Identification of subjects at risk of proximal advanced neoplasia for colorectal cancer screening

European Journal of Cancer (2015) 51, 37– 44

Available at www.sciencedirect.com

ScienceDirect journal homepage: www.ejcancer.com

Identification of subjects at risk of proximal advanced neoplasia for colorectal cancer screening q Martin C.S. Wong a,b, Jessica Y.L. Ching a, Victor C.W. Chan a, Thomas Y.T. Lam a, Arthur K.C. Luk a, Sunny H. Wong a, Siew C. Ng a, Simon S.M. Ng a, Justin C.Y. Wu a, Francis K.L. Chan a, Joseph J.Y. Sung a,⇑ a Institute of Digestive Disease, Chinese University of Hong Kong, 7/F, Lui Che Woo Clinical Science Building, Prince of Wales Hospital, Shatin, NT, Hong Kong Special Administrative Region b School of Public Health and Primary Care, Chinese University of Hong Kong, 4/F, School of Public Health and Primary Care, Prince of Wales Hospital, Shatin, NT, Hong Kong Special Administrative Region

Received 13 June 2014; received in revised form 2 October 2014; accepted 3 October 2014 Available online 6 November 2014

KEYWORDS Colorectal cancer screening Proximal advanced neoplasia Colonoscopy Flexible sigmoidoscopy

Abstract Flexible sigmoidoscopy (FS) and colonoscopy are two commonly used screening tools for colorectal cancer (CRC), and FS mainly detects distal lesions. Colonoscopy resource is limited, yet there is no definite evidence on when flexible sigmoidoscopy is suitable as a screening alternative. This study evaluated the optimal cut-off score from a validated risk stratification system which best predicts proximal advanced neoplasia (PAN) by comparing the sensitivity, specificity and relative risk of PAN according to various cut-off scores. 5819 asymptomatic subjects aged between 50 and 70 years (average age 57.7 years, standard deviation (SD) 4.9) received colonoscopy between 2008 and 2014 in Hong Kong. Their prevalence of PAN was evaluated according to a prediction tool for colorectal neoplasia based on age, gender, smoking status, family history of CRC, body mass index (BMI) and diabetes (ranging from 0 to 6). One binary logistic regression model was performed with PAN as the outcome variable and the risk score as the variable tested for association. In multivariate regression analysis, risk score P3 was associated with significantly higher risk of PAN (3.4–9.1%; AOR = 3.18–8.09, p < 0.001) when compared with those scoring 0. Risk scores 0–2 were associated with either insignificant or lower risks of PAN compared to the overall risk. Applying FS for screening those who scored 0–2 and colonoscopy for those who scored P3 led to a very small proportion of PAN being missed (1.60%), whilst maintaining a high level of specificity

Source of support: The Hong Kong Jockey Club Charities Trust. ⇑ Corresponding author at: 7/F, Lui Che Woo Clinical Science Building, Prince of Wales Hospital, Shatin, NT, Hong Kong Special Administrative Region. Tel.: +852 3943 8600; fax: +852 2603 7301. E-mail address: [email protected] (J.J.Y. Sung). http://dx.doi.org/10.1016/j.ejca.2014.10.003 0959-8049/Ó 2014 Elsevier Ltd. All rights reserved.

38

M.C.S. Wong et al. / European Journal of Cancer 51 (2015) 37–44

(81.9%). Clinicians may use this scoring system to inform subjects and facilitate their choice between colonoscopy and FS. Ó 2014 Elsevier Ltd. All rights reserved.

1. Introduction Globally, colorectal cancer (CRC) is one of the most common malignancies. In 2012, it accounted for around 10% of all cancers and more than 8% of cancer-specific mortality worldwide [1]. In the past decades, many Asia Pacific countries were increasingly affected, gradually catching up the incidence figures of the Western countries [2]. It leads to a substantial public health burden in terms of morbidity and mortality. CRC screening tests have been proven effective to reduce mortality by up to 33%, 40% and 56% using faecal occult blood tests, flexible sigmoidoscopy (FS) and colonoscopy, respectively [3,4]. Guidelines endorsed these three tests as suitable screening options [5,2]. In 2014, an updated Asia Pacific Consensus Recommendation for CRC screening proposed that colonoscopy is the preferred test of choice in high risk individuals [6]. Colonoscopy is also the predominant mode of screening in the US [7] among many other countries. However, colonoscopy is a labour-intensive, invasive and expensive procedure relative to other screening modalities. Its demand is increasing, and endoscopic capacity constraints induce a prohibitive challenge to population-based CRC screening [8,9]. It also requires a high level of expertise – which implies that it might not be suitable as a first-line test in resource deprived regions. On the other hand, FS represents an attractive option as it is an office-based procedure requiring minimal bowel preparation and no sedation. The reach of the FS endoscope allows examination of a portion of the descending colon. To deal with the increasing demand for colonoscopy, subjects at lower risks of proximal lesions – and hence more suited for FS – will need to be identified. Our research team has previously developed and validated a scoring system predicting the presence of colorectal neoplasia [10]. We have also evaluated the risk factors associated with advanced neoplasia [11], and also the association between distal hyperplastic polyps and proximal lesions. However, whether this scoring system could effectively predict the presence of proximal advanced neoplasia (PAN) remains unknown. The objective of this study is to evaluate the optimal cut-off score from this risk stratification system which best predicts PAN by comparing the sensitivity, specificity and relative risk of PAN according to various cut-off scores. The primary purpose is to evaluate whether this existing validated scoring system for any colorectal advanced neoplasia [10] can also be used to stratify

