Association of endoscopist colonoscopy quality measures with follow-up colonoscopy outcomes after positive stool tests (mt-sDNA or FIT): Retrospective cross-sectional analysis of data from the New Hampshire Colonoscopy Registry

Abstract

Background

Negative colonoscopies following positive stool tests could result from stool test characteristics or from the quality of endoscopist performance. We used New Hampshire Colonoscopy Registry data to examine the association between endoscopist detection rates and polyp yield in colonoscopies performed for positive FIT or mt-sDNA tests to evaluate the degree to which positive stool tests followed by negative colonoscopy (‘false positives’) vary with endoscopist quality. Additionally, we investigated the frequency of significant polyps in the sub-group of highest quality colonoscopies following positive stool tests.

Methods

We compared the frequencies of negative colonoscopies, and of specific polyps following positive stool tests across quartiles of endoscopist adenoma detection rate (ADR) and clinically significant serrated polyp detection rate (CSSDR).

Results

Our sample included 864 mt-sDNA+ and 497 FIT+ patients. We found a significantly lower frequency of negative colonoscopies following positive stool tests among endoscopists with higher ADR and CSSDR, particularly in the two highest quartiles. Additionally, detection of any adenoma after a positive stool test for endoscopists in the 4^th ADR quartile was 63.3% (FIT+) and 62.8% (mt-sDNA+). Among endoscopists in the 4^th CSSDR quartile, SSLs were found in 29.2% of exams following a positive mt-sDNA, and in 13.5% following FIT+ exams.

Conclusions

The frequency of negative colonoscopies after positive stool tests was significantly higher in exams performed by endoscopists with low ADR and CSSDR. Our results also suggest a benchmark target of at least 40% for ADR in patients with mt-sDNA+ or FIT+ tests, and 20% for SSLs in mt-sDNA+ patients.

Introduction

Stool-based tests such as the fecal immunochemical test (FIT) and the multi-target stool DNA (mt-sDNA) test provide important screening options for average risk individuals. These tests allow early detection of colorectal cancer (CRC) and also help to detect precancerous polyps which can then be removed at follow-up colonoscopy. Current evidence suggests that stool tests are effective; positive tests are associated with higher rates of neoplastic findings on follow-up colonoscopy, compared to primary colonoscopy.^1–4 Both stool-based tests have been endorsed by several societies for CRC screening.^5–7 However, not all colonoscopies after a positive stool test detect findings in the colon. These negative follow-up colonoscopies could reflect characteristics of the stool tests (“false positive” results), or the negative results could be a reflection of sub-optimal endoscopist performance in finding polyps which are present. There are limited data on the proportion of negative follow-up colonoscopies done by high versus low performing endoscopists, which could be helpful in distinguishing the stool test characteristics from the effect of endoscopist quality on follow-up colonoscopy outcomes.

The effectiveness of colonoscopy, whether done for positive stool tests or as the primary screening or surveillance test for CRC, depends on endoscopist skill; wide variation in polyp detection among endoscopists has been demonstrated.⁸ For this reason, national gastroenterological societies have recommended quality measures to ensure optimal polyp detection,⁹ including benchmarks for adenoma detection (ADR) and serrated polyp detection (SDR).^{8, 10, 11} Higher adenoma and serrated polyp detection rates have been shown to be protective for post-colonoscopy CRC.^10–14

Because the rate of polyp findings after positive stool test results is substantially higher than for average-risk screening colonoscopies, new benchmarks would help endoscopists more accurately gauge whether they are missing polyps in stool-test-positive patients. To date, professional societies and meta-analyses have suggested a higher ADR threshold for physicians performing colonoscopy in FIT+ individuals.^{5, 15} However, given the limited evidence on polyp detection in colonoscopies following positive stool tests, suggested thresholds vary. The U.S. Multi-Society Task Force on CRC prevention proposed an ADR benchmark of 35% for women and 45% for men in the FIT+ asymptomatic population.⁵

In addition, little data exist concerning the impact of endoscopist ADR or SDR on polyp detection in colonoscopies following positive stool tests. Specifically, the frequency of positive findings in follow-up colonoscopies performed by high-performing endoscopists (those with high ADR and SDR) would provide clinically relevant evidence to inform aspirational targets for polyp detection in stool-test-positive patients.

