Adenoma Detection Rates Calculated Using All Exams Are Associated with Lower Risk for Post Colonoscopy Colorectal Cancer: Data From the New Hampshire Colonoscopy Registry

Abstract

Introduction

We used New Hampshire Colonoscopy Registry (NHCR) data to examine the association between post-colonoscopy colorectal cancer (PCCRC) risk and an adenoma detection rate (ADR) which was calculated using exams with all indications, as compared to ADR restricted to only screening exams.

Methods

Our cohort study included NHCR patients with an index exam and at least one follow-up event, either a colonoscopy or a CRC diagnosis. Our outcome, PCCRC was any CRC diagnosed ≥ 6 months after an index exam. The exposure variable was endoscopist-specific ADR (ADR-A), calculated for all indications, divided into quintiles. We also compared the ADR-A to a screening ADR (ADR-S). Cox regression was used to model the hazard of PCCRC on ADR, controlling for age, sex, and other covariates.

Results

In 32,535 patients, a lower hazard for PCCRC (n=178) was observed for ADR-A’s ≥ 23%, as compared to ADR-A’s <23% (Reference) (23%-<29%: HR=0.56, 95%CI:0.36-0.87;29%-<34%: HR=0.60, 95% CI:0.38-0.94; 34%-<44%: HR=0.43,95% CI: 0.29-0.65; and ≥44%: HR=0.32, 95% CI: 0.16-0.63). The highest quartile of ADR-A (42%+)(HR=0.41 95%CI:0.23-0.75) had a similar protection from PCCRC as the highest quartile of ADR-S (35%+)(HR=0.38 95%CI:0.21-0.70). We observed 95% CIs for ADR’s were 28% narrower (median=0.72;IQR:0.10) for endoscopists when using ADR-A versus ADR-S.

Discussion

Our data demonstrating lower PCCRC risk in exams performed by endoscopists with higher ADR’s calculated with all exams helps to validate ADR-A as a quality measure. ADR-A may also increase precision of the calculated ADR. Endoscopists should strive for a higher ADR-A with 44% as an aspirational target.

Background

Colonoscopy prevents colorectal cancer (CRC) by detection and resection of precursors known as polyps. Effective CRC prevention with colonoscopy depends on optimal adenoma detection. The adenoma detection rate (ADR) is an important endoscopist-specific quality measure. Higher ADRs have been shown to be associated with lower risks for post colonoscopy colorectal cancer (CRC) in patients having the exam performed by physicians with these higher rates.^{1, 2} ADR has been calculated by dividing the number of complete screening colonoscopies with adequate bowel preparation where at least one conventional adenoma is detected by the total number of complete screening colonoscopies with adequate bowel preparation.^{3, 4} The recommended ADR benchmark, which had been 25% for screening colonoscopy, was recently raised to 35% in conjunction with lowering the age to begin measurement to 45 years and expanding the procedure indications to be included in ADR measurement.^{5, 6}

The original ADR, which was proposed in 2002, included all indications in the calculation.³ However, this ADR calculation was modified in 2006 to include only screening exams.⁴ Some limitations to this ADR calculation include the ability of endoscopists to game the system by changing the exam indication based on findings.⁷ Another concern about the screening ADR is that some endoscopists have a lower volume of exams, limiting the ability to provide accurate assessment of the physician’s detection performance. Increasing the number of total exams included in ADR measurement narrows the confidence interval around the calculated ADR. Therefore, it increases the precision of the calculated ADR. The recently updated American College of Gastroenterology (ACG)/American Society of Gastrointestinal Endoscopy (ASGE) quality metrics recommend an overall ADR of 35% calculated using screening, surveillance and diagnostic colonoscopies for reasons other than positive screening tests (e.g. fecal tests), as opposed to only screening colonoscopies.⁵

Validation of detection rates by demonstrating lower post colonoscopy CRC rates for endoscopists who have achieved the benchmark ADR is important.^{1, 8} A recent study using data from an HMO in California demonstrated that ADR using all exams can predict post CRC similarly well to ADR using screening exams.⁹ However, these data included data from centers which had a high proportion of colonoscopies in FIT+ patients and may not be representative of ADRs in other populations. We used data from the New Hampshire Colonoscopy Registry (NHCR) to examine the association between ADR calculated from all exams, regardless of indication, with risk for post colonoscopy CRC. We also compared outcomes to ADR calculated from screening colonoscopies.

Methods

Population

Our sample consisted of patients enrolled in the NHCR with an initial colonoscopy and at least one follow-up event, whether a colonoscopy or a CRC diagnosis, in the New Hampshire State Cancer Registry, which collects data from NH and other states (VT, MA, ME) as described in previous publications.^{10, 11} Patients who have a colonoscopy in NH are invited to participate in the NHCR. Those who consent give permission to have data collected from index and subsequent colonoscopies. An NHCR Patient Questionnaire is completed prior to colonoscopy. Information collected includes patient demographics, health behavior, and personal and family history of polyps and CRC. Endoscopists and/or endoscopy nurses complete the NHCR Colonoscopy Procedure Form during or immediately after colonoscopy. Bowel preparation quality is assessed for every colonoscopy, “based on the segment (after cleaning all colon segments) which has the worst quality of preparation” following detailed descriptions of each preparation quality option noted on the NHCR Procedure form. Trained NHCR abstractors match polyp-level pathology data to each finding recorded on the NHCR Procedure Form¹². All data collection and study procedures were approved by the Committee for the Protection of Human Subjects at Dartmouth College (CPHS#00015834).

Analyzed Sample

Our primary analysis included all patients with an index colonoscopy in the NHCR database plus at least one follow-up colonoscopy in the NHCR performed 6 months or longer after the index exam, or a diagnosis of CRC at least 6 months after the index colonoscopy as recorded in the NH State Cancer Registry. In addition, as sensitivity analysis, we have examined our data using a 12 month cutoff for PCCRC. In our analysis, follow-up time was months to first follow-up colonoscopy or diagnosis of CRC (in NHCR or New Hampshire State Cancer Registry data). Exclusion criteria included a personal history of inflammatory bowel disease, familial genetic cancer syndromes, or CRC diagnosed at index or within 6 months of the index exam. Excluded genetic syndromes included Lynch syndrome, polyposis syndromes and patients who had serrated polyposis syndromes.

In addition, we also examined data from all patients with index colonoscopies including those with no follow-up event, whether a colonoscopy or a CRC diagnosis in the New Hampshire State Cancer Registry. Because there is a lag in NH State Cancer Registry data availability, we required that index exams in this sample took place at least 12 months before the NHCR-NH State Cancer Registry linkage date to allow for CRC diagnosis and reporting of CRC to the NH State Cancer Registry.

Outcomes

Our main outcome was post colonoscopy colorectal cancer (PCCRC), which was defined as any CRC diagnosed 6 months or longer after index exam.

Exposure variable

Our exposure of interest was endoscopist-specific ADR detection rates calculated separately for all exams and screening colonoscopies. For each endoscopist we calculated the all exam ADR (ADR-A) and screening ADR (ADR-S) by dividing the number of complete colonoscopies in patients 45 years or older with an adequate bowel preparation and at least one conventional adenoma by the total number of complete exams with adequate bowel preparation performed by the endoscopist.

The specific calculations for our analysis were:

ADR-A= The proportion of all colonoscopies in patients 45 years or older with an adequate bowel preparation performed by an endoscopist regardless of indication with at least one adenoma divided by the total number of colonoscopies in patients 45 years or older with an adequate bowel preparation performed by that endoscopist regardless of indications.

ADR-S= The proportion of screening colonoscopies in patients 45 years or older with an adequate bowel preparation performed by an endoscopist with at least one adenoma divided by the total number of screening colonoscopies in patients 45 years or older with an adequate bowel preparation performed by that endoscopist.

For the ADR-A we included diagnostic, screening and surveillance exams and excluded those performed for FOBT/FIT positive tests, IBD or genetic syndromes. To examine the association between ADR-A and post colonoscopy CRC, we divided ADR-A into categories based on quintiles of endoscopist level ADR-A detection rates^{13, 14}. We calculated ADR-S for each endoscopist using screening colonoscopies. To compare ADR-A and ADR-S, we divided these rates into quartiles⁹ based on the endoscopists’ respective ADRs.

Covariates

Potential covariates in our models included age, sex, index findings of clinically significant serrated polyps¹¹ (traditional serrated adenomas, sessile serrated polyps, large hyperplastic polyps or proximal hyperplastic polyps > 5 mm) or conventional advanced adenomas (large (≥ 1 cm), with villous elements or high grade dysplasia), index exam surveillance indication, bowel preparation quality (poor preparation; yes versus no), family history of CRC, and whether there was >1 surveillance exam after the index exam.

Statistical analysis

Means and standard deviations were calculated for continuous variables while frequencies and percentages were derived for proportions. We used the chi squared test for trend and Fisher’s exact test to evaluate categorical variables. T-tests were used to compare continuous variables and Mann-Whitney U tests were used for continuous variables with non-Gaussian distributions. Pearson Correlation was used to examine the strength of the association between 2 variables. Cox’s proportional hazards regression was used to model the hazard of PCCRC on the detection rates controlling for age, sex, indication of index exam, index exam findings, bowel prep quality, and having more than 1 surveillance exam. We calculated follow-up time from the date of the patient’s index exam until the time of second colonoscopy or CRC diagnosis, at least 6 months after index exam. For the secondary analysis, as described above, we excluded index exams that took place less than 12 months before the date of our NHCR-NH State Cancer Registry linkage to allow for adequate time for CRC diagnosis and reporting of the tumor. Thus, a patient with only an index exam and no follow up event would need a 12 month or longer follow-up time before the linkage with the NH State Cancer Registry in order to be included. We also calculated the ratio of the ADR-S 95% CI to that for ADR-A by dividing the 95% CI for the ADR-S by that for ADR-A. To further explore the association between ADR-A and PCCRC, we examined the ADR-A by deciles. All statistics were performed using SPSS 29 (IBM).

Results

Our sample included 32,535 exams performed by 152 endoscopists with 178 post colonoscopy CRCs (PCCRC) diagnosed 6 months or longer after the index exam. Index exams were performed between 10/2004 and 12/2020 and the follow up exams between 1/2006 and 5/2023. When examining PCCRC by time, 14 CRCs were diagnosed between 6 and 12 months after the index, 40 diagnosed between 12 and 36 months and 124 diagnosed 36 months or longer.

Factors associated with PCCRC are shown in Table 1. As compared to patients with no PCCRC, those with PCCRC were more likely to be older and have a shorter time to follow up. Both ADR-A and ADR-S of endoscopists who performed the index colonoscopy were lower in PCCRC patients as compared to patients without PCCRC. These data are shown in Table 1.

Table 1.

Characteristics of the patients and colonoscopies with and without PCCRC.

	Index exams with follow-up event
Characteristic	Follow up with no PCCRC	Patients with PCCRC	P value (With PCCRC versus no PCCRC)
Patients	N=32,357	N=178
Sex (% male)	47.9% (15,483)	44.4% (79)	0.35
Age (average, ±S.D.)	57.9 (9.5)	65.1 (10.7)	< 0.001
BMI (average, ±S.D.)	28.5 (6.1)	28.3 (6.5)	0.74
Exams
Bowel Prep (% poor)	2.1% (665)	2.8% (4)	0.86
Follow-up time to 1st exam or CRC diagnosis (mean months, ±S.D.)	66.5 (33.2)	56.5 (34.5)	< 0.001
ADR-A detection rate (average, +S.D.)	34.4 (9.1)	29.5 (8.7)	< 0.001
ADR-S detection rate (average, +S.D.)	29.1 (8.5)	23.4 (8.0)	< 0.001

When divided into quintiles, higher ADR-A was associated with lower unadjusted risks for PCCRC. After adjusting for covariates, ADR-A was also associated with lower PCRCC risk hazard ratios. These data are shown in Table 2. We also examined our data using a cutoff of 12 months and observed similar results (see Supplemental Table 1). We examined the characteristics of the endoscopists as stratified by ADR-A quintiles. As shown in Table 3, endoscopists with higher ADR-A were more likely to be gastroenterologists, have higher completion rates and higher serrated polyp detection rates as compared to those with lower ADR-As.