screening participants according to their risk of proximal neoplasia. If so, the same scoring system can be used by primary care professionals to risk stratify subjects according to their overall risk of colorectal neoplasia and also that of proximal neoplasia. The underlying framework assumes that FS is a suitable screening option when the screening participants do not have proximal lesions. 2. Materials and methods The details of the study setting have been described elsewhere [12–17]. In short, a bowel cancer screening centre has been established in Hong Kong in 2008, and invited eligible Hong Kong residents aged 50– 70 years who were asymptomatic of CRC to participate in a free CRC screening programme via media announcement. The study was approved by the Clinical Research Ethics Committee of the Chinese University of Hong Kong (protocol CRE-2008.404). All participants provided informed consent for the study. 2.1. Study design Self-referred screening participants for CRC screening were recruited. The eligibility criteria for this study were (i) age 50–70 years; (ii) absence of existing or previous symptoms suggestive of CRC such as haematochezia, tarry stool, anorexia or change in bowel habit in the past 4 weeks, or weight loss of greater than 5 kg in the past 6 months; and (iii) not having received any CRC screening tests in the past 5 years. Exclusion criteria included personal history of CRC, colonic adenoma, diverticular disease, inflammatory bowel disease, prosthetic heart valve or vascular graft surgery. Participants with medical conditions which were contraindications for colonoscopy, like cardiopulmonary insufficiency and the use of double antiplatelet therapies were also excluded. 2.2. Study logistics Eligible participants were invited to complete a selfadministered questionnaire, including information on their socio-demographic details, past medical history and use of medications for chronic diseases. The completeness of questionnaires was checked by our centre’s staff, and trained volunteers supported survey completion for illiterate subjects by reading the questions as exactly printed in the questionnaire. The body weight

M.C.S. Wong et al. / European Journal of Cancer 51 (2015) 37–44

was measured with the participant wearing light clothing without wearing shoes, using reliable weight scales which were regularly calibrated over time by the healthcare staff in the centre. The screening participants chose between a yearly Faecal Immunochemical Test (FIT) for up to 4 years and one direct colonoscopy. The total population of Hong Kong in mid-2010 was 1.72 million according to the statistics of the Government Census. We have recruited 10,732 asymptomatic screening participants in 2008–2014. Among them, 54.6% received FIT and 45.4% received colonoscopy. An additional 1147 subjects received both FIT and colonoscopy. This study included all participants who received colonoscopy in the study period 2008–2014.

39

body mass index (BMI) < 25 kg/m2 (0), BMI P 25 kg/ m2 (1), diabetes (1), no diabetes (0). The prediction tool was developed from the first 2000 screening participants enrolled in this programme as a derivation cohort, and the subsequent 3220 subjects as a validation cohort. The variables used in the scoring system were selected from the derivation cohort by regression analysis, which were then tested in the validation cohort. The scoring system ranges from 0 to 6, and a participant’s score was based on the total of all the scores allocated to each individual parameter. The covariates included alcohol drinking, concomitant comorbidities, and the use of aspirin and other non-steroidal anti-inflammatory drugs. 2.5. Statistical analyses

2.3. Colonoscopy procedure Each study participant was explained about the detailed procedure of colonoscopy before the scheduled colonoscopy appointment. Polyethylene Glycol (KleanPrepÒ, Helsinn Birex Pharmaceuticals Ltd., Ireland) was used as a standardised bowel preparation regime, and was given to each participant before they left the centre. Colonoscopy was conducted by experienced colonoscopists in an endoscopy centre in a major hospital. Before the procedure, all subjects received a standardised sedation regime consisting of Midazolam 2.5 mg (Groupe Panpharma, France). Meperidine 25 mg (Martindale Pharmaceuticals, United Kingdom) was administered intravenously. According the subject’s level of discomfort, further doses of midazolam and meperidine were supplemented. Air insufflation was used for all the colonoscopies. The endoscopists aimed for a withdrawal time of P6 min, which is in accordance with the current quality indicators for colonoscopy [18]. All lesions were removed and biopsied as deemed appropriate by the endoscopists. The biopsied specimens were examined by gross and microscopic evaluation in a certified, accredited laboratory.

The absolute risks of PAN and proximal neoplasia were calculated according to the sociodemographic and clinical variables. The distribution of the neoplastic lesions in the proximal and distal colon was also presented. Univariate analysis was conducted between the presence of PAN and each covariate consecutively, and variables found to be significant were included in a binary logistic regression model. To evaluate the cutoff score which could best predict PAN, Youden’s index [19] and the Area Under the Curve (AUC) of each cutoff score were delineated. The Youden’s index is used for the evaluation of overall discriminative power of a diagnostic procedure. It aims to determine the optimal cutoff point for maximising the difference between sensitivity, the probability of correctly classifying diseased individuals and (1-specificity), the probability of incorrectly classifying health individuals. Under most screening protocols, a distal adenoma on sigmoidoscopy would prompt referral to colonoscopy. Hence, a sensitivity analysis was conducted where we excluded those without distal adenomas and the above analyses were performed again. All p-values <0.05 in the multivariate regression analysis were regarded as statistically significant.

2.4. Outcome variables and covariates

3. Results

The primary outcome was the presence of proximal advanced neoplasia (PAN), irrespective of distal findings. PAN is defined as CRC, any colorectal adenoma or sessile serrated polyp which has a size of P10 mm in diameter, high grade dysplasia, villous or tubulovillous histologic characteristics, or any combination thereof in the proximal colon. ‘Proximal’ refers to a location in the colon which is proximal to the splenic flexure. The independent variable was the scoring system developed from our previous study [10]. The scoring included the following parameters: age 50–55 years (0), 56–70 years (1), male gender (1), female gender (0), family history of CRC in a first-degree relative present (1) or absent (0), current or ex-smoker (1), non-smoker (0),

3.1. Participant characteristics A total of 5819 screening participants received colonoscopy. Their average age was 57.7 years (standard deviation (SD) 4.9) and 46.9% were male (Table 1). The proportion of subjects who had incomplete colonoscopy performed was 0.5%. There were 43 subjects (0.88%) who suffered from bleeding after colonoscopy, and one subject (0.017%) who had perforation during the colonoscopy procedure requiring hospital stay. All these complications were uneventful after appropriate management. The proportion of subjects found to have proximal neoplasia (PN) and PAN was 18.4% and 2.4%, respectively. Among those diagnosed to have advanced neoplasia, 42.6% of the

40

M.C.S. Wong et al. / European Journal of Cancer 51 (2015) 37–44

Table 1 Characteristics of participants (N = 5819).

Age (years), mean ± standard deviation (SD) BMI (kg/m2), mean ± SD Risk scores, n (Row%) @ 0 1 2 3 4 5 Gender, male, n (%) Current smoking, n (%) Alcohol consumption, n (%) Diabetes mellitus, n (%) Family history present for a first-degree relative, n (%) Hypertension, n (%) IHD/heart disease, n (%) COAD, n (%) Stroke, n (%) Cirrhosis, n (%) GERD, n (%) Use of non-steroidal anti-inflammatory drugs, n (%) Use of aspirin, n (%)

Advanced neoplasia** or CRC, n (Row%) CRC, n (Row%) Advanced neoplasia, n (Row%)

Proximal neoplasia* (N = 1073)

Proximal advanced neoplasia* (N = 140)

Total (N = 5819)

58.95 ± 5.03 24.04 ± 3.27

59.16 ± 5.09 24.09 ± 3.40

57.71 ± 4.89 23.53 ± 3.18

80 (10.1%) 275 (13.9%) 388 (20.3%) 257 (29.3%) 54 (31.2%) 3 (27.3%) 649 (60.5%) 127 (11.8%) 129 (12.0%) 102 (23.4%) 172 (21.1%) 293 (21.9%) 24 (24.5%) 6 (0.6%) 6 (0.6%) 1 (0.1%) 54 (5.0%) 39 (3.6%) 24 (2.4%)

9 (1.1%) 40 (2.0%) 44 (2.3%) 34 (3.9%) 9 (5.2%) 1 (9.1%) 80 (57.1%) 15 (10.7%) 16 (11.4%) 8 (5.7%) 31 (22.1%) 45 (32.1%) 5 (3.6%) 3 (2.1%) 0 (0%) 0 (0%) 12 (8.6%) 4 (2.9%) 4 (2.9%)

