to December 31, 2000, a total of 1299 patients with gastric carcinoma underwent surgery at the Department of Surgery, Korea University Medical Center. Of these patients, 175 were 40 years or younger, and they were classified as the young group; the remaining 1124 patients were classified as the older group. We then compared the clinical and pathological features between the young group and the older group according to sex. The features compared were operative method (gastric resection, gastrojejunal bypass, and no surgical procedure [opening and closure]), tumor location, histologic differentiation, curability, depth of invasion, lymph node metastasis, and TNM stage. A survival analysis was performed, and the 10-year survival rate was compared according to age and sex. Finally, we performed multivariate analysis to determine the independent prognostic factors by Cox proportional hazards regression model. For the statistical analysis, SPSS 13.0 for Windows (SPSS Inc, Chicago, Illinois) was used. The clinical variables were analyzed with the 2 test and the independent paired t test. The survival rates were calculated by the Kaplan-Meier method, and outcomes of treatment were compared with the log-rank test. P.05 was considered statistically significant. RESULTS

SURVIVAL ANALYSIS When the 10-year survival rate was compared between the 2 age groups in all patients, the young group showed a better overall survival than the older group; however, this difference was not statistically significant (P =.10). When the 10-year survival rate was compared between sexes, female patients showed a better survival rate than males, but this difference was not statistically significant (P =.08). When all patients were divided into 4 groups according to age and sex and the 10-year survival rates were compared, for male patients the overall 10-year survival rate was best for the young male group (62.5%) and worst for the older males (44.6%), and this difference was statistically significant (P = .03) (Figure 1). In contrast, however, the survival rate for female patients in the older group was higher than that of the young group, but this difference was not statistically significant (56.2% vs 51.9%; P =.33) (Figure 2). When the survival rates of patients who underwent curative resection were compared, there was a significant difference between the young male and older male patients in favor of the young group (P =.03) (Figure 3). However, the survival result was reversed between the older and young female patients, although the difference was not significant (P =.30) (Figure 4). On univariate survival analysis, the statistically significant prognostic factors were sex, age, tumor differentiation, and cancer stage. By Cox proportional hazards model, cancer stage and sex were proved to be statistically significant (Table 2).
COMMENT

DEMOGRAPHIC CHARACTERISTICS Of 1299 patients, 175 (13.5%) (100 male [7.7%] and 75 female [5.8%]) were classified in the young group, and the male to female ratio was 1.33. The remaining 1124 patients (86.5%) (765 male [58.9%] and 359 female [27.6%]) were classified as the older group, and the male to female ratio was 2.13. In total, 865 (66.6%) were male and 434 (33.4%) were female, and the sex ratio was 1.99. CLINICAL AND PATHOLOGICAL FINDINGS IN THE YOUNG GROUP For both sexes, the young group had more undifferentiated tumors (63.0% in males and 93.3% in females) than the older group (40.1% in males and 58.2% in females), and this difference was statistically significant (P .001) (Table 1). For male patients, the histologic differentiation was the only statistically significant factor (P .001). For females, both histologic differentiation and operative methods were significantly different between the young and the older groups (P .001 and P =.008, respectively). For female patients, no surgical procedure (opening and closure) was performed in 8 patients (10.7%) in the young group and 11 patients (3.1%) in the older group because of multiple seeding of the peritoneum. Other variables, such as tumor location, curability, depth of invasion, lymph node metastasis, and overall TNM stage, were not statistically significantly different between the 2 groups for either sex (Table 1).

