- Email: [email protected]
Application of a conversion factor to estimate the adenoma detection rate from the polyp detection rate
ORIGINAL ARTICLE: Clinical Endoscopy
Application of a conversion factor to estimate the adenoma detection rate from the polyp detection rate Dawn L. Francis, MD, MHS, Daniel T. Rodriguez-Correa, PhD, Anna Buchner, MD, PhD, Gavin C. Harewood, MD, MBA, MSc, Michael Wallace, MD, MPH Rochester, Minnesota; Philadelphia, Pennsylvania; Jacksonville, Florida, USA; Dublin, Ireland
Background: The adenoma detection rate (ADR) is a quality benchmark for colonoscopy. Many practices find it difficult to determine the ADR because it requires a combination of endoscopic and histologic findings. It may be possible to apply a conversion factor to estimate the ADR from the polyp detection rate (PDR). Objective: To create a conversion factor that can be used to accurately estimate the ADR from the PDR. Design: This was a retrospective study of colonoscopies performed by board-certified gastroenterologists to determine the average adenoma to polyp detection rate quotient (APDRQ) for all endoscopists, individually and as a group. Setting: Academic group practice. Intervention: The group average APDRQ was used as a conversion factor for the endoscopist’s PDR to estimate the ADR. Main Outcome Measurements: The strength of the relationship between the estimated ADR and the actual ADR determined by Pearson’s correlation coefficient. Results: A total of 3367 colonoscopies performed by 20 staff gastroenterologists were included. The average ADR for all indications, all patient age groups, and both sexes was 0.17 (range 0.09-0.27, standard deviation 0.05). The average APDRQ was 0.64 (range 0.46-1.00, standard deviation 0.13). The correlation between the estimated ADR and the actual ADR was 0.85 (95% CI, 0.65-0.93, P ⫽ .000001). Limitations: Retrospective study in 1 practice setting with all patient types. Conclusions: The use of a conversion factor can accurately estimate the ADR from the PDR. Further study is needed to determine whether such a conversion factor can be applied to different practice settings and patient groups. (Gastrointest Endosc 2011;73:493-7.)
The prevention of colorectal cancer (CRC) by colonoscopy depends on the identification and removal of precancerous polyps, the most common of which are adenomas. Unfortunately, colonoscopy does not completely prevent CRC, with recent studies suggesting that it reduces mortality by 60% to 70%1 rather than the touted 90%, as previously reported.2,3 Although the rea-
sons for this are likely multifactorial, one reason is missed adenomas. The current information regarding the adenoma miss rate is disappointing, with miss rates for adenomas larger than 10 mm in diameter as high as 12% to 17%.4,5 It is well recognized that there is variability among endoscopists in their ability to detect and remove adeno-
Abbreviations: ADR, adenoma detection rate; APDRQ, adenoma to polyp detection rate quotient; CRC, colorectal cancer; PDR, polyp detection rate.
Current affiliations: Division of Gastroenterology and Hepatology (D.L.F.), Mayo Clinic, Rochester, Minnesota, USA, Mayo Medical School (D.T.R.-C.), Rochester, Minnesota, USA, Division of Gastroenterology and Hepatology (A.B.), University of Pennsylvania, Philadelphia, Pennsylvania, USA, Division of Gastroenterology and Hepatology (G.C.H.), Beaumont Hospital, Dublin, Ireland, Division of Gastroenterology and Hepatology (M.W.), Mayo Clinic, Jacksonville, Florida, USA.
DISCLOSURE: All authors disclosed no financial relationships relevant to this publication. Copyright © 2011 by the American Society for Gastrointestinal Endoscopy 0016-5107/$36.00 doi:10.1016/j.gie.2011.01.005 Received September 16, 2010. Accepted January 3, 2011.
www.giejournal.org
Reprint requests: Dawn L. Francis, MD, MHS, Mayo Clinic Rochester, 200 First Street SW, Rochester, MN 55905.
Volume 73, No. 3 : 2011 GASTROINTESTINAL ENDOSCOPY 493
Estimation of adenoma detection rate
mas during colonoscopy.6,7 Several different variables have been identified that are associated with adenoma detection rate: withdrawal time,8,9 quality of colon preparation,10,11 patient age and sex,12 and the time of day that the colonoscopy is performed.13,14 As such, an individual endoscopist’s adenoma detection rate (ADR) has been promoted as a quality benchmark by both the U.S. MultiSociety Task Force on Colorectal Cancer15 and a joint American College of Gastroenterology/American Society for Gastrointestinal Endoscopy task force.16 The calculation of the ADR for an individual endoscopist is not an easy task, especially for large-volume endoscopy practices and practices that do not automatically link histologic and endoscopic findings. The calculation requires the number of polyps found at colonoscopy and the histology of those polyps as determined by a pathologist. For many practices, this information is not easily combined because endoscopy information and histology are often in different locations within the medical record. As a result, the calculation of the ADR often requires a “hand” audit, a time-consuming and labor-intensive undertaking. It is unknown whether there is a conversion factor that could be applied to the polyp detection rate (PDR) to estimate the ADR for individual endoscopists. If such a factor exists, it would allow easy conversion from the PDR, a number readily accessible by endoscopic data only, to the ADR. The ADR is always a proportion of the PDR; that is, adenomas are always a subset of polyps that are removed. With this in mind, a possible conversion factor is the average proportion of the PDR that is the ADR among a diverse group of endoscopists. The aim of this study was to determine whether the average ADR to PDR quotient, ie, the proportion of colonoscopies in which at least 1 adenoma is detected (ADR) divided by the proportion of colonoscopies in which at least 1 polyp is detected (PDR), could be used to estimate accurately the ADR from the PDR.
