Higher Sessile Serrated Lesion Detection Rates Calculated Using All Examinations Are Associated With Lower Risk for Postcolonoscopy Colorectal Cancer: Data From the New Hampshire Colonoscopy Registry

Abstract

INTRODUCTION:

Calculating detection rates using data from colonoscopies for all indications, rather than screening examinations, is simpler and can mitigate gaming by endoscopists. We hypothesized that calculating sessile serrated lesion detection rates (SSLDR-A) using all examinations may also be a quality metric for predicting postcolonoscopy colorectal cancer (PCCRC) risk.

METHODS:

The cohort included New Hampshire Colonoscopy Registry 115,762 patients with an index colonoscopy. The primary outcome was PCCRC, defined as colorectal cancer (CRC) diagnosed ≥6 months after the index colonoscopy. The exposure variables were endoscopist-specific SSLDR-A (using all examinations) and SSLDR-S (using screening examinations), stratified into quintiles. Cox regression was used to model the hazard of PCCRC on SSLDR, adjusting for relevant covariates, such as patient age and sex.

RESULTS:

There were 177 PCCRCs diagnosed in 115,762 patients with index colonoscopies. Higher SSLDR-A and SSLDR-S rates were associated with lower PCCRC risks. After adjusting for covariates, we observed that higher SSLDR-A rates were associated with lower hazard ratios (HRs) as compared with the reference group (SSLDR-A: <1.5%; HR = 1.0 vs SSLDR-A: 1.5% to <3.0%; HR = 0.53, 95% CI 0.35–0.79; SSLDR-A: 3.0% to <5.0%; HR = 0.59, 95% CI 0.38–0.92; SSLDR: 5.0% to <8.0%; HR = 0.44, 95% CI 0.28–0.70; and SSLDR: 8.0+%; HR = 0.20, 95% CI 0.08–0.46). The highest quintile of SSLDR-A (8.0%+) (HR = 0.20, 95% CI 0.08–0.46) and SSLDR-S (8.0%+) (HR = 0.20, 95% CI 0.09–0.44) provided similar protection from PCCRC.

DISCUSSION:

These findings demonstrate that colonoscopies performed by endoscopists with higher SSLDR-A are associated with a lower risk of PCCRC, validating SSLDR-A as a quality metric. Furthermore, our data suggest that endoscopists should aim for an SSLDR-A of 6% and have an aspirational SSLDR-A of 8.0% or higher.

INTRODUCTION

Colonoscopies prevent colorectal cancer (CRC) through the detection and resection of precancerous polyps known as adenomas or serrated polyps. Optimal detection is important for colonoscopy to be effective in CRC prevention. The adenoma detection rate (ADR), which is an important colonoscopy quality indicator, is inversely associated with postcolonoscopy colorectal cancer (PCCRC) incidence (1,2).

Although ADR had been calculated using screening colonoscopies alone, excluding diagnostic and surveillance procedures, recent recommendations from the joint American College of Gastroenterology (ACG)/American Society of Gastrointestinal Endoscopy (ASGE) task force suggest calculating ADR using all examinations (3–6). There are many benefits to using data from colonoscopies performed for all indications (ADR-A), rather than screening examinations alone (ADR-S), including a larger sample size, which can provide a more accurate assessment of an endoscopist's overall performance. It also mitigates gaming by endoscopists to manipulate their metrics (7).

In addition to adenomas, serrated polyp detection is also important to the prevention of CRC because serrated lesions may account for approximately one-third of all CRCs and a substantial portion of postcolonoscopy interval cancers. Serrated polyps include hyperplastic polyps (HPs), sessile serrated lesions (SSLs), and traditional serrated adenomas of any size. As opposed to HPs, SSLs can develop dysplasia and progress to CRC and thus are considered significant lesions. A sessile serrated lesion detection rate for screening examinations (SSLDR-S) is defined as the number of screening colonoscopies with at least 1 SSL divided by the total number of screening colonoscopies for each endoscopist. An analysis of New Hampshire Colonoscopy Registry (NHCR) data suggests that a SSLDR-S of 6% may offer the most protection from PCCRC (8).