Table 2.

Unadjusted risks and Adjusted Hazard Ratios for Post Colonoscopy Colorectal Cancer as stratified by quintiles of endoscopist all exam adenoma detection rates (ADR-A)

	ADR-A
	< 23	23 - < 29	29 - < 34	34 - < 44	44 and higher
N	4130	5507	4668	13399	4831
PCCRC (N)	47	35	31	54	11
PCCRC (%)*	1.14%	0.64%	0.66%	0.40%	0.23%
Odd ratio (unadjusted)	1.0 (reference)	0.56 (0.36-0.86)	0.58 (0.37-0.92)	0.35 (0.24-0.52)	0.20 (0.10-0.38)
HR (95% CI)	1.0 (Reference)	0.56 (0.36-0.87)	0.60 (0.38-0.94)	0.43 (0.29-0.65)	0.32 (0.16-0.63)

^*p < 0.001 (Chi Square for trend)

Table 3.

Characteristics of endoscopists stratified by quintiles of endoscopist all exam adenoma detection rates (ADR-A)

	ADR-A
	< 23	23 - < 29	29 - < 34	34 - < 44	44 & higher
Endoscopists #	29	30	18	46	29
Gastroenterologists (N (%))*	10 (34.5%)	13 (43.3%)	13 (72.2%)	39 (84.8%)	21 (72.4%)
Completion rate**	97.3%	97.0%	98.3%	99.0%	99.3%
SSL detection rate**	1.64%	1.89%	4.17%	5.54%	8.94%

^*p<0.001 (Chi Square for trend)

^**p<0.001

To compare ADR-A to ADR-S, we divided both rates into quartiles. As shown in Table 4 both had similar unadjusted risks and hazard ratios associated with PCCRC. In addition, ADR-A and ADR-S quartiles had a similar predictive ability for PCCRC (area under the curve: ADR-A: 0.74; 95% CI: 0.70-0.77 and ADR-S: 0.74; 95% CI: 0.70-0.77). From the ROC curve for ADR-A, we observed that the optimal cut off point or threshold value was at an ADR of approximately 35%. Finally, the Pearson correlation coefficient for ADR-A and ADR-S for endoscopists was 0.83 (p < 0.001).

Table 4.

Unadjusted risks and Adjusted Hazard Ratios for Post Colonoscopy Colorectal Cancer for quartiles of endoscopist adenoma detection rates (ADR-A and ADR-S)

ADR-A
	< 25	25 - < 34	34 - < 42	42+
N	6243	8062	11690	6540
PCCRC (N)	59	54	50	15
PCCRC (%)*	0.95%	0.67%	0.43%	0.23%
HR (95% CI)	1.0 (Reference)	0.85 (0.58-1.24)	0.59 (0.40-0.87)	0.41 (0.23-0.75)
Endoscopists (N)	37	40	37	38
ADR-S
	< 21	21 - < 29	29 - < 35	35+
N	6050	12020	7316	7149
PCCRC (N)	56	84	23	15
PCCRC (%)*	0.93%	0.70%	0.31%	0.21%
HR (95% CI)	1.0 (Reference)	0.81 (0.58-1.14)	0.47 (0.28-0.77)	0.38 (0.21-0.70)
Endoscopists (N)	39	38	35	40

^*p <0.001 (Chi Square for trend)

Table 5 demonstrates the shift of endoscopists when detection rates were calculated by ADR-A as opposed to ADR-S. There 21 endoscopists who shifted to a lower quintile (those below and to the left of the diagonal) and 21 who moved to a higher quintile (above and to the right of the diagonal). There was an almost twofold higher (median=1.90; IQR=0.52) number of colonoscopies for each endoscopist for the ADR-A calculation as opposed to the ADR-S calculation.. In addition, The median ratio of the ADR-A 95% CI to the ADR-S 95% CI was 0.72 (IQR:0.10). In other words, on average, the 95% CI for ADR-A were 29% smaller (mean=0.71; S.D.: 0.09).