METHODS After receiving approval from our institutional review board, we retrospectively evaluated colonoscopies at the Mayo Clinic Rochester over a 6-month period. We included colonoscopies performed by staff gastroenterologists, and the colonoscopies were performed with or without fellows and for all indications. We excluded procedures in which the cecum was not reached or the colonic preparation was considered inadequate to detect polyps that were smaller than 5 mm. We excluded endoscopists who had performed fewer than 200 colonoscopies in the preceding year. The Mayo Clinic Rochester has a proprietary electronic database that is used to enter procedural information in real-time called the Mayo Electronic Record for Gastrointestinal Endoscopy. Some of the items that are recorded in 494 GASTROINTESTINAL ENDOSCOPY Volume 73, No. 3 : 2011
Francis et al
Take-home Message ● ●
A conversion factor can accurately estimate the adenoma detection rate (ADR) from the polyp detection rate. An estimated ADR cannot take the place of an actual ADR calculation for determining compliance with established quality benchmarks, but may allow some providers and practices, who are not otherwise able to calculate an actual ADR, an accurate estimation.
this database include patient identifiers, the indication for the procedure, the performing endoscopist, the presence of a gastroenterology fellow, insertion and withdrawal times during colonoscopy, and the presence, size, and number of polyps. Polyp histology is not entered into this database. Therefore, histologic information was obtained from the electronic records separately. The PDR was considered to be number of colonoscopies in which at least 1 polyp was detected divided by the total number of colonoscopies and was calculated for each endoscopist. Similarly, the ADR is the number of colonoscopies in which at least 1 adenomatous polyp is detected divided by the total number of colonoscopies for each endoscopist.
Statistical analysis We determined the adenoma to polyp detection rate quotient (APDRQ) for all individual endoscopists and the weighted average for the group as a whole. Endoscopist A. ADR (proportion of colonoscopies in which at least 1 adenoma is detected) ⫼ PDR ⫽ APDRQ. ADR (0.25) ⫼ PDR (0.40) ⫽ 0.625 We then took the weighted group average (APDRQ) and used it as a conversion factor for each endoscopist’s PDR. Group average APDRQ ⫽ 0.64 Endoscopist A PDR ⫽ 0.40 PDR (0.40) ⫻ group average APDRQ (0.64) ⫽ 0.256 (estimated ADR) We used Pearson’s correlation coefficient to evaluate the strength of the relationship between the estimated and actual ADR. We used linear regression analysis to determine whether there were any endoscopist characteristics that we could identify that would predict the variability of the APDRQ among endoscopists. The following endoscopist characteristics were evaluated as independent (predictor) variables in the regression model: experience (⬍ or ⬎5 years since completion of fellowship training), average withdrawal time, colonoscopy completion rate, annual colonoscopy volume (preceding year), and average patient age. www.giejournal.org
Francis et al
Estimation of adenoma detection rate
Endoscopist
PDR
Actual ADR
Estimated ADR
1
0.40
0.25
0.26
2
0.23
0.08
0.15
3
0.29
0.19
0.19
4
0.31
0.19
0.20
5
0.15
0.13
0.09
6
0.28
0.18
0.18
We used linear regression analysis to evaluate whether an endoscopist or patient characteristic could be identified that would predict the variability of the APDRQ. We tested endoscopist experience (⬍5 years out of fellowship vs ⬎5 years out of fellowship), the total number of colonoscopies performed by that endoscopist in the preceding year, the average withdrawal time for average risk, screening colonoscopies for patients in whom no polyp was found, the cecal intubation rate for the preceding year, and the average age of the patient. None of these factors were found to significantly predict an individual endoscopists’ average APDRQ (Table 2). Only average patient age approached significance (P ⫽ .08).
