ORIGINAL ARTICLE
Development and Validation of a Risk Stratification-based Screening Model for Predicting Colorectal Advanced Neoplasia in Korea Dong Hyun Kim, MD,* Jae Myung Cha, MD, PhD,* Hyun Phil Shin, MD, PhD,* Kwang Ro Joo, MD, PhD,* Joung Il Lee, MD, PhD,* and Dong Il Park, MD, PhDw
Goals: To develop and validate a risk stratification-based screening model for predicting colorectal advanced neoplasia in Korea. Background: Colorectal advanced neoplasia is the relevant finding of screening colonoscopy. Risk estimation for advanced neoplasia may be helpful to improve compliance and to develop more costeffective approaches toward screening. Study: We developed Korean Colorectal Screening (KCS) score by optimizing and adjusting Asia-Pacific Colorectal Screening (APCS) score to predict advanced neoplasia in an asymptomatic Korean population who received screening colonoscopies from September 2006 to September 2009. Moreover, we validated the KCS score in another Korean cohort who received screening colonoscopies from October 2009 to February 2011. We also assessed the predictive power and diagnostic performance of both KCS and APCS scores. Results: There were 3561 subjects in the derivation cohort and 1316 subjects in the validation cohort, with a prevalence of advanced neoplasia of 4.7% and 4.3%, respectively. After a multivariate analysis, KCS was developed as 0 to 8 points comprising of age, sex, body mass index, smoking, and family history of CRC. Using KCS scores to stratify the validation cohort, the prevalences of advanced neoplasia in the 3 risk tiers (average, moderate, and high) were 2.0%, 3.7%, and 10.9%, respectively. Moderate-risk and high-risk tiers showed 2.1- and 6.5-fold increased prevalences, respectively, of advanced neoplasia compared with average risk tier. In addition, KCS score showed relatively good discriminative power (ROC = 0.681) and higher sensitivity compared with APCS score for the high-risk tier. Conclusions: KCS score may be clinically simple and useful for assessing advanced neoplasia risk in Korea. However, racial disparity should be considered in risk stratification-based screening in each country. Key Words: colorectal neoplasm, colonoscopy, risk assessment, screening
(J Clin Gastroenterol 2015;49:41–49)
C
olorectal cancer (CRC) is one of the most common malignancies whose incidence is steadily increasing in many Asian countries.1 In Korea, CRC is the third most common cancer and accounts for 12.8% of all cancers. The age-standardized incidence rate of CRC in Korea increased to 36.2/100,000 in 2010 from 30.8/100,000 in 2008.2 In general, CRC usually develops through an adenoma-carcinoma sequence, and adenomas with advanced features have been shown to exhibit higher risk.3 Screening for CRC, along with the removal of these premalignant lesions, is also well known to reduce CRC-related mortality and morbidity.4–6 Therefore, early detection and removal of premalignant lesions by CRC screening has been advocated in many countries.7,8 However, limited resources have been an obstacle to expand CRC screening.9–11 In this regard, risk stratification models may help make CRC screening more effective. Recently, the Asia-Pacific Working Group on CRC developed the Asia-Pacific Colorectal Screening (APCS) score, based on the age, sex, family history of CRC, and smoking status of selected asymptomatic Asian subjects.12 However, the APCS score was limited in that it did not include the body mass index (BMI); furthermore, it was based on a heterogenous population comprising multiple races from many Asian countries. Therefore, this scoring system does not take into account the findings that CRC risk varies significantly with ethnicity,1,13 dietary,14–17 and other factors. Therefore, an optimized risk stratification-based screening model, based on the CRC epidemiology and resource limitations of each country, may be useful in predicting CRC risk. The aim of this study was to develop and validate a risk stratification-based screening model predictive of colorectal advanced neoplasia in asymptomatic Korean subjects.
METHODS Study Population of Derivation Cohort Received for publication August 1, 2013; accepted January 22, 2014. From the *Department of Internal Medicine, Kyung Hee University School of Medicine; and wDepartment of Internal Medicine, Sungkyunkwan University School of Medicine, Seoul, Republic of Korea. All authors were involved in study concept and design, acquisition/ analysis/interpretation of data, and drafting and critical revision of the manuscript. The authors declare that they have nothing to disclose. Reprints: Jae Myung Cha, MD, PhD, Department of Internal Medicine, Kyung Hee University School of Medicine, Kyung Hee University Hospital at Gang Dong, 149 Sangil-dong, Gangdong-gu, Seoul 134-727, Republic of Korea (e-mail: [email protected]). Copyright r 2014 by Lippincott Williams & Wilkins
J Clin Gastroenterol
Volume 49, Number 1, January 2015
We developed a clinical risk score for CRC screening based on consecutive asymptomatic subjects who underwent a screening colonoscopy, at Health Promotion Center of Kyung Hee University Hospital in Gangdong, Seoul, Korea, as a part of medical check-up between September 2006 and September 2009. From this colonoscopy database, subjects were retrospectively identified as eligible if they were between 30 and 75 years of age, were asymptomatic, and had undergone a first-time screening colonoscopy. Potential subjects were excluded if they met any of the following criteria: (1) aged below 30 or above 75 years; (2) previous colorectal examinations (including www.jcge.com |
41
Kim et al
colonoscopy, sigmoidoscopy, or barium enema); (3) incomplete colonoscopy because of poor bowel preparation or cecal intubation failure; (4) history of CRC or inflammatory bowel disease; (5) history of colorectal surgery; or (6) symptoms or signs indicating the need for a colonoscopy (eg, changes in bowel habits, hematochezia, unexplained weight loss, or a positive fecal occult blood test). To avoid the potential protective effects of medications, subjects with a history of anti-inflammatory drug use (NSAIDs, aspirin, or statins) for >1 year on a regular basis were also excluded. Smoking habits, alcohol consumption, past medical history, and family history of CRC in a first-degree relative were determined by interviews conducted by well-trained nurses. A patient was defined as a current smoker if they consumed at least 1 pack per week; consumption of any amount of alcohol exceeding 140 g per week was considered a positive history of alcohol use. Diabetes mellitus (DM) was defined as a fasting glucose level of Z126 mg/dL or use of hypoglycemic agents and/or insulin. Height and body weight, used to calculate BMI, were routinely measured by trained nurses. This study was conducted according to the Declaration of Helsinki and was approved by the Institutional Review Board of Kyung Hee University Hospital at Gang Dong (KHNMC IRB 2012-105).
KCS Score Development The KCS score was developed in the same manner as the APCS score.12 In summary, a univariate analysis was carried out on the study population using the Pearson w2 method to examine the association between clinical risk factors and colorectal neoplasia. Variables associated with any neoplasia and advanced neoplasia in the univariate analyses (P < 0.15) were entered into multivariate logistic regression models. Risk factors that retained significance in multivariate analyses were incorporated into the risk score model. For each risk factor, weights were assigned using respective adjusted odds ratios (ORs) obtained by logistic regression. To simplify the risk score as much as possible, adjusted ORs were divided by 2 and then rounded to the nearest whole number. The final risk score was calculated by summing the scores of all individual risk factors. The KCS score was divided into 3 risk tiers: score 0 to 1, “average risk (AR)”; 2 to 3, “moderate risk (MR)”; and 4 to 8, “high risk (HR).” Score validities were assessed by receiver operating characteristics (ROC) analysis, and the predicted risks of advanced neoplasia were assessed by comparing risk tier groups.
KCS Score Validation for the Validation Cohort The sample size estimation for the validation cohort was based on methodology of APCS development.12 In the derivation cohort of the current study, the prevalences of individual risk factors ranged from 3.6% to 60.4% (mean 40%), and the prevalence of colorectal advanced neoplasia was 4.7%. We assumed the estimated prevalence of individual risk factors to be 40% and the estimated prevalence of advanced neoplasia to be 5.0% in the validation cohort, as recently implemented quality controls of screening colonoscopies in Korea will potentially increase the detection rate of advanced neoplasia. On the basis of these assumptions, at least 1300 asymptomatic subjects were required for a power of 80% to detect a risk factor with an OR of 2 at P < 0.05 level of significance.
42 | www.jcge.com
J Clin Gastroenterol
Volume 49, Number 1, January 2015
The validation cohort was an independent cohort, containing 1316 asymptomatic subjects undergoing screening colonoscopy between October 2009 and February 2011, and their data were retrospectively collected and analyzed to validate the KCS score. The colonoscopy and study protocols used for subjects in the validation cohort were identical to those used for subjects in the derivation cohort. Each subject in the validation group was assigned a personal risk score, as calculated by the KCS scoring method. The performance of the KCS score to predict the risk of advanced neoplasia was evaluated by determining the ORs for each risk tier, and the validity of the KCS score was assessed by ROC analysis.
APCS Score and Validation APCS score was also validated in our study population. The APCS score has a range of 0 to 7 points and is calculated as the sum of individual risk factors either present or absent in the individual: age below 50 years (0), between 50 and 69 years (2), above 70 years (3), male sex (1), female sex (0), family history of CRC in a first-degree relative present (2) or absent (0), current or past smoker (1), and nonsmoker (0).12 The APCS score stratifies patients into 3 risk tiers: score 0 to 1, AR; 2 to 3, MR; and 4 to 7, HR. After calculating the APCS scores, the predicted risks of advanced neoplasia were assessed by comparing the risk tier groups, and the validity of the APCS score was evaluated by ROC analysis. The diagnostic performances of both the KCS and APCS scores were also compared. With respect to diagnostic performance, sensitivity is the probability of subjects with advanced neoplasia who had a high risk on score model, whereas specificity is the probability of subjects without advanced neoplasia who do not have a high risk.
Colonoscopy All patients underwent a 3-day dietary restriction and drank 4 L of polyethylene glycol solution for 12 hours before the colonoscopy, until clear rectal fluid was evacuated. Potential adverse events were explained to the patients before the colonoscopy, and all patients provided written consent before the procedure. For conscious sedative colonoscopy, an individualized dose of midazolam and/ or propofol was given by the gastroenterologist based on the age, weight, and general condition of the patient. An antispasmodic medication (10 mg intravenous hyoscine methobromide) was also administered to subjects when no contraindications were present. All examinations were performed using a standard video colonoscope (EC-590 ZW/L, Fujinon Inc., Saitama, Japan) in a standardized manner by experienced gastroenterologists. The quality of the bowel preparation was assessed as excellent, good, fair, or poor.18 All detected polyps were biopsied, except for multiple hyperplastic polyps in the rectosigmoid colon showing endoscopic features consistent with a gross hyperplastic appearance (small size, sessile shape, pale color, and type 1 or 2 pit pattern).19 According to the regulations of our hospital, all detected polyps were imaged and all polyp characteristics, such as shape, size, number, and location, were documented. Shapes were classified according to the Paris classification system.20 In addition, polyp sizes were estimated by comparing the polyps to open biopsy forceps (Olympus FB-28U-1; Amori Olympus Co. Ltd, Amori, Japan). All excised polyps were evaluated histologically by r
2014 Lippincott Williams & Wilkins
J Clin Gastroenterol
Volume 49, Number 1, January 2015
gastrointestinal pathologists. Colorectal adenoma was defined as an adenoma, irrespective of its grade or villous components, whereas advanced neoplasia was defined as a colorectal carcinoma or advanced adenoma (diameter Z10 mm, highgrade dysplasia, or >25% villous features).12,21
Risk Stratification Model for Advanced Neoplasia in Korea
were performed using the Statistical Package for the Social Sciences, version 18.0 (SPSS, Chicago, IL).
