| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Original Article |
1 Clínica de Neoplasias Gástricas, Gastroenterology Department, Instituto Nacional de Cancerología, San Fernando 22, México D.F. 14080, México
2 Medical Oncology Department, Instituto Nacional de Cancerología, Mexico City, México
3 Pathology Department, Instituto Nacional de Cancerología, Mexico City, México
4 Computed Tomography, Ultrasound and Magnetic Resonance Department, Instituto Nacional de Cancerología, Mexico City, México
5 Surgery Division, Instituto Nacional de Cancerología, Mexico City, México
6 Hematology Department, Instituto Nacional de Cancerología, Mexico City, México
7 Unidad de Investigación Biomédica en Cáncer, Instituto Nacional de Cancerología/Instituto de Biomédicas. Universidad Nacional Autónoma de México, Mexico City, México
Correspondence: Address correspondence and reprint requests to: Luis F. Oñate-Ocaña, M.D.; E-mail: lfonate{at}gmail.com
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIAL AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Methods: A cohort treated from January 1987 to December 2002 was studied. Relevant clinical, pathological and therapeutic variables were recorded. Kaplan–Meier and Cox’s methods were used to define prognostic factors associated with cancer-related survival.
Results: One thousand and twenty-three patients were included. Serum albumin did impact survival, showing a dose-response effect. This effect was present after adjustment for other prognostic factors, including Tumor-Node-Metastasis (TNM) stage, surgical resection and type of lymphadenectomy. In multivariate analysis, TNM stage [Stage Ia and Ib Hazard Ratio [HR] 1, Stage II HR 1.6 (95% confidence interval [CI], 0.56–4.7), Stage IIIa HR 4.4 (95% CI 1.7–11.3), Stage IIIb HR 5.6 (95% CI 2.6–17.2), Stage IV HR 6.8 (95% CI 2.7–17.5), high albumin HR 1, medium albumin HR 1.2 (95% CI 0.8–1.7), low albumin HR 1.2 (95% CI 0.8–1.8), very low albumin HR 1.8 (95% CI 1.3–2.6), D2 dissection HR 1, D1 dissection HR 1.9 (95% CI 1.3–2.97), and no resection HR 3.7 (95% CI 2.4–5.7)] were the most significant prognostic factors associated to survival (model P = 0.00001).
Conclusion: Pretherapeutic serum albumin level is a significant prognostic factor, which should be evaluated along with other well-defined prognostic factors in decisions concerning therapy for GC.
Key Words: Stomach neoplasm • Adenocarcinoma • Prognostic factor • Serum albumin
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIAL AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Complete resection is the only potentially curative treatment. However, long-term survival after gastrectomy for advanced GC remains poor: only 30% of the cases can be resected and only 2% achieve 5-year survival.4
Establishing prognosis is an essential part of oncology. Prognostic factor analysis was used to develop the Tumor-Node-Metastases staging system (TNM), which is a framework comprising these three fundamental prognostic factors allowing the definition of six risk groups.5 However, some patients with localized disease develop recurrences, and conversely, some patients with advanced disease respond to treatment and survive for a long time.
Identifying more variables, which characterize those patients likely to have either good or poor outcomes in these risk groups would help direct future research towards more refined treatments for target populations. In practice, prognostic variables are the tools for determining which patients are candidates for moderate or extreme treatment measures.
Thus, as we increase the number of significant prognostic variables, the prediction power of a prediction system for GC also increases.6,7
The idea that serum albumin may act as a prognostic factor is not new. Strong evidence suggests that an inverse relationship exists between the serum albumin level and mortality among the general population,8,9 and in cohorts with benign10–14 and malignant diseases.15–18 Accordingly, our group observed an association between the pretherapeutic serum albumin level and the survival of patients suffering from GC4,19 as well as with operative morbidity after gastrectomy for GC.20
The aim of the present study was to define the prognostic role of pretherapeutic serum albumin level among patients with GC.
|
MATERIAL AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIAL AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Endoscopic ultrasound was performed on patients with small tumors who presented morphology compatible with early gastric cancer. Staging laparoscopy was performed on patients with bulky tumors or where CT scan indicated the suspicion of loco-regional invasion.
The 2002 edition of the American Joint Committee on Cancer staging system was used.5 In the case of patients undergoing surgical resection, the histo-pathological pTNM staging was used. Among those patients with unresectable tumors, the CT classification was applied, as described by Moss21 and all clinical information was used to define the clinical TNM staging, including laparoscopy or laparotomy. Patients with insufficient or inadequate data for defining the cTNM stage (particularly abdominal CT scan) or with missing pretherapeutic serum albumin values were excluded from the study.
