Endoscopy 2020; 52(03): 220-226
DOI: 10.1055/a-1041-2945
Original article
© Georg Thieme Verlag KG Stuttgart · New York

How to identify patients who are less likely to have metachronous neoplasms after a colon cancer: a predictive model

Leonardo Frazzoni
 1   Department of Medical and Surgical Sciences, S. Orsola-Malpighi Hospital, Bologna, Italy
,
Liboria Laterza
 2   Endoscopy Service, AUSL Reggio Emilia, Reggio Emilia, Italy
,
Alessandro Mussetto
 3   Division of Gastroenterology, S. Maria delle Croci Hospital, Ravenna, Italy
,
Rocco Maurizio Zagari
 1   Department of Medical and Surgical Sciences, S. Orsola-Malpighi Hospital, Bologna, Italy
,
Cristina Trovato
 4   Division of Endoscopy, European Institute of Oncology, IRCCS, Milan, Italy
,
Mario De Bellis
 5   Gastroenterology and Endoscopy Unit, Department of Abdominal Oncology, Istituto Nazionale Tumori – IRCCS – Fondazione G. Pascale, Naples, Italy
,
Silvia Paggi
 6   Division of Digestive Endoscopy and Gastroenterology, Valduce Hospital, Como, Italy
,
Stefania Piccirelli
 7   Digestive Endoscopy Unit, Fondazione Poliambulanza, Brescia, Italy
,
Luigi Ricciardiello
 1   Department of Medical and Surgical Sciences, S. Orsola-Malpighi Hospital, Bologna, Italy
,
Paola Cesaro
 7   Digestive Endoscopy Unit, Fondazione Poliambulanza, Brescia, Italy
,
Cristiano Spada
 7   Digestive Endoscopy Unit, Fondazione Poliambulanza, Brescia, Italy
,
Giulia Dal Piaz
 3   Division of Gastroenterology, S. Maria delle Croci Hospital, Ravenna, Italy
,
Marina La Marca
 1   Department of Medical and Surgical Sciences, S. Orsola-Malpighi Hospital, Bologna, Italy
,
Fabio Fabbian
 2   Endoscopy Service, AUSL Reggio Emilia, Reggio Emilia, Italy
,
Laura Petrella
 8   Department of Statistics, University of Bologna, Bologna, Italy
,
Veronica Smania
 1   Department of Medical and Surgical Sciences, S. Orsola-Malpighi Hospital, Bologna, Italy
,
Pietro Marone
 5   Gastroenterology and Endoscopy Unit, Department of Abdominal Oncology, Istituto Nazionale Tumori – IRCCS – Fondazione G. Pascale, Naples, Italy
,
Fabiana Tatangelo
 9   Division of Pathology and Cytology, Istituto Nazionale Tumori – IRCSS – Fondazione Pascale, Naples, Italy
,
Franco Bazzoli
 1   Department of Medical and Surgical Sciences, S. Orsola-Malpighi Hospital, Bologna, Italy
,
Franco Radaelli
 6   Division of Digestive Endoscopy and Gastroenterology, Valduce Hospital, Como, Italy
,
Alessandro Repici
10   Digestive Endoscopy Unit, Division of Gastroenterology, Humanitas Clinical and Research Center, Rozzano, Italy
,
Cesare Hassan
11   Endoscopy Unit, Nuovo Regina Margherita Hospital, Rome, Italy
,
Michele Scagliarini
 8   Department of Statistics, University of Bologna, Bologna, Italy
,
Lorenzo Fuccio
 1   Department of Medical and Surgical Sciences, S. Orsola-Malpighi Hospital, Bologna, Italy
› Institutsangaben
Weitere Informationen

Corresponding author

Lorenzo Fuccio, MD
Gastroenterology Unit, DIMEC
S. Orsola-Malpighi Hospital
University of Bologna
Via Massarenti 9
40138 Bologna
Italy   
Fax: +39-051-2149112   

Publikationsverlauf

submitted 23. Mai 2019

accepted after revision 08. Oktober 2019

Publikationsdatum:
13. Dezember 2019 (online)

 

Abstract

Background Patients with prior colon cancer have increased risk of metachronous colorectal neoplasms; therefore, endoscopic surveillance is indicated. Current recommendations are not risk-stratified. We investigated predictive factors for colorectal neoplasms to build a model to spare colonoscopies for low-risk patients.