7
0.20
0.15
0.13
DISCUSSION
8
0.41
0.21
0.26
9
0.19
0.12
0.12
10
0.24
0.16
0.15
11
0.15
0.15
0.10
12
0.39
0.25
0.25
13
0.22
0.13
0.14
14
0.28
0.16
0.18
15
0.36
0.26
0.23
16
0.37
0.27
0.24
17
0.28
0.15
0.16
18
0.26
0.13
0.15
19
0.30
0.14
0.17
20
0.31
0.18
0.18
TABLE 1. Actual polyp detection rate, actual adenoma detection rate, and estimated adenoma detection rate using the conversion factor of 0.64
ADR, Adenoma detection rate; PDR, polyp detection rate.
RESULTS A total of 3367 colonoscopies performed by 20 staff gastroenterologists were included. The average patient age was 64, and 54% were male. The average PDR for all indications was 0.28, whereas the average ADR for all indications was 0.17 (range 0.09-0.27, standard deviation 0.05). The weighted group average APDRQ was 0.64 (range 0.46-1.00, standard deviation 0.13). We then used 0.64 as a constant multiplier for each endoscopists’ PDR to estimate the ADR: the estimated ADR for endoscopist 1: 0.64 ⫻ 0.40 ⫽ 0.26; actual ADR for endoscopist A ⫽ 0.25. The actual PDR, actual ADR, and estimated ADR of each endoscopist are listed in Table 1. We used Pearson’s correlation coefficient to evaluate the strength of the relationship between the estimated and actual ADR, and it was 0.85 (95% CI, 0.65-0.93, P ⫽ .000001) (Fig. 1). www.giejournal.org
These data suggest that it is possible to estimate accurately the ADR from PDR by using a conversion factor. Moreover, this multiplier seems to remain stable regardless of endoscopist-related variables. We believe that this is an important finding because it will allow easier evaluation of colonoscopy quality in many practices without the requirement of laborious chart review. We firmly believe in the promotion of quality benchmarks for colonoscopy and recognize that the easier it is to gather the data for evaluation, the more likely it is that practices will measure and improve the quality of colonoscopies performed. Certainly the best marker of quality in colonoscopy is the CRC detection and miss rate. Given the relative rarity of CRC and the difficulty in tracking miss rates, the ADR was developed as a surrogate marker. A recent study shows that ADR is, in fact, a good surrogate for the CRC miss rate in that endoscopists with a low ADR also had a high CRC miss rate.17 One potential criticism of this study is that we have taken the accepted surrogate (ie, ADR) and diluted it even further—in essence, a surrogate for a surrogate. Although this notion is valid, we believe that the current limitation of endoscopy databases to interface endoscopic findings with pathology records is pervasive and an obstacle for many practices to track ADR, including our own. This study provides a ready solution until pathology and endoscopic data are automatically and uniformly linked. Most institutions and database vendors recognize that the calculation of the ADR must be automated with endoscopy tracking systems, and this should be aggressively pursued, but it will likely be years before this is accomplished on a broad scale. In the meantime, this conversion factor concept may be used to generate a thumbnail estimation of the ADR for providers who would otherwise have no ability to know their own ADR. This study was not performed to evaluate what a “good” ADPRQ is, but rather to create a robust conversion factor for ADR estimation. It is beyond the scope of this study to conclude whether it is better to remove more polyps, including small ones, with a lower ADPRQ or to remove fewer polyps and having more of them be larger Volume 73, No. 3 : 2011 GASTROINTESTINAL ENDOSCOPY 495
Estimation of adenoma detection rate
Francis et al
Figure 1. Estimated versus actual ADR.
TABLE 2. Linear regression analysis to evaluate influence of endoscopist and patient characteristics on the adenoma to polyp detection rate quotient Variables
Estimate
Pr > |t|
Intercept
0.459
0.621
Cecal intubation rate
⫺0.017
0.167
Withdrawal time
⫺0.016
0.396
Experience (⬎5 y)
0.061
0.517
Average patient age
0.030
0.083
None of these factors was found to significantly predict an individual endoscopist’s average adenoma to polyp detection rate quotient.
adenomas. One study suggests that the more polyps that are removed, the lower the ADPRQ but the higher number of total adenomas removed.18 The importance of removing very small adenomas remains to be defined. A distinct drawback in using the PDR to estimate the ADR is the risk that endoscopists may perform more biopsies of lesions than they would otherwise to falsely inflate their rates, so-called gaming the system. This may be especially tempting in pay-for-performance models of reimbursement. We chose to include all completed colonoscopies with adequate bowel preparations and for all indications because we believe that the APDRQ is determined primarily by endoscopist characteristics rather than patient characteristics. Further, including all these colonoscopies for analysis provides a more robust statistical model. We acknowledge, however, that the inclusion of all colonoscopy indications and patient groups (eg, young and old) may also be a limitation of this study design. It is unknown whether the APDRQ varies among different patient populations. Factors that most strongly influence an endosco496 GASTROINTESTINAL ENDOSCOