| Colorectal cancer (CRC) has a lifetime risk of 6%, and accounts for about 15% of all cancer cases in the developed countries. The incidence of CRC in the developing Asian countries has been increasing. Detection of CRC is thus of importance for the optimal outcome. Colonoscopy is a good method of choice for identifying patients with CRC. However, not all patients are willing to accept colonoscopy due to the discomforts and possible contraindications associated with the procedure. In addition, if all patients with low gastrointestinal(LGI) symptoms prefer colonoscopy, endoscopy workload and medical cost are greatly increased. Thus it may be necessary to identify patients having a high risk of CRC for preferred colonoscopy. Fecal occult blood test (FOBT) and doublecontrast barium enema have been used as screening tools for CRC, but are of limited values due to low sensitivity. Gastrointestinal tumor markers, mainly including Carcinoembryoni cantigen (CEA), CA19-9 and CA72-4, are commonly used in clinic. These tumor markers are elevated in some CRC patients, but as mass screening for CRC, they all are limited because of low sensitivities4-6. In clinical practice, two methods including selective screening and combining diagnostic test results may be used to improve diagnostic test accuracy. Now, there have been few studies to explore these two methods to improve diagnostic values of gastrointestinal tumor markers forC RC. Thus, we evaluated the values of CEA in combination with CA19-9 and CA72-4 in subjects with LGI symptoms by a subgroup analysis based on the presenting symptoms and patients’ age.Study subjectsThe current study included all in-patients seeking medical help for lower GI tract symptoms at the Departments of Gastroenterology, Nanfang Hospital and the First Affiliated Hospital of Guangzhou Medical College, during the period from February 2008 to February 2012. Serum levels of tumor marker CEA, CA19-9 and CA72-4 were measured in all included subjects. Subjects with one or more of the following clinical features were excluded:1) diagnosis of hepatocarcinoma or pancreatic carcinoma at the time of admission,2) no definitive diagnosis or non-gastrointestinal diseases,3) diagnosis of CRC was made, but the subjects underwent chemotherapy or surgery. The diagnosis of CRC was made based on histological examination of biopsy specimens in all cases.Data Collection and definitionsThe clinical data included age, gender, and symptoms, including hematochezia, abdominal pain, abdominal distention, diarrhea, constipation, tenesmus, and weight loss. Hematochezia is defined as the passage of fresh blood through the rectum. Weight loss is defined as a reduction of total body mass and treated as a dichotomized variable. Constipation is defined as difficulty passing stools. In our clinical practice, the cut-off values for serum tumor markers were:>37 U/mL for CA19-9,>6.9U/mL for CA72-4, and >5.0 ug/L for CEA, as previously described. A serum marker level above the cut-off value was considered positive.MethodsThe clinical features of patients with CRC vs. those with benign LGI diseases were compared. The risk factors for CRC were examined by step-forward multivariate logistic regression. Subgroup analysis was carried out based on the age (<40 years,≤50 years,>50 years) and the presenting symptom (s). Further, an additional subgroup analysis to determine the trend of changes of diagnostic value with age increasing was carried only based on 10-year age increment. The receiver operating characteristic (ROC) curve was used to analyze the sensitivity, specificity, and diagnostic accuracy of CEA in combination with CA19-9 and CA72-4.Statistical AnalysisThe three tumor markers were logarithmically transformed prior to analysis. For step-forward multivariate regression, variables with P<0.10 in univariate analysis were included. P≤0.05 and odds ratio (OR)≥1 were considered statistically significant.The ROC curve was obtained based on the predicted probability of logistic regression model. The cut-off values were chosen based on the highest Youden Index (the sum of sensitivity plus specificity minus 1). Patients with a probability by the model greater than the cut-off value were predicted to have CRC, which was referred to the method described by Liu J, et al. All statistical analyses were performed using SPSS-13.0 (SPSS Inc, Chicago, IL). A two-sided P value of <0.05 was considered statistically significant.ResultsThe current study included 948 CRC patients and 789 patients with benign LGI diseases (Table 1). Abdominal pain, weight loss and hematochezia were the most common symptoms in both groups of patients. Weight loss had a sensitivity of 40.1% and a specificity of 79.2%. Hematochezia had a sensitivity of 54.5% and a specificity of 77.2%. Serum CEA, CA19-9 and CA72-4 levels were elevated in 33.0%, 23. 