Methods This was a multicenter, retrospective study including patients who underwent surgery for colon cancer in 2001 – 2008 (derivation cohort) and 2009 – 2013 (validation cohort). A predictive model for neoplasm occurrence at second surveillance colonoscopy was developed and validated.

Results 421 and 203 patients were included in derivation and validation cohort, respectively. At second surveillance colonoscopy, 112 (26.6 %) and 55 (27.1 %) patients had metachronous neoplasms in derivation and validation groups; three cancers were detected in the latter. History of left-sided colon cancer (OR 1.64, 95 %CI 1.02 – 2.64), ≥ 1 advanced adenoma at index colonoscopy (OR 1.90, 95 %CI 1.05 – 3.43), and ≥ 1 adenoma at first surveillance colonoscopy (OR 2.06, 95 %CI 1.29 – 3.27) were independently predictive of metachronous colorectal neoplasms at second surveillance colonoscopy. For patients without such risk factors, diagnostic accuracy parameters were: 89.3 % (95 %CI 82.0 %-94.3 %) and 78.2 % (95 %CI 65.0 %-88.2 %) sensitivity, and 28.5 % (95 %CI 23.5 %-33.9 %) and 33.8 % (95 %CI 26.2 %-42.0 %) specificity in derivation and validation group, respectively. No cancer would be missed.

Conclusions Patients with prior left-sided colon cancer or ≥ 1 advanced adenoma at index colonoscopy or ≥ 1 adenoma at first surveillance colonoscopy had a significantly higher risk of neoplasms at second surveillance colonoscopy; patients without such factors had much lower risk and could safely skip the second surveillance colonoscopy. A prospective, multicenter validation study is needed.


Introduction

Globally, colorectal cancer (CRC) is the third most commonly diagnosed cancer in males and the second in females, with 1.4 million new cases and almost 694 000 deaths estimated to have occurred in 2012, and with a 5-year survival rate of 65 % [1]. Patients with a prior history of curative colon resection for cancer are at increased risk of developing recurrent and/or metachronous neoplasms [2]. Thus, colonoscopy-based surveillance protocols have been established in order to prolong survival by diagnosing recurrent and metachronous cancers at a curable stage, and to prevent metachronous cancer by detecting and removing precancerous lesions [3] [4]. Current guidelines recommend performing surveillance colonoscopy 1 year after surgery; if results are negative, the interval to the next colonoscopy should be 3 years and if negative again, an interval of 5 years is proposed; subsequent colonoscopies should occur at 5-year intervals [3] [4]. However, current surveillance recommendations are mostly based on outdated studies [5] [6] [7], despite the fact that treatment modalities for colon cancer have evolved over time, and the data might not stand the test of time. Thus, it may be argued that current surveillance protocols could entail a considerable waste of resources, and attempts to stratify the risk of metachronous neoplasms may result in more cost-effective strategies.

In a recent multicenter retrospective study conducted in 441 patients with a history of colonic resection for cancer, we found that patients with a prior left-sided colon cancer were at significantly increased risk of having metachronous colorectal adenomas at the second surveillance colonoscopy than patients with a history of right-sided colon cancer [8]. However, this study had several limitations as it did not consider additional potentially relevant information, such as the findings at the index colonoscopy (i. e. the examination performed at the cancer diagnosis) and at the first surveillance colonoscopy.

The aim of the present study was to identify predictive factors of metachronous neoplasms in the residual colon at the second postoperative colonoscopy in patients with a history of colon cancer. We developed and validated a predictive model to identify low-risk patients who could safely skip the second surveillance colonoscopy.


Methods

This multicenter retrospective study was performed at seven institutions in Italy (Bologna, Brescia, Como, Milan, Naples, Ravenna, and Reggio Emilia). Consecutive patients with a diagnosis of colon carcinoma who had undergone surgical resection from 1 January 2001 to 31 December 2008 were eligible to be included in the derivation cohort. Patients who underwent surgical resection from 1 January 2009 to 31 December 2013 were eligible to be included in the validation group. Given the retrospective design, not all centers included patients for the entire length of the two time frames considered in the derivation and validation cohorts.