Our goal was to investigate the associations between established endoscopist quality measures (ADR and SDR for potentially precancerous serrated polyps) and colonoscopy findings following a mt-sDNA+ or FIT+ test in community practice. Our primary aim was to investigate the hypothesis that there would be increased frequencies of negative colonoscopies (‘false positives’), in exams performed by endoscopists with lower polyp detection rates (as established in primary screening colonoscopies) and to assess the degree to which detection rates differ between low and high ADR and SDR quartiles. Our secondary aim was to use data from high quality colonoscopies to provide evidence informing the establishment of both detection rate benchmarks and aspirational targets for colonoscopies after positive stool tests.

Methods

Population

Prior to colonoscopy, patients from endoscopy practices across New Hampshire complete a New Hampshire Colonoscopy Registry (NHCR) Patient Information Form, providing demographic, lifestyle, medical and family history data. Data from colonoscopies are collected on the NHCR Procedure Form, which is completed by endoscopists and/or endoscopy nurses during or immediately after colonoscopy. Data include colonoscopy indication, completion status, bowel preparation quality, and the location, size, and resection methods for all polyps. Pathology outcomes are abstracted by trained NHCR staff who match data from pathology reports to polyp-level findings recorded on the Procedure Form.¹⁶

All data collection and study procedures were approved by the Committee for the Protection of Human Subjects at Dartmouth College (CPHS#00015834) in accordance with the Belmont Report and the US Common Rule.

Study cohorts

Using an IRB-approved protocol, Exact Sciences Laboratories, LLC (Madison, Wisconsin), provided the NHCR with patient identifiers of all patients with mt-sDNA+ tests within the NHCR catchment area (New Hampshire, Vermont, Maine and Massachusetts) through 12/31/2022. Patients with positive stool test results were referred by their primary care providers to NH endoscopists. While all patients were asymptomatic, some had stool tests for off label indications, including a personal or family history of CRC or a personal history of polyps (Table 1).

Table 1:

Characteristics of exams after positive mt-sDNA and FIT tests¹

	Mt-sDNA+		FIT+
	n=864		n=497
	N/Mean	%/SD	N/Mean	%/SD
Age (continuous)	66.3	8.0	66.3	8.9
Age <65 years	362	41.9	218	43.9
Patient sex female	534	61.8	250	50.3
Patient at increased risk for CRC 2	190	22.0	131	26.4
Race, Caucasian	660	96.9	395	96.8
BMI (continuous)	28.9	7.7	28.7	7.1
Smoking status
Never smoker	319	45.1	202	48.0
Former smoker	295	41.7	182	43.2
Current smoker	94	13.3	37	8.8
Self-reported health
Good to excellent	630	90.4	366	87.4
Fair	62	8.9	46	11.0
Poor	5	0.7	7	1.7
Aspirin use at least once a week	145	22.7	138	35.8
NSAIDs use at least once a week	72	10.7	44	10.8
Aspirin/NSAID use at least once a week	194	31.0	164	42.8
Blood thinner use	58	8.5	44	10.4
Time from positive mt-sDNA test to exam (in months, continuous)	2.60	2.8	n/a	n/a
Prior history of endoscopy	435	50.3	308	62.0
Patient history of neoplastic findings	112	13.0	79	15.9
1st degree family history of colorectal cancer	88	13.2	68	17.1

^1:Missing data: Race (20%), BMI (25%), smoking (17%), health status (18%), aspirin use (25%), NSAID use (21%), aspirin or NSAID use (26%), blood thinner use (19%), patient history of neoplastic findings (0%), first degree family history of colorectal cancer (22%)

^2:Increased risk: first degree family member with a history of CRC, self-reported history of neoplasia, prior neoplasia recorded in the NHCR database, or surveillance indication for colonoscopy.