1% and 16.5% of the CRC patients and in 2.6%,2.7% and 5.7% of the patients with benign LGI diseases, respectively. Elevation of all three markers (CEA, CA19-9 and CA72-4) was noted in 47.6%of the CRC patients and only 10.3% of the patients with benign LGI diseases. Elevation of at least two serum tumor markers was seen in 20.5%of the CRC patients, and only in 0.6% of the patients with benign LGI diseases. Although the sensitivities for all three serum tumor makers were low, their specificities were remarkably high, especially when the serum levels of two or three markers were simultaneously elevated (99-100%) (Table3-1)Risk factors associated with CRCFactors for the multivariate analysis (P<0.10 in the univariate analysis) included:gender, age, hematochezia, abdominal pain, abdominal distention, diarrhea, constipation, weight loss, lnCA72-4, lnCA19-9 and lnCEA (Table3-2). The multivariate logistic regression revealed the following risk factors for CRC:male gender, age, hematochezia, diarrhea, weight loss, lnCA72-4, lnCA19-9 and lnCEA. Two dichotomized variables (hematochezia and weight loss) had odds ratios over 2, and they were included in the subsequent analysis, which also included four continuous variables (age, lnCA72-4, lnCA19-9 and lnCEA)Diagnostic accuracy-subgroup analysis based on the presenting symptoms and age.The 1737 subjects were divided based on symptoms as well as age(<40years, ≤50 years,>50years) for a subgroup analysis (Table 3). For patients with hematochezia and weight loss, the overall AUC for CEA in combination with CA19-9 and CA72-4 was 0.79, AUC was 0.93 in those under 40 years of age, 0.86 in those under 50 years old, and 0.74 in subjects at over 50 years of age. In particular, it should be pointed out that only CEA contributed to the logistic regression model to identify patients with CRC. Thus, we only need to consider the use of CEA alone in these patients. For patients with weight loss, the overall AUC was 0.82,AUC was 0.88 in those under 40 years of age,0.85 in those under 50 years old, and 0.77 in subjects over 50 years of age. The overall analysis in patients with hematochezia and without weight loss indicated that combing CEA with CA19-9 and CA72-4 does not improve diagnostic value, with an AUC of 0.72. However, irrespective of age, the diagnostic value remains not improved. The results of the improved diagnostic accuracies referred above, with AUCs over 0.80, by the combination of CEA with CA19-9 and CA72-4 were summarized in Table 4. The sensitivities and specificities were calculated by the use of cut-off values set at the highest of Youden Index (sensitivity plus specificity minus 1). In the total patients in our study, the sensitivities and specificities for CEA alone and the combination were 49.7%、86.3% and 51.8%、87.4%, respectively. In patients with hematochezia and under 50 years old, the sensitivity was increased to 68.3%, but the specificity was markedly decreased to 66.7%. However, sensitivities and specificities for the combination were relatively improved in some subgroups of patients, as shown in Table3-4.Diagnostic accuracy-age subgroup analysisFor patients under 21 years of age as well as patients over 80 years of age, the AUCs was <0.70 for both CEA alone and the combination of the three markers, with wide 95% CIs (probably due to small sample size) (Fig3-1). For subjects at 30-80 years of age, the AUC did not change significantly over the age for either marker combination or CEA alone, but combining CEA with CA19-9 and CA72-4 increased AUC in subjects at 20-30 years of age (0.89 vs.0.71 for CEA alone, p=0.006), and an AUC of 0.96 was achieved with the three marker combination in male patients at 20-30 years ofage.ConclusionThrough this research we found that the symptoms in patients with lower LGI diseases included in this study are the most common clinical manifestation, We also found that the combined test had good diagnostic values in identifying patients with CRC from patients with LGI symptoms, Further, it may be more easy to be the majority of the patients received. However, There may be statistical bias due to the smaller sample size, attempt to extrapolate the findings to the general population should be taken with caution, further study should be necessary and more sensitive and specific diagnostic markers should be identified in the future. |
PDF Full Text Request