Surgery is the only potentially curative modality for patients with localized gastric tumors, and it has been shown that curative resection offers the only chance of longterm survival in patients with this disease.1,13 In this study, we performed curative resection in 82% of the patients in the young group and in 79% of the patients in the older group; however, this difference was not statistically significant. A comparative study in Japan demonstrated different features of gastric cancer between patients younger than 40 years and patients older than 70 years.13 In our study, we chose 40 years of age as the cutoff between the young and older groups. Generally, the average perimenopausal age range is 45 to 55 years. In our analyses, the average age of the young female group was about 34 years and that of the older group was 59 years. This suggests the possibility of sex hormone differences between the 2 age groups in our study. Also, we performed additional survival analyses by moving the cutoff age by 10 years, but there were no survival differences or other relevant findings, so we kept the cutoff value at 40 years in our study. Earlier reports showed a higher incidence of gastric cancer in males and in patients with advanced age but a higher proportion of females among younger patients than older patients. Our study supports this finding of a relatively higher proportion of female patients in the young group.4,7,14-16 In our series, the male to female ratio for
WWW.ARCHSURG.COM

(REPRINTED) ARCH SURG/ VOL 143 (NO. 11), NOV 2008 1063

2008 American Medical Association. All rights reserved. Downloaded From: http://archsurg.jamanetwork.com/ on 04/03/2013

Table 1. Comparison of Clinicopathological Factors Between Young and Older Patients With Gastric Carcinoma
No. (%) Variable No. of patients Age, mean (SD), y Operative method Gastric resection Bypass O&C Tumor location Upper third Middle third Lower third Entire Histologic differentiation Well differentiated Undifferentiated Curability R0 resection R1, R2 resection Depth of invasion, T stage T1 T2 T3 T4 Lymph node metastasis Negative Positive Overall UICC stage I II III IV No. of patients Age, mean (SD), y Operative method Gastric resection Bypass O&C Tumor location Upper third Middle third Lower third Entire Histologic differentiation Well differentiated Undifferentiated Curability R0 resection R1, R2 resection Depth of invasion, T stage T1 T2 T3 T4 Lymph node metastasis Negative Positive Overall UICC stage I II III IV Young Group Male Patients (n = 865) 100 (11.6) 35.07 (3.87) 91 (91.0) 2 (2.0) 7 (7.0) 11 (11.0) 33 (33.0) 54 (54.0) 2 (2.0) 37 (37.0) 63 (63.0) 83 (83.0) 17 (17.0) 39 (39.0) 15 (15.0) 38 (38.0) 8 (8.0) 55 (55.0) 45 (45.0) 46 (46.0) 10 (10.0) 25 (25.0) 19 (19.0) Female Patients (n = 434) 75 (17.3) 33.92 (4.77) 66 (88.0) 1 (1.3) 8 (10.7) 8 (10.7) 34 (45.3) 29 (38.7) 4 (5.3) 5 (6.7) 70 (93.3) 61 (81.3) 14 (18.7) 18 (24.0) 16 (21.3) 40 (53.3) 1 (1.3) 34 (45.3) 41 (54.7) 33 (44.0) 10 (13.3) 24 (32.0) 8 (10.7) Older Group 765 (88.4) 58.86 (8.92) 679 (88.8) 55 (7.2) 31 (4.1) .84 89 (11.6) 232 (30.3) 435 (56.9) 9 (1.2) .001 a 458 (59.9) 307 (40.1) .26 597 (78.0) 168 (22.0) 225 (29.4) 124 (16.2) 341 (44.6) 75 (9.8) .40 383 (50.1) 382 (49.9) .23 275 (35.9) 94 (12.3) 205 (26.8) 191 (25.0) 359 (82.7) 59.00 (9.70) 334 (93.0) 14 (3.9) 11 (3.1) .08 31 (8.6) 132 (36.8) 189 (52.6) 7 (1.9) .001 a 150 (41.8) 209 (58.2) .96 291 (81.1) 68 (18.9) 110 (30.6) 67 (18.7) 167 (46.5) 15 (4.2) .59 175 (48.7) 184 (51.3) .88 164 (45.7) 51 (14.2) 99 (27.6) 45 (12.5) P Value

.001 a .07

.001 a .008 a

Abbreviations: O & C, open and closure; UICC, International Union Against Cancer. a P.05, statistically significant.