However, no data are available on a new metric of SSLDR calculated across all examinations (SSLDR-A), similar to ADR-A. SSLDR-A is calculated by dividing the number of colonoscopies for all examinations with at least 1 SSL by the total number of colonoscopies performed for each endoscopist. We hypothesized that calculating sessile serrated lesion detection rates (SSLDR-A) using all examinations may be a quality metric for predicting PCCRC risk. Using NHCR data, we examined the association between SSLDR-A and PCCRC risk. We also compared SSLDR-A with all screening examinations to determine the most appropriate measure of endoscopist performance.

METHODS

Population

In New Hampshire, patients who undergo colonoscopy are invited to participate in the NHCR, which collects longitudinal data on all colonoscopies performed. Before the procedure, patients complete a questionnaire capturing patient demographics, medical history, as well as personal and family history of polyps and CRC. Endoscopy staff complete the NHCR Procedure Form during or immediately after colonoscopy. Bowel preparation quality for every colonoscopy is determined by identifying the worst-prepared segment after cleaning, using NHCR form criteria. Since 2004, the NHCR Procedure Form has provided 4 specific categories of bowel preparation: excellent (only scattered, tiny particles and/or clear liquid—100% visualization possible throughout colon), good (easily removable small amounts of particles and/or liquid—very unlikely to impair visualization throughout colon), fair (residual feces and/or nontransparent fluid—possibly impairing visualization), and poor (feces and/or nontransparent fluid—definitely impairing visualization). Trained personnel then abstract the data, linking pathology results to corresponding procedural findings. Data collection was approved by the Committee for the Protection of Human Subjects at Dartmouth College (CPHS#00015834).

Analyzed sample

The primary analysis included all patients with an index colonoscopy in the NHCR. Exclusion criteria included a personal history of inflammatory bowel disease, familial genetic cancer syndromes (including Lynch syndrome, polyposis syndromes, and serrated polyposis syndromes), or CRC diagnosed at index or within 6 months of the index examination.

Outcomes

The primary outcome was PCCRC, defined as CRC diagnosed in either the NHCR or the NH Cancer Registry ≥6 months after the index colonoscopy.

Exposure variable

The primary exposure variables were endoscopist-specific SSLDR-A (calculated using all examinations) and SSLDR-S (calculated using screening examinations only). The SSLDRs were calculated using all colonoscopies in the database that met the respective criteria for each rate. SSLDR-A and SSLDR-S were calculated for each endoscopist by dividing the total number of colonoscopies performed in patients 45 years or older with adequate bowel preparation and at least 1 sessile serrated lesion by the total number of complete examinations with adequate bowel preparation. SSLDR-A included all diagnostic, screening, and surveillance examinations, excluding procedures performed for positive fecal immunochemical tests or fecal occult blood test, inflammatory bowel disease, or genetic syndromes. SSLDR-S was calculated similarly, but only screening colonoscopies were used in the calculations. We stratified endoscopists' SSLDR-S and SSLDR-A into quintiles. We used the lowest quintile as the reference group as has been performed in similar previously published articles (1,2).

Covariates

Covariates adjusted for in these models included age, sex, findings of clinically significant serrated polyps on index colonoscopy or conventional advanced adenomas, the indication of the index examination surveillance, quality of bowel preparation, and family history of CRC.

Statistical analysis

Means and SDs were calculated for continuous variables. Frequencies and percentages were derived for categorical variables. Categorical variables were compared using the χ² test for trend and the Fisher exact test. Continuous variables were analyzed using T tests or Mann-Whitney U tests, depending on the distribution. Pearson correlation was used to assess the strength of the association between 2 continuous variables. Cox proportional hazards regression was used to model the hazard of PCCRC in relation to SSLDR, adjusting for relevant covariates as described above. We censored the individuals at the second colonoscopy or at the time of diagnosis of CRC in the NHCR or the NH Cancer Registry. For those without a follow-up colonoscopy, the time for censoring was the time of linkage to the NH State Cancer Registry at least 18 months after index colonoscopy, ensuring adequate time for diagnosis and reporting to the state registry. We also examined the risk for PCCRC for ADR calculated for all examinations as stratified by SSLDR.