All mt-sDNA tests were ordered in the context of usual care and reported to NHCR investigators as positive mt-sDNA tests, exactly as the results are reported to the ordering clinicians. We included all patients with these positive stool tests ordered as part of usual clinical care and a record in the NHCR of a complete colonoscopy with adequate bowel preparation. Exclusion criteria were age < 50 years (since the national CRC screening guideline prior to 2021 recommended average-risk screening starting at age 50), Inflammatory Bowel Disease (IBD), or a genetic syndrome such as Lynch Syndrome. If two or more colonoscopies were performed within 12 months of each other and the initial exam was incomplete or had poor bowel preparation or the subsequent exam was indicated for polypectomy or completion of polypectomy of a known polyp, the colonoscopy outcomes were merged and treated as a single exam. After this merge, patients with no complete exam with adequate bowel preparation were excluded. Exams with missing outcome data or for which endoscopist ADR or CSSDR could not be calculated were excluded from this analysis, and complete exams with missing bowel preparation were assumed to have adequate bowel preparation due to the extremely low rates of poor bowel prep in complete exams in the NHCR database. Our study cohort included 864 individuals with an mt-sDNA+ result and a subsequent colonoscopy in the NHCR database and 497 individuals with positive FIT tests during the same time interval and a subsequent colonoscopy in the NHCR database.

Outcomes

Study outcomes included negative colonoscopies (exams without findings requiring follow-up within 10 years as per the USMSTF, which includes all normal colonoscopies and those with diminutive rectosigmoid hyperplastic polyps only),¹⁷ as well as colonoscopies with adenomas, advanced adenomas, adenocarcinoma, and/or sessile serrated lesions.

Other factors examined

Patient variables were derived from the NHCR Patient Information Form and included demographic factors (e.g. age, sex, race) health behaviors (e.g. smoking status, Body Mass Index, overall health status), use of aspirin/nonsteroidal anti-inflammatory drug (NSAID) or blood thinners, and history of prior colonoscopy.

Calculating exposure variables: Endoscopist ADR and SDR

Endoscopist ADR was calculated by dividing the total number of an endoscopist’s screening colonoscopies with at least one adenoma by the total number of screening colonoscopies performed by that endoscopist. Endoscopists were then divided into quartiles based on their ADR. Endoscopist serrated polyp detection was calculated for clinically significant serrated polyps (CSSPs), which include any SSL or traditional serrated adenoma (TSA), any hyperplastic polyp (HP) 1 cm or larger and those HPs 5 mm or larger in the proximal colon. CSSDRs were calculated in the same manner as ADRs, and endoscopists were then divided into quartiles based on their CSSDR. Colonoscopies following a positive stool test (mt-sDNA or FIT) were excluded from these calculations of endoscopist ADR and CSSDR.

Statistical and analytic approach

Proportions of adenomas, SSLs, and advanced adenomas were calculated for each quartile of ADR and CSSDR by dividing the total number of exams performed by endoscopists within each ADR or CSSDR quartile with at least one polyp by the total number of exams in that quartile. Proportions were compared across quartiles using the Cochran-Armitage test for trend.

We also suggest potential ADR and CSSDR benchmarks for colonoscopy following a positive stool test based on serrated and adenoma detection rates of endoscopists. For ADR we used quartile 2 (Q2) since that quartile had endoscopists with ADRs of 25.7% or higher, which is consistent with the current ADR benchmark of 25%. Since there are no benchmarks for SDR, we used a CSSDR of 9% based on our recently published paper showing optimal protection from post colonoscopy CRC in endoscopists with this CSSDR or higher.¹⁰ Since Q3 for CSSDR had endoscopists with a rate of 8.7% or higher, we used this quartile to suggest benchmarks for SSL. We also suggested aspirational targets based on quartile 4, the highest performing endoscopists. All of our suggested benchmarks were rounded to a multiplier of 5 to make the targets easier for endoscopists to remember.