The following inclusion criteria had to be satisfied: i) previous proximal or distal colon cancer (considering the splenic flexure as the border between proximal and distal colon); ii) availability of the index colonoscopy report; iii) availability of reports of the first and second surveillance colonoscopies, conducted after the surgical intervention; iv) complete colonoscopy to the cecum or ileocolonic anastomosis, explicitly defining the quality of bowel cleansing as adequate; in detail, bowel cleansing was reported according to the Aronchick scale and the Boston Bowel Preparation Scale, and was judged as adequate when the Aronchick scale was “good” or “fair”, or the Boston score was ≥ 2 for each colonic segment; v) age ≥ 18 years at the time of the diagnosis of colon carcinoma. All participating centers adopted the same surveillance recommendations [9].

Patients had to undergo a perioperative cleansing colonoscopy, either at the time of diagnosis or performed within 6 months after the surgical resection. In cases where the cleansing colonoscopy was performed after the resection, this colonoscopy was not considered as the first surveillance colonoscopy.

Patients with colonic resection for diseases other than colon cancer, of those with rectal resection or diagnosis of hereditary cancer predisposing syndromes (i. e. familial adenomatous polyposis, or Lynch syndrome) were excluded from the study.

The following data were extracted for each patient: sex, age at diagnosis, colon cancer staging, site of colon cancer (i. e. proximal or distal to the splenic flexure), number and location of adenomas found during index colonoscopy and during the first two surveillance colonoscopies after the surgical intervention.

The primary outcome of the study was the occurrence of metachronous colorectal neoplasms (i. e. adenoma, advanced adenoma or cancer) at the second surveillance colonoscopy. Advanced adenoma was defined if one of the following was satisfied: i) ≥ 1 cm in size, ii) tubulovillous or villous histology, iii) high grade dysplasia [10].

A predictive model was developed to derive the probability of finding ≥ 1 metachronous neoplasm at the second surveillance colonoscopy in the derivation cohort, then validated in the validation group following the Transparent Reporting of a multivariable prediction model for Individual Prognosis or Diagnosis (TRIPOD) recommendations [11] [12]. In order to develop the predictive model, patients with cancer at the first surveillance colonoscopy were excluded from the analysis as they would restart their surveillance protocol after surgery for recurrent CRC. The TRIPOD checklist is available in Table 1s in the online-only supplementary material.

The study was approved by the Institutional Review Board of the coordinating center (S. Orsola-Malpighi Hospital, University of Bologna, Bologna, Italy; approved: 05/12/2015; protocol number: 1538/2015) and, thereafter, by the Ethics Committee of each participating center.

Statistical analysis

Results are presented as absolute frequency and percentage with 95 % confidence interval (CI) for categorical variables, and mean with standard deviation (SD) or median with interquartile range (IQR) for normally or non-normally distributed continuous variables, respectively. A multivariate logistic regression analysis was performed in order to identify predictive factors of neoplasms at the second surveillance colonoscopy. Odds ratios (ORs) and 95 %CIs were estimated for endoscopic findings at the index and first surveillance colonoscopies, adjusted for age, sex, and stage of index cancer. A predictive model was subsequently developed. The diagnostic accuracy of the model was explored by computing sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) for absence of risk factors and presence of each of them. The model derived in the derivation cohort was therefore validated in the validation cohort. Analyses were conducted using R statistical software (The R Project for Statistical Computing, Vienna, Austria) and STATA software (Stata Corp., College Station, Texas, USA).


Sample size

The sample size estimation was based on deriving a predictive model for metachronous neoplasm occurrence at the second surveillance colonoscopy. Estimating that a model based on logistic regression would increase the probability of finding ≥ 1 metachronous neoplasm from 22 % to 35 % (OR 1.91), with 80 % power and one-sided 5 % alpha level, we computed a sample size of 373 patients in the derivation cohort. We assumed a binomial distribution of covariate and R2 = 0.2. Sample size calculation was conducted using G*power v3.1 for Mac [13] [14].