(REPRINTED) ARCH SURG/ VOL 143 (NO. 11), NOV 2008 1064

WWW.ARCHSURG.COM

2008 American Medical Association. All rights reserved. Downloaded From: http://archsurg.jamanetwork.com/ on 04/03/2013

1.0

1.0

0.8 Young male group

0.8

Young male curative resection group

Overall Survival

0.6

Overall Survival

0.6

0.4

Older male group

0.4

Older male curative resection group

0.2 P = .03 0.0 0 2 4 6 8 10 12

0.2 P = .03 0.0 0 2 4 6 8 10 12

Time, y

Time, y

Figure 1. Survival in young and older groups of male patients with gastric cancer.

Figure 3. Survival in male patients with gastric cancer who underwent curative resection.

1.0

1.0

Older female curative resection group 0.8 0.8

Overall Survival

Older female group

0.6 Young female curative resection group 0.4

Overall Survival

0.6

Young female group 0.4

0.2 P = .30 0.2 0.0 P = .33 0.0 0 2 4 6 8 10 12 0 2 4 6 8 10 12

Time, y

Time, y

Figure 4. Survival in female patients with gastric cancer who underwent curative resection.

Figure 2. Survival in young and older groups of female patients with gastric cancer.

all patients 40 years or younger was 1.33, and it was 2.13 for the remaining patients. The reason for this high number of female patients in the young group is not known. Nevertheless, male patients might have more frequent and prolonged exposure to environmental carcinogens than females, which might explain the male preponderance among the older patients.17 In the present study, comparison of histologic differentiation showed that the young group had more poorly differentiated tumor types in both sexes, this finding being in agreement with studies by several other groups.2,4,15,18 Nakamura et al15 analyzed the pathological findings and prognosis of gastric carcinoma in young and elderly patients and found that 45.5% of early gastric carcinomas in elderly patients were well-differentiated adenocarcinomas, whereas a greater incidence of poorly differentiated adenocarcinoma was identified in the young patients. In agreement with other studies, young age at onset of gas-

Table 2. Cox Regression Analysis of Prognostic Factors in Patients With Curative Surgery
Prognostic Variables Age 40 y vs 40 y Sex, male vs female Cancer stage I II III IV RR (95% CI) 0.85 (0.60-1.21) 0.77 (0.60-0.99) 1 [Reference] 0.03 (0.02-0.05) 0.19 (0.13-0.29) 0.38 (0.27-0.53) P Value .37 .04 a .001 a

Abbreviations: CI, confidence interval; RR, relative risk. a P.05, statistically significant.

tric cancer appears to be associated with the diffuse type by the Lauren classification.4,19 In our study, we could not comment on the Lauren classification; nevertheless, our findings suggest that the diffuse type would be more prominent in young patients with gastric cancer.
WWW.ARCHSURG.COM