RESULTS
Statistical Analysis
Characteristics of the Derivation and Validation Cohorts
All data are presented as mean ± SD for continuous variables and as numbers (percentages) of subjects for categorical variables. Categorical variables, including candidate risk factors such as age, sex, BMI, smoking status, alcohol consumption, DM status, and family history of CRC in a first-degree relative, were compared using the Pearson w2 test. Multiple logistic regressions was used to control for confounding variables and to determine the independent effects of risk factors for either any type of neoplasia, or advanced neoplasia. ORs were calculated for dichotomized endpoints; values are presented with modelbased lower and upper 95% confidence intervals (CIs). The predictive powers of both score models were analyzed by determining the areas under the corresponding ROC curves, including the 95% CIs. The Hosmer-Lemeshow goodness-of-fit statistic was used to assess reliability of the model. To compare the statistical ability of the APCS and the KCS scores to predict the risk of advanced neoplasia, ROC analysis and the McNemar tests were performed. All P-values were 2-tailed, a P-value 1 year (n = 159); and insufficient patient data (n = 39). After filtering, 3561 asymptomatic subjects were enrolled in the study cohort. The mean age was 51.3 ± 9.0 years and the mean BMI was 23.8 ± 3.1 kg/ m2. In the derivation population, 2152 patients were male patients (60.4%), 34.4% were obese, 47.7% were smokers, and 3.6% had a family history of CRC in a first-degree relative. With respect to pathologic findings, 841 (23.6%) were found to have any neoplasia, of which 156 patients (4.4% of the total) had advanced neoplasia and 13 patients (0.4% of the total) had CRC. The characteristics of the study population with any neoplasia and advanced neoplasia are shown in Table 1.
FIGURE 1. Study enrollment. A, In this study, 3561 asymptomatic subjects of the 3970 subjects retrospectively enrolled were placed in the derivation cohort. B, 1316 asymptomatic subjects from an independent cohort were also retrospectively enrolled after exclusion of 89 subjects and included in the validation cohort. These asymptomatic subjects also had undergone screening colonoscopies. The study protocols for the derivation and validation cohorts were identical. r
2014 Lippincott Williams & Wilkins
www.jcge.com |
43
J Clin Gastroenterol
Kim et al
Volume 49, Number 1, January 2015
TABLE 1. Characteristics of the Derivation Cohort According to Risk Factors
Any Neoplasia (n = 841, 23.6%)
Advanced Neoplasia (n = 169, 4.7%)
All Subjects, n (%)
n (%)
P
n (%)
P
2152 (60.4) 1409 (39.6)
639 (76.0) 202 (24.0)
< 0.001
142 (84.0) 27 (16.0)
< 0.001
1513 (42.5) 2048 (57.5)
225 (26.8) 616 (73.2)
< 0.001
23 (13.6) 146 (86.4)
< 0.001
2372 (66.6) 1189 (33.4)
487 (57.9) 354 (42.1)
< 0.001
92 (55.1) 77 (44.9)
0.001
127 (3.6) 3434 (96.4)
37 (4.4) 804 (95.6)
0.136
10 (5.9) 159 (94.1)
0.091
1863 (52.3) 1698 (47.7)
318 (37.8) 523 (62.2)
< 0.001
46 (27.2) 123 (72.8)
< 0.001
2429 (68.2) 1132 (31.8)
504 (59.9) 337 (40.1)
< 0.001
91 (53.8) 78 (46.2)
< 0.001
3409 (95.7) 152 (4.3)
792 (94.2) 49 (5.8)
0.011
158 (93.5) 11 (6.5)
0.140
Sex Male Female Age (y) < 50 Z50 BMI (kg/m2) < 25 Z25 FHx of CRC Present Absent Smoking Never Current or ex Alcohol No Yes Diabetes No Yes
BMI indicates body mass index; CRC, colorectal cancer; FHx, family history.
For the validation cohort, 1316 asymptomatic subjects were enrolled after exclusion of 89 subjects because of the following reasons: aged below 30 or above 75 years (n = 38); prior colonoscopy, sigmoidoscopy, or barium enema (n = 9); poor bowel preparation (n = 10); history of CRC (n = 5); history of colorectal surgery (n = 5); regular use of NSAIDs, aspirin, or statins for >1 year (n = 16); and insufficient patient data (n = 6). The mean age was 49.6 ± 9.9 years and the mean BMI was 23.8 ± 3.2 kg/m2. In the validation population, 557 patients were male patients (42.3%), 36.3% were obese, 19.0% were smokers, and 2.9% had a family history of CRC in a first-degree relative. With respect to pathologic findings, 405 (30.8%) were found to have any neoplasia, of which 56 patients (4.3% of the total) had advanced neoplasia.
Univariate and Multivariate Analysis of Predictors for Colorectal Neoplasia in the Derivation Cohort Univariate and multivariate analyses were performed for each risk factor. Multivariate logistic regression showed that patients aged 50 years or above, male sex, BMIZ25 kg/m2, positive family history of CRC in a first-degree relative, and smoking status were significant risk factors for any neoplasia, with ORs (95% CI) of 2.6 (2.2 to 3.1), 1.8 (1.4 to 2.3), 1.3 (1.1 to 1.6), 1.6 (1.1 to 2.4), and 1.6 (1.2 to 1.8), respectively (Table 2). Age above 50 years, male sex, positive family history in a first-degree relative, and smoking status were also significant risk factors for advanced neoplasia, with ORs (95% CI) of 5.3 (3.4 to 8.3), 1.9 (1.1 to 3.4), 2.2 (1.1 to 4.4), and 2.0 (1.3 to 3.2); however, BMI only showed the trend of a risk factor (P = 0.092), with an OR of 1.3 (1.0 to 1.8) (Table 2). The Hosmer-Lemeshow goodness-of-fit statistic was P = 0.81 for the derivation study cohort.
44 | www.jcge.com
Results of the KCS Score in the Derivation Cohort To derive the KCS score, points were assigned for each risk factor of advanced neoplasia as follows: age below 50 years (0), between 50 and 69 years (2), 70 years or above(4), male sex (1), female sex (0), family history of colorectal cancer in a first-degree relative present (1) or absent (0), nonsmoker (0), current or past smoker (1), and BMI < 25 kg/m2 (0) and Z25 kg/m2 (1). One point was assigned for the presence of obesity as it was a significant risk factor for colorectal neoplasia in the previous literature,22,23 although it did not reach significance for advanced neoplasia in our study population. The KCS score ranges from 0 to 8 points, and the frequency distribution of KCS scores by subject is shown in Table 3. According to the KCS score stratification, the AR tier contained 683 (19.2%) subjects, the MR tier contained 1762 (49.5%) subjects, and the HR tier contained 1116 (31.3%) subjects. The prevalences of advanced neoplasia in the 3 tiers were: 0.7% (95% CI, 0.15%-1.46%) in the AR tier, 2.2% (95% CI, 1.48%-2.84%) in the MR tier, and 11.3% (95% CI, 9.50%-13.17%) in the HR tier. Subjects in the MR and HR tiers had 3.0-fold (95% CI, 1.2-7.6) and 17.3-fold (95% CI, 7.0-42.4) increased rates of advanced neoplasia, respectively, compared with subjects in the AR tier. The Hosmer-Lemeshow goodness-of-fit statistic had a P-value of 0.51.
Risk Stratification of KCS and APCS Scores in the Validation Cohort According to their KCS score, 32.6%, 52.1%, and 15.3% of the subjects were placed into the AR, MR, and HR tiers, respectively (Table 4). The prevalences of advanced neoplasia in the 3 tiers were 1.9%, 3.8%, and 10.9%, respectively. Subjects in the MR and HR tiers had 2.1-fold (95% CI, 0.9-4.6) and 6.6-fold (95% CI, 2.8-14.8) r
2014 Lippincott Williams & Wilkins
J Clin Gastroenterol
Volume 49, Number 1, January 2015
Risk Stratification Model for Advanced Neoplasia in Korea
TABLE 2. Univariate and Multivariate Predictors of Colorectal Neoplasia in the Derivation Cohort
Univariate Analysis Risk Factors Predictors of any neoplasia Sex Age (y) 50-69 Z70 BMIZ25 kg/m2 FHx of CRC Smoking Alcohol Diabetes Predictors of advanced neoplasia Sex Age (y) 50-69 Z70 BMIZ25 kg/m2 FHx of CRC Smoking Alcohol Diabetes
Multivariate Analysis
OR (95% CI)
P
OR (95% CI)
P
2.5 (2.1-3.0)
< 0.001
1.8 (1.4-2.3)
< 0.001
2.4 4.1 1.6 1.3 2.2 1.6 1.6
(2.0-2.9) (2.6-6.4) (1.4-1.9) (0.9-2.0) (1.8-2.5) (1.4-1.9) (1.1-2.2)
< 0.001 < 0.001 < 0.001 0.134 < 0.001 < 0.001 0.023
2.6 4.8 1.3 1.6 1.6 1.1 1.0
(2.2-3.1) (3.0-7.6) (1.1-1.6) (1.1-2.4) (1.2-1.8) (0.9-1.4) (0.7-1.5)
< 0.001 < 0.001 0.001 0.027 0.001 0.187 0.811
3.6 (2.4-5.5)
< 0.001
1.9 (1.1-3.4)
0.022
4.8 9.1 1.7 1.8 3.1 1.9 1.6
< 0.001 < 0.001 0.001 0.096 < 0.001 < 0.001 0.143
(3.1-7.5) (4.3-19.4) (1.3-2.3) (0.9-3.4) (2.2-4.4) (1.4-2.6) (0.9-3.0)
5.3 11.8 1.3 2.2 2.0 1.3 0.9
(3.4-8.3) (5.4-25.7) (1.0-1.8) (1.1-4.4) (1.3-3.2) (0.9-1.8) (0.5-1.8)
< 0.001 < 0.001 0.092 0.030 0.004 0.170 0.876
BMI indicates body mass index; CI, confidence interval; CRC, colorectal cancer; FHx, family history; OR, odds ratio.
increased rates of advanced neoplasia, compared with subjects in the AR tier (P < 0.001). The Hosmer-Lemeshow goodness-of-fit statistic had a P-value of 0.48. By using their APCS scores to stratify the subjects, 37.6%, 53.5%, and 8.9% were included in the AR, MR, and HR tiers, respectively (Table 4). The prevalences of advanced neoplasia in the 3 tiers were 2.4%, 4.0%, and
13.7%, respectively. When compared with subjects in the AR tier, those in the MR and HR tiers had 1.7-fold (95% CI, 0.83.3) and 6.4-fold (95% CI, 2.9-13.9) increased rates of advanced neoplasia, respectively (P < 0.001). The Hosmer-Lemeshow goodness-of-fit statistic had a P-value of 0.13.