The ratio of the 95% confidence interval (CI) for SSLDR-S to that for SSLDR-A was also calculated. We also examined the shift in ranking of the endoscopists when using SSLDR-A versus SSLDR-S. All statistical analyses were conducted using SPSS version 29 (IBM, Armonk, NY).

RESULTS

The cohort included 115,762 patients in the NHCR who underwent an index colonoscopy. Procedures were performed by 126 endoscopists, with at least 50 examinations each. Index examinations were conducted between October 2004 and December 2020. There were 177 PCCRCs diagnosed. Both higher SSLDR-A and SSLDR-S rates were associated with lower PCCRC risks (Table 1). The mean time for follow-up for those who were not diagnosed with-PCCRC was 64.12 months (SD 32.69; 95% CI 63.93–64.31), while the mean time to development of PCCRC was 56.81 months (SD 34.41; 95% CI 51.70–61.91).

Table 1.

Characteristics of the patients and colonoscopies with and without PCCRC

Characteristic	Patients with no PCCRC	Patients with PCCRC	P value
Patients	N = 115,585	N = 177
Sex (% male)	46.2% (53,213)	44.1% (78)	0.58
Age (avg ± SD)	58.2 (11.0)	65.1 (10.6)	<0.001
First degree relative with CRC	17.4% (20,145)	19.2% (34)	0.53
Follow-up (avg months ± SD)	64.1 (32.7)	56.8 (34.4)	0.0001

CRC, colorectal cancer; PCCRC, postcolonoscopy colorectal cancer.

After adjusting for covariates, we observed that higher SSLDR-A rates were associated with lower hazard ratios (HRs) as compared with the reference group (SSLDR-A: <1.5%; HR = 1.0 vs SSLDR-A: 1.5 to <3.0%; HR = 0.53, 95% CI 0.35–0.79; SSLDR-A: 3.0 to <5.0%; HR = 0.59, 95% CI 0.38–0.92; SSLDR: 5.0% to <8.0%; HR = 0.44, 95% CI 0.28–0.70; and SSLDR: 8.0+% HR = 0.20, 95% CI 0.08–0.46). The highest quintile of SSLDR-A (8.0%+) (HR = 0.20, 95% CI 0.08–0.46) and SSLDR-S (8.0%+) (HR = 0.20, 95% CI 0.09–0.44) provided similar protection from PCCRC and the highest protection overall (Table 2). When examining the exposure variables as continuous, we observed that for each 1% increase in the SSLDR-A, there was a 13.0% decrease in the HR for PCCRC (HR 0.87; 95% CI, 0.82–0.93). This was similar to that for SSLDR (HR 0.87; 95% CI, 0.82–0.93).

Table 2.

Unadjusted risks and adjusted hazard ratios for postcolonoscopy colorectal cancer for quintiles of endoscopist SSLDR-S and SSLDR-A

SSLDR-S
	<1.0	1.0 to <2.5	2.5 to <4.5	4.5 to <8.0	8.0+
N	16,710	31,986	20,105	29,141	17,820
PCCRC (N)	62	53	28	27	7
PCCRC (%)a	0.37	0.17	0.14	0.09	0.04
HR (95% CI)	1.0 (reference)	0.42 (0.29–0.62)	0.59 (0.37–0.94)	0.38 (0.24–0.60)	0.20 (0.09–0.44)
Endoscopists (N)	25	28	22	26	25

SSLDR-A
	<1.5	1.5 to <3.0	3.0 to <5.0	5.0 to <8.0	8.0+
N	25,972	21,172	23,421	28,768	16,429
PCCRC (N)	79	36	30	26	6
PCCRC (%)a	0.30	0.17	0.13	0.09	0.04
HR (95% CI)	1.0 (reference)	0.53 (0.35–0.79)	0.59 (0.38–0.92)	0.44 (0.28–0.70)	0.20 (0.08–0.46)
Endoscopists (N)	21	30	22	29	24

CI, confidence interval; HR, hazard ratio; PCCRC, postcolonoscopy colorectal cancer; SSLDR-A, all-examination sessile serrated lesion detection rates; SSLDR-S, screening sessile serrated lesion detection rate.

