Clinical features of post-colonoscopy colorectal cancer and real-world multi-scale correlation analysis

Abstract

BACKGROUND

Post-colonoscopy colorectal cancer (PCCRC) remains an important issue in endoscopic examination. This study investigates the clinical features of PCCRC and possible correlation factors.

AIM

To explore clinical features of PCCRC and correlation factors.

METHODS

A retrospective cohort analysis enrolled patients diagnosed with colorectal cancer (CRC) via colonoscopy at West China Hospital, Sichuan University, between January 1, 2022, and December 30, 2024. Demographic data, tumor characteristics, endoscopic findings, and miss records were extracted from electronic medical records and telephone follow-ups. An exploratory analysis was performed to identify causes of missed diagnosis during endoscopy.

RESULTS

Among 5411 colonoscopies in 2047 CRC patients, 66 prior examinations (27 colonoscopies in 17 non-PCCRC patients; 39 colonoscopies in 25 PCCRC patients) failed to establish diagnosis. The overall miss rate was 1.2%, with a PCCRC rate of 0.7%. Compared to the non-PCCRC group, advanced age was significantly associated with PCCRC (P = 0.006). The most common location that occurred PCCRC was sigmoid colon. PCCRC cases had higher rate of prior CRC surgery (41.0%). For endoscopists, PCCRC cases with CRC surgery increased the risk of judgement error. Insertion time demonstrated a positive correlation with missed diagnosis risk, whereas withdrawal time exhibited a negative correlation.

CONCLUSION

The incidence of PCCRC remains significant. Beyond tumor characteristics, endoscopist proficiency and procedural factors critically impact detection accuracy.

Key Words: Post-colonoscopy colorectal cancer; Gastrointestinal endoscopy; Colorectal neoplasms; Colonoscopy quality; Endoscopist

Core Tip: The incidence of post-colonoscopy colorectal cancer (CRC) remains significant. Patients with advanced age or prior CRC surgical history are associated with an elevated risk of post-colonoscopy CRC and may warrant shorter surveillance intervals for follow-up colonoscopy. The proficiency of the endoscopist can also influence the incidence of post-colonoscopy CRC. Enhancing technical proficiency among endoscopists and prolonging withdrawal time during colonoscopy may reduce post-colonoscopy CRC risk.

INTRODUCTION

Colorectal cancer (CRC) is the third most common cancer and the second leading cause of cancer-related deaths worldwide[1]. Post-colonoscopy CRC (PCCRC) refers to colorectal carcinomas diagnosed within a defined interval after a complete colonoscopy in which no malignancy was detected during the index examination[2]. According to the World Endoscopy Organization criteria, PCCRC encompasses cancers diagnosed 6-36 months post-procedure[3]. This metric has emerged as a crucial quality indicator for colonoscopy, as high-quality examinations should prevent CRC through early detection[4].

Reported PCCRC rates vary across studies, with pooled estimates ranging from 0.7% to 9.0%[5-7]. This heterogeneity reflects disparities in colonoscopy quality, including operator expertise, technological resources, and patient preparation[8]. Notably, PCCRC represents not only a healthcare quality deficit but also correlates with adverse prognoses; studies have demonstrated significantly higher cancer-specific mortality in patients with PCCRC than in their non-PCCRC counterparts[9].

With expanding global CRC screening initiatives, reducing PCCRC incidence has become a critical objective in both clinical practice and public health[8,10]. However, dedicated research remains limited. This retrospective cohort study investigated the clinical characteristics of PCCRC to provide clinicians with diagnostic insights and inform them about procedural optimization.

MATERIALS AND METHODS

Study design and data source

This retrospective cohort study was conducted at West China Hospital of Sichuan University. The study protocol was approved by the hospital’s Institutional Review Board. Given the retrospective nature of the study, the Institutional Review Board granted a waiver for the requirement of informed consent.