792 (13.8%) 1985 (34.5%) 1911 (33.2%) 878 (15.3%) 173 (3.0%) 11 (2.0%) 2729 (46.9%) 460 (7.9%) 562 (9.7%) 435 (7.5%) 817 (14.0%) 1340 (23.0%) 98 (1.7%) 41 (0.7%) 38 (0.7%) 8 (0.1%) 304 (5.2%) 270 (4.6%) 139 (2.4%)

Proximal colon

All sites

140 (42.6%) 8 (33.3%) 132 (43.3%)

329 (5.6%) 24 (0.4%) 305 (5.2%)

BMI, body mass index; IHD, ischaemic heart disease; COAD, chronic obstructive pulmonary disease; CRC, colorectal cancer; GERD, gastrooesophageal reflux disease. * Neoplasia refers to colorectal adenoma, advanced neoplasia or CRC. Hyperplastic polyps were not included in the analysis. ** Advanced Neoplasia is defined as CRC, any colorectal adenoma or sessile serrated polyp which has a size of P10 mm in diameter, high grade dysplasia, villous or tubulovillous histologic characteristics or any combination thereof in the proximal colon. @ 69 missing due to missing BMI.

lesions were found in the proximal colon. The risk of having PN and PAN according to the risk scores showed an increasing trend (Table 1). 3.2. The distribution of neoplastic lesions When the distribution of the lesions according to the colonic site was analysed (Table 2), it was found that those who had both proximal and distal neoplasia were older, had higher BMI; and more were male; ever smokers; alcohol drinkers; and reported to have diabetes, hypertension or ischaemic heart disease than other groups. They were less likely to be users of non-steroidal anti-inflammatory drugs (NSAIDs) (Table 2). The differences in distribution of these neoplastic lesions were not statistically significant and were only a trend. 3.3. Factors associated with the presence of proximal advanced neoplasia From univariate analysis, it was found that the risk scores were significantly associated with PAN (Table 3).

The score of the individuals was derived from their age, gender, family history, smoking, BMI and self-reported diabetes. From binary logistic regression modelling, participants who scored 1 or 2 were not significantly associated with higher risk of PAN compared to those with score zero. There is a substantially higher risk of PAN among those with scores P3 (Adjusted Odds Ratios (AOR) ranged from 3.177 to 8.091). 3.4. Optimal cut-off value for prediction of PAN The prevalence of PAN among all subjects was 2.4%. The absolute risk of PAN among those who scored 0 (1.1%, p = 0.033) was lower than the overall risk, whilst those who scored 1 (2.0%, p = 0.360) or 2 (2.3%, p = 0.864) had statistically similar risk of PAN when compared with the risk of all subjects (Table 4). Subjects who scored P3 had significantly higher risk than the overall risk (range 3.9–9.1%). The Youden’s index was maximal (0.140) when the score P3 was used as a cutoff to predict PAN. The Area Under the Curve (AUC) was 0.60 (95% confidence interval (CI) 0.55–0.65). The

M.C.S. Wong et al. / European Journal of Cancer 51 (2015) 37–44

41

Table 2 The distribution of the neoplastic lesions (N = 5819).

Age (years), mean ± standard deviation (SD) BMI (kg/m2), mean ± SD Gender, male, n (%) Current smoking, n (%) Alcohol consumption, n (%) Diabetes mellitus, n (%) Family history present for a firstdegree relative, n (%) Hypertension, n (%) IHD/heart disease, n (%) COAD, n (%) Stroke, n (%) Cirrhosis, n (%) GERD, n (%) Use of non-steroidal antiinflammatory drugs, n (%) Use of aspirin, n (%)

Normal (n = 3949)

Proximal neoplasia only*,** (n = 711)

Proximal and distal neoplasia*,** (n = 362)

Distal neoplasia only*,** (n = 797)

57.26 ± 4.76

58.49 ± 4.92

59.86 ± 5.12

58.26 ± 4.97

23.34 ± 3.14 1664 (42.1%) 255 (6.5%) 335 (8.5%) 250 (6.3%) 515 (13.0%)

23.8 ± 3.21 395 (55.6%) 70 (9.8%) 81 (11.4%) 62 (8.7%) 120 (16.9%)

24.49 ± 3.36 254 (70.2%) 57 (15.7%) 48 (13.3%) 40 (11.0%) 52 (14.4%)