Results

Our sample included 864 patients with positive mt-sDNA tests and 497 patients with FIT+ stool tests. Demographic and behavioral characteristics are shown in Table 1. We also present patient characteristics across quartiles for ADR and CSSDR in the mt-sDNA+ (Supplemental Tables 1a and 1b) and FIT+ cohorts (Supplemental Tables 1c and 1d). There were no significant differences across the detection rate quartiles except for the following: mt-sDNA+ ADR quartiles (BMI, history of prior exams, family history of CRC and time to exam from positive stool test), mt-sDNA+ CSSDR quartiles (history of prior exams) and FIT+ CSSDR (age). All colonoscopies in our analysis were performed between February 2015-June 2023.

Of the 65 endoscopists stratified by ADR (Table 2A) and CSSDR (Table 2B), 56 had completed colonoscopies after a positive mt-sDNA test and 63 had completed colonoscopies after a positive FIT test. ADR ranged from 11.0% to 51.9% and CSSDR from 1.6% to 26.8% across all 4 quartiles, with median ADRs of 22% (Q1), 29% (Q2), 36% (Q3) and 44% (Q4) and median CSSDRs of 4.3% (Q1), 7.6% (Q2), 11.4% (Q3) and 17.1% (Q4).

Table 2a:

Characteristics of Adenoma Detection Rate (ADR) quartiles

	Q1	Q2	Q3	Q4
Detection rate range (%)*	11.0-25.7	25.7-31.6	32.0-39.3	41.0-51.9
Endoscopist ADR (median, %)	21.5	29.0	35.8	44.4
NHCR, Endoscopists in quartile (#)	17	16	16	16
mt-sDNA+ cohort      Endoscopists in quartile (#)	15	11	15	15
Exams in quartile (#)	117	166	234	347
FIT cohort      Endoscopists in quartile (#)	16	16	15	16
Exams in quartile (#)	95	106	127	169

Table 2b:

Characteristics of Clinically Significant Serrated Polyp Detection Rate (CSSDR) quartiles

	Q1	Q2	Q3	Q4
Detection rate range (%)	1.6-6.4	6.5-8.6	8.7-13.6	13.9-26.8
Endoscopist CSSDR (median, %)	4.3	7.6	11.4	17.1
NHCR, Endoscopists in quartile (#)	17	16	16	16
mt-sDNA+ cohort      Endoscopists in quartile (#)	14	15	12	15
Exams in quartile (#)	128	230	191	315
FIT cohort      Endoscopists in quartile (#)	17	15	15	16
Exams in quartile (#)	102	104	128	163

^*The maximum value in ADR Q1 is 25.71 and the minimum value for ADR Q2 is 25.74

The number and percentage of negative colonoscopies following mt-sDNA (Table 3A) and FIT (Table 3B) tests varied significantly across quartiles of ADR and CSSDR, with lower frequencies of negative colonoscopies following positive stool tests (i.e. fewer false positives) among endoscopists with higher ADR and CSSDR. Following positive mt-sDNA tests, the proportion of negative colonoscopies was 17.6% higher in the lowest quartile of ADR than in the highest quartile of ADR, with a similar difference (19.5%) for CSSDR quartiles. Following positive FIT tests, the proportion of negative colonoscopies was 25.8% higher in ADR Q1 than in ADR Q4, and 16.6% higher in CSSDR Q1 than in CSSDR Q4.

Table 3:

Negative and positive colonoscopy according to the US Multi-Society Task Force on Colorectal Cancer (USMSTF)^*, by endoscopist quality measure quartiles