Surgical resection was performed using Japanese standards.22 However, from 1995 onwards, a modified D2 lymphadenectomy was performed. Splenectomy was indicated only in tumors located in the greater curvature of the proximal third of the stomach. Pancreatosplenectomy was performed only when GC presented direct pancreatic invasion.
Surgical morbidity was defined as any complication that occurred as the result of the gastrectomy during the 30-day period after the operation, even if it occurred after the hospital discharge. Surgical mortality was defined as any death caused by surgical morbidity.
Patients were followed-up after surgery by means of clinical history and clinical examination, blood cell counts, liver function tests and carcinoembryonic antigen, every 3–4-month-interval during the first 3 years after surgery. We performed chest X-rays and upper GI tract endoscopy each year. CT scan was performed only in the event of clinical or laboratory abnormalities. Recurrences were defined as those diagnosed by CT scan, endoscopic biopsy, cytology study or laparotomy findings. Biopsy was not required for defining a recurrence.
Variables
Serum albumin was always measured by means of the bromocresol purple method23 using different fully automated equipments throughout the 16-year period of this study. These reagents and equipments were convenience-validated and standardized in our central clinical laboratory.
Independent variables analyzed included: age (in years), sex (coded as female = 0, male = 1), hemoglobin (g/dl), lymphocyte count (cell/mm3), serum albumin (g/dl), location in the stomach (distal = 1, medium third = 2, proximal = 3), Moss CT classification (MI = 1, MII = 2, MIII = 3, MIV = 4), degree of differentiation (well = 1, moderately = 2, poor = 3), Bormann Classification (BI = 1, BII = 2, BIII = 3, BIV = 4), Laurén classification (intestinal = 0, diffuse/mixed = 1), TNM stage (Stage Ia–Ib = 1, Stage II = 2, Stage IIIa = 3, Stage IIIb = 4, Stage IV = 5), tumor size (cm), Gastrectomy (No resection = 0, subtotal = 1, total = 2), lymph node dissection (No resection = 0, D0/D1 = 1, D2 = 2), surgical intention (No resection = 0, palliative intention = 1, curative intention = 2), residual status after resection (R0 = 0, R1 = 1, R2 = 2, no resection = 3), operative bleeding (ml), adjuvant chemotherapy (no = 0, yes = 1) and adjuvant chemoradiation (no = 0, yes = 1).
Serum albumin was analyzed as a continuous variable and was also categorized dividing the cohort into quartiles (coded as 1, 2, 3 and 4, respectively).24
Statistical Analysis
One-way Anova was used to compare the mean serum albumin values between two or more groups of patients.
Survival times were calculated from the date of diagnosis of GC until the last visit recorded in the clinical charts. Only death, which was associated with GC, was considered as event (including surgery- or chemotherapy-associated mortality). All patients either alive or death due to other causes than GC were censored. Those patients who were lost to follow-up were located by phone.
Bivariate and stratified survival analyses were carried out using the method described by Kaplan and Meier25 and differences in survival were calculated using the Logrank method.26
The Cox proportional hazards model was used for multivariate analysis in order to define prognostic factors determining survival.27 All variables with a probability value of 0.2 or less in the bivariate analysis were used in the multivariate analysis. Interaction terms and proportionality of hazards were checked as previously described.27
Any probability value of 0.05 or less was considered as significant. Two-tailed statistics were taken into consideration in all cases. The program SPSS for Windows version 10 was used to make computations (SPSS, Inc. Chicago, IL 1999).
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIAL AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
The mean age of this group was 54.6 years (SD 14.4). Demographical and clinical data referring to the cohort in association with preoperative serum albumin levels is shown in Table 1
.
|
The TNM stage classification was clinical in 666 patients (65.1 %) who underwent palliative treatments (no surgical resection). Surgical pathology pTNM staging was recorded among 357 patients for whom surgical resection was possible (34.9%).
Major surgical morbidity after gastrectomy was recorded in 74 cases (20.7%) and surgical mortality was found in 24 cases (6.7%).
Ninety-three patients (35%) with surgical resections with curative intention presented recurrence and 173 did not (65%).
The use of neoadjuvant chemotherapy, adjuvant chemotherapy, adjuvant chemo-radiotherapy or palliative chemotherapy did not affect the prognosis.