23.84 ± 3.20 416 (52.2%) 78 (9.8%) 98 (12.3%) 83 (10.4%) 130 (16.3%)

847 (21.4%) 61 (1.5%) 25 (0.6%) 27 (0.7%) 5 (0.1%) 209 (5.3%) 187 (4.7%)

179 (25.2%) 14 (2.0%) 5 (0.7%) 6 (0.8%) 1 (0.1%) 38 (5.3%) 27 (3.8%)

114 (31.5%) 10 (2.8%) 1 (0.3%) 0 (0%) 0 (0%) 16 (4.4%) 12 (3.3%)

200 (25.1%) 13 (1.6%) 10 (1.3%) 5 (0.6%) 2 (0.3%) 41 (5.1%) 44 (5.5%)

86 (2.2%)

15 (2.1%)

9 (2.5%)

29 (3.6%)

*

Neoplasia refers to colorectal adenoma, advanced neoplasia or CRC. Hyperplastic polyps were not included in the analysis. BMI, body mass index; IHD, ischaemic heart disease; COAD, chronic obstructive pulmonary disease; GERD, gastro-oesophageal reflux disease. ** Advanced Neoplasia is defined as colorectal cancer, any colorectal adenoma or sessile serrated polyp which has a size of P10 mm in diameter, high grade dysplasia, villous or tubulovillous histologic characteristics or any combination thereof in the proximal colon.

specificity of detecting PAN sharply increased from 48.6% to 81.9% when the cut-off score changed from P2 to P3. Hypothetically, using FS to screen all the study participants irrespective of their score could lead to 2.41% of PAN being missed. Assuming colonoscopy

as a gold standard which can detect all proximal and distal lesions, its use to screen all participants could lead to no missed PAN lesions but 97.6% of the colonoscopic procedures could be replaced by FS without reduction in diagnostic yield. Using FS for those who scored 0–2

Table 3 Univariate and multivariate predictors of proximal advanced neoplasia.* Unadjusted (univariate analysis) Risk factors Risk scores 0 1

Adjusted (multivariate regression analysis) Odds ratio (OR) (95% confidence interval (CI))

pValue

B coefficient

Standard error (SE)

OR (95% CI)

0.001

0.003

Referent 1.789 (0.864–3.705)

0.117

0.541

0.372

2

2.050 (0.996–4.220)

0.051

0.645

0.371

3

3.505 (1.670–7.354)

0.001

1.156

0.382

4

4.774 (1.867–12.212)

0.001

1.438

0.484

5

8.700 (1.005–75.295)

0.049

2.091

1.103

Hypertension

1.604 (1.119–2.299)

0.010

0.360

0.189

IHD/heart disease Alcohol COAD Stroke Cirrhosis GERD Use of non-steroidal antiinflammatory drugs Use of aspirin

2.225 1.213 3.251 NA NA 1.730 0.599

(0.946–3.162) (0.220–1.630)

0.087 0.473 0.052 NA NA 0.075 0.315

– – – NA NA – –

1.208 (0.440–3.312)

0.714

–

(0.890–5.559) (0.715–2.057) (0.991–10.659)

pValue

– – – NA NA – –

Referent 1.718 (0.828– 3.562) 1.905 (0.921– 3.940) 3.177 (1.502– 6.721) 4.212 (1.629– 10.886) 8.091 (0.931– 70.295) 1.433 (0.990– 2.074) – – – NA NA – –

– – – NA NA – –

–

–

–

0.146 0.082 0.002 0.003 0.058 0.057

BMI, body mass index; IHD, ischaemic heart disease; COAD, chronic obstructive airway disease; GERD, gastro-oesophageal reflux disease. **NA, no subjects with stroke or cirrhosis had proximal advanced neoplasia. * Proximal neoplasia refers to colorectal adenoma, advanced neoplasia or colorectal cancer.

42

M.C.S. Wong et al. / European Journal of Cancer 51 (2015) 37–44

Table 4 The relative risk of proximal advanced neoplasia (PAN) compared to overall risk. Risk score

Absolute risk of PAN (%)

% of missed PAN

Sensitivity (%)

Specificity (%)