3A) Exams after positive mt-sDNA test
	Adenoma detection rate
	Q1		Q2		Q3		Q4		P
	N	%	N	%	N	%	N	%
Negative colonoscopy	84	71.8	119	71.7	144	61.5	188	54.2	<0.001
Positive colonoscopy	33	28.2	47	28.3	90	38.5	159	45.8
	Clinically Significant Serrated Polyp detection rate
	Q1		Q2		Q3		Q4		P
	N	%	N	%	N	%	N	%
Negative colonoscopy	92	71.9	159	69.1	119	62.3	165	52.4	<0.001
Positive colonoscopy	36	28.1	71	30.9	72	37.7	150	47.6
3B) Exams after positive FIT test
	Adenoma detection rate
	Q1 (95 exams)		Q2 (106 exams)		Q3 (127 exams)		Q4 (169 exams)		P
	N	%	N	%	N	%	N	%
Negative colonoscopy	83	87.4	80	75.5	88	69.3	104	61.5	<0.001
Positive colonoscopy	12	12.6	26	24.5	39	30.7	65	38.5
	Clinically Significant Serrated Polyp detection rate
	Q1 (102 exams)		Q2 (104 exams)		Q3 (128 exams)		Q4 (163 exams)		P
	N	%	N	%	N	%	N	%
Negative colonoscopy	84	82.4	82	78.8	81	63.3	108	66.3	<0.001
Positive colonoscopy	18	17.6	22	21.2	47	36.7	55	33.7

^*USMSTF positive colonoscopy: Any adenocarcinoma, tubular adenoma, villous adenoma, tubulovillous adenoma, high grade dysplasia, traditional serrated adenoma, or sessile serrated lesion.

Results for detection of specific types of polyp across quartiles for mt-sDNA+ and FIT+ patients are shown in Tables 4A (mt-sDNA) and 4B (FIT). In colonoscopies following mt-sDNA+ tests, the proportion of exams with adenomas trended significantly higher from Q1 to Q4 for both endoscopist ADRs (48.7 to 62.8%) (P <0.001) and CSSDRs (46.9% to 66.7%) (p <0.001). The proportion of exams with SSLs also had a significant increasing trend from ADR Q1 to Q4 (12.8% to 26.2%) (p <0.001) and CSSDR Q1 to Q4 (10.2% to 29.2%) (p <0.001).

Table 4:

Positive predictive values for selected outcomes by endoscopist quality measure quartiles

4A) Exams after positive mt-sDNA test
	Adenoma detection rate
	Q1		Q2		Q3		Q4		P
	N	%	N	%	N	%	N	%
Any adenoma	57	48.7	81	48.8	140	59.8	218	62.8	<0.001
Any sessile serrated lesion	15	12.8	24	14.5	52	22.2	91	26.2	<0.001
Any advanced adenoma	18	15.4	28	16.9	43	18.4	77	22.2	0.060
Any adenocarcinoma	3	2.6	3	1.8	1	0.4	7	2.0	0.773
	Clinically Significant Serrated Polyp detection rate
	Q1		Q2		Q3		Q4		P
	N	%	N	%	N	%	N	%
Any adenoma	60	46.9	117	50.9	109	57.1	210	66.7	<0.001
Any sessile serrated lesion	13	10.2	43	18.7	34	17.8	92	29.2	<0.001
Any advanced adenoma	22	17.2	31	13.5	35	18.3	78	24.8	0.004
Any adenocarcinoma	3	2.3	4	1.7	1	0.5	6	1.9	0.761
4B) Exams after positive FIT test
	Adenoma detection rate
	Q1		Q2		Q3		Q4		P
	N	%	N	%	N	%	N	%
Any adenoma	34	35.8	45	42.5	62	48.8	107	63.3	<0.001
Any sessile serrated lesion	5	5.3	7	6.6	12	9.4	28	16.6	0.001
Any advanced adenoma	8	8.4	19	17.9	23	18.1	37	21.9	0.011
Any adenocarcinoma	0	0.0	3	2.8	2	1.6	4	2.4	0.321
	Clinically Significant Serrated Polyp detection rate
	Q1		Q2		Q3		Q4		P
	N	%	N	%	N	%	N	%
Any adenoma	38	37.3	41	39.4	80	62.5	89	54.6	<0.001
Any sessile serrated lesion*	4	3.9	7	6.7	19	14.8	22	13.5	0.004
Any advanced adenoma	13	12.7	16	15.4	27	21.1	31	19.0	0.132
Any adenocarcinoma	1	1.0	3	2.9	3	2.3	2	1.2	0.911

Among FIT+ patients, there was a higher yield of adenomas and SSLs across endoscopist ADR quartile (Table 4B; adenomas, 35.8% to 63.3% (p <0.001); SSLs, 5.3% to 16.6%) (p<0.001). There was also a higher yield of adenomas and SSLs across endoscopist CSSDRs (adenomas,37.3% to 54.6%) (p<0.001); SSLs, 3.9% to 13.5%) (p=0.004).