^aP < 0.001 (χ² for trend).

Table 3 summarizes the shift in endoscopist detection rates when comparing SSLDR-A and SSLDR-S. Most (74%) endoscopists remained in the same quintile (Table 3). However, 13 (10.3%) shifted to a higher quintile, while 20 (15.9%) moved to a lower quintile. When we examined the percentage of surveillance examinations for each endoscopist, we observed no difference between those who did not change quintile rank (%surveillance = 27.8%), those who moved down a quintile or 2 (31.3%) or those who moved up (27.2%).

Table 3.

Endoscopists' detection rates as stratified by quintiles when calculating detection rates as SSLDR-A compared with SSLDR-S

SSLDR-A
		<1.5	1.5 to <3.0	3.0 to <5.0	5.0 to <8.0	8.0+	Total SSLDR-A
SSLDR-S	<1.0	19	2	0	0	0	21
	1.0 to <2.5	6	19	4	1	0	30
	2.5 to <4.5	0	6	13	3	0	22
	4.5 to <8.0	0	1	5	20	3	29
	8.0+	0	0	0	2	22	24
	Total SSLDR-S	25	28	22	26	25	126

SSLDR-A, all-examination sessile serrated lesion detection rates; SSLDR-S, screening sessile serrated lesion detection rate.

Bold represents SSLDR-A and SSLDR-S are in the same quintile, italic represents SSLDR-A and SSLDR-S are in different quintiles.

To examine the impact of patients without a follow-up event, we conducted a sensitivity analysis by including patients with and without follow-up at 12 months before the last linkage data between the NHCR and NH State Cancer Registry.

There were more than twice the number of total colonoscopies as compared with screening examinations (mean = 2.06; SD = 0.66). In addition, the proportion of screening to total colonoscopies ranged from 16 to 85%. The 95% CI was narrower (median = 73%; interquartile range [IQR]: 0.09) for endoscopists when calculating SSLDR-A vs SSLDR-S. By contrast, when calculating the 95% CI for SSLDR-S compared with SSLDR-A, the CI increased nearly 2-fold (median = 137%; IQR 0.17).

We also stratified another metric ADR-A by SSLDR of 8% and observed that those examinations performed by an endoscopist with an adequate ADR ≥ 35% but an SSLDR < 8% had a higher HR for PCCRC than those performed by an endoscopists with an ADR ≥ 35% and an SSLDR ≥ 8%. These data are given in Table 4.

Table 4.

Association between ADR-A of 35% as stratified by SSLDR-A of 8%

	ADR-A < 35 + SSLDR-A < 8%	ADR-A > 35 + SSLDR-A < 8%	ADR-A > 35 + SSLDR-A > 8%
N	66,978	32,355	16,429
PCCRC (N)	126	45	6
PCCRC (%)a	0.19	0.14	0.04
HR (95% CI)	3.36 (1.46–7.77)	2.50 (1.05–5.96)	1.0 (reference)

ADR, adenoma detection rate; HR, hazard ratio; PCCRC, postcolonoscopy colorectal cancer; SSLDR-A, all-examination sessile serrated lesion detection rates.

^a<0.001.

DISCUSSION

Our analysis was designed to examine the association between a new metric for SSL detection using all examinations and the incidence of postcolonoscopy CRC. We observed that higher SSLDR-A rates were associated with lower HR as compared with the reference group with an SSLDR-A of <1.5%. The highest quintile of SSLDR-A (8.0%+) (HR = 0.20, 95% CI 0.08–0.46) had the highest overall protection. These findings demonstrate that colonoscopies performed by endoscopists with higher SSLDR-A are associated with a lower risk of PCCRC, validating SSLDR-A as a quality measure.