Youden’s index

p Value

P0 P1 P2 P3 P4 P5

1.1 2.0 2.3 3.9 5.2 9.1

0.0 0.15 0.84 1.60 2.18 2.41

100 93.4 64.2 32.1 7.3 0.7

0 13.9 48.6 81.9 96.9 99.8

0.000 0.074 0.128 0.140 0.042 0.006

0.360 0.864 0.015 0.038 0.028

Proximal advanced neoplasia is defined as CRC, any colorectal adenoma or sessile serrated polyp which has a size of P10 mm in diameter, high grade dysplasia, villous or tubulovillous histologic characteristics or any combination thereof in the proximal colon. The absolute risk of PAN refers to the proportion of subjects found to have PAN among all subjects with the risk score specified in the row. The % of missing PAN refers to the proportion of subjects who would have missed the detection of PAN if colonoscopy procedures were performed only among subjects with risk scores specified in the row, and that flexible sigmoidoscopy was performed among subjects with risk score lower than that in the specified row. The Youden’s index refers to the overall discriminative power when the score was used as a cut-off for prediction of PAN. The p values represent comparison of the relative risk of PAN between each individual score versus the risk among all subjects (2.4%) by Chi-square tests or Fisher’s exact tests.

and colonoscopy for those who scored P3 could lead to smaller proportion of PAN being missed (1.60%), whilst maintaining a high level of specificity (81.9%). Nevertheless, the sensitivity of using the cut-off score at P3 was modest (32.1%) (Table 4). A sensitivity analysis by excluding subjects with distal adenoma (17.0%) led to the same conclusions. 4. Discussion 4.1. Major findings and implications to clinical practice This study evaluated the relative risk of PAN according to a previously validated scoring system predicting colorectal neoplasia, and found that those who scored P3 had significantly higher risk of PAN. Despite a limited predictive ability, the cut-off score of 3 had the optimal overall discriminative power to stratify those at higher risk of PAN. Therefore, the scoring system could be used to stratify screening participants into those who justify for colonoscopy (e.g. score P3). In practices where colonoscopy service is limited, those who prefer earlier screening yet scored low (e.g. <3) could opt for FS due to the low risk of PAN. The use of this individualised stratification strategy is expected to better inform physicians and patients on their choice of screening options based on one’s risk. These findings also bear significant health service implications. For instance, forecasting models from decision analysis and surveys conducted by the authoritative institutes affirm that healthcare resource is lacking to accommodate the growing demand of colonoscopy in the United States, as well as other Asia Pacific countries [8,20,21]. Re-orienting the service capacity to the use of FS could potentially relieve the substantial healthcare resources, including colonoscopists, required for population screening. Randomised controlled trials have tested the efficacy of FS in the UKFS and US PLCO trials – with reduction of CRC incidence and mortality by up to 23% and 31%, respectively [22,23]. Whilst FS has traditionally been performed by gastroenterologists or surgeons,

studies have demonstrated that adequately trained family physicians and nurse endoscopists can perform screening FS as safely and effectively as gastroenterologists or surgeons [24–27]. 4.2. Relationship with literature This study is unique as it is the first to suggest using a single composite score based on recognised risk factors to stratify individuals into those suitable for colonoscopy versus FS. From a thorough literature review, there is a scarcity of studies which have devised validated risk stratification systems for PAN. Imperiale et al. developed a risk stratification model for individuals with very low risk of PAN for whom FS alone is sufficient to act as the primary screening tool [28]. The clinical index consists of a seven-point scoring system based on age, sex and distal findings on FS. Each of these three variables has been identified as an independent predictor of PAN in the derivation cohort [29]. Nevertheless, Levitzky and colleagues evaluated the validity of this clinical index, and found that the index has limited ability to discriminate low from intermediate-risk white, black and Hispanic patients for PAN [30]. From another large scale study involving more than 10,000 adults, it was concluded that the risk of PAN is a function of age and gender [31]. Yet another study conducted by Levin et al. demonstrated that age, family history and distal findings were predictors of PAN [32]. There are other studies which evaluated the predictors of colorectal neoplasia [33–36] with similar conclusions – yet the major objectives target at studying independent predictors only. Our study did not rely on distal findings as one of the predictive factors since its major research objective is to evaluate a cut-off score where FS could be used as an alternative screening tool among subjects who have not undergone either colonoscopy or FS before. 4.3. Study limitations This study prospectively enrolled a large number of asymptomatic subjects who underwent colonoscopy.