Diagnostic Performances of the KCS and APCS Scores in the Validation Cohort TABLE 3. Score and Risk Tier Distribution of the APCS and KCS Score in the Derivation Cohort
No. Subjects (%) APCS score 0 1 2 3 4 5 6 7 Total KCS score 0 1 2 3 4 5 6 7 8 Total
515 225 1474 373 902 43 29
No. Any Neoplasia (%)
(14.5) (6.3) (41.4) (10.5) (25.3) (1.2) (0.8) 0 3561 (100)
(4.0) (3.6) (33.9) (13.6) (40.7) (2.5) (1.8) 0 841 (100)
453 230 1060 702 665 395 43 13
26 27 190 172 232 164 22 8
(12.7) (6.5) (29.8) (19.7) (18.7) (11.1) (1.2) (0.4) 0 3561 (100)
34 30 285 114 342 21 15
(3.1) (3.2) (22.6) (20.5) (27.6) (19.5) (2.6) (1.0) 0 841 (100)
No. Advanced Neoplasia (%) 4 3 30 22 97 7 6
(2.4) (1.8) (17.8) (13.0) (57.4) (4.1) (3.6) 0 169 (100) 4 1 22 21 62 48 7 4
(2.4) (0.6) (13.0) (12.4) (36.7) (28.4) (4.1) (2.4) 0 169 (100)
APCS indicates Asia-Pacific Colorectal Screening; KCS, Korean Colorectal Screening.
r
2014 Lippincott Williams & Wilkins
Comparison of the diagnostic performances of the KCS and APCS scores using the McNemar test revealed that the KCS score exhibited statistically higher sensitivity
TABLE 4. Prevalence of Advanced Neoplasia by Risk Tier in the Validation Cohort According to KCS and APCS Score
Risk Tier
No. Subjects (%)
KCS score Average 429 (0-1) Moderate 686 (2-3) High (4-8) 201 Total 1316 APCS score Average 495 (0-1) Moderate 704 (2-3) High (4-7) 117 Total 1316
Advanced Neoplasia Number (%) (95% CI)
RR (95% CI)
(32.6)
8 (1.9) (0.55-3.17)
Reference
(52.1)
26 (3.8) (2.33-5.25)
2.1 (0.9-4.6)
(15.3) (100)
22 (10.9) (6.53-15.37) 56 (4.3) (3.15-5.36)
6.5 (2.8-14.8)
(37.6)
12 (2.4) (1.04-3.80)
Reference
(53.5)
28 (4.0) (2.51-5.45)
1.7 (0.8-3.3)
(8.9) (100)
16 (13.7) (7.30-20.06) 56 (4.3) (3.15-5.36)
6.4 (2.9-13.9)
APCS indicates Asia-Pacific Colorectal Screening; CI, confidence interval; KCS, Korean Colorectal Screening.
www.jcge.com |
45
J Clin Gastroenterol
Kim et al
TABLE 5. Diagnostic Performance of the APCS and KCS Score for Detection of Advanced Colorectal Neoplasia in the Validation Cohort
KCS (%)
APCS (%)
P
39.3 85.8
28.6 92.0
0.031 < 0.001
85.7 33.4
78.6 38.3
0.125 < 0.001
HR vs. MR + AR Sensitivity Specificity MR + HR vs. AR Sensitivity Specificity
APCS indicates Asia-Pacific Colorectal Screening; AR, average risk; HR, high risk; KCS, Korean Colorectal Screening; MR, moderate risk.
compared with the APCS score when the HR tier was compared with MR plus AR tier group (39.3% vs. 28.6%, P = 0.031) (Table 5). In contrast, the APCS score showed statistically higher specificity compared with the KCS score when the HR tier was compared with MR plus AR tier group (P < 0.001). By ROC analysis, KCS and APCS scores were 0.681 (95% CI, 0.608-0.763) and 0.676 (95% CI, 0.600-0.751), respectively, indicating relatively good discrimination, but no significant differences were noted (P = 0.86) (Fig. 2).
DISCUSSION In this study, we developed the KCS score comprising age, sex, BMI, smoking, and family history of CRC to predict the risk of advanced neoplasia after modifying the APCS score and validated the KCS score through a validation cohort whose members underwent screening
FIGURE 2. ROC curves for the KCS (solid line) and APCS (dotted line) score models. The curves for each score showed a similar pattern, and the difference between them was not statistically significant (0.675 vs. 0.670; P = 0.69). APCS indicates Asia-Pacific Colorectal Screening; KCS, Korean Colorectal Screening; ROC, receiver operating characteristic analysis.
46 | www.jcge.com
Volume 49, Number 1, January 2015
colonoscopies. In our study, KCS score showed good discriminatory power measured by ROC and could be applied to identify the high-risk subjects for advanced neoplasia. Subjects in the HR tier showed a 6.0-fold increase in their chance of having advanced neoplasia detected, according to their KCS scores, when compared with subjects in the AR tier. We developed the KCS score to overcome certain limitations of the APCS score, thereby making it more suitable for use in the Korean population. For example, BMI was not included as a risk factor contributing to the APCS score; furthermore, a variety of populations with substantial variations in their risk factors were included in the study population used to derive the APCS score.12 Compared with the APCS score, the KCS score model includes BMI as a risk factor, as well as assigns different points for some risk factors based on a Korean derivation cohort (that is, 4 points are assigned in the KCS vs. 3 points in the APCS for patients aged 70 years or older, and 1 point is assigned in the KCS vs. 2 points in the APCS for a family history of CRC). As sensitivity is more important than specificity in screening for significant diseases such as CRC or prevalent diseases such as advanced neoplasia, the KCS score, which showed higher sensitivity, may be a more appropriate option compared with the APCS score, at least in Korea. Although the APCS score displayed higher specificity, the proportion of subjects stratified into the HR tier was only 8.7% in our validation cohort. Therefore, our data indicate that the KCS score is optimized for use with an asymptomatic Korean population. CRC screening has been shown to reduce CRC incidence and mortality24–26; however, implementation of CRC screening is often hampered by limited resources, lack of awareness in the target population, insufficient advocacy by healthcare professionals, and poor patient compliance.27–32 Risk stratification-based screening, including KCS, may improve compliance of current CRC screening methods by providing motivation for those who are assigned high-risk scores. According to the “Health Belief” model, perceived risk is positively associated with both CRC screening utilization and intention to be screened. Therefore, compliance of CRC screening may be increased when participants perceive themselves as highly susceptible to colorectal advanced neoplasia.33–39 In addition, risk stratification-based screening may enable more cost-effective approaches by prioritizing CRC screening in the high-risk group, especially in countries with limited resources. Risk stratification-based screening should be adjusted after careful consideration of CRC epidemiology and the available health resources of each country. For example, less invasive and inexpensive screening processes, such as fecal blood screening, may be recommended for individuals in the AR tier, whereas more thorough screening processes, such as colonoscopy screening, may be recommended for those in the HR tier.40 In the present study, 39.3% of advanced neoplasia could be detected in the HR group, which accounts for only 15.6% of total population. Furthermore, 85.7% of advanced neoplasia could be detected with colonoscopy screening in the MR and HR groups, which accounts for 67.4% of the total population. Until now, few efforts to develop risk prediction models for advanced neoplasia have been reported. Most risk stratification-based screening models have been developed in the West. Imperiale et al41 proposed a clinical index to estimate the risk of proximal advanced neoplasia; this index was based on age, sex, and the distal findings from r
2014 Lippincott Williams & Wilkins
J Clin Gastroenterol
Volume 49, Number 1, January 2015
sigmoidoscopy. However, this index requires an initial sigmoidoscopy to calculate the index score. Betes et al42 described an index taking into account age, sex, and BMI to stratify subjects by risk for advanced neoplasia; however, this scoring system did not include smoking status or family history. Lin et al40 proposed an index to stratify subjects into a high-risk group for colonoscopy screening based on age, sex, and family history, without including smoking status or BMI, and indicating different screening tools according to the various risk groups. In the East, 2 risk stratification-based screening models have also been suggested. Cai et al43 proposed a risk stratification tool for Chinese patients of average risk based on their age, sex, smoking status, DM status, and dietary habits. Although this tool showed good discriminatory power (ROC = 0.74), it was limited in the sense that dietary habits are relatively subjective and complex to measure, in particular aspects such as duration and quantity of the subject’s diet. In the recently reported APCS scoring system, the risk of advanced neoplasia was evaluated based on age, sex, family history, and smoking status.12 However, the APCS score showed some limitations, as described above, and was based on colonoscopy data with low adenoma detection rate (ADR). As the risk of CRC varies significantly with ethnicity1,13 and diet,14–17 racial disparity should be considered in risk stratification-based screening. However, only 44.4% of the subjects in the derivation cohort for the APCS score were Korean patients.12,44 Moreover, Korean patients showed a 2-fold higher risk of colorectal neoplasia than Chinese patients in a multinational, multicenter study,44 and age was a stronger risk factor for advanced neoplasia in Korean patients than in Chinese patients.43,45,46 In this regard, it is difficult to develop universal risk stratificationbased screening models appropriate for all countries; therefore, the discriminatory power of each risk stratification-based screening model should be validated for each target population. There are several advantages to our study. First, this is the first study reporting country-specific modifications to the APCS score. Moreover, our score has a clinical implication in that its risk factors are convenient to measure and are easily applied. Furthermore, all risk factors used in the KCS score have been shown to be relevant risk factors for advanced neoplasia in previous studies.22,43,46–54 Second, our study was based on a large number of asymptomatic Korean patients: 3561 asymptomatic subjects were included in the derivation cohort and 1316 asymptomatic subjects were included in the validation cohort. Third, data collected in this study were also of high quality, despite the study’s retrospective design. The level of experience of the colonoscopists used in our study is reflected by their 100% cecal intubation rate, their 24% ADR in the derivation cohort, and 31% ADR in the validation cohort. These results are improved compared with previous studies of colonoscopy quality indicators.55 Furthermore, subjects with poor bowel preparations were excluded from data analysis in our study. Considering the low ADR (18.7%) found in a previous study of many Asian countries,12 our risk score model is the first one to be based on high-quality colonoscopy data. Finally, this study may improve compliance and costeffectiveness of CRC screening considering a clinical risk score, as CRC screening is currently underused despite its apparent effectiveness.56 Adults older than 50 years of age in the HR tier may be more encouraged to undergo CRC screening, as they exhibit an approximately 3.9-fold r
2014 Lippincott Williams & Wilkins
Risk Stratification Model for Advanced Neoplasia in Korea
increased risk of advanced neoplasia compared with subjects in the MR tier (relative risk, 3.50; 95% CI, 1.80-6.82; P < 0.001). Furthermore, adults younger than 50 years of age may be recommended to undergo CRC screening if their scores place them in either the MR or HR tiers, as these groups had an approximately 2.3-fold increased risk of advanced neoplasia compared with the AR tier (relative risk, 2.53; 95% CI, 0.98-6.51; P = 0.054). We concede that one of the limitations of our study is that it was a single-center study; furthermore, participants were more likely to be health-conscious and economically affluent, which may limit the generalization of our findings. As another limitation, the retrospective nature of the study design somewhat limited the data that could be collected. However, considering the extensive collection of personal information by well-trained nurses before health screening in our hospital, such data limitations were likely minimal. Third, the diagnostic performance of our risk score model was suboptimal; however, at the time we designed the model, accepted criteria for CRC screening scores to predict advanced neoplasia were not yet available. Furthermore, the diagnostic performance of our model should be interpreted differently in each country. As risk stratification-based screening models can be optimized for each country by taking into account differences in colorectal neoplasia prevalences and health resources available for screening, our study may help physicians in other countries develop appropriate risk stratification-based screening models for their particular countries. Finally, asymptomatic subjects younger than 50 years of age were included in this study. Although CRC screening is not recommended for subjects of this age, we included their data in the model derivation to determine the value of using a risk stratification model for this young population. In summary, the KCS score is a clinically simple and useful parameter for predicting advanced neoplasia in asymptomatic Korean patients. However, racial disparity should be considered in risk stratification-based screening in individual countries. Risk stratification-based screening deserves to be explored further for its potential benefits and cost-effectiveness in each country. REFERENCES 1. Sung JJ, Lau JY, Goh KL, et al. Increasing incidence of colorectal cancer in Asia: implications for screening. Lancet Oncol. 2005;6:871–876. 2. Jung KW, Park S, Kong HJ, et al. Cancer statistics in Korea: incidence, mortality, survival, and prevalence in 2009. Cancer Res Treat. 2012;44:11–24. 3. Atkin WS, Morson BC, Cuzick J. Long-term risk of colorectal cancer after excision of rectosigmoid adenomas. N Engl J Med. 1992;326:658–662. 4. Winawer SJ, Zauber AG, Ho MN, et al. Prevention of colorectal cancer by colonoscopic polypectomy. The National Polyp Study Workgroup. N Engl J Med. 1993;329:1977–1981. 5. Citarda F, Tomaselli G, Capocaccia R, et al. Efficacy in standard clinical practice of colonoscopic polypectomy in reducing colorectal cancer incidence. Gut. 2001;48:812–815. 6. Zauber AG, Winawer SJ, O’Brien MJ, et al. Colonoscopic polypectomy and long-term prevention of colorectal-cancer deaths. N Engl J Med. 2012;366:687–696. 7. Levin B, Lieberman DA, McFarland B, et al. Screening and surveillance for the early detection of colorectal cancer and adenomatous polyps, 2008: a joint guideline from the American Cancer Society, the US Multi-Society Task Force on Colorectal Cancer, and the American College of Radiology. Gastroenterology. 2008;134:1570–1595.