Results

Study population

Derivation cohort

A total of 431 patients with prior curative surgery for colon cancer between 2001 and 2008 were included ([Fig. 1a]). A total of 10 patients (2.3 %) had cancer at the first surveillance colonoscopy and were excluded from the analysis, giving 421 patients in the derivation cohort.

Zoom
Fig. 1 Study flow chart. a Derivation cohort. b Validation cohort.

The time interval between surgery and first surveillance colonoscopy was 365 days (IQR 273 – 504). A history of left-sided colon cancer was documented in 253 patients (60.1 %). Mean age was 62.3 years (SD 9.2), and 224 patients (53.2 %) were female. At the index and first surveillance colonoscopies, 171 (40.6 %) and 136 (32.3 %) patients had ≥ 1 adenoma, of whom 61 (35.7 %; 14.5 % of whole group) and 21 (15.4 %; 5.0 % of whole group) had ≥ 1 advanced adenoma, respectively. At the second surveillance colonoscopy, no CRCs were discovered, whereas ≥ 1 adenoma was found in 112 patients (26.6 %), of whom 22 patients (19.6 %; 5.2 % of whole cohort) had ≥ 1 advanced adenoma ([Table 1]). The time interval between surgery and the second surveillance colonoscopy was 960 days (IQR 726 – 1386).

Table 1

Patient characteristics and findings at index, first, and second surveillance colonoscopies in the derivation and validation cohorts.

Patient characteristics

Derivation cohort (n = 421)

Validation cohort (n = 203)

Female sex, n (%)

224 (53.2)

104 (51.2)

Age, mean (SD), years

62.3 (9.2)

63.4 (11.3)

TNM stage, n (%)

  • Stage I

147 (34.9)

80 (39.4)

  • Stage II

147 (34.9)

71 (35.0)

  • Stage III

123 (29.2)

50 (24.6)

  • Stage IV

4 (1.0)

2 (1.0)

History of left-sided colon cancer, n (%)

253 (60.1)

104 (51.2)

Index colonoscopy

  • ≥ 1 adenoma at index colonoscopy, n (%)

171 (40.6)

70 (34.5)

  • ≥ 1 advanced adenoma at index colonoscopy, n (%)

61 (14.5)

35 (17.2)

First surveillance colonoscopy

  • Interval between surgery and first surveillance colonoscopy, median (IQR), days

365 (273 – 504)

388 (335 – 500)

  • ≥ 1 adenoma at first surveillance colonoscopy, n (%)

136 (32.3)

70 (34.5)

  • ≥ 1 advanced adenoma at first surveillance colonoscopy, n (%)

21 (5.0)

25 (12.3)

Second surveillance colonoscopy

  • Interval between surgery and second surveillance colonoscopy, median (IQR), days

960 (726 – 1386)

1088 (803 – 1444)

  • ≥ 1 neoplasm at second surveillance colonoscopy, n (%)

112 (26.6)

55 (27.1)

  • ≥ 1 advanced adenoma at second surveillance colonoscopy, n (%)

22 (5.2)

20 (9.9)

  • ≥ 1 cancer at second surveillance colonoscopy, n (%)

0 (0)

3 (1.5)

SD, standard deviation; IQR, interquartile range.


Validation cohort

Between 2009 and 2013, a total of 209 patients with prior curative surgery for colon cancer were included ([Fig. 1b]). Six patients (2.9 %) had cancer at the first surveillance colonoscopy and were excluded from the analysis, giving 203 patients in the validation cohort. The time interval between surgery and the first surveillance colonoscopy was 388 days (IQR 335 – 500). A history of left-sided colon cancer was reported in 104 patients (51.2 %). Mean age was 63.4 years (SD 11.3), and 104 patients (51.2 %) were female. At index and first surveillance colonoscopies, 70 (34.5 %) and 70 (34.5 %) patients had ≥ 1 adenoma, of whom 35 (50.0 %; 17.2 % of whole group) and 25 (35.7 %; 12.3 % of whole group) had ≥ 1 advanced adenoma, respectively. At the second surveillance colonoscopy, one metachronous colon cancer and two anastomotic recurrences were found ([Table 1]). In addition, ≥ 1 adenoma was found in 55 patients (27.1 %), of whom 20 patients (36.3 %; 9.9 % of whole group) had ≥ 1 advanced adenoma. The time interval between surgery and the second surveillance colonoscopy was 1088 days (IQR 803 – 1444).