Study population

Patients meeting the following criteria were included: (1) Those diagnosed with CRC via endoscopic histopathological biopsy at West China Hospital between January 1, 2022, and December 30, 2024; and (2) Those with documented colonoscopy records within the ten years preceding CRC diagnosis. The exclusion criteria were as follows: (1) Incomplete medical records; and (2) Indeterminate localization of the anorectal lesions. A total of 2047 histologically confirmed CRC cases and 5411 colonoscopies were included and categorized into three cohorts.

Confirmed diagnosis cohort

A total of 2005 patients (97.9%) with 5345 prior colonoscopies (98.8%) had colorectal neoplasia successfully detected.

Non-PCCRC cohort

Seventeen patients (0.8%) with 27 prior colonoscopies (0.5%) had missed lesions that did not meet the 6- to 36-month interval criterion.

PCCRC cohort

Twenty-five patients (1.2%) with 39 prior colonoscopies (0.7%) had missed lesions that met the 6- to 36-month interval criterion.

Definitions

Based on criteria established in prior research: (1) PCCRC: A case in which CRC was diagnosed 6-36 months after a colonoscopy that failed to detect colorectal neoplasia; (2) Non-PCCRC: A case in which CRC was diagnosed after a colonoscopy that failed to detect colorectal neoplasia, but the interval between the negative colonoscopy and diagnosis fell outside the 6-36 months window; (3) Confirmed diagnosis: A case with pathologically confirmed CRC with no history of negative colonoscopy results before diagnosis; (4) Overall miss-rate: Calculated as the sum of PCCRC and non-PCCRC cases divided by the total number of eligible patients and expressed as a percentage; (5) PCCRC rate: Calculated as the number of PCCRC cases divided by the total number of eligible patients, expressed as a percentage; (6) High miss-rate endoscopist: An endoscopist who performed > 2000 colonoscopies during the three-year study period and whose PCCRC rate was within the highest quartile; (7) Zero-miss endoscopist: An endoscopist who performed > 2000 colonoscopies during the study period and had zero PCCRC cases; and (8) Low miss-rate endoscopist: An endoscopist who performed > 2000 colonoscopies during the study period but was not classified as either high miss-rate or zero-miss.

All colonoscopy reports and corresponding imaging data were meticulously reviewed to ascertain the reason for the missed lesion and were categorized as follows: (1) Exposure error: Lesions were missed due to inadequate visualization of the colonic mucosa during index colonoscopy; (2) Judgement error: Lesions visible in archived endoscopic images but not recognized or diagnosed as neoplastic lesions during the initial procedure; (3) Biopsy error: Lesions documented, reported, and biopsied during the index colonoscopy with pathology results interpreted as non-neoplastic; and (4) Poor preparation: Visualization hampered by poor bowel preparation quality.

Data collection

The following data were collected for analysis: (1) Demographics: Age and sex; (2) Endoscopic data: Number of prior colonoscopies and details of index colonoscopies (date, endoscopic diagnosis, and availability of archived high-definition images); (3) Tumor characteristics: Location, histological type, and differentiation grade; and (4) Endoscopist characteristics: Total lifetime colonoscopy volume, number of polyps ≥ 5 mm detected, mean insertion time, mean withdrawal time, and mean number of photographic documentation images per procedure. All data were cross-verified by two independent researchers to ensure completeness and accuracy, thereby minimizing the risk of data entry errors.

Statistical analysis

Continuous variables were described using the mean (SD), and categorical variables were presented as frequencies and percentages. For between-group comparisons of continuous variables, the Welch t-test or analysis of variance was used. For between-group comparisons of categorical variables, Fisher’s exact test was employed when the expected frequency was < 5; otherwise, the χ² test was used. Post-hoc power analysis revealed a statistical power of 100%. For the confirmed diagnosis, non-PCCRC, and PCCRC groups, multinomial logistic regression was performed using PCCRC as the reference group to analyze the associations between various clinical characteristics. For endoscopists categorized as zero-miss, low miss-rate, and high miss-rate, ordinal logistic regression was used to evaluate the factors influencing miss rates. All statistical analyses were conducted using the R software (version 5.4.1).