Notably, endoscopists with a SSLDR-A of 8.0% or higher had the lowest hazard ratio for PCCRC compared with those in the lowest quintile, supporting this as an aspirational benchmark. Recently published data regarding SSL prevalence support 8% as a reasonable benchmark for endoscopists to achieve, especially when performing high-quality colonoscopy (9,10). Although many of these studies were conducted in screening populations, we observed little difference between our screening SSLDR-S and the new SSLDR-A based on a novel calculation using all examinations regardless of indication. Although previous studies have proposed SSLDR targets, previous literature did not examine the impact of using all examinations for calculating SSLDR or provide a benchmark for SSLDR-A (5,9,11–14).

There are few published data examining the association between endoscopist serrated detection rates and PCCRC. Previous work by our group demonstrated that higher CSSDRs correlate with reduced PCCRC risk, supporting the role of serrated lesion detection metrics in quality assessment (15). A Dutch cohort study reported the median proximal serrated polyp detection rates of 11.9% (IQR 8.3–15.8) (16). They observed that for each percentage point increase in proximal serrated polyp detection rate resulted in a 7% lower risk of interval postcolonoscopy colorectal cancer.

Furthermore, we observed that SSLDR-S and SSLDR-A performed similarly overall in predicting PCCRC risk. Among endoscopists in the highest quintiles, hazard ratios were nearly identical (SSLDR-S HR = 0.20, 95% CI 0.09–0.44 vs SSLDR-A HR = 0.20, 95% CI 0.08–0.46). We also observed that most endoscopists (73.8%) remained in the same quintile rank, but about one-quarter moved to a different quintile. Among the 126 endoscopists, 31 (24.6%) differed by 1 quintile between SSLDR-S and SSLDR-A, while only 2 (1.9%) differed by more than 1 quintile. This is likely due to a variation in screening-to-diagnostic case mix, patient characteristics, bowel preparation quality distributions, or individual practice patterns, among other potential contributors. We examined the proportion of surveillance examinations among endoscopists and observed no differences between those who moved up or down a quintile or those who did not change quintiles. However, the numbers were too small to provide meaningful outcomes. Overall differences were minimal and consistent with previous ADR analyses comparing all versus screening-only examinations, further validating the reliability of SSLDR-A in performance stratification. These data suggest that calculating SSLDR based on all examinations comparably ranks endoscopists based on SSL detection.

A key advantage of using SSLDR-A is its inclusivity of all examinations regardless of indication. This also simplifies the calculation by eliminating the need to distinguish between examination types. It avoids potential misclassification by examination indication and reduces endoscopist-level biases. In addition, including all colonoscopies enhances statistical power and generalizability. The markedly narrower CI when using all examinations likely reflects the larger sample size, indicating the data have high precision and low variability. These are some reasons why the ACG/ASGE guidelines were recently updated in agreement with using all examinations when calculating ADR. This approach can similarly be applied to SSLDR.

We also examined the impact of SSLDR-A on ADR-A about protection from PCCRC. We observed that those examinations performed by an endoscopist with an adequate ADR > 35% but an SSLDR < 8% had a higher HR for PCCRC than those performed by an endoscopists with an ADR > 35% and an SSLDR > 8%. These data suggest that endoscopists should meet benchmarks for ADR and SSLDR.

Our study used a large data set of endoscopists and colonoscopies, with detailed information on each resected polyp. Thus, the NHCR database allowed us to calculate detection rates at the endoscopist level, which facilitated the nuanced performance assessment and measurement of physician-specific variability about serrated polyp detection. We recognize that the NHCR population lacks racial and ethnic diversity, limiting the generalizability of our findings. Validation in more national databases with more diverse populations is warranted. Another limitation of our study is that CRC stage and potential mortality were not reported in the study. Finally, we excluded patients with inflammatory bowel disease (IBD), positive occult blood stool tests, and genetic syndromes. Patients with positive fecal immunochemical test (FIT) or positive fecal occult blood test represent 0.6% of our total population in the NHCR, those with IBD represent 6.9%, and those with genetic cancer syndromes represent 0.2%. However, the recent recommendations from the ACG/ASGE suggest that ADRs in FIT-positive patients be measured separately (6). We do acknowledge that our data may not be generalizable to endoscopist practices with high prevalence of examinations for patients with IBD or genetic syndromes.