M.C.S. Wong et al. / European Journal of Cancer 51 (2015) 37–44

There are a few limitations which should be addressed here. Firstly, the study participants were all self-referred via media invitation, and they could represent a community of more health-conscious subjects when compared with the general public. Also, it was assumed that FS could reach the proximal descending colon – which might be too optimistic especially among non-specialists who do not perform the procedure on a regular basis [31]. In addition, the scoring system proposed relied on the absolute risks of PAN and the probability of not missing PAN for stratifying screening strategies, and the system was not found to be a sensitive tool to detect PAN. The proportion of missed PAN (1.6%) may be perceived by critics as high in large-scale population screening programmes, and the low sensitivity despite the high specificity should be taken into account when these findings were applied in screening practices. Hence the overall performance of the prediction system in terms of the AUC is only modest. Independent validation of the prediction tool is required in future studies. In summary, this study evaluated the optimal cut-off score from a previously validated scoring system which could inform decision making in clinical consultations with respect to the choice of colonoscopy versus FS based on the risk of PAN. Future studies are needed to further explore additional risk factors which could further enhance the performance of the prediction score, as well as the cost-effectiveness of applying this stratified approach based on this tool. Role of the funding source None. Conflict of interest statement None declared. Acknowledgement We acknowledged the support of the Hong Kong Jockey Club Charities Trust for the full funding support of this project. References [1] GLOBOCAN. Cancer fact sheet. Colorectal cancer incidence, mortality and prevalence worldwide in 2012, Available at: ; 2012 [accessed 30.05.14]. [2] Sung JJY, Lau JYW, Goh KL, Leung WK, for the on Asia Pacific Working Group on Colorectal Cancer. Increasing incidence of colorectal cancer in Asia: implications for screening. Lancet Oncol 2005;6:871–6. [3] Mandel JS, Church TR, Ederer F, Bond JH. Colorectal cancer mortality: effectiveness of biennial screening for fecal occult blood. J Natl Cancer Inst 1999;91:434–7. [4] Nishihara R, Wu K, Lochhead P, Morikawa T, Liao X, Qian ZR, et al. Long-term colorectal-cancer incidence and mortality after lower endoscopy. N Engl J Med 2013;369:1095–105.

43

[5] Levin B, Lieberman DA, McFarland B, Smith RA, Brooks D, Andrews KS, 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. CA Cancer J Clin 2008;58: 130–60. [6] Sung JJ, Chan FK, Chiu HM, the Asia Pacific Working Group on Colorectal Cancer Screening. An updated Asia Pacific consensus recommendations on colorectal cancer screening. Gut 2014. http://dx.doi.org/10.1136/gutjnl-2013-306503. [7] Klabunde CN, Lanier D, Nadel MR, McLeod C, Yuan G, Vernon SW. Colorectal cancer screening by primary care physicians: recommendations and practices, 2006–2007. Am J Prev Med 2009;37:8–16. [8] Seeff LC, Manninen DL, Dong FB, Chattopadhyay SK, Nadel MR, Tangka FK, et al. Is there endoscopic capacity to provide colorectal cancer screening to the unscreened population in the United States? Gastroenterology 2004;127:1661–9. [9] Seeff LC, Richards TB, Shapiro JA, Nadel MR, Manninen DL, Given LS, et al. How many endoscopies are performed for colorectal cancer screening? Results from CDC’s survey of endoscopic capacity. Gastroenterology 2004;127:1670–7. [10] Wong MCS, Lam TYT, Tsoi KKF, Hirai HW, Chan VC, Ching JY, et al. A validated tool to predict colorectal neoplasia and inform screening choice for asymptomatic subjects. Gut 2014;63:1130–6. [11] Wong MCS, Lam TYT, Tsoi KKF, Chan VC, Hirai HW, Ching JY, et al. Predictors of colorectal advanced neoplasia for colorectal cancer screening. Am J Prev Med 2014;46:433–9. [12] Wong MC, Hirai HW, Luk AK, Lam TY, Ching JY, Griffiths SM, et al. The knowledge of colorectal cancer symptoms and risk factors among 10,078 screening participants: are high risk individuals more knowledgeable? PLoS One 2013;8:e60366. [13] Wong MC, Ching JY, Hirai HH, Lam TY, Griffiths SM, Chan FK, et al. Perceived obstacles of colorectal cancer screening and their associated factors among 10,078 Chinese participants. PLoS One 2013;8:e70209. [14] Wong MCS, John GK, Hirai HW, Lam TY, Luk AK, Ching JY, et al. Changes in the choice of colorectal cancer screening tests in primary care settings from 7845 prospectively collected surveys. Cancer Causes Control 2012;23:1541–8. [15] Ng SC, Ching JYL, Chan VCW, Wong MC, Suen BY, Hirai HW, et al. Diagnostic accuracy of fecal immunochemical test for screening individuals with a family history of colorectal cancer. Aliment Pharmacol Ther 2013;38:835–41. [16] Wong MCS, Ching JY, Lam TY, Luk AK, Hirai HW, Griffiths SM, et al. Prospective cohort study of compliance with faecal immunochemical tests for colorectal cancer screening in Hong Kong. Prev Med 2013;57:227–31. [17] Wong MCS, Tsoi KKF, Ng SSM, Lou VW, Choi SY, Ling KW, et al. A comparison of the acceptance of immunochemical faecal occult blood test and colonoscopy in colorectal cancer screening: a prospective study among Chinese. Aliment Pharmacol Ther 2010;32:74–82. [18] Rex DK, Petrini JL, Baron TH, Chak A, Cohen J, Deal SE, et al. Quality indicators for colonoscopy. Gastrointest Endosc 2006;63:S16–28. [19] Youden WJ. Index for rating diagnostic tests. Washington, DC: The National Bureau of Standards; 1949. [20] Vijan S, Inadomi J, Hayward RA, Hofer TP, Fendrick AM. Projections of demand and capacity for colonoscopy related to increasing rates of colorectal cancer screening in the United States. Aliment Pharmacol Ther 2004;20:507–15. [21] 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;115:129–33.