RESULTS

Baseline characteristics

This study included 5411 endoscopies from 2047 patients with pathologically confirmed colorectal tumors stratified into three groups based on missed diagnoses and PCCRC. Age distribution differed significantly among the groups (P = 0.008). The PCCRC group had a significantly higher proportion of patients aged ≥ 60 years (71.79%) compared with the other groups. Tumor location varied: The rectum was the most common site in the missed diagnosis group (58.15%), whereas the sigmoid colon was the most frequently missed location in the PCCRC group. The PCCRC group also had a significantly higher proportion of anastomotic-site tumors (20.51%). The detailed baseline characteristics are summarized in Table 1. Cases of missed CRC are presented in Figure 1.

Figure 1 Cases of missed colorectal cancer. A: Four months after rectal cancer surgery, colonoscopy revealed polypoid changes near the anastomotic site. Post-polypectomy pathology indicated tubular adenoma with high-grade intraepithelial neoplasia; B: Three months after polypectomy, colonoscopy showed no abnormalities; C: Four months later, follow-up colonoscopy confirmed adenocarcinoma at the anastomotic site; D: Five months after rectal cancer surgery, follow-up colonoscopy suggested an anastomotic fistula at the anastomotic site, but no biopsy was performed; E: Two years later, colonoscopy indicated swelling of the anastomotic mucosa with undetermined nature; biopsy pathology revealed moderate chronic mucosal inflammation; F: Nine months later, follow-up colonoscopy confirmed adenocarcinoma at the anastomotic site; G: During the initial colonoscopy, no lesions were identified in the rectum; H: Two years later, repeat colonoscopy confirmed adenocarcinoma in the rectum.

Table 1 Baseline characteristics, n (%).

Characteristic	Confirmed diagnosis	Non-PCCRC	PCCRC	P value
Sex				0.3551
Male	3306 (61.85)	20 (74.07)	26 (66.67)
Female	2039 (38.15)	7 (25.93)	13 (33.33)
Age				0.0081
< 60	2523 (47.20)	18 (66.67)	11 (28.21)
≥ 60	2822 (52.80)	9 (33.33)	28 (71.79)
Location				< 0.0012
Ileocecal region	130 (2.43)	1 (3.70)	4 (10.26)
Ascending colon	314 (5.87)	1 (3.70)	3 (7.69)
Hepatic flexure	125 (2.34)	1 (3.70)	0 (0.00)
Transverse colon	122 (2.28)	4 (14.81)	1 (2.56)
Splenic flexure	21 (0.39)	0 (0.00)	0 (0.00)
Descending colon	197 (3.69)	3 (11.11)	4 (10.26)
Sigmoid colon	1266 (23.69)	6 (22.22)	12 (30.77)
Rectum	3108 (58.15)	10 (37.04)	7 (17.95)
Anastomotic site	62 (1.16)	1 (3.70)	8 (20.51)
Sedation				0.0461
Present	3019 (56.48)	18 (66.67)	29 (74.36)
Absent	2326 (43.52)	9 (33.33)	10 (25.64)
Bowl preparation				0.0042
Excellent	590 (11.04)	0 (0.00)	1 (2.56)
Good	2118 (39.63)	14 (51.85)	15 (38.46)
Inadequate	927 (17.34)	11 (40.74)	9 (23.08)
Poor	1015 (18.99)	2 (7.41)	11 (28.21)
Unclear	695 (13.00)	0 (0.00)	3 (7.69)
Adenoma3				0.7741
Present	1264 (65.12)	15 (60.00)	12 (70.59)
Absent	677 (34.88)	10 (40.00)	5 (29.41)

¹Pearson’s χ² test.

²Simulated P value.

³n was calculated as the total number of endoscopic procedures performed (n = 5411), while n was calculated based on the number of unique patients (n = 2047).

PCCRC: Post-colonoscopy colorectal cancer.