We present suggestions for potential benchmarks for polyp detection in patients with positive stool tests. Based on our data we suggest at least 40% adenoma detection rates for both mt-sDNA and FIT positive patients. In addition, we suggest sessile serrated lesion detection rates of 20% and 15% for mt-sDNA and FIT positive patients respectively. These data are shown in Table 5 as well as suggestions for aspirational targets.

Table 5.

Adenoma and Serrated polyp benchmarks and aspirational targets for patients with positive stool tests

	Mt-sDNA+		FIT+
	*Benchmark	**Aspirational	*Benchmark	**Aspirational
Adenoma detection rate	40%	60%	40%	60%
Sessile serrated lesion detection rate	20%	30%	15%	15%

^*Benchmarks based on 25% ADR for adenomas and 9% CSSDR for sessile serrated lesions

^**Aspirational targets based on adenoma detection in the 4^th quartile of ADR and sessile serrated lesion detection in the 4^th quartile of CSSDR

To assess the potential impact of endoscopists in the lowest quartile on our results, we excluded quartile 1 and compared quartiles 2-4 using the Cochran-Armitage test for trend. Except for SSL detection in FIT+ tests across clinically significant serrated polyp detection rate quartiles, we observed no differences in statistical significance compared to the results for all 4 quartiles (see supplementary table 3 A and B).

Furthermore, to address the impact of having a prior colonoscopy on polyp detection across quartiles among mt-sDNA+ patients, we used logistic regression models with polynomial contrasts for trends in detection rate quartiles, adjusting for patient age, sex, BMI, previous colonoscopy, family history of first degree relative with CRC, and time from positive test to colonoscopy. We found significant linear trends across detection rate quartiles for each outcome, supporting our earlier results, except in the “any adenoma” outcome across ADR quartiles. In other words, higher quartiles were still associated with higher polyp yields (Supplementary Table 4).

We also examined our data by classifying our endoscopists by SSLDR and observed similar results for that when stratifying by CSSDR. These data along with that of the endoscopists are shown in Supplementary Tables 5 and 6. Finally, in Supplementary Table 7, we present the positive predictive values (as shown in Table 4) for 2 additional outcomes, advanced adenoma/CRC and advanced serrated lesion (SSL>=1cm, SSL with dysplasia, TSA).

Discussion

The first goal of our analysis was to examine polyp detection in patients having follow-up colonoscopies after positive stool tests. Past research has shown significant variability in endoscopist quality, and furthermore that endoscopists with adequate ADRs may have low SDRs, ^{8, 10, 11} suggesting that serrated polyp detection may require different skills and abilities than conventional adenoma detection. Therefore, we evaluated trends in both negative colonoscopy and polyp detection after positive stool tests across both endoscopist ADR and CSSDR.

We found that the proportion of negative follow-up colonoscopies was significantly higher in exams done by endoscopists with low ADR and CSSDR compared to high detecting endoscopists. Correspondingly, the rate of clinically-important polyp outcomes increased with higher ADR and CSSDR. Furthermore, as shown in supplemental tables 1A–1D, the risk profiles of the patients in ADR and CSSDR quartiles did not vary across quartiles in clinically meaningful ways, suggesting that the trend toward lower polyp yield in both ADR and CSSDR quartile 1 may reflect lower polyp detection, rather than lower polyp prevalence.

These findings have several implications. The difference across ADR and CSSDR quartiles in the rate of negative exams (as defined by the USMSTF) implies that important polyps may have gone undetected in the lowest quartiles of ADR and CSSDR. These exams would therefore be incorrectly classified as ‘false positives’, and patients would be assigned inappropriately long follow-up intervals. For that reason, it is essential to account for endoscopist detection ability when assessing the accuracy of stool tests using data from follow-up colonoscopies. Our results suggest that some of the “false positive” outcomes attributed to stool tests might actually be a reflection of sub-optimal endoscopist performance.