To our knowledge, this is the first study to establish a data-driven SSLDR benchmark based on PCCRC outcomes using all indications rather than screening examinations alone. Our findings support the use of SSLDR-A as a practical quality indicator. Unlike ADR-A and ADR-S which had different cutoffs for quintiles, the 2 SSLDRs were similar about quintile cutoffs. The decrease in PCCRC was most pronounced in the fourth quintile, which included SSLDR-A 5.0 to <8%. Therefore, our data suggest that endoscopists should achieve an SSLDR-A of 6% similar to the current recommended benchmark for SSLDR-S (6). An aspirational SSLDR-A target of ≥8.0% should be considered for quality assessment and used to inform future society guideline recommendations.

CONFLICTS OF INTEREST

Guarantor of the article: Joseph C. Anderson, MD, MHCDS, FACG.

Specific author contributions: Conception and design: J.C.A., R.H., D.K.R., T.A.M., L.F.B., C.I.M. Data curation: J.C.A., T.A.M., L.F.B. Analysis and interpretation of the data: J.C.A., T.A.M., C.I.M., L.F.B., R.H., D.K.R. Drafting of the article: J.C.A., R.H., D.K.R., L.F.B., T.A.M. Critical revision of the article for important intellectual content: J.C.A., R.H., D.K.R., L.F.B., T.A.M., C.I.M. Statistical analysis: J.C.A., R.H., D.K.R., T.A.M., L.F.B. Obtained funding: J.C.A., C.I.M., L.F.B. Technical or material support: L.F.B. Final approval of the article: J.C.A., R.H., D.K.R., L.F.B., C.I.M., T.A.M.

Financial support: This analysis was funded in part by a clinical grant from the American College of Gastroenterology in 2023 (JCA). This project was also supported in part by the Centers for Disease Control and Prevention's National Program of Cancer Registries, cooperative agreement 5U58DP003930 awarded to the New Hampshire Department of Health and Human Services, Division of Public Health Services, Bureau of Public Health Statistics and Informatics, Office of Health Statistics and Data Management. Research reported in this publication was supported by the National Cancer Institute of the National Institutes of Health under Award Number R01CA243449 (L.F.B./C.I.M.). The funding agencies had no role in the study design, interpretation of results, or writing of the manuscript. The contents of this work do not represent the views of the Department of Veterans Affairs or the United States Government. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health, the Centers for Disease Control and Prevention, or the New Hampshire Department of Health and Human Services.

Potential competing interests: D.K.R.: consultant for Boston Scientific, Olympus Corporation, Braintree Laboratories, Sebela Pharmaceuticals, Laborie, Medtronic; research support - Boston Scientific, Sebela Pharmaceuticals, Medtronic, Olympus Corporation, Erbe, Laborie; ownership interest - Dova Health Intelligence.

Study Highlights.

WHAT IS KNOWN

• ✓ The serrated pathway may account for a large proportion of cancers.

• ✓ Higher endoscopist specific sessile serrated detection rates calculated using screening exams (SSLDR-S) are associated with a lower risk for post colonoscopy colorectal cancer incidence.

WHAT IS NEW HERE

• ✓ Colonoscopies performed by endoscopists with higher SSLDRs using all exams (SSLDR-A) are associated with a lower risk of PCCRC.

• ✓ These data validate SSLDR-A as a quality metric.

• ✓ Our data suggest that endoscopists should aim for an SSLDR-A of 6% and have an aspirational SSLDR-A of 8.0% or higher.

ABBREVIATIONS:

ACG

American College of Gastroenterology

ADR

Adenoma detection rate

ASGE

American Society of Gastrointestinal Endoscopy

FIT

Fecal immunochemical test

FOBT

Fecal occult blood test

HPs

hyperplastic polyps

HR

Hazard Ratio

IBD

Inflammatory bowel disease

NHCR

New Hampshire Colonoscopy Registry

PCCRC

postcolonoscopy colorectal cancer

SSL

Sessile serrated lesions

SSLDR-A

sessile serrated lesion detection rates using all exams

SSLDR-S

sessile serrated lesion detection rates using screening exams