Clinical features of postoperative PCCRC

A total of 16 cases (25.6%) in the PCCRC group had a history of prior CRC surgery. Regarding missed diagnosis causes, “inadequate visualization” ranked highest in both groups (73.91% vs 62.50%, P = 0.260), while “failure to recognize” occurred exclusively in the surgery group (18.75%). A statistically significant difference was observed in tumor location between the groups (P < 0.001), with sigmoid colon lesions predominating in the non-surgery group (47.83%) and anastomotic site involvement being the most common in the surgery group (56.25%). The mean diagnostic delay was significantly longer in patients without surgical history (22 ± 10 months vs 16 ± 8 months, P = 0.039). The detailed PCCRC characteristics are presented in Table 2.

Table 2 Clinical characteristics of post-colonoscopy colorectal cancer, n (%).

Characteristic	CRC surgery history		P value
	Absent (n = 23)	Present (n = 16)
Sex			0.1071
Male	13 (56.52)	13 (81.25)
Female	10 (43.48)	3 (18.75)
Age			0.7342
< 60	6 (26.09)	5 (31.25)
≥ 60	17 (73.91)	11 (68.75)
Reason			0.2602
Poor preparation	4 (17.39)	2 (12.50)
Exposure error	17 (73.91)	10 (62.50)
Biopsy error	2 (8.70)	1 (6.25)
Judgement error	0 (0.00)	3 (18.75)
Location			< 0.0012
Ileocecal region	1 (4.35)	3 (18.75)
Ascending colon	3 (13.04)	0 (0.00)
Transverse colon	1 (4.35)	0 (0.00)
Descending colon	2 (8.70)	2 (12.50)
Sigmoid colon	11 (47.83)	1 (6.25)
Rectum	5 (21.74)	1 (6.25)
Anastomotic site	0 (0.00)	9 (56.25)
Differentiation			0.1172
Well-differentiated	1 (4.35)	4 (25.00)
Moderately differentiated	15 (65.22)	6 (37.50)
Poorly differentiated	7 (30.43)	6 (37.50)
Missed diagnosis duration (month), mean ± SD	22 ± 10	16 ± 8	0.0393

¹Pearson’s χ² test.

²Fisher’s exact test.

³Welch two sample t-test.

CRC: Colorectal cancer.

Influencing factors for PCCRC: Non-PCCRC vs PCCRC and confirmed diagnosis vs PCCRC

Age ≥ 60 years was significantly associated with reduced likelihood of non-PCCRC compared to younger patients [non-PCCRC vs PCCRC, odds ratio (OR) = 0.22, 95% confidence interval (CI): 0.08-0.64, P = 0.006]. Lesions at the anastomotic site were associated with decreased odds of confirmed diagnosis (confirmed diagnosis vs PCCRC, OR = 0.07, 95%CI: 0.03-0.18, P < 0.001). Conversely, rectal lesions were associated with increased odds of confirmed diagnosis (confirmed diagnosis vs PCCRC, OR = 3.64, 95%CI: 1.42-9.36, P = 0.007). The detailed analysis results are presented in Table 3.

Table 3 Multinomial regression analysis, post-colonoscopy colorectal cancer vs confirmed diagnosis and non-post-colonoscopy colorectal cancer.

Outcome	Characteristic	n	Event n	OR	95%CI	P value
Non-PCCRC	Sex
	Female	2059	7	Reference	Reference	-
	Male	3352	20	1.47	0.49-4.47	0.493
	Age
	< 60	2552	18	Reference	Reference	-
	≥ 60	2859	9	0.22	0.08-0.64	0.006
	Location
	Sigmoid colon	1284	6	Reference	Reference	-
	Ileocecal region	135	1	0.41	0.04-4.67	0.471
	Ascending colon	318	1	0.61	0.05-7.40	0.696
	Hepatic flexure	126	1	2.97 × 109	1.01 × 109, 8.71 × 109	< 0.001
	Transverse colon	127	4	8.50	0.75, 96.48	0.084
	Splenic flexure	21	0	0.01	0.01-0.01	< 0.001
	Descending colon	204	3	1.62	0.26-9.92	0.603
	Rectum	3125	10	2.45	0.61-9.84	0.207
	Anastomotic site	71	1	0.31	0.03-3.25	0.327
	Sedation
	Absent	2345	9	Reference	Reference	-
	Present	3066	18	0.48	0.14-1.59	0.229
	Bowl preparation
	Good	2147	14	Reference	Reference	-
	Excellent	591	0	0.00	-	-
	Inadequate	947	11	1.06	0.33-3.41	0.917
	Poor	1028	2	0.21	0.04-1.17	0.075
	Unclear	698	0	0.00	0.00-0.00	< 0.001
Confirmed diagnosis	Sex
	Female	2059	2039	Reference	Reference	-
	Male	3352	3306	0.83	0.42-1.65	0.598
	Age
	< 60	2552	2523	Reference	Reference	-
	≥ 60	2859	2822	0.50	0.24-1.02	0.056
	Location
	Sigmoid colon	1284	1266	Reference	Reference	-
	Ileocecal region	135	130	0.31	0.10-1.01	0.052
	Ascending colon	318	314	1.10	0.30-4.04	0.880
	Hepatic flexure	126	125	1.95 × 109	6.64 × 108, 5.73 × 109	< 0.001
	Transverse colon	127	122	1.34	0.17-10.48	0.782
	Splenic flexure	21	21	1.19 × 109	1.19 × 105, 1.19 × 105	< 0.001
	Descending colon	204	197	0.50	0.16-1.57	0.232
	Rectum	3125	3108	3.64	1.42-9.36	0.007
	Anastomotic site	71	62	0.07	0.03-0.18	< 0.001
	Sedation
	Absent	2345	2326	Reference	Reference	-
	Present	3066	3019	0.64	0.29-1.41	0.269
	Bowl preparation
	Good	2147	2118	Reference	Reference	-
	Excellent	591	590	2.12	0.26-17.42	0.483
	Inadequate	947	927	0.63	0.27-1.47	0.286
	Poor	1028	1015	0.74	0.32-1.70	0.485
	Unclear	698	695	1.78	0.50-6.40	0.374

CI: Confidence interval; OR: Odds ratio; PCCRC: Post-colonoscopy colorectal cancer.

Factors associated with endoscopist missed diagnosis multivariable regression analysis revealed that an increase in insertion time increases the risk of missed diagnosis (adjusted OR = 0.98, 95%CI: 0.95-1.00, P = 0.045). Increase in withdrawal time reduces the risk of missed diagnosis (adjusted OR = 1.02, 95%CI: 1.01-1.05, P = 0.011). The endoscopist’s mean number of images captured per procedure and polyp detection rate (PDR) (≥ 5 mm) were not significantly associated with their miss rate. The detailed analysis results are presented in Table 4.

Table 4 Univariate and multivariate analyses of influencing factors among doctors (ordered logistic regression).

Characteristic	Univariable				Multivariable
	n	OR	95%CI	P value	n	OR	95%CI	P value
Withdrawal time (second)	29	0.99	0.97-1.01	0.224	29	0.98	0.95-1.00	0.045
Insertion time (second)	29	1.02	1.01-1.04	0.016	29	1.02	1.01-1.05	0.011
Standardized count of detected polyps (> 5 mm)	29	1.00	1.00-1.00	0.184	29	1.00	1.00-1.00	0.640
Picture	29	1.06	0.98-1.15	0.156	29	1.01	0.90-1.13	0.903

CI: Confidence interval; OR: Odds ratio.