(REPRINTED) ARCH SURG/ VOL 143 (NO. 11), NOV 2008 1065

2008 American Medical Association. All rights reserved. Downloaded From: http://archsurg.jamanetwork.com/ on 04/03/2013

Earlier reports showed that the prognosis of patients with early onset of gastric carcinoma was poor, with a short survival potential, especially in patients who presented with advanced gastric carcinoma.15,20,21 Delay in diagnosis and more aggressive biological behavior of gastric carcinoma in young patients have been suggested as possible causes of the poor prognosis.6-8 In a few reports, however, the prognosis of patients with early-onset gastric cancer who underwent gastrectomy was better than that of older patients.22,23 In our cases, the 10-year survival rates were 58.1% in the young group and 48.1% in the older group. When curative resection was performed, the 10-year survival rates were 64.8% and 53.2% in the young and older groups, respectively. Therefore, the survival for the young patients with gastric cancer in our study was superior to that of the older group; however, these differences were not statistically significant. Although we did not match the survival outcome with the stage of disease, the findings indicate better outcome with surgical resection in younger patients with gastric cancer. By contrast, recent reports show no difference in surgical outcomes between older and younger patients with gastric cancer2-5,21; therefore, age does not appear to be an independent risk factor. However, when the survival was analyzed between the age groups by sex, we found that the young group had a better survival outcome for males and the older group had a better survival outcome for females. Yamachika et al24 reported that expression of the intestinal trefoil factor was correlated with a poor prognosis in patients with gastric cancer. Intestinal trefoil factor expression appears to be induced by estrogens in breast cancer cells.25 Although women are more likely to express intestinal trefoil factor, its expression in men has also been correlated with more aggressive tumor phenotype and significantly poorer survival. Furthermore, some studies have reported pregnancy as a poor prognostic factor for patients with gastric cancer.26,27 Hormonal differences, therefore, appeared to be an important factor for prognosis in patients with gastric cancer. Harrison et al28 proposed that estradiol is trophic to gastric cancer cell lines in in vitro experiments and suggested that the estradiol receptor D5 antibody could be an independent negative prognostic factor for patients with gastric cancer.11 Matsui et al10 immunohistologically investigated sex hormone receptors in 107 patients with gastric cancer who underwent curative resection and found that the estrogen receptorpositive rate was slightly higher in young females and in patients with poorly differentiated gastric cancer. Others proposed a suppressive effect of sex hormones on the spreading of stomach cancer; Furukawa et al26 suggested that female sex hormones or their analogues may play a suppressive role in carcinogenesis or progression of stomach cancer, and that pregnancy and delivery may accelerate growth of stomach cancer cells. In a 2004 report, Lindblad et al29 proposed a reduced risk of gastric cancer in a male cohort exposed to estrogens, thus supporting the hypothesis that estrogens may play a preventive role in gastric cancer. The foregoing studies strongly suggest that sex hormones, especially estrogens, affect the invasiveness and progression of gastric cancer, and that our results are consistent with these findings. In multivariate analyses, the independent prognostic survival factor was

sex. Also, the survival shown in this study was not diseaseadjusted but overall survival. Therefore, we hypothesized that the survival difference in male patients is due to the impact of age, but in female patients it may be due to a hormonal effect. In Korea, gastric cancer is a leading cause of cancerrelated death, and several studies have shown epidemiologic factors associated with the prognosis of gastric cancer, especially in regard to age. Our surgical outcomes for gastric cancer may reflect those of other cancer treatment institutions in Korea because of the large number of cases (1299) and the long-term follow-up period of approximately 10 years.
CONCLUSIONS

In this study, we performed comparative and survival analyses of the prognostic effect of age and sex, and the present findings strongly suggest that a single age factor does not appear to be meaningful for epidemiologic evaluation of gastric carcinomas in regard to prognosis. When the clinical and pathological factors were compared between young and older patients with gastric cancer, tumor differentiation was the only significant factor. The prognosis for patients with gastric cancer was the best for the young male group and the poorest in the older male group. For female patients with gastric cancer, the prognosis for the older group was better than that for the young group. There may be important hormonal influences on the prognosis of gastric cancer, and further investigation is needed for a deeper understanding of hormonal influences on gastric cancer. Accepted for Publication: May 4, 2007. Correspondence: Sung-Soo Park, MD, PhD, Department of Surgery, Korea University College of Medicine, Korea University Anam Hospital, Anam-dong 5-ga Seongbuk-gu, Seoul, 136-705, Korea (kugspss@korea .ac.kr). Author Contributions: Study concept and design: J.-H. Kim, S.-S. Park, C.-S. Kim, and Mok. Acquisition of data: J.-H. Kim, Boo, J.-M. Park, S.-S. Park, and S.-J. Kim. Analysis and interpretation of data: J.-H. Kim and S.-S. Park. Drafting of the manuscript: J.-H. Kim, Boo, J.-M. Park, S.-S. Park, and Mok. Critical revision of the manuscript for important intellectual content: J.-H. Kim, S.-S. Park, S.-J. Kim, and C.-S. Kim. Statistical analysis: J.-H. Kim and S.-S. Park. Obtained funding: J.-H. Kim and S.-S. Park. Administrative, technical, and material support: J.-H. Kim, S.-S. Park, and Mok. Study supervision: J.-H. Kim and S.-S. Park. Financial Disclosure: None reported.
REFERENCES
1. Mitsudomi T, Matsusaka T, Wakasugi K, et al. A clinicopathological study of gastric cancer with special reference to age of the patients: an analysis of 1,630 cases. World J Surg. 1989;13(2):225-230. 2. Kim DY, Ryu SY, Kim YJ, Kim SK. Clinicopathological characteristics of gastric carcinoma in young patients. Langenbecks Arch Surg. 2003;388(4):245-249. 3. Kim DY, Joo JK, Ryu SY, Park YK, Kim YJ, Kim SK. Clinicopathologic characteristics of gastric carcinoma in elderly patients: a comparison with young patients. World J Gastroenterol. 2005;11(1):22-26.