Predictive model development

A history of left-sided colon cancer (OR 1.64, 95 %CI 1.02 to 2.64), ≥ 1 advanced adenoma at the index colonoscopy (OR 1.90, 95 %CI 1.05 to 3.43), and ≥ 1 adenoma at the first surveillance colonoscopy (OR 2.06, 95 %CI 1.29 to 3.27) were independently associated with an increased risk of metachronous colorectal neoplasms at second surveillance colonoscopy ([Table 2]). In order to exclude a possible multicollinearity between age and stage of index cancer, we excluded from the statistical model one of these two variables at a time; we found that the association between aforesaid variables and outcome remained statistically significant. We found that the presence of ≥ 1 adenoma at the index colonoscopy, differently from advanced adenoma, was not significantly associated with the outcome (52/112 [46.4 %] and 119/309 [38.5 %] of patients with and without neoplasms at second surveillance colonoscopy, respectively; OR 1.25, 95 %CI 0.79 to 1.98).

Table 2

Characteristics of patients with and without metachronous neoplasms at the second surveillance colonoscopy in the derivation cohort.

Neoplasms at second surveillance colonoscopy

Beta coefficient
(95 %CI)

OR (95 %CI)

Absent
(n = 309)

Present
(n = 112)

Intercept

–1.52 (–3.17 to 0.13)

0.22 (0.04 to 1.14)

Female sex, n (%)

162 (52.4)

62 (55.4)

0.24 (–0.43 to 0.48)

1.02 (0.65 to 1.61)

Age, mean (SD), years

62.2 (9.4)

62.7 (8.7)

0.01 (–0.02 to 0.03)

1.00 (0.98 to 1.03)

TNM stage, n (%)

  • Stage I

96 (31.1)

51 (45.5)

  • Stage II

114 (36.9)

33 (29.5)

–0.48 (–1.02 to 0.05)

0.62 (0.36 to 1.05)

  • Stage III

96 (31.1)

27 (24.1)

 – 0.57 ( – 1.13 to – 0.01)

0.57 (0.32 to 1.00)

  • Stage IV

3 (1.0)

1 (0.9)

–0.29 (–2.63 to 2.04)

0.75 (0.07 to 7.68)

History of left-sided colon cancer, n (%)

175 (56.6)

78 (69.6)

0.50 (0.02 to 0.97)

1.64 (1.02 to 2.64)

≥ 1 advanced adenoma at index colonoscopy, n (%)

37 (12.0)

24 (21.4)

0.64 (0.05 to 1.23)

1.90 (1.05 to 3.43)

≥ 1 adenoma at first surveillance colonoscopy, n (%)

86 (27.8)

50 (44.6)

0.72 (0.26 to 1.18)

2.06 (1.29 to 3.27)

OR, odds ratio; CI, confidence interval; SD, standard deviation.

Beta coefficient, OR, and 95 %CI computed by a multivariable logistic regression model adjusted for all variables in the table.

We defined a patient with a history of right-sided colon cancer, no advanced adenomas at the index colonoscopy, and no adenomas at the first surveillance colonoscopy as a “low-risk” patient. When we excluded this patient group from the second surveillance colonoscopy, the diagnostic accuracy parameters of our model were as follows: 89.3 % (95 %CI 82.0 % to 94.3 %) and 78.2 % (95 %CI 65.0 % to 88.2 %) sensitivity, and 28.5 % (95 %CI 23.5 % to 33.9 %) and 33.8 % (95 %CI 26.2 % to 42.0 %) specificity, in the derivation and validation groups, respectively ([Table 3]). Considering a 26.6 % and 27.1 % prevalence of neoplasms at the second surveillance colonoscopy in the derivation and validation group, we obtained 88.0 % (95 %CI 80.0 % to 93.6 %) and 80.6 % (95 %CI 68.6 % to 89.6 %) NPV and 31.2 % (95 %CI 26.1 % to 36.5 %) and 30.5 % (95 %CI 23.0 % to 38.8 %) PPV, in the two cohorts, respectively. Three out of 22 (13.6 %) and 3 out of 20 (15.0 %) advanced adenomas would be missed in the derivation and validation cohorts, respectively, but no cancer would be missed in the validation cohort.