The median survival of the cohort was 2.54 years (95% CI 1.7–3.4). The all-stage 2- and 5-year survival was 53 and 42.6%, respectively.
The mean survival times of cases with stages I (including Ia and Ib) and II were 16.5 years (95% CI 14.5–18.6) and 11.9 years (95% CI 10.2–13.5), respectively (median not reached). The median survival times of cases with stages IIIa, IIIb and IV were 3.47 years (95% CI 1.5–5.4), 2.07 years (95% CI 0.87–3.3) and 0.74 years (95% CI 0.58–0.89), respectively (P < 0.00001).
The median survival times of 666 cases who underwent non-surgical palliation and 177 who underwent resection with D1 lymphadenectomy were 0.65 years (95% CI 0.57–0.73) and 3.08 years (95% CI 1.06–5.1), respectively. The mean survival time for 180 patients who underwent D2 lymphadenectomy was 10.7 years (median not reached, 95% CI 9.45–12.01) (P < 0.00001).
The median survival time of 91 cases of palliative resections was 1.32 years (95% CI 0.49–2.14). The mean survival time for the 266 cases of curative resections was 10.2 years (median not reached, 95% CI 10.8–13.7) (P < 0.00001).
Categorized pretherapeutic serum albumin groups (medium, low and very low albumin) presented median survival times of 1.44 years (95% CI 0.34–2.54), 1.96 years (95% CI 0.36–3.56) and 2.62 years (95% CI 0.79–4.45), respectively. The group of high albumin presented a mean survival of 10.68 years (median not reached, 95% CI 8.93–12.42), respectively (P < 0.00001). Survival curves by albumin levels according to these four groups are depicted in Fig. 1.
|
and 3 and demonstrated that the association between albumin levels and survival persisted after adjustment for different strata of TNM stages and for the intention of surgical resection.
|
|
. Those estimators obtained for 357 patients, who underwent to surgical resection (curative or palliative) and for 266 patients which underwent to curative gastrectomy only are listed in Tables 3 and 4, respectively.
|
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIAL AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
This inverse relationship meets the current criteria for causality: a strong and temporal relationship between pretherapeutic serum albumin levels and survival, which shows a clear dose–response effect, and biological plausibility, ubiquity and reproducibility are indicated in the medical literature.
A critical gap for demonstrating causality, however, is the absence of clinical trials that show that raising albumin levels by means of intravenous infusion or by hyperalimentation decreases the excess risk of mortality in GC.
The potential of serum albumin as a prognostic factor in GC is emphasized because the technique to measure it is simple, cheap and widely available. Moreover, serum albumin measurement is part of liver function tests, and for most patients it is measured even before cancer is diagnosed and before the staging protocol or the design of a treatment.
Recognition of albumin as a risk or prognostic factor is not new. Previous studies have indicated an important association between low serum albumin and excess mortality in healthy populations of both western8 and eastern countries.9
This association has been reported also in patients with liver cirrhosis and is the basis for Child’s classification, which has deep therapeutic implications.10 Serum albumin is also strongly associated with survival in patients with hemodyalisis for end-stage renal disease,11 it has been reported among patients with community-acquired pneumonia12 and in patients with idiopathic pulmonary arterial hypertension13 among other non-neoplastic diseases.
Serum albumin has been found to be an independent predictor of survival in the critical care setting. The third version of the Acute Physiology and Chronic Health Evaluation (APACHE III) is a multivariate logistic model for quantifying the risk of mortality after admittance to the intensive care unit and includes a broad range of physiologic alterations and co-morbidity conditions. The APACHE III model includes serum albumin as a covariate for the first time and this new model is more predictive than the second version.14
In the field of oncology, the prognostic role of serum albumin has been described for a long list of malignant diseases. Serum albumin is used in the prognostic model for patients with hepatocellular carcinoma,15 in a prognostic score for patients with colorectal cancer liver metastases,16 in the International Staging System for Multiple Myeloma,17 in survival prediction for women with breast cancer18 and in many other tumors.
In the specific case of GC, an association between serum albumin and survival has been reported by our group among a retrospective cohort of patients with all TNM-stages4 and we also reported this association after gastrectomy for GC.19 On the other hand, serum albumin is a prognostic factor strongly associated with operative morbidity after gastrectomy for GC.20
Recently, an association between serum albumin with operative mortality, resectability and survival among patients with adenocarcinoma localized in the gastric cardia has been described.28 However, the authors did not present results regarding other locations in the stomach, nor did they provide enough data to support a clear dose-response effect.