Table 5.

Endoscopists’ detection rates when calculating detection rates as ADR-A as opposed to ADR-S

		ADR-A
		< 25	25 - < 34	34 - < 42	42+	Total ADR-A
ADR-S	< 21	31	8	0	0	39
	21 - <29	6	23	8	1	38
	29 - <35	0	8	23	4	35
	35+	0	1	6	33	40
	Total ADR-S	37	40	37	38	152

We also examined ADR-A by decile and observed that the highest decile (46%), which was similar to the highest quintile (44%) had the lowest PCCRC risk (Supplemental Table 2).

To examine the impact of patients without a follow up event, we conducted a sensitivity analysis by including those NHCR patients (with and without follow up) at 12 months prior to the last linkage date between the NHCR and the NH State Cancer Registry. There were 110626 patients (46.0% male, average age 57.9 (±S.D. 9.5)) had an average time to censoring of 64.1 months (S.D.+32.8). As in the primary analysis, higher ADR-As were associated with lower unadjusted risks of PCCRC and HRs: ADR-As<23% (Reference) (23%-<29%: HR=0.79, 95% CI:0.51-1.23; 29%-<34%: HR=0.92,95% CI:0.58-1.45; 34%-<44%: HR=0.64, 95% CI:0.43-0.95; and ≥44%: HR=0.46, 95% CI:0.24-0.90).

Discussion

Adenoma detection rates are important quality measures which have shown to be associated with lower post colonoscopy CRC rates.^{1, 2} While the original proposed ADR included all exams in the calculation, the modified ADR used only screening colonoscopies.^{4, 15} Published data, including those from our group, have demonstrated that surveillance exams have higher detection rates than screening colonoscopies.^16–19 However, there are published data which suggest that using exams for all indications to calculate endoscopists’ ADRs can result in rates which are similar to those using screening exams. Ladabaum et al demonstrated that ADRs calculated with all exams had lower variability than ADR using screening exams.¹⁹ In addition, using data from US veterans, Kaltenbach et al demonstrated that ADR for screening exams were similar to that using all exams.¹⁷ The authors concluded that measuring ADR regardless of indication may be adequate for monitoring quality. Given these findings, the next step would be to validate ADR-A with respect to protection from post colonoscopy CRC.

The goal of our analysis was to examine the association between an ADR calculated using all exams (ADR-A) and the incidence of post colonoscopy colorectal cancer in patients undergoing colonoscopy. We observed that endoscopists with ADR-As of 44% or higher had the lowest post colonoscopy CRC risk hazard ratio relative to the lowest quintile. Since post colonoscopy CRC incidence is the most important outcome for CRC prevention with colonoscopy, these data validate ADR-A as a quality metric.

We also observed that ADR-A performed similarly to the ADR calculated using only screening exam data (ADR-S). Specifically in the highest quartiles, the post colonoscopy CRC risk hazard ratio was similar for the ADR calculated with screening exams (HR=0.38 95% CI 21-0.70), and the ADR calculated with all exams (HR=0.41 95%CI: 0.23-0.75). A recent study from Kaiser Permanente demonstrated that an all-exam ADR had similar protection from post colonoscopy CRC to a screening ADR.⁹ The PCCRC risk hazard ratios were similar to that observed in our study. A major difference between their analysis and our study was that our population did not include any patients who had a positive stool test for blood. Therefore, our data are more applicable for those centers who do not use occult blood testing as a primary tool for screening for colorectal cancer.

Another analysis using data from veterans examined the association between a global or indication-agnostic ADR and metachronous advanced neoplasia which included advanced adenomas and CRC.²⁰ They divided the ADR into quintiles ADR <19.7% (quintile 1), >19.7%–32.2% (quintile 2), >32.2%–39.3% (quintile 3), >39.3%–47.0% (quintile 4), and >47.0% (quintile 5). There was an increased risk for metachronous advanced neoplasia in the 3 lowest ADR quintiles but not the 2 highest quintiles. These data suggest that endoscopists should aim for an ADR-A of 40% or higher.