Table 3

Sensitivity and specificity for finding ≥ 1 neoplasm at the second surveillance colonoscopy, according to model-derived scenarios in the derivation and validation cohorts.

True positive

True negative

False negative

False positive

Sensitivity (95 %CI), %

Specificity (95 %CI), %

Derivation cohort

Absence of risk factors

100

88

12

221

89.3 (82 to 94.3)

28.5 (23.5 to 33.9)

History of left-sided colon cancer

63

195

49

114

56.3 (46.6 to 65.6)

63.1 (57.5 to 68.5)

≥ 1 advanced adenoma at index colonoscopy

57

207

55

102

50.9 (41.3 to 60.5)

67 (61.4 to 72.2)

≥ 1 adenoma at first surveillance colonoscopy

43

238

69

71

38.4 (29.4 to 48.1)

77 (71.9 to 81.6)

Validation cohort

Absence of risk factors

43

50

12

98

78.2 (65 to 88.2)

33.8 (26.2 to 42)

History of left-sided colon cancer

31

85

24

63

56.4 (42.3 to 69.7)

57.4 (49 to 65.5)

≥ 1 advanced adenoma at index colonoscopy

29

94

26

54

52.7 (38.8 to 66.3)

63.5 (55.2 to 71.3)

≥ 1 adenoma at first surveillance colonoscopy

23

111

32

37

41.8 (28.7 to 55.9)

75 (67.2 to 81.7)

CI, confidence interval.



Discussion

In this study, we found that patients with prior colon cancer who underwent cleansing colonoscopy had an occurrence of premalignant metachronous lesions of 32.3 % and 34.5 % at the first surveillance colonoscopy, and of 26.6 % and 27.1 % at the second surveillance colonoscopy, in the derivation and validation cohorts, respectively. Of note, the rate of metachronous CRCs was 2.4 % and 2.9 % at the first surveillance colonoscopy and decreased to 0 % and 1.5 % at the second surveillance colonoscopy in the two groups. We identified the following risk factors for metachronous neoplasms at the second surveillance colonoscopy: i) history of left-sided colon cancer, ii) having ≥ 1 advanced adenoma at the index colonoscopy, and iii) having ≥ 1 adenoma at the first surveillance colonoscopy. In an attempt at risk stratification, we provided a rule-out strategy to select patients who could safely skip the second surveillance colonoscopy. Indeed, if “low-risk” patients did not undergo the second colonoscopy, our model excluded a colorectal neoplasm with both sensitivity and NPV of about 90 % in the derivation cohort and about 80 % in the validation cohort. However, the model had low specificity and PPV. Nonetheless, we were much more interested in finding a “rule-out” strategy with high sensitivity and NPV in order to exclude patients at low risk of metachronous neoplasms from undergoing a second surveillance colonoscopy ([Fig. 2]).

Zoom
Fig. 2 Profiles of patients at high and low risk of metachronous neoplasms at the second surveillance colonoscopy, according to the predictive model. A risk-stratified strategy of endoscopic surveillance is proposed.

Current guidelines recommend performing a surveillance colonoscopy 1 year after surgery in order to detect anastomotic recurrence early and at a curable stage, as well as to identify precancerous and cancerous metachronous lesions [3] [15]. Published evidence supports the role of endoscopic postoperative surveillance and has shown that performing at least one surveillance colonoscopy in the first 5 years after surgery significantly reduces mortality [5] [6] [7]. Both a meta-analysis of randomized controlled trials (RCTs) and a recent RCT failed to show improved survival in patients undergoing more frequent colonoscopies [16] [17]. However, more recent series have shown that the actual risk of detecting metachronous cancer at subsequent examinations could be much lower [18]. Furthermore, a recent systematic review with meta-analysis of 27 endoscopy-based studies showed that most metachronous CRCs were detected during the first 2 – 3 years after surgery for primary cancer, with a substantial decrease in the incidence after 36 months [19]. These findings were confirmed by our results, as the rate of metachronous CRCs decreased from about 2 % – 3 % at first surveillance colonoscopy to 0 % – 1 % at second surveillance colonoscopy in the two cohorts. These findings can be at least partly explained by an increased detection of premalignant lesions at previous colonoscopies, probably as a result of an increased awareness by the endoscopist, better bowel cleansing, and improved performance of endoscopes. Indeed, better bowel cleansing has been associated with a higher adenoma detection rate [20], which inversely correlates with CRC occurrence and mortality [21] [22]. These findings have led the European Society of Gastrointestinal Endoscopy to recommend performing high-quality colonoscopy [23], which plays a crucial role in the surveillance setting, probably much more than shorter colonoscopy intervals.