We also found that the proportions of exams with adenomas and SSLs after positive stool tests trended significantly higher in exams performed by endoscopists with higher ADR, CSSDR as well as SSLDR. This variation suggests that there is room for improvement in polyp detection following positive stool tests; establishing appropriately high detection rate benchmarks for these colonoscopies is important for ensuring optimal CRC prevention.

Another goal of our analysis was to propose specific benchmark detection rates as well as aspirational target detection rates for ADR and SDR following positive stool tests. Benchmark detection rates identify an acceptable level of detection in practice, whereas aspirational targets are intended to reflect an ideal scenario where nearly all polyps are detected.

Suggested benchmarks are based on detection levels among endoscopists who have demonstrated the quality of their clinical care by meeting the current ADR standard of 25% for screening exams in average-risk patients (i.e., excluding patients with positive stool tests); conveniently, the 2^nd quartile of ADR in our NHCR endoscopist population starts at 25.7%, approximating the current benchmark. In patients with preceding positive mt-sDNA, the endoscopists in this quartile had an adenoma detection rate of 48.8%. For FIT+ patients, the corresponding adenoma detection rate in the 2nd ADR quartile was 42.5%. For the sake of simplicity, this could suggest a potential ADR benchmark of at least 40% in colonoscopies for both FIT and mt-sDNA positive patients. A recent meta-analysis of 34 studies which included 2,655,345 individuals having colonoscopy after a positive stool test observed a pooled ADR of 47.8%,¹⁵ which is similar to our own prior findings³ and also supported by the results of the present study.

Although there are no established benchmarks for serrated detection rates, previous NHCR research has shown that a CSSDR of at least 9% offered the most protection from post-colonoscopy CRC;¹⁰ 9% has therefore been proposed as a reasonable benchmark. The 3^rd screening CSSDR quartile included CSSDRs of 8.7% to 13.6%. Endoscopists within this quartile detected sessile serrated lesions in 17.8% of exams after mt-sDNA+ and in 14.8% of exams after FIT+. Rounding off, this suggests potential CSSDR benchmarks of 20% and 15% after positive mt-sDNA and FIT tests, respectively. A recent paper suggested 20% as an SSL benchmark in patients with mt-sDNA+ stool tests.¹⁸

In order to set aspirational targets for both adenomas and serrated polyps after positive stool tests, we used the rates of polyp findings achieved by the highest-detecting endoscopists as an approximation for the ideal detection scenario. The frequency of mt-sDNA+ exams with adenomas was highest, at 62.8%, for the 4^th quartile of screening ADR. The highest rate of adenomatous findings after positive FIT, which was also approximately 60% (63.3%), was also in the 4^th screening ADR quartile. These results suggest 60% as an aspirational target for adenoma detection in patients with positive stool tests. NHCR endoscopists in the 4^th ADR quartile who achieved this goal had a screening ADR of at least 41%, which is significantly higher than the current 25% benchmark for screening patients,⁹ but potentially achievable for an even wider group of endoscopists over time. This aspirational target of 60% is in line with past findings for FIT+ patients, including a 66% adenoma detection rate observed in a FIT+ population from the Netherlands.¹⁹

Our results also suggest an aspirational SSL detection rate target of 30% for colonoscopies after positive mt-sDNA tests, based on the SSL frequency of 29.2% in exams performed by endoscopists in the 4^th screening CSSDR quartile. The rate of SSL detection after a positive FIT was more modest (approximately 15%). This is not surprising because mt-sDNA includes methylated markers associated with serrated lesions, such as BMP3, and thus has higher sensitivity for these lesions than FIT, which does not include these markers.^{2, 3, 20–22}

To examine the impact of the low performing endoscopists on our results, we excluded the lowest quartiles for ADR and CSSDR and observed an increasing PPV for any adenoma or SSL but not for advanced adenomas. These data suggest that there may be a threshold at which endoscopists’ ability to detect advanced adenomas, which are mostly larger than 1 cm, is adequate. However, endoscopists with higher detection rates above this threshold will detect a greater number of smaller adenomas.