(REPRINTED) ARCH SURG/ VOL 143 (NO. 11), NOV 2008 1066

WWW.ARCHSURG.COM

2008 American Medical Association. All rights reserved. Downloaded From: http://archsurg.jamanetwork.com/ on 04/03/2013

4. Bani-Hani KE. Clinicopathological comparison between young and old age patients with gastric adenocarcinoma. Int J Gastrointest Cancer. 2005;35(1): 43-52. 5. Kokkola A, Sipponen P. Gastric carcinoma in young adults. Hepatogastroenterology. 2001;48(42):1552-1555. 6. Maconi G, Kurihara H, Panizzo V, et al; Italian Research Group for Gastric Cancer. Gastric cancer in young patients with no alarm symptoms: focus on delay in diagnosis, stage of neoplasm and survival. Scand J Gastroenterol. 2003;38 (12):1249-1255. 7. Simsa J, Leffler J, Hoch J, Linke Z, Pa dr R. Gastric cancer in young patientsis there any hope for them? Acta Chir Belg. 2004;104(6):673-676. 8. Llanos O, Butte JM, Crovari F, Duarte I, Guzma n S. Survival of young patients after gastrectomy for gastric cancer. World J Surg. 2006;30(1):17-20. 9. Arai T, Esaki Y, Inoshita N, et al. Pathologic characteristics of gastric cancer in the elderly: a retrospective study of 994 surgical patients. Gastric Cancer. 2004; 7(3):154-159. 10. Matsui M, Kojima O, Kawakami S, Uehara Y, Takahashi T. The prognosis of patients with gastric cancer possessing sex hormone receptors. Surg Today. 1992; 22(5):421-425. 11. Harrison JD, Jones JA, Ellis IO, Morris DL. Oestrogen receptor D5 antibody is an independent negative prognostic factor in gastric cancer. Br J Surg. 1991; 78(3):334-336. 12. Kominea A, Konstantinopoulos PA, Kapranos N, et al. Androgen receptor (AR) expression is an independent unfavorable prognostic factor in gastric cancer. J Cancer Res Clin Oncol. 2004;130(5):253-258. 13. Maehara Y, Emi Y, Tomisaki S, et al. Age-related characteristics of gastric carcinoma in young and elderly patients. Cancer. 1996;77(9):1774-1780. 14. Ramos-De la Medina A, Salgado-Nesme N, Torres-Villalobos G, Medina-Franco H. Clinicopathologic characteristics of gastric cancer in a young patient population. J Gastrointest Surg. 2004;8(3):240-244. 15. Nakamura T, Yao T, Niho Y, Tsuneyoshi M. A clinicopathological study in young patients with gastric carcinoma. J Surg Oncol. 1999;71(4):214-219. 16. Lim S, Lee HS, Kim HS, Kim YI, Kim WH. Alteration of E-cadherinmediated adhesion protein is common, but microsatellite instability is uncommon in young age gastric cancers. Histopathology. 2003;42(2):128-136.