When comparing the two ADRs, we observed that 21 out of 152 (13.8%) endoscopists had a lower quartile when using the all-exam ADR as opposed to the screening ADR. A similar number, 21 (13.8%), moved to a higher quartile. These data are similar to that observed in the Kaiser Permanente analysis.⁹ We also observed that using ADR-A resulted in a significantly narrower 95% confidence interval. So even though there were doctors moving up or moving down in quintile, the overall measurements are more accurate and therefore reflect performance more precisely.

Our data have implications for practicing endoscopists. They validate the use of all-exam ADR as a quality measure; using all-exam ADR could mitigate the likelihood of an endoscopist “gaming” the system by changing the indication depending upon exam findings – for example, changing an exam’s indication from diagnostic to screening for a patient with an adenoma and no previous exams.⁷ An all-exam ADR also simplifies ADR calculation by obviating the need for differentiating by the exam indication. Finally, it increases the volume of colonoscopies for each endoscopist, allowing for a more precise measurement of the endoscopist’s detection rate. These are the major reasons why the recent ACG/ASGE quality metric guidelines endorsed the use of an all-exam ADR.⁵

With respect to ADR-A goals for endoscopists, our data suggest that higher ADR-As are associated with lower PCCRC risks. The current benchmark as per the recent ACG/ASGE latest recommendations on quality indicators for colonoscopy is 35%.⁵ This benchmark falls within our 4th quintile of ADR-A which is associated with a lower HR than the lowest quintile. This cutoff of 35% was also the optimal point on the ROC curve. While the lowest HR was observed for the highest quintile of an ADR-A of 44% or higher, we think that this is an aspirational target. We recognize that only 1/5 our endoscopists achieved an ADR-A 44% or greater. Thus, with respect to operationalizing based on our data, we would recommend that endoscopists achieve an ADR-A of at least 35%. Those who do not achieve this benchmark, should undergo remedial training to increase their detection. Potential steps to improve detection, as highlighted in the ACG/ASGE recommendations, include training in meticulous technique (which requires a withdrawal time of 8 minutes or longer), double examination of the proximal colon, and consideration of mucosal exposure devices or tools for highlighting flat lesions.^{5, 21}

Our data suggest that endoscopist training and endoscopic proficiency plays a large role in higher detection. Specifically, we observed that a higher percentage of the higher detectors were gastroenterologists. Furthermore, the endoscopists with higher ADR-As had significantly higher completion and serrated polyp detection rates.

Limitations in our analysis include the collection of data from a single state which has a low degree of racial diversity. While it is important to point out that there is significant socioeconomic and urban versus rural diversity in our population, we acknowledge that our findings should be validated in other populations. Another limitation is that we excluded IBD exams from our analysis, as per the recommendations of the new ACG/ASGE colonoscopy quality recommendations.⁵ Therefore our findings may not be generalizable to practices in which endoscopists are IBD focused.

In summary, we observed that an all-exam ADR of 44% or higher was associated with the lowest hazard ratio. For practicing endoscopists, our data also support the use of 35% a benchmark for ADR-A. The ADR-A metric also performed similarly to an ADR calculated using screen exams. Therefore, our data validate the widespread use of an all-exam ADR, calculated with all screening and surveillance colonoscopies plus diagnostic colonoscopies other than for positive fecal tests, as an important quality metric.

WHAT IS KNOWN

• Adenoma detection rates (ADR) using screening exams have been inversely correlated with post colonoscopy colorectal cancer (CRC)

• Recent ACG/ASGE guidelines have recommended calculating ADR using all exam indications as opposed to only screening

WHAT IS NEW HERE

• ADR-A using all exams was inversely correlated with PCCRC risk

• ADR-A had a similar association with PCCRC as ADR-S using screening exams

• ADR-A was associated with more precise measurement of detection rates as evident by a narrower ADR.