DISCUSSION

PCCRC, defined as CRC diagnosed within a specified interval following colonoscopy examination, serves as a direct indicator of colonoscopy quality and effectiveness. Large-scale epidemiological studies have revealed significant geographical variations in PCCRC rates[11]. A Swedish nationwide cohort study (2001-2010) encompassing 16319 patients with CRC reported a PCCRC rate of 7.9%[12]. United Kingdom hospital statistics (2003-2009) indicated a substantially higher proportion of PCCRC (12.1%) among 67202 CRC cases[13]. Conversely, an Australian study (2011-2018) documented a lower PCCRC rate (within 3 years) of 2.16%[14]. In the present study, the PCCRC incidence was 0.7%, comparable to the rates (0.7%-1.7%) reported for high-definition colonoscopy in a Japanese study[6].

The prevalence of PCCRC is high in certain patient populations. Multiple studies have consistently demonstrated an elevated PCCRC risk among older adults, women, and those with a history of inflammatory bowel disease (IBD) or diverticular disease[4,5,14]. A United Kingdom multicenter study revealed that 61% of IBD-associated CRCs were classified as PCCRC, a proportion significantly higher than that in the general CRC population[15]. Swedish registry data indicated a markedly increased PCCRC risk in IBD[16]. In this study, compared to the confirmed diagnosis and non-PCCRC groups, the probability of PCCRC occurrence was significantly higher in individuals ≥ 60 years of age. This suggests that shortening the surveillance interval for colonoscopy in patients of advanced age may be warranted.

Tumor location is also a significant influencing factor[9,17]. A meta-analysis revealed a significantly higher incidence of proximal PCCRC (9.7%) than of distal PCCRC (5.4%)[7]. In the present study, missed neoplasia were predominantly located in the sigmoid colon, which might be attributed to limited PCCRC data or racial group differences. Furthermore, a history of prior intestinal surgery is also a significant risk factor for PCCRC, which may be explained by the following reasons: First, altered anatomical structures may impair endoscopic visualization and reduce CRC detection rates. Second, endoscopists may lack adequate knowledge about anastomotic site lesions. An important reason for missed PCCRC in these cases was a judgment error; a review of endoscopic images confirmed that the lesions were photographed but incorrectly identified as benign.

Colonoscopy quality-related factors, such as PDR, may affect PCCRC rates. Multiple studies have reported an inverse correlation between adenoma detection rate or PDR and the risk of PCCRC[8,18-20]. However, in our study, the PDR for polyps ≥ 5 mm did not significantly impact the endoscopist miss rate. This may be because the PCCRC rate was relatively low at our center. Colonoscopy withdrawal time influenced the quality of colonoscopy[21]. In our study, longer insertion time was associated with a higher risk of PCCRC, potentially reflecting lower endoscopist proficiency. Conversely, a longer withdrawal time was associated with a lower PCCRC risk, indicating that an adequate mucosal inspection time effectively reduces PCCRC occurrence. The average number of images captured per procedure showed no significant difference between groups, likely because all exceeded the minimum threshold of 40 images. Although image quality (anatomical coverage and clarity) may hold greater research value than quantity alone, data for such analyses were not available in this study.

The strengths of this study lie in its multidimensional approach, integrating data on patient and tumor characteristics as well as endoscopist-related factors to comprehensively analyze potential contributors to PCCRC. Notably, our findings highlight the significant impact of the complex intestinal environment at anastomotic sites resulting from prior surgery on PCCRC risk. Furthermore, the use of consecutive patient data from a single high-volume center ensures consistency in data quality and management. Moving forward, we plan to conduct prospective, multicenter studies to further validate our conclusions, and will incorporate more comprehensive clinical data such as history of familial adenomatous polyposis and IBD to enhance our research.

This study has some limitations that warrant a cautious interpretation of the results. First, as this was a single-center retrospective study, selection bias may have compromised the reliability of our conclusions. Second, the subjective classification of missed lesions lacks standardized diagnostic criteria or blinded expert reviews, which introduce a potential misclassification bias. Third, the small sample size precluded meaningful subgroup analyses such as assessing the impact of sedation or specific lesion locations on PCCRC.

CONCLUSION

This study identified key clinical characteristics associated with PCCRC. Enhancing endoscopist training, standardizing examination protocols, and utilizing advanced endoscopic technologies are essential for achieving high-quality examinations.