17. Ershler WB, Longo DL. The biology of aging: the current research agenda. Cancer. 1997;80(7):1284-1293. 18. Choi JH, Chung HC, Yoo NC, et al. Gastric cancer in young patients who underwent curative resection: comparative study with older patients. Am J Clin Oncol. 1996;19(1):45-48. 19. Theuer CP, de Virgilio C, Keese G, et al. Gastric adenocarcinoma in patients 40 years of age or younger. Am J Surg. 1996;172(5):473-476. 20. Lo SS, Kuo HS, Wu CW, et al. Poorer prognosis in young patients with gastric cancer? Hepatogastroenterology. 1999;46(28):2690-2693. 21. Medina-Franco H, Heslin MJ, Cortes-Gonzalez R. Clinicopathological characteristics of gastric carcinoma in young and elderly patients: a comparative study. Ann Surg Oncol. 2000;7(7):515-519. 22. Wang JY, Hsieh JS, Huang CJ, Huang YS, Huang TJ. Clinicopathologic study of advanced gastric cancer without serosal invasion in young and old patients. J Surg Oncol. 1996;63(1):36-40. 23. Yokota T, Takahashi N, Teshima S, et al. Early gastric cancer in the young: clinicopathological study. Aust N Z J Surg. 1999;69(6):443-446. 24. Yamachika T, Werther JL, Bodian C, et al. Intestinal trefoil factor: a marker of poor prognosis in gastric carcinoma. Clin Cancer Res. 2002;8(5):1092-1099. 25. May FE, Westley BR. Expression of human intestinal trefoil factor in malignant cells and its regulation by oestrogen in breast cancer cells. J Pathol. 1997; 182(4):404-413. 26. Furukawa H, Iwanaga T, Hiratsuka M, et al. Gastric cancer in young adults: growth accelerating effect of pregnancy and delivery. J Surg Oncol. 1994; 55(1):3-6. 27. Maeta M, Yamashiro H, Oka A, Tsujitani S, Ikeguchi M, Kaibara N. Gastric cancer in the young, with special reference to 14 pregnancy-associated cases: analysis based on 2,325 consecutive cases of gastric cancer. J Surg Oncol. 1995; 58(3):191-195. 28. Harrison JD, Watson S, Morris DL. The effect of sex hormones and tamoxifen on the growth of human gastric and colorectal cancer cell lines. Cancer. 1989; 63(11):2148-2151. 29. Lindblad M, Ye W, Rubio C, Lagergren J. Estrogen and risk of gastric cancer: a protective effect in a nationwide cohort study of patients with prostate cancer in Sweden. Cancer Epidemiol Biomarkers Prev. 2004;13(12):2203-2207.

INVITED CRITIQUE

etrospective analyses of single-center patient data sets serve really only 1 purpose: to generate hypotheses for prospective studies. By convention, the hypotheses are generated from the statistically significant differences found in the retrospective analysis. In this analysis of nearly 1300 patients with gastric cancer treated in 1 university setting in Korea, the authors have taken a nonsignificant finding (that overall survival of young women is lower than that of older women, with a P value of .33) and generated a hypothesis that sex hormones such as estrogens contribute to the survival differences. While it is possible that there is a real difference in survival but it was not significant because of small sample size (type II error), there are also many other potential confounders, not the least of which are the lack of objective data concerning the hormonal status of the women and the lack of any numeric difference when the age break point was moved 10 years up. The analysis was based on overall survival, not diseasespecific survival. If based on the latter, would the significant survival advantage of young men disappear (ie, are the older men dying of other causes)? The adminis-

tration of adjuvant chemotherapy was not part of the analysis. Perhaps there are some biases within the institution beyond pathological stage that influence the administration of such depending on age and functional status of the patients or the patients fertility concerns. The perceived interaction of age and sex, if used to generate hypotheses, should generate more than one. Two alternate hypotheses could include the balance of sex hormones that influences survival (less testosterone in the older male patients), or interaction of age and functional status that drives adjuvant therapy decision making. Given the data presented, there is a need for a larger, multicenter analysis before generating the hypotheses that would be worthy of prospective evaluation. Leigh Neumayer, MD, MS Correspondence: Dr Neumayer, Department of Surgery, University of Utah School of Medicine, Integrated Breast Program, Huntsman Cancer Institute, 1950 Circle of Hope, Room 6540, Salt Lake City, UT 84112 (leigh .neumayer@hsc.utah.edu).

(REPRINTED) ARCH SURG/ VOL 143 (NO. 11), NOV 2008 1067

WWW.ARCHSURG.COM

2008 American Medical Association. All rights reserved. Downloaded From: http://archsurg.jamanetwork.com/ on 04/03/2013