www.jcge.com |
47
Kim et al
8. Sung JJ, Lau JY, Young GP, et al. Asia Pacific consensus recommendations for colorectal cancer screening. Gut. 2008;57:1166–1176. 9. Levin TR. Colonoscopy capacity: can we build it? Will they come? Gastroenterology. 2004;127:1841–1844. 10. Brown ML, Klabunde CN, Mysliwiec P. Current capacity for endoscopic colorectal cancer screening in the United States: data from the National Cancer Institute Survey of Colorectal Cancer Screening Practices. Am J Med. 2003;115: 129–133. 11. Rex DK, Lieberman DA. Feasibility of colonoscopy screening: discussion of issues and recommendations regarding implementation. Gastrointest Endosc. 2001;54:662–667. 12. Yeoh KG, Ho KY, Chiu HM, et al. The Asia-Pacific Colorectal Screening score: a validated tool that stratifies risk for colorectal advanced neoplasia in asymptomatic Asian subjects. Gut. 2011;60:1236–1241. 13. Leung WK, Ho KY, Kim WH, et al. Colorectal neoplasia in Asia: a multicenter colonoscopy survey in symptomatic patients. Gastrointest Endosc. 2006;64:751–759. 14. Otani T, Iwasaki M, Yamamoto S, et al. Alcohol consumption, smoking, and subsequent risk of colorectal cancer in middleaged and elderly Japanese men and women: Japan Public Health Center-based prospective study. Cancer Epidemiol Biomarkers Prev. 2003;12:1492–1500. 15. Ho JW, Lam TH, Tse CW, et al. Smoking, drinking and colorectal cancer in Hong Kong Chinese: a case-control study. Int J Cancer. 2004;109:587–597. 16. Toyomura K, Yamaguchi K, Kawamoto H, et al. Relation of cigarette smoking and alcohol use to colorectal adenomas by subsite: the self-defense forces health study. Cancer Sci. 2004;95:72–76. 17. Tsong WH, Koh WP, Yuan JM, et al. Cigarettes and alcohol in relation to colorectal cancer: the Singapore Chinese Health Study. Br J Cancer. 2007;96:821–827. 18. Rostom A, Jolicoeur E. Validation of a new scale for the assessment of bowel preparation quality. Gastrointest Endosc. 2004;59:482–486. 19. Kudo S, Hirota S, Nakajima T, et al. Colorectal tumours and pit pattern. J Clin Pathol. 1994;47:880–885. 20. Participants in the Paris Workshop. The Paris endoscopic classification of superficial neoplastic lesions: esophagus, stomach, and colon: November 30 to December 1, 2002. Gastrointest Endosc. 2003;58:S3–43. 21. Lieberman DA, Rex DK, Winawer SJ, et al. Guidelines for colonoscopy surveillance after screening and polypectomy: a consensus update by the US Multi-Society Task Force on Colorectal Cancer. Gastroenterology. 2012;143:844–857. 22. Kim SE, Shim KN, Jung SA, et al. An association between obesity and the prevalence of colonic adenoma according to age and gender. J Gastroenterol. 2007;42:616–623. 23. Kim BC, Shin A, Hong CW, et al. Association of colorectal adenoma with components of metabolic syndrome. Cancer Causes Control. 2012;23:727–735. 24. Atkin WS, Cuzick J, Northover JM, et al. Prevention of colorectal cancer by once-only sigmoidoscopy. Lancet. 1993;341:736–740. 25. Winawer SJ, Flehinger BJ, Schottenfeld D, et al. Screening for colorectal cancer with fecal occult blood testing and sigmoidoscopy. J Natl Cancer Inst. 1993;85:1311–1318. 26. Mandel JS, Church TR, Bond JH, et al. The effect of fecal occult-blood screening on the incidence of colorectal cancer. N Engl J Med. 2000;343:1603–1607. 27. Gimeno-Garcia AZ, Quintero E, Nicolas-Perez D, et al. Impact of an educational video-based strategy on the behavior process associated with colorectal cancer screening: a randomized controlled study. Cancer Epidemiol. 2009;33: 216–222. 28. Domati F, Travlos E, Cirilli C, et al. Attitude of the Italian general population towards prevention and screening of the most common tumors, with special emphasis on colorectal malignancies. Intern Emerg Med. 2009;4:213–220.
48 | www.jcge.com
J Clin Gastroenterol
Volume 49, Number 1, January 2015
29. Cai SR, Zhang SZ, Zhu HH, et al. Barriers to colorectal cancer screening: a case-control study. World J Gastroenterol. 2009;15:2531–2536. 30. Emmons K, Puleo E, McNeill LH, et al. Colorectal cancer screening awareness and intentions among low income, sociodemographically diverse adults under age 50. Cancer Causes Control. 2008;19:1031–1041. 31. Kamposioras K, Mauri D, Alevizaki P, et al. Cancer screening in Greece. Guideline awareness and prescription behavior among Hellenic physicians. Eur J Intern Med. 2008;19: 452–460. 32. Inadomi JM. Taishotoyama Symposium Barriers to colorectal cancer screening: economics, capacity and adherence. J Gastroenterol Hepatol. 2008;23(suppl 2):S198–S204. 33. Frame PS, Kowulich BA. Stool occult blood screening for colorectal cancer. J Fam Pract. 1982;15:1071–1075. 34. Farrands PA, Hardcastle JD, Chamberlain J, et al. Factors affecting compliance with screening for colorectal cancer. Community Med. 1984;6:12–19. 35. Kelly RB, Shank JC. Adherence to screening flexible sigmoidoscopy in asymptomatic patients. Med Care. 1992;30: 1029–1042. 36. Price JH. Perceptions of colorectal cancer in a socioeconomically disadvantaged population. J Community Health. 1993;18: 347–362. 37. Lewis SF, Jensen NM. Screening sigmoidoscopy. Factors associated with utilization. J Gen Intern Med. 1996;11:542–544. 38. Strecher VJ, Rosenstock IM. The health belief model. In: Glanz K, Lewis FM, Rimer BK, eds. Health Behavior and Health Education. San Francisco, CA: Jossey-Bass; 1997:41–59. 39. Watts BG, Vernon SW, Myers RE, et al. Intention to be screened over time for colorectal cancer in male automotive workers. Cancer Epidemiol Biomarkers Prev. 2003;12: 339–349. 40. Lin OS, Kozarek RA, Schembre DB, et al. Risk stratification for colon neoplasia: screening strategies using colonoscopy and computerized tomographic colonography. Gastroenterology. 2006;131:1011–1019. 41. Imperiale TF, Wagner DR, Lin CY, et al. Using risk for advanced proximal colonic neoplasia to tailor endoscopic screening for colorectal cancer. Ann Intern Med. 2003;139: 959–965. 42. Betes M, Munoz-Navas MA, Duque JM, et al. Use of colonoscopy as a primary screening test for colorectal cancer in average risk people. Am J Gastroenterol. 2003;98:2648–2654. 43. Cai QC, Yu ED, Xiao Y, et al. Derivation and validation of a prediction rule for estimating advanced colorectal neoplasm risk in average-risk Chinese. Am J Epidemiol. 2012;175: 584–593. 44. Byeon JS, Yang SK, Kim TI, et al. Colorectal neoplasm in asymptomatic Asians: a prospective multinational multicenter colonoscopy survey. Gastrointest Endosc. 2007;65: 1015–1022. 45. Chen HM, Weng YR, Jiang B, et al. Epidemiological study of colorectal adenoma and cancer in symptomatic patients in China between 1990 and 2009. J Dig Dis. 2011;12:371–378. 46. Choe JW, Chang HS, Yang SK, et al. Screening colonoscopy in asymptomatic average-risk Koreans: analysis in relation to age and sex. J Gastroenterol Hepatol. 2007;22:1003–1008. 47. Pariente A, Milan C, Lafon J, et al. Colonoscopic screening in first-degree relatives of patients with “sporadic” colorectal cancer: a case-control study. The Association Nationale des Gastroenterologues des Hopitaux and Registre Bourguignon des Cancers Digestifs (INSERM CRI 9505). Gastroenterology. 1998;115:7–12. 48. Sturmer T, Glynn RJ, Lee IM, et al. Lifetime cigarette smoking and colorectal cancer incidence in the Physicians’ Health Study I. J Natl Cancer Inst. 2000;92:1178–1181. 49. Sung JJ, Chan FK, Leung WK, et al. Screening for colorectal cancer in Chinese: comparison of fecal occult blood test, flexible sigmoidoscopy, and colonoscopy. Gastroenterology. 2003;124:608–614. r
2014 Lippincott Williams & Wilkins
J Clin Gastroenterol
Volume 49, Number 1, January 2015
50. Botteri E, Iodice S, Raimondi S, et al. Cigarette smoking and adenomatous polyps: a meta-analysis. Gastroenterology. 2008; 134:388–395. 51. Nguyen SP, Bent S, Chen YH, et al. Gender as a risk factor for advanced neoplasia and colorectal cancer: a systematic review and meta-analysis. Clin Gastroenterol Hepatol. 2009;7:e1–e3. 52. Jacobs ET, Ahnen DJ, Ashbeck EL, et al. Association between body mass index and colorectal neoplasia at follow-up colonoscopy: a pooling study. Am J Epidemiol. 2009;169: 657–666.
r
2014 Lippincott Williams & Wilkins
Risk Stratification Model for Advanced Neoplasia in Korea
53. Fu Z, Shrubsole MJ, Smalley WE, et al. Lifestyle factors and their combined impact on the risk of colorectal polyps. Am J Epidemiol. 2012;176:766–776. 54. Ng SC, Lau JY, Chan FK, et al. Increased risk of advanced neoplasms among asymptomatic siblings of patients with colorectal cancer. Gastroenterology. 2013;144:544–550. 55. Rex DK, Petrini JL, Baron TH, et al. Quality indicators for colonoscopy. Am J Gastroenterol. 2006;101:873–885. 56. Vernon SW. Participation in colorectal cancer screening: a review. J Natl Cancer Inst. 1997;89:1406–1422.