Given the abovementioned considerations, and the costs associated with colon cancer endoscopic surveillance, it may seem reasonable to rationalize endoscopic surveillance by stratifying the risk of developing subsequent colorectal neoplasms, allowing the creation of customized surveillance programs. Data on the association between site of colon cancer and occurrence of metachronous CRC are conflicting [24]; however, the present study confirmed our previous finding that patients with prior left-sided colon cancer have an increased risk of adenomas in the residual colon [8]. This fact may have at least two explanations. First, right-sided colon cancer is more frequently associated with microsatellite instability, having been associated with better prognosis and reduction of recurrence risk [25] [26]. Second, right colectomy implies the resection of the terminal ileum and the ileocecal valve, which may be related to an accelerated bowel transit [27], thus reducing the contact time of potential carcinogenic substances within the residual colon. Our finding that advanced adenoma(s) at the index colonoscopy was a risk factor confirmed previous findings by Moon et al., who demonstrated an increased risk of metachronous adenomas in a cohort of 503 patients with prior surgery for CRC [28].

Therefore, the strength of our model relies principally on four factors. First, variables in our model are consistent with the published literature. Second, data can be easily extracted from the index and first surveillance colonoscopy reports. Third, the validation is remarkable as we applied a temporal approach, which is regarded as the strongest method [12], and the validating cohort was nearly as large as half the derivation group. Fourth, we decided to include in the composite end point of our model not only advanced adenomas or cancer, but also adenomas, both for consistency and to better define the low-risk patient. Thus, our model seems appealing as it is not time-consuming and it could save a considerable amount of resources.

Our study has some limitations. First, the retrospective design might have hampered our findings and we cannot exclude a selection bias, as the sample size is relatively small in contrast to the pathology volume of the centers and the study duration of more than 10 years. However, this can be partly explained by the fact that patients had to undergo all of the three colonoscopies at the same center, and that we included only complete colonoscopies with bowel cleansing explicitly reported as adequate. Second, the temporal validation, although being the most robust method is based on data derived from the same centers that constituted the derivation cohort. Third, we had no information on the third surveillance colonoscopy and therefore we could not assess the occurrence of metachronous neoplasms in “low-risk” patients.

In conclusion, we found that patients with prior left-sided colon cancer or ≥ 1 advanced adenoma at the index colonoscopy or ≥ 1 adenoma at the first surveillance colonoscopy were significantly more likely to have neoplasms at the second surveillance colonoscopy. Patients without such factors (i. e. patients with a history of right-sided colon cancer, no advanced adenoma at the index colonoscopy, and no adenoma at the first surveillance colonoscopy) had a substantially lower risk and could safely skip the second surveillance colonoscopy in view of cost-effectiveness. Nevertheless, a prospective, multicenter validation study is needed.



Competing interests

The authors declare that they have no conflict of interest.

Table 1s


Corresponding author

Lorenzo Fuccio, MD
Gastroenterology Unit, DIMEC
S. Orsola-Malpighi Hospital
University of Bologna
Via Massarenti 9
40138 Bologna
Italy   
Fax: +39-051-2149112   


Zoom
Fig. 1 Study flow chart. a Derivation cohort. b Validation cohort.
Zoom
Fig. 2 Profiles of patients at high and low risk of metachronous neoplasms at the second surveillance colonoscopy, according to the predictive model. A risk-stratified strategy of endoscopic surveillance is proposed.