A potential pitfall of our methodology is that there was a great heterogeneity in the automated equipments used for serum albumin measurement during the period of this study. A total of 37 different automated equipments have been used between 1987 and 2002 in our institution. All these equipments measure serum albumin based on the bromocresol purple method, which has been described as the more efficient and precise.23 However, all have different reference standards and the punctual serum albumin values showed a systematic variation associated with the automated equipment and reagents used.
This fact does not invalidate our results because theoretically the effect of this random variation would be to decrease the strength of the risk ratio for the serum albumin-survival association.
In any case, we made a statistical adjustment to our results based on the grouping of these reference standards, which revealed no effect on survival (data not shown). Moreover, the selection of these reference standards was random and they were not associated with survival or with the TNM stage.
The multivariate analysis presented in this work revealed strong evidence supporting the independent value of serum albumin as a prognostic factor, particularly in the case of the analysis performed on the subgroup of patients who underwent surgical resection of the stomach. We believe that the risk ratios for serum albumin are particularly well defined in this surgical subgroup with precise TNM staging and long-term follow-up.
Currently, prognosis of patients with GC is estimated based on the TNM staging system of the American Joint Committee on Cancer and it is useful for the planning of a tailored treatment for each patient.5
The TNM staging system is a construct made up of the three major prognostic factors in GC: the invasion through the stomach wall, the number of lymph node with metastases and the presence of distant metastases. However, none of many well recognized prognostic factors (like patient-related, treatment-related, morphology or molecular factors) are included in the system.29
Integrating these prognostic factors into a well-designed model would help to predict survival with greater accuracy than the current TNM staging system and would probably permit better treatment design.
An effort to integrate these variables in a nomogram has yielded a valid model, which is a more accurate predictor of survival than the TNM staging system.6
Accurate prediction can help in the selection of an apt surgical approach or the best adjuvant treatment as well as for individual patient counseling and for the follow-up schedule.6,7 Moreover, it may also play a role in design of future trials and for identifying subsets of patients who fall within known AJCC stages, who have different prognoses, and who may have different responses to novel chemotherapy drugs or combinations.
The association between albumin and cancer-related survival may be caused by confounding factors. Serum albumin level is related to the balance between albumin synthesis and catabolism.30 Many biological mechanisms could be responsible for this association in patients with GC.
Even after adjusting for potential confounders such as comorbidity, there can still be residual confounding. Disease, by inducing anorexia and malnutrition, can result in reduction in serum albumin.31 This reduction may vary with cancer progression, from preclinical to terminal stage.
Elevation of cytokines represents another potential mechanism through which disease severity could confound the relationship between serum albumin and GC-related survival.
C-reactive protein, tumor necrosis factor and interleukin-1 are involved in the pathogenesis of many diseases, including cancer, and can decrease serum albumin concentrations.30–32 Moreover, tumor necrosis factor and interleukin-2 have been found to be elevated in patients with GC and are associated to poor prognosis.33
Serum interleukin-6 is increased in acute or chronic inflammation situations, it is associated to hypoal-buminemia34 and it is associated to the development of lymph node and liver metastases in GC.35
A potentially important protective role of serum albumin is enhancement of removal of reactive oxygen species, which are implicated in the pathogenesis of many diseases, including cancer.36,37
Regardless of the mechanism of hipoalbuminemia, one fact is clear, the less the albumin value, the worst the prognosis.
Based on our data the pretherapeutic serum albumin level is a significant prognostic factor in GC among Mexican patients. This association meets the current criteria for causality and should be evaluated in a prospective setting in different institutions to demonstrate its ubiquity before it can be integrated to the TNM staging system.
Accordingly, serum albumin level could be used in clinical trials to better define the baseline risk in patients with GC. Visceral protein depletion, inflammatory response markers, nutritional therapy and anabolic strategies status should be studied together with GC-specific therapy.
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
Received for publication May 22, 2006. Accepted for publication May 23, 2006.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIAL AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
 |
|
 |
|
  C. Kunisaki, H. Makino, R. Takagawa, T. Oshima, Y. Nagano, T. Kosaka, H. A. Ono, Y. Otsuka, H. Akiyama, Y. Ichikawa, et al. Tumor Diameter as a Prognostic Factor in Patients with Gastric Cancer Ann. Surg. Oncol., July 1, 2008; 15(7): 1959 - 1967. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