www.jcge.com |
49
Development and Validation of a Risk Stratification-based Screening Model for Predicting Colorectal Advanced Neoplasia in Korea Dong Hyun Kim, MD,* Jae Myung Cha, MD, PhD,* Hyun Phil Shin, MD, PhD,* Kwang Ro Joo, MD, PhD,* Joung Il Lee, MD, PhD,* and Dong Il Park, MD, PhDw
Goals: To develop and validate a risk stratification-based screening model for predicting colorectal advanced neoplasia in Korea. Background: Colorectal advanced neoplasia is the relevant finding of screening colonoscopy. Risk estimation for advanced neoplasia may be helpful to improve compliance and to develop more costeffective approaches toward screening. Study: We developed Korean Colorectal Screening (KCS) score by optimizing and adjusting Asia-Pacific Colorectal Screening (APCS) score to predict advanced neoplasia in an asymptomatic Korean population who received screening colonoscopies from September 2006 to September 2009. Moreover, we validated the KCS score in another Korean cohort who received screening colonoscopies from October 2009 to February 2011. We also assessed the predictive power and diagnostic performance of both KCS and APCS scores. Results: There were 3561 subjects in the derivation cohort and 1316 subjects in the validation cohort, with a prevalence of advanced neoplasia of 4.7% and 4.3%, respectively. After a multivariate analysis, KCS was developed as 0 to 8 points comprising of age, sex, body mass index, smoking, and family history of CRC. Using KCS scores to stratify the validation cohort, the prevalences of advanced neoplasia in the 3 risk tiers (average, moderate, and high) were 2.0%, 3.7%, and 10.9%, respectively. Moderate-risk and high-risk tiers showed 2.1- and 6.5-fold increased prevalences, respectively, of advanced neoplasia compared with average risk tier. In addition, KCS score showed relatively good discriminative power (ROC = 0.681) and higher sensitivity compared with APCS score for the high-risk tier. Conclusions: KCS score may be clinically simple and useful for assessing advanced neoplasia risk in Korea. However, racial disparity should be considered in risk stratification-based screening in each country. Key Words: colorectal neoplasm, colonoscopy, risk assessment, screening
(J Clin Gastroenterol 2015;49:41–49)
C
olorectal cancer (CRC) is one of the most common malignancies whose incidence is steadily increasing in many Asian countries.1 In Korea, CRC is the third most common cancer and accounts for 12.8% of all cancers. The age-standardized incidence rate of CRC in Korea increased to 36.2/100,000 in 2010 from 30.8/100,000 in 2008.2 In general, CRC usually develops through an adenoma-carcinoma sequence, and adenomas with advanced features have been shown to exhibit higher risk.3 Screening for CRC, along with the removal of these premalignant lesions, is also well known to reduce CRC-related mortality and morbidity.4–6 Therefore, early detection and removal of premalignant lesions by CRC screening has been advocated in many countries.7,8 However, limited resources have been an obstacle to expand CRC screening.9–11 In this regard, risk stratification models may help make CRC screening more effective. Recently, the Asia-Pacific Working Group on CRC developed the Asia-Pacific Colorectal Screening (APCS) score, based on the age, sex, family history of CRC, and smoking status of selected asymptomatic Asian subjects.12 However, the APCS score was limited in that it did not include the body mass index (BMI); furthermore, it was based on a heterogenous population comprising multiple races from many Asian countries. Therefore, this scoring system does not take into account the findings that CRC risk varies significantly with ethnicity,1,13 dietary,14–17 and other factors. Therefore, an optimized risk stratification-based screening model, based on the CRC epidemiology and resource limitations of each country, may be useful in predicting CRC risk. The aim of this study was to develop and validate a risk stratification-based screening model predictive of colorectal advanced neoplasia in asymptomatic Korean subjects.
METHODS Study Population of Derivation Cohort Received for publication August 1, 2013; accepted January 22, 2014. From the *Department of Internal Medicine, Kyung Hee University School of Medicine; and wDepartment of Internal Medicine, Sungkyunkwan University School of Medicine, Seoul, Republic of Korea. All authors were involved in study concept and design, acquisition/ analysis/interpretation of data, and drafting and critical revision of the manuscript. The authors declare that they have nothing to disclose. Reprints: Jae Myung Cha, MD, PhD, Department of Internal Medicine, Kyung Hee University School of Medicine, Kyung Hee University Hospital at Gang Dong, 149 Sangil-dong, Gangdong-gu, Seoul 134-727, Republic of Korea (e-mail: [email protected]). Copyright r 2014 by Lippincott Williams & Wilkins
J Clin Gastroenterol
Volume 49, Number 1, January 2015
We developed a clinical risk score for CRC screening based on consecutive asymptomatic subjects who underwent a screening colonoscopy, at Health Promotion Center of Kyung Hee University Hospital in Gangdong, Seoul, Korea, as a part of medical check-up between September 2006 and September 2009. From this colonoscopy database, subjects were retrospectively identified as eligible if they were between 30 and 75 years of age, were asymptomatic, and had undergone a first-time screening colonoscopy. Potential subjects were excluded if they met any of the following criteria: (1) aged below 30 or above 75 years; (2) previous colorectal examinations (including www.jcge.com |
41
Kim et al
colonoscopy, sigmoidoscopy, or barium enema); (3) incomplete colonoscopy because of poor bowel preparation or cecal intubation failure; (4) history of CRC or inflammatory bowel disease; (5) history of colorectal surgery; or (6) symptoms or signs indicating the need for a colonoscopy (eg, changes in bowel habits, hematochezia, unexplained weight loss, or a positive fecal occult blood test). To avoid the potential protective effects of medications, subjects with a history of anti-inflammatory drug use (NSAIDs, aspirin, or statins) for >1 year on a regular basis were also excluded. Smoking habits, alcohol consumption, past medical history, and family history of CRC in a first-degree relative were determined by interviews conducted by well-trained nurses. A patient was defined as a current smoker if they consumed at least 1 pack per week; consumption of any amount of alcohol exceeding 140 g per week was considered a positive history of alcohol use. Diabetes mellitus (DM) was defined as a fasting glucose level of Z126 mg/dL or use of hypoglycemic agents and/or insulin. Height and body weight, used to calculate BMI, were routinely measured by trained nurses. This study was conducted according to the Declaration of Helsinki and was approved by the Institutional Review Board of Kyung Hee University Hospital at Gang Dong (KHNMC IRB 2012-105).
KCS Score Development The KCS score was developed in the same manner as the APCS score.12 In summary, a univariate analysis was carried out on the study population using the Pearson w2 method to examine the association between clinical risk factors and colorectal neoplasia. Variables associated with any neoplasia and advanced neoplasia in the univariate analyses (P < 0.15) were entered into multivariate logistic regression models. Risk factors that retained significance in multivariate analyses were incorporated into the risk score model. For each risk factor, weights were assigned using respective adjusted odds ratios (ORs) obtained by logistic regression. To simplify the risk score as much as possible, adjusted ORs were divided by 2 and then rounded to the nearest whole number. The final risk score was calculated by summing the scores of all individual risk factors. The KCS score was divided into 3 risk tiers: score 0 to 1, “average risk (AR)”; 2 to 3, “moderate risk (MR)”; and 4 to 8, “high risk (HR).” Score validities were assessed by receiver operating characteristics (ROC) analysis, and the predicted risks of advanced neoplasia were assessed by comparing risk tier groups.
KCS Score Validation for the Validation Cohort The sample size estimation for the validation cohort was based on methodology of APCS development.12 In the derivation cohort of the current study, the prevalences of individual risk factors ranged from 3.6% to 60.4% (mean 40%), and the prevalence of colorectal advanced neoplasia was 4.7%. We assumed the estimated prevalence of individual risk factors to be 40% and the estimated prevalence of advanced neoplasia to be 5.0% in the validation cohort, as recently implemented quality controls of screening colonoscopies in Korea will potentially increase the detection rate of advanced neoplasia. On the basis of these assumptions, at least 1300 asymptomatic subjects were required for a power of 80% to detect a risk factor with an OR of 2 at P < 0.05 level of significance.
42 | www.jcge.com
J Clin Gastroenterol
Volume 49, Number 1, January 2015
The validation cohort was an independent cohort, containing 1316 asymptomatic subjects undergoing screening colonoscopy between October 2009 and February 2011, and their data were retrospectively collected and analyzed to validate the KCS score. The colonoscopy and study protocols used for subjects in the validation cohort were identical to those used for subjects in the derivation cohort. Each subject in the validation group was assigned a personal risk score, as calculated by the KCS scoring method. The performance of the KCS score to predict the risk of advanced neoplasia was evaluated by determining the ORs for each risk tier, and the validity of the KCS score was assessed by ROC analysis.
APCS Score and Validation APCS score was also validated in our study population. The APCS score has a range of 0 to 7 points and is calculated as the sum of individual risk factors either present or absent in the individual: age below 50 years (0), between 50 and 69 years (2), above 70 years (3), male sex (1), female sex (0), family history of CRC in a first-degree relative present (2) or absent (0), current or past smoker (1), and nonsmoker (0).12 The APCS score stratifies patients into 3 risk tiers: score 0 to 1, AR; 2 to 3, MR; and 4 to 7, HR. After calculating the APCS scores, the predicted risks of advanced neoplasia were assessed by comparing the risk tier groups, and the validity of the APCS score was evaluated by ROC analysis. The diagnostic performances of both the KCS and APCS scores were also compared. With respect to diagnostic performance, sensitivity is the probability of subjects with advanced neoplasia who had a high risk on score model, whereas specificity is the probability of subjects without advanced neoplasia who do not have a high risk.
Colonoscopy All patients underwent a 3-day dietary restriction and drank 4 L of polyethylene glycol solution for 12 hours before the colonoscopy, until clear rectal fluid was evacuated. Potential adverse events were explained to the patients before the colonoscopy, and all patients provided written consent before the procedure. For conscious sedative colonoscopy, an individualized dose of midazolam and/ or propofol was given by the gastroenterologist based on the age, weight, and general condition of the patient. An antispasmodic medication (10 mg intravenous hyoscine methobromide) was also administered to subjects when no contraindications were present. All examinations were performed using a standard video colonoscope (EC-590 ZW/L, Fujinon Inc., Saitama, Japan) in a standardized manner by experienced gastroenterologists. The quality of the bowel preparation was assessed as excellent, good, fair, or poor.18 All detected polyps were biopsied, except for multiple hyperplastic polyps in the rectosigmoid colon showing endoscopic features consistent with a gross hyperplastic appearance (small size, sessile shape, pale color, and type 1 or 2 pit pattern).19 According to the regulations of our hospital, all detected polyps were imaged and all polyp characteristics, such as shape, size, number, and location, were documented. Shapes were classified according to the Paris classification system.20 In addition, polyp sizes were estimated by comparing the polyps to open biopsy forceps (Olympus FB-28U-1; Amori Olympus Co. Ltd, Amori, Japan). All excised polyps were evaluated histologically by r
2014 Lippincott Williams & Wilkins
J Clin Gastroenterol
Volume 49, Number 1, January 2015
gastrointestinal pathologists. Colorectal adenoma was defined as an adenoma, irrespective of its grade or villous components, whereas advanced neoplasia was defined as a colorectal carcinoma or advanced adenoma (diameter Z10 mm, highgrade dysplasia, or >25% villous features).12,21
Risk Stratification Model for Advanced Neoplasia in Korea
were performed using the Statistical Package for the Social Sciences, version 18.0 (SPSS, Chicago, IL).