A limitation of this real-world study derives from the low racial diversity found within the New Hampshire population, which limits the generalizability of our findings. However, NHCR data does capture the substantial rural/urban, ethnic and socioeconomic diversity present within New Hampshire. We also acknowledge that while our proposed benchmarks reflect our data, further validation in other populations is needed. In addition, we acknowledge that our analysis may lack sufficient precision to differentiate potential ADR benchmarks for FIT and mt-sDNA.

Strengths of our analysis include the large number of exams performed by 63 endoscopists in the FIT+ group and 56 endoscopists in the mt-sDNA+ group, with most endoscopists included in both groups. In addition, we compared patient characteristics across the ADR and CSSDR quartiles to ensure that there were no differences in patient or exam characteristics between the groups which could account for variation in polyp detection. Factors such as smoking and BMI have been strongly associated with serrated and adenomatous polyp prevalence.²³ In our analyses, we observed no clinically meaningful differences in patient characteristics across the quartiles which could explain differences in detection. For example, while there was a significant trend in patient age across FIT+ CSSDR quartiles, the highest CSSDR quartile was younger on average than the lowest quartile. Similarly, there was a significant decreasing trend in the proportion of mt-sDNA+ patients with family history of CRC across ADR quartiles. If anything, these trends would suggest that higher detection rate quartiles contain patients at a slightly lower risk for polyps, supporting our contention that increased polyp yield in quartiles 3 and 4 is related to endoscopist quality, rather than underlying polyp prevalence.

In summary, we observed that after positive stool tests, negative follow-up colonoscopies were significantly more common among endoscopists with lower ADR and CSSDR. This suggests that the frequency of ‘false positive’ stool test results may not accurately reflect the performance of the stool test itself, and that endoscopist quality improvement could enhance the performance of stool tests for screening. Additionally, based on our investigation of adenoma and serrated polyp detection among high performing endoscopists (in the top quartiles of ADR and CSSDR), we propose benchmarks of at least 40% (with 60% aspirational detection) for adenoma detection following positive mt-sDNA or FIT, benchmarks of 20% (with 30% aspirational detection) for sessile serrated lesions detection following positive mt-sDNA, and a benchmark of 15% for serrated polyp detection following a positive FIT. Such targets are helpful in guiding clinicians with respect to expected polyp yield in patients with positive stool tests, helping to minimize rates of missed lesions and improve quality. Additional evidence from other populations will strengthen and further clarify our findings.

Study Highlights.

What is known

• Stool based CRC screening tests can increase polyp yield at colonoscopy.

• Polyp detection is higher in endoscopists with superior performance.

• Positive stool tests followed by a negative colonoscopy are often described as ‘false positive‘ tests.

What is new here

• Higher endoscopist detection rates are significantly associated with fewer negative colonoscopies after positive stool tests.

• Some ‘false positive’ stool tests may reflect lower polyp detection rather than lower polyp prevalence.

• After positive stool tests, NHCR data supports adenoma detection benchmarks of at least 40% for both mt-sDNA+ and FIT+, and sessile serrated lesion detection benchmarks of 20% (mt-sDNA+) and 15% (FIT+).

Abbreviations:

AA

Advanced Adenoma

ADR

Adenoma Detection Rate

BMI

Body Mass Index

BMP3

Bone morphogenetic protein 3

CI

confidence interval

CRC

colorectal cancer

CSSDR

Clinically Significant Serrated polyp Detection Rate

FIT

fecal immunochemical test

HgB

hemoglobin

HP

hyperplastic polyp

IBD

inflammatory bowel disease

IRB

Institutional review board

mt-sDNA

multi-target stool DNA

NHCR

New Hampshire Colonoscopy Registry

NSAID

nonsteroidal anti-inflammatory drug

OR

odds ratio

SD

standard deviation

SSL

sessile serrated lesion

TSA

traditional serrated adenoma

USMSTF

United States Multi-Society Task Force on Colorectal Cancer