44

M.C.S. Wong et al. / European Journal of Cancer 51 (2015) 37–44

[22] Atkin WS, Edwards R, Kralj-Hans I, Wooldrage K, Hart AR, Northover JM, et al. Once-only flexible sigmoidoscopy screening in prevention of colorectal cancer: a multicentre randomised controlled trial. Lancet 2010;375:1624–33. [23] Schoen RE, Pinsky PF, Weissfeld JL, Yokochi LA, Church T, Laiyemo AO, et al. Colorectal-cancer incidence and mortality with screening flexible sigmoidoscopy. N Engl J Med 2012;366: 2345–57. [24] Shaukat MS, Ramirex FC. The utilization of flexible sigmoidoscopy by family practitioners after residency training. Gastrointest Endosc 2000;52:45–7. [25] Williams J, Russell I, Dural D, Cheung WY, Farrin A, Bloor K, et al. Effectiveness of nurse delivered endoscopy: findings from randomised multi-institution nurse endoscopy trial (MINuET). BMJ 2009;338:b231. [26] Limoges-Gonalez M, Singh N, Al-Juburi Amar, Tseng D, Inadomi J, Rossaro L. Comparisons of screening colonoscopy performed by a nurse practitioner and gastroenterologists. Gastroenterol Nurs 2011;34:210–6. [27] Wallace MB, Kemp JA, Meyer F, Horton K, Reffel A, Christiansen CL, et al. Screening for colorectal cancer with flexible sigmoidoscopy by nonphysician endoscopists. Am J Med 1999;107:214–8. [28] Imperiale TF, Wagner DR, Lin CY, Larkin GN, Rogge JD, Ransohoff DF. Using risk for advanced proximal colonic neoplasia to tailor endoscopic screening for colorectal cancer. Ann Intern Med 2003;139:959–65. [29] Imperiale TF, Wagner DR, Lin CY, Larkin GN, Rogge JD, Ransohoff DF. Risk of advanced proximal neoplasms in asymp-

[30]

[31]

[32]

[33]

[34]

[35]

[36]

tomatic adults according to the distal colorectal findings. N Engl J Med 2000;343:169–74. Levitzky BE, Brown CC, Heeren TC, Schroy 3rd PC. Performance of a risk index for advanced proximal colorectal neoplasia among a racially/ethnically diverse patient population (Risk index for advanced proximal neoplasia). Am J Gastroenterol 2011;106:1099–106. Imperiale TF, Glowinski EA, Lin-Cooper C, Ransohoff DF. Tailoring colorectal cancer screening by considering risk of advanced proximal neoplasia. Am J Med 2012;125:1181–7. Levin TR, Palitz A, Grossman S, Conell C, Finkler L, Ackerson L, et al. Predicting advanced proximal colonic neoplasia with screening sigmoidoscopy. JAMA 1999;281:1611–7. Lieberman DA, Prindiville S, Weiss DG, Willett W, VA Cooperative Study Group 380. Risk factors for advanced colonic neoplasia and hyperplastic polyps in asymptomatic individuals. JAMA 2003;290:2959–67. Schoen RE, Corle D, Cranston L, Weissfeld JL, Lance P, Burt R, et al. Is colonoscopy needed for the nonadvanced adenoma found on sigmoidoscopy? The Polyp Prevention Trial. Gastroenterology 1998;115:533–41. Wallace MB, Kemp JA, Trnka YM, Donovan JM, Farraye FA. Is colonoscopy indicated for small adenomas found by screening flexible sigmoidoscopy? Ann Intern Med 1998;129:273–8. Rex DK, Lehman GA, Ulbright TM, Smith JJ, Pound DC, Hawes RH, et al. Colonic neoplasia in asymptomatic persons with negative fecal occult blood tests: influence of age, gender, and family history. Am J Gastroenterol 1993;88:825–31.