RESULTS
Statistical Analysis
Characteristics of the Derivation and Validation Cohorts
All data are presented as mean ± SD for continuous variables and as numbers (percentages) of subjects for categorical variables. Categorical variables, including candidate risk factors such as age, sex, BMI, smoking status, alcohol consumption, DM status, and family history of CRC in a first-degree relative, were compared using the Pearson w2 test. Multiple logistic regressions was used to control for confounding variables and to determine the independent effects of risk factors for either any type of neoplasia, or advanced neoplasia. ORs were calculated for dichotomized endpoints; values are presented with modelbased lower and upper 95% confidence intervals (CIs). The predictive powers of both score models were analyzed by determining the areas under the corresponding ROC curves, including the 95% CIs. The Hosmer-Lemeshow goodness-of-fit statistic was used to assess reliability of the model. To compare the statistical ability of the APCS and the KCS scores to predict the risk of advanced neoplasia, ROC analysis and the McNemar tests were performed. All P-values were 2-tailed, a P-value 1 year (n = 159); and insufficient patient data (n = 39). After filtering, 3561 asymptomatic subjects were enrolled in the study cohort. The mean age was 51.3 ± 9.0 years and the mean BMI was 23.8 ± 3.1 kg/ m2. In the derivation population, 2152 patients were male patients (60.4%), 34.4% were obese, 47.7% were smokers, and 3.6% had a family history of CRC in a first-degree relative. With respect to pathologic findings, 841 (23.6%) were found to have any neoplasia, of which 156 patients (4.4% of the total) had advanced neoplasia and 13 patients (0.4% of the total) had CRC. The characteristics of the study population with any neoplasia and advanced neoplasia are shown in Table 1.
FIGURE 1. Study enrollment. A, In this study, 3561 asymptomatic subjects of the 3970 subjects retrospectively enrolled were placed in the derivation cohort. B, 1316 asymptomatic subjects from an independent cohort were also retrospectively enrolled after exclusion of 89 subjects and included in the validation cohort. These asymptomatic subjects also had undergone screening colonoscopies. The study protocols for the derivation and validation cohorts were identical. r
2014 Lippincott Williams & Wilkins
www.jcge.com |
43
J Clin Gastroenterol
Kim et al
Volume 49, Number 1, January 2015
TABLE 1. Characteristics of the Derivation Cohort According to Risk Factors
Any Neoplasia (n = 841, 23.6%)
Advanced Neoplasia (n = 169, 4.7%)
All Subjects, n (%)
n (%)
P
n (%)
P
2152 (60.4) 1409 (39.6)
639 (76.0) 202 (24.0)
< 0.001
142 (84.0) 27 (16.0)
< 0.001
1513 (42.5) 2048 (57.5)
225 (26.8) 616 (73.2)
< 0.001
23 (13.6) 146 (86.4)
< 0.001
2372 (66.6) 1189 (33.4)
487 (57.9) 354 (42.1)
< 0.001
92 (55.1) 77 (44.9)
0.001
127 (3.6) 3434 (96.4)
37 (4.4) 804 (95.6)
0.136
10 (5.9) 159 (94.1)
0.091
1863 (52.3) 1698 (47.7)
318 (37.8) 523 (62.2)
< 0.001
46 (27.2) 123 (72.8)
< 0.001
2429 (68.2) 1132 (31.8)
504 (59.9) 337 (40.1)
< 0.001
91 (53.8) 78 (46.2)
< 0.001
3409 (95.7) 152 (4.3)
792 (94.2) 49 (5.8)
0.011
158 (93.5) 11 (6.5)
0.140
Sex Male Female Age (y) < 50 Z50 BMI (kg/m2) < 25 Z25 FHx of CRC Present Absent Smoking Never Current or ex Alcohol No Yes Diabetes No Yes
BMI indicates body mass index; CRC, colorectal cancer; FHx, family history.
For the validation cohort, 1316 asymptomatic subjects were enrolled after exclusion of 89 subjects because of the following reasons: aged below 30 or above 75 years (n = 38); prior colonoscopy, sigmoidoscopy, or barium enema (n = 9); poor bowel preparation (n = 10); history of CRC (n = 5); history of colorectal surgery (n = 5); regular use of NSAIDs, aspirin, or statins for >1 year (n = 16); and insufficient patient data (n = 6). The mean age was 49.6 ± 9.9 years and the mean BMI was 23.8 ± 3.2 kg/m2. In the validation population, 557 patients were male patients (42.3%), 36.3% were obese, 19.0% were smokers, and 2.9% had a family history of CRC in a first-degree relative. With respect to pathologic findings, 405 (30.8%) were found to have any neoplasia, of which 56 patients (4.3% of the total) had advanced neoplasia.
Univariate and Multivariate Analysis of Predictors for Colorectal Neoplasia in the Derivation Cohort Univariate and multivariate analyses were performed for each risk factor. Multivariate logistic regression showed that patients aged 50 years or above, male sex, BMIZ25 kg/m2, positive family history of CRC in a first-degree relative, and smoking status were significant risk factors for any neoplasia, with ORs (95% CI) of 2.6 (2.2 to 3.1), 1.8 (1.4 to 2.3), 1.3 (1.1 to 1.6), 1.6 (1.1 to 2.4), and 1.6 (1.2 to 1.8), respectively (Table 2). Age above 50 years, male sex, positive family history in a first-degree relative, and smoking status were also significant risk factors for advanced neoplasia, with ORs (95% CI) of 5.3 (3.4 to 8.3), 1.9 (1.1 to 3.4), 2.2 (1.1 to 4.4), and 2.0 (1.3 to 3.2); however, BMI only showed the trend of a risk factor (P = 0.092), with an OR of 1.3 (1.0 to 1.8) (Table 2). The Hosmer-Lemeshow goodness-of-fit statistic was P = 0.81 for the derivation study cohort.
44 | www.jcge.com
Results of the KCS Score in the Derivation Cohort To derive the KCS score, points were assigned for each risk factor of advanced neoplasia as follows: age below 50 years (0), between 50 and 69 years (2), 70 years or above(4), male sex (1), female sex (0), family history of colorectal cancer in a first-degree relative present (1) or absent (0), nonsmoker (0), current or past smoker (1), and BMI < 25 kg/m2 (0) and Z25 kg/m2 (1). One point was assigned for the presence of obesity as it was a significant risk factor for colorectal neoplasia in the previous literature,22,23 although it did not reach significance for advanced neoplasia in our study population. The KCS score ranges from 0 to 8 points, and the frequency distribution of KCS scores by subject is shown in Table 3. According to the KCS score stratification, the AR tier contained 683 (19.2%) subjects, the MR tier contained 1762 (49.5%) subjects, and the HR tier contained 1116 (31.3%) subjects. The prevalences of advanced neoplasia in the 3 tiers were: 0.7% (95% CI, 0.15%-1.46%) in the AR tier, 2.2% (95% CI, 1.48%-2.84%) in the MR tier, and 11.3% (95% CI, 9.50%-13.17%) in the HR tier. Subjects in the MR and HR tiers had 3.0-fold (95% CI, 1.2-7.6) and 17.3-fold (95% CI, 7.0-42.4) increased rates of advanced neoplasia, respectively, compared with subjects in the AR tier. The Hosmer-Lemeshow goodness-of-fit statistic had a P-value of 0.51.
Risk Stratification of KCS and APCS Scores in the Validation Cohort According to their KCS score, 32.6%, 52.1%, and 15.3% of the subjects were placed into the AR, MR, and HR tiers, respectively (Table 4). The prevalences of advanced neoplasia in the 3 tiers were 1.9%, 3.8%, and 10.9%, respectively. Subjects in the MR and HR tiers had 2.1-fold (95% CI, 0.9-4.6) and 6.6-fold (95% CI, 2.8-14.8) r
2014 Lippincott Williams & Wilkins
J Clin Gastroenterol
Volume 49, Number 1, January 2015
Risk Stratification Model for Advanced Neoplasia in Korea
TABLE 2. Univariate and Multivariate Predictors of Colorectal Neoplasia in the Derivation Cohort
Univariate Analysis Risk Factors Predictors of any neoplasia Sex Age (y) 50-69 Z70 BMIZ25 kg/m2 FHx of CRC Smoking Alcohol Diabetes Predictors of advanced neoplasia Sex Age (y) 50-69 Z70 BMIZ25 kg/m2 FHx of CRC Smoking Alcohol Diabetes
Multivariate Analysis
OR (95% CI)
P
OR (95% CI)
P
2.5 (2.1-3.0)
< 0.001
1.8 (1.4-2.3)
< 0.001
2.4 4.1 1.6 1.3 2.2 1.6 1.6
(2.0-2.9) (2.6-6.4) (1.4-1.9) (0.9-2.0) (1.8-2.5) (1.4-1.9) (1.1-2.2)
< 0.001 < 0.001 < 0.001 0.134 < 0.001 < 0.001 0.023
2.6 4.8 1.3 1.6 1.6 1.1 1.0
(2.2-3.1) (3.0-7.6) (1.1-1.6) (1.1-2.4) (1.2-1.8) (0.9-1.4) (0.7-1.5)
< 0.001 < 0.001 0.001 0.027 0.001 0.187 0.811
3.6 (2.4-5.5)
< 0.001
1.9 (1.1-3.4)
0.022
4.8 9.1 1.7 1.8 3.1 1.9 1.6
< 0.001 < 0.001 0.001 0.096 < 0.001 < 0.001 0.143
(3.1-7.5) (4.3-19.4) (1.3-2.3) (0.9-3.4) (2.2-4.4) (1.4-2.6) (0.9-3.0)
5.3 11.8 1.3 2.2 2.0 1.3 0.9
(3.4-8.3) (5.4-25.7) (1.0-1.8) (1.1-4.4) (1.3-3.2) (0.9-1.8) (0.5-1.8)
< 0.001 < 0.001 0.092 0.030 0.004 0.170 0.876
BMI indicates body mass index; CI, confidence interval; CRC, colorectal cancer; FHx, family history; OR, odds ratio.
increased rates of advanced neoplasia, compared with subjects in the AR tier (P < 0.001). The Hosmer-Lemeshow goodness-of-fit statistic had a P-value of 0.48. By using their APCS scores to stratify the subjects, 37.6%, 53.5%, and 8.9% were included in the AR, MR, and HR tiers, respectively (Table 4). The prevalences of advanced neoplasia in the 3 tiers were 2.4%, 4.0%, and
13.7%, respectively. When compared with subjects in the AR tier, those in the MR and HR tiers had 1.7-fold (95% CI, 0.83.3) and 6.4-fold (95% CI, 2.9-13.9) increased rates of advanced neoplasia, respectively (P < 0.001). The Hosmer-Lemeshow goodness-of-fit statistic had a P-value of 0.13.
Diagnostic Performances of the KCS and APCS Scores in the Validation Cohort TABLE 3. Score and Risk Tier Distribution of the APCS and KCS Score in the Derivation Cohort
No. Subjects (%) APCS score 0 1 2 3 4 5 6 7 Total KCS score 0 1 2 3 4 5 6 7 8 Total
515 225 1474 373 902 43 29
No. Any Neoplasia (%)
(14.5) (6.3) (41.4) (10.5) (25.3) (1.2) (0.8) 0 3561 (100)
(4.0) (3.6) (33.9) (13.6) (40.7) (2.5) (1.8) 0 841 (100)
453 230 1060 702 665 395 43 13
26 27 190 172 232 164 22 8
(12.7) (6.5) (29.8) (19.7) (18.7) (11.1) (1.2) (0.4) 0 3561 (100)
34 30 285 114 342 21 15
(3.1) (3.2) (22.6) (20.5) (27.6) (19.5) (2.6) (1.0) 0 841 (100)
No. Advanced Neoplasia (%) 4 3 30 22 97 7 6
(2.4) (1.8) (17.8) (13.0) (57.4) (4.1) (3.6) 0 169 (100) 4 1 22 21 62 48 7 4
(2.4) (0.6) (13.0) (12.4) (36.7) (28.4) (4.1) (2.4) 0 169 (100)
APCS indicates Asia-Pacific Colorectal Screening; KCS, Korean Colorectal Screening.
r
2014 Lippincott Williams & Wilkins
Comparison of the diagnostic performances of the KCS and APCS scores using the McNemar test revealed that the KCS score exhibited statistically higher sensitivity
TABLE 4. Prevalence of Advanced Neoplasia by Risk Tier in the Validation Cohort According to KCS and APCS Score
Risk Tier
No. Subjects (%)
KCS score Average 429 (0-1) Moderate 686 (2-3) High (4-8) 201 Total 1316 APCS score Average 495 (0-1) Moderate 704 (2-3) High (4-7) 117 Total 1316
Advanced Neoplasia Number (%) (95% CI)
RR (95% CI)
(32.6)
8 (1.9) (0.55-3.17)
Reference
(52.1)
26 (3.8) (2.33-5.25)
2.1 (0.9-4.6)
(15.3) (100)
22 (10.9) (6.53-15.37) 56 (4.3) (3.15-5.36)
6.5 (2.8-14.8)
(37.6)
12 (2.4) (1.04-3.80)
Reference
(53.5)
28 (4.0) (2.51-5.45)
1.7 (0.8-3.3)
(8.9) (100)
16 (13.7) (7.30-20.06) 56 (4.3) (3.15-5.36)
6.4 (2.9-13.9)
APCS indicates Asia-Pacific Colorectal Screening; CI, confidence interval; KCS, Korean Colorectal Screening.
www.jcge.com |
45
J Clin Gastroenterol
Kim et al
TABLE 5. Diagnostic Performance of the APCS and KCS Score for Detection of Advanced Colorectal Neoplasia in the Validation Cohort
KCS (%)
APCS (%)
P
39.3 85.8
28.6 92.0
0.031 < 0.001
85.7 33.4
78.6 38.3
0.125 < 0.001
HR vs. MR + AR Sensitivity Specificity MR + HR vs. AR Sensitivity Specificity
APCS indicates Asia-Pacific Colorectal Screening; AR, average risk; HR, high risk; KCS, Korean Colorectal Screening; MR, moderate risk.
compared with the APCS score when the HR tier was compared with MR plus AR tier group (39.3% vs. 28.6%, P = 0.031) (Table 5). In contrast, the APCS score showed statistically higher specificity compared with the KCS score when the HR tier was compared with MR plus AR tier group (P < 0.001). By ROC analysis, KCS and APCS scores were 0.681 (95% CI, 0.608-0.763) and 0.676 (95% CI, 0.600-0.751), respectively, indicating relatively good discrimination, but no significant differences were noted (P = 0.86) (Fig. 2).
DISCUSSION In this study, we developed the KCS score comprising age, sex, BMI, smoking, and family history of CRC to predict the risk of advanced neoplasia after modifying the APCS score and validated the KCS score through a validation cohort whose members underwent screening
FIGURE 2. ROC curves for the KCS (solid line) and APCS (dotted line) score models. The curves for each score showed a similar pattern, and the difference between them was not statistically significant (0.675 vs. 0.670; P = 0.69). APCS indicates Asia-Pacific Colorectal Screening; KCS, Korean Colorectal Screening; ROC, receiver operating characteristic analysis.
46 | www.jcge.com
Volume 49, Number 1, January 2015
colonoscopies. In our study, KCS score showed good discriminatory power measured by ROC and could be applied to identify the high-risk subjects for advanced neoplasia. Subjects in the HR tier showed a 6.0-fold increase in their chance of having advanced neoplasia detected, according to their KCS scores, when compared with subjects in the AR tier. We developed the KCS score to overcome certain limitations of the APCS score, thereby making it more suitable for use in the Korean population. For example, BMI was not included as a risk factor contributing to the APCS score; furthermore, a variety of populations with substantial variations in their risk factors were included in the study population used to derive the APCS score.12 Compared with the APCS score, the KCS score model includes BMI as a risk factor, as well as assigns different points for some risk factors based on a Korean derivation cohort (that is, 4 points are assigned in the KCS vs. 3 points in the APCS for patients aged 70 years or older, and 1 point is assigned in the KCS vs. 2 points in the APCS for a family history of CRC). As sensitivity is more important than specificity in screening for significant diseases such as CRC or prevalent diseases such as advanced neoplasia, the KCS score, which showed higher sensitivity, may be a more appropriate option compared with the APCS score, at least in Korea. Although the APCS score displayed higher specificity, the proportion of subjects stratified into the HR tier was only 8.7% in our validation cohort. Therefore, our data indicate that the KCS score is optimized for use with an asymptomatic Korean population. CRC screening has been shown to reduce CRC incidence and mortality24–26; however, implementation of CRC screening is often hampered by limited resources, lack of awareness in the target population, insufficient advocacy by healthcare professionals, and poor patient compliance.27–32 Risk stratification-based screening, including KCS, may improve compliance of current CRC screening methods by providing motivation for those who are assigned high-risk scores. According to the “Health Belief” model, perceived risk is positively associated with both CRC screening utilization and intention to be screened. Therefore, compliance of CRC screening may be increased when participants perceive themselves as highly susceptible to colorectal advanced neoplasia.33–39 In addition, risk stratification-based screening may enable more cost-effective approaches by prioritizing CRC screening in the high-risk group, especially in countries with limited resources. Risk stratification-based screening should be adjusted after careful consideration of CRC epidemiology and the available health resources of each country. For example, less invasive and inexpensive screening processes, such as fecal blood screening, may be recommended for individuals in the AR tier, whereas more thorough screening processes, such as colonoscopy screening, may be recommended for those in the HR tier.40 In the present study, 39.3% of advanced neoplasia could be detected in the HR group, which accounts for only 15.6% of total population. Furthermore, 85.7% of advanced neoplasia could be detected with colonoscopy screening in the MR and HR groups, which accounts for 67.4% of the total population. Until now, few efforts to develop risk prediction models for advanced neoplasia have been reported. Most risk stratification-based screening models have been developed in the West. Imperiale et al41 proposed a clinical index to estimate the risk of proximal advanced neoplasia; this index was based on age, sex, and the distal findings from r
2014 Lippincott Williams & Wilkins
J Clin Gastroenterol
Volume 49, Number 1, January 2015
sigmoidoscopy. However, this index requires an initial sigmoidoscopy to calculate the index score. Betes et al42 described an index taking into account age, sex, and BMI to stratify subjects by risk for advanced neoplasia; however, this scoring system did not include smoking status or family history. Lin et al40 proposed an index to stratify subjects into a high-risk group for colonoscopy screening based on age, sex, and family history, without including smoking status or BMI, and indicating different screening tools according to the various risk groups. In the East, 2 risk stratification-based screening models have also been suggested. Cai et al43 proposed a risk stratification tool for Chinese patients of average risk based on their age, sex, smoking status, DM status, and dietary habits. Although this tool showed good discriminatory power (ROC = 0.74), it was limited in the sense that dietary habits are relatively subjective and complex to measure, in particular aspects such as duration and quantity of the subject’s diet. In the recently reported APCS scoring system, the risk of advanced neoplasia was evaluated based on age, sex, family history, and smoking status.12 However, the APCS score showed some limitations, as described above, and was based on colonoscopy data with low adenoma detection rate (ADR). As the risk of CRC varies significantly with ethnicity1,13 and diet,14–17 racial disparity should be considered in risk stratification-based screening. However, only 44.4% of the subjects in the derivation cohort for the APCS score were Korean patients.12,44 Moreover, Korean patients showed a 2-fold higher risk of colorectal neoplasia than Chinese patients in a multinational, multicenter study,44 and age was a stronger risk factor for advanced neoplasia in Korean patients than in Chinese patients.43,45,46 In this regard, it is difficult to develop universal risk stratificationbased screening models appropriate for all countries; therefore, the discriminatory power of each risk stratification-based screening model should be validated for each target population. There are several advantages to our study. First, this is the first study reporting country-specific modifications to the APCS score. Moreover, our score has a clinical implication in that its risk factors are convenient to measure and are easily applied. Furthermore, all risk factors used in the KCS score have been shown to be relevant risk factors for advanced neoplasia in previous studies.22,43,46–54 Second, our study was based on a large number of asymptomatic Korean patients: 3561 asymptomatic subjects were included in the derivation cohort and 1316 asymptomatic subjects were included in the validation cohort. Third, data collected in this study were also of high quality, despite the study’s retrospective design. The level of experience of the colonoscopists used in our study is reflected by their 100% cecal intubation rate, their 24% ADR in the derivation cohort, and 31% ADR in the validation cohort. These results are improved compared with previous studies of colonoscopy quality indicators.55 Furthermore, subjects with poor bowel preparations were excluded from data analysis in our study. Considering the low ADR (18.7%) found in a previous study of many Asian countries,12 our risk score model is the first one to be based on high-quality colonoscopy data. Finally, this study may improve compliance and costeffectiveness of CRC screening considering a clinical risk score, as CRC screening is currently underused despite its apparent effectiveness.56 Adults older than 50 years of age in the HR tier may be more encouraged to undergo CRC screening, as they exhibit an approximately 3.9-fold r
2014 Lippincott Williams & Wilkins
Risk Stratification Model for Advanced Neoplasia in Korea
increased risk of advanced neoplasia compared with subjects in the MR tier (relative risk, 3.50; 95% CI, 1.80-6.82; P < 0.001). Furthermore, adults younger than 50 years of age may be recommended to undergo CRC screening if their scores place them in either the MR or HR tiers, as these groups had an approximately 2.3-fold increased risk of advanced neoplasia compared with the AR tier (relative risk, 2.53; 95% CI, 0.98-6.51; P = 0.054). We concede that one of the limitations of our study is that it was a single-center study; furthermore, participants were more likely to be health-conscious and economically affluent, which may limit the generalization of our findings. As another limitation, the retrospective nature of the study design somewhat limited the data that could be collected. However, considering the extensive collection of personal information by well-trained nurses before health screening in our hospital, such data limitations were likely minimal. Third, the diagnostic performance of our risk score model was suboptimal; however, at the time we designed the model, accepted criteria for CRC screening scores to predict advanced neoplasia were not yet available. Furthermore, the diagnostic performance of our model should be interpreted differently in each country. As risk stratification-based screening models can be optimized for each country by taking into account differences in colorectal neoplasia prevalences and health resources available for screening, our study may help physicians in other countries develop appropriate risk stratification-based screening models for their particular countries. Finally, asymptomatic subjects younger than 50 years of age were included in this study. Although CRC screening is not recommended for subjects of this age, we included their data in the model derivation to determine the value of using a risk stratification model for this young population. In summary, the KCS score is a clinically simple and useful parameter for predicting advanced neoplasia in asymptomatic Korean patients. However, racial disparity should be considered in risk stratification-based screening in individual countries. Risk stratification-based screening deserves to be explored further for its potential benefits and cost-effectiveness in each country. REFERENCES 1. Sung JJ, Lau JY, Goh KL, et al. Increasing incidence of colorectal cancer in Asia: implications for screening. Lancet Oncol. 2005;6:871–876. 2. Jung KW, Park S, Kong HJ, et al. Cancer statistics in Korea: incidence, mortality, survival, and prevalence in 2009. Cancer Res Treat. 2012;44:11–24. 3. Atkin WS, Morson BC, Cuzick J. Long-term risk of colorectal cancer after excision of rectosigmoid adenomas. N Engl J Med. 1992;326:658–662. 4. Winawer SJ, Zauber AG, Ho MN, et al. Prevention of colorectal cancer by colonoscopic polypectomy. The National Polyp Study Workgroup. N Engl J Med. 1993;329:1977–1981. 5. Citarda F, Tomaselli G, Capocaccia R, et al. Efficacy in standard clinical practice of colonoscopic polypectomy in reducing colorectal cancer incidence. Gut. 2001;48:812–815. 6. Zauber AG, Winawer SJ, O’Brien MJ, et al. Colonoscopic polypectomy and long-term prevention of colorectal-cancer deaths. N Engl J Med. 2012;366:687–696. 7. Levin B, Lieberman DA, McFarland B, et al. Screening and surveillance for the early detection of colorectal cancer and adenomatous polyps, 2008: a joint guideline from the American Cancer Society, the US Multi-Society Task Force on Colorectal Cancer, and the American College of Radiology. Gastroenterology. 2008;134:1570–1595.
www.jcge.com |
47
Kim et al
8. Sung JJ, Lau JY, Young GP, et al. Asia Pacific consensus recommendations for colorectal cancer screening. Gut. 2008;57:1166–1176. 9. Levin TR. Colonoscopy capacity: can we build it? Will they come? Gastroenterology. 2004;127:1841–1844. 10. Brown ML, Klabunde CN, Mysliwiec P. Current capacity for endoscopic colorectal cancer screening in the United States: data from the National Cancer Institute Survey of Colorectal Cancer Screening Practices. Am J Med. 2003;115: 129–133. 11. Rex DK, Lieberman DA. Feasibility of colonoscopy screening: discussion of issues and recommendations regarding implementation. Gastrointest Endosc. 2001;54:662–667. 12. Yeoh KG, Ho KY, Chiu HM, et al. The Asia-Pacific Colorectal Screening score: a validated tool that stratifies risk for colorectal advanced neoplasia in asymptomatic Asian subjects. Gut. 2011;60:1236–1241. 13. Leung WK, Ho KY, Kim WH, et al. Colorectal neoplasia in Asia: a multicenter colonoscopy survey in symptomatic patients. Gastrointest Endosc. 2006;64:751–759. 14. Otani T, Iwasaki M, Yamamoto S, et al. Alcohol consumption, smoking, and subsequent risk of colorectal cancer in middleaged and elderly Japanese men and women: Japan Public Health Center-based prospective study. Cancer Epidemiol Biomarkers Prev. 2003;12:1492–1500. 15. Ho JW, Lam TH, Tse CW, et al. Smoking, drinking and colorectal cancer in Hong Kong Chinese: a case-control study. Int J Cancer. 2004;109:587–597. 16. Toyomura K, Yamaguchi K, Kawamoto H, et al. Relation of cigarette smoking and alcohol use to colorectal adenomas by subsite: the self-defense forces health study. Cancer Sci. 2004;95:72–76. 17. Tsong WH, Koh WP, Yuan JM, et al. Cigarettes and alcohol in relation to colorectal cancer: the Singapore Chinese Health Study. Br J Cancer. 2007;96:821–827. 18. Rostom A, Jolicoeur E. Validation of a new scale for the assessment of bowel preparation quality. Gastrointest Endosc. 2004;59:482–486. 19. Kudo S, Hirota S, Nakajima T, et al. Colorectal tumours and pit pattern. J Clin Pathol. 1994;47:880–885. 20. Participants in the Paris Workshop. The Paris endoscopic classification of superficial neoplastic lesions: esophagus, stomach, and colon: November 30 to December 1, 2002. Gastrointest Endosc. 2003;58:S3–43. 21. Lieberman DA, Rex DK, Winawer SJ, et al. Guidelines for colonoscopy surveillance after screening and polypectomy: a consensus update by the US Multi-Society Task Force on Colorectal Cancer. Gastroenterology. 2012;143:844–857. 22. Kim SE, Shim KN, Jung SA, et al. An association between obesity and the prevalence of colonic adenoma according to age and gender. J Gastroenterol. 2007;42:616–623. 23. Kim BC, Shin A, Hong CW, et al. Association of colorectal adenoma with components of metabolic syndrome. Cancer Causes Control. 2012;23:727–735. 24. Atkin WS, Cuzick J, Northover JM, et al. Prevention of colorectal cancer by once-only sigmoidoscopy. Lancet. 1993;341:736–740. 25. Winawer SJ, Flehinger BJ, Schottenfeld D, et al. Screening for colorectal cancer with fecal occult blood testing and sigmoidoscopy. J Natl Cancer Inst. 1993;85:1311–1318. 26. Mandel JS, Church TR, Bond JH, et al. The effect of fecal occult-blood screening on the incidence of colorectal cancer. N Engl J Med. 2000;343:1603–1607. 27. Gimeno-Garcia AZ, Quintero E, Nicolas-Perez D, et al. Impact of an educational video-based strategy on the behavior process associated with colorectal cancer screening: a randomized controlled study. Cancer Epidemiol. 2009;33: 216–222. 28. Domati F, Travlos E, Cirilli C, et al. Attitude of the Italian general population towards prevention and screening of the most common tumors, with special emphasis on colorectal malignancies. Intern Emerg Med. 2009;4:213–220.
48 | www.jcge.com
J Clin Gastroenterol
Volume 49, Number 1, January 2015
29. Cai SR, Zhang SZ, Zhu HH, et al. Barriers to colorectal cancer screening: a case-control study. World J Gastroenterol. 2009;15:2531–2536. 30. Emmons K, Puleo E, McNeill LH, et al. Colorectal cancer screening awareness and intentions among low income, sociodemographically diverse adults under age 50. Cancer Causes Control. 2008;19:1031–1041. 31. Kamposioras K, Mauri D, Alevizaki P, et al. Cancer screening in Greece. Guideline awareness and prescription behavior among Hellenic physicians. Eur J Intern Med. 2008;19: 452–460. 32. Inadomi JM. Taishotoyama Symposium Barriers to colorectal cancer screening: economics, capacity and adherence. J Gastroenterol Hepatol. 2008;23(suppl 2):S198–S204. 33. Frame PS, Kowulich BA. Stool occult blood screening for colorectal cancer. J Fam Pract. 1982;15:1071–1075. 34. Farrands PA, Hardcastle JD, Chamberlain J, et al. Factors affecting compliance with screening for colorectal cancer. Community Med. 1984;6:12–19. 35. Kelly RB, Shank JC. Adherence to screening flexible sigmoidoscopy in asymptomatic patients. Med Care. 1992;30: 1029–1042. 36. Price JH. Perceptions of colorectal cancer in a socioeconomically disadvantaged population. J Community Health. 1993;18: 347–362. 37. Lewis SF, Jensen NM. Screening sigmoidoscopy. Factors associated with utilization. J Gen Intern Med. 1996;11:542–544. 38. Strecher VJ, Rosenstock IM. The health belief model. In: Glanz K, Lewis FM, Rimer BK, eds. Health Behavior and Health Education. San Francisco, CA: Jossey-Bass; 1997:41–59. 39. Watts BG, Vernon SW, Myers RE, et al. Intention to be screened over time for colorectal cancer in male automotive workers. Cancer Epidemiol Biomarkers Prev. 2003;12: 339–349. 40. Lin OS, Kozarek RA, Schembre DB, et al. Risk stratification for colon neoplasia: screening strategies using colonoscopy and computerized tomographic colonography. Gastroenterology. 2006;131:1011–1019. 41. Imperiale TF, Wagner DR, Lin CY, et al. Using risk for advanced proximal colonic neoplasia to tailor endoscopic screening for colorectal cancer. Ann Intern Med. 2003;139: 959–965. 42. Betes M, Munoz-Navas MA, Duque JM, et al. Use of colonoscopy as a primary screening test for colorectal cancer in average risk people. Am J Gastroenterol. 2003;98:2648–2654. 43. Cai QC, Yu ED, Xiao Y, et al. Derivation and validation of a prediction rule for estimating advanced colorectal neoplasm risk in average-risk Chinese. Am J Epidemiol. 2012;175: 584–593. 44. Byeon JS, Yang SK, Kim TI, et al. Colorectal neoplasm in asymptomatic Asians: a prospective multinational multicenter colonoscopy survey. Gastrointest Endosc. 2007;65: 1015–1022. 45. Chen HM, Weng YR, Jiang B, et al. Epidemiological study of colorectal adenoma and cancer in symptomatic patients in China between 1990 and 2009. J Dig Dis. 2011;12:371–378. 46. Choe JW, Chang HS, Yang SK, et al. Screening colonoscopy in asymptomatic average-risk Koreans: analysis in relation to age and sex. J Gastroenterol Hepatol. 2007;22:1003–1008. 47. Pariente A, Milan C, Lafon J, et al. Colonoscopic screening in first-degree relatives of patients with “sporadic” colorectal cancer: a case-control study. The Association Nationale des Gastroenterologues des Hopitaux and Registre Bourguignon des Cancers Digestifs (INSERM CRI 9505). Gastroenterology. 1998;115:7–12. 48. Sturmer T, Glynn RJ, Lee IM, et al. Lifetime cigarette smoking and colorectal cancer incidence in the Physicians’ Health Study I. J Natl Cancer Inst. 2000;92:1178–1181. 49. Sung JJ, Chan FK, Leung WK, et al. Screening for colorectal cancer in Chinese: comparison of fecal occult blood test, flexible sigmoidoscopy, and colonoscopy. Gastroenterology. 2003;124:608–614. r
2014 Lippincott Williams & Wilkins
J Clin Gastroenterol
Volume 49, Number 1, January 2015
50. Botteri E, Iodice S, Raimondi S, et al. Cigarette smoking and adenomatous polyps: a meta-analysis. Gastroenterology. 2008; 134:388–395. 51. Nguyen SP, Bent S, Chen YH, et al. Gender as a risk factor for advanced neoplasia and colorectal cancer: a systematic review and meta-analysis. Clin Gastroenterol Hepatol. 2009;7:e1–e3. 52. Jacobs ET, Ahnen DJ, Ashbeck EL, et al. Association between body mass index and colorectal neoplasia at follow-up colonoscopy: a pooling study. Am J Epidemiol. 2009;169: 657–666.
r
2014 Lippincott Williams & Wilkins
Risk Stratification Model for Advanced Neoplasia in Korea
53. Fu Z, Shrubsole MJ, Smalley WE, et al. Lifestyle factors and their combined impact on the risk of colorectal polyps. Am J Epidemiol. 2012;176:766–776. 54. Ng SC, Lau JY, Chan FK, et al. Increased risk of advanced neoplasms among asymptomatic siblings of patients with colorectal cancer. Gastroenterology. 2013;144:544–550. 55. Rex DK, Petrini JL, Baron TH, et al. Quality indicators for colonoscopy. Am J Gastroenterol. 2006;101:873–885. 56. Vernon SW. Participation in colorectal cancer screening: a review. J Natl Cancer Inst. 1997;89:1406–1422.
www.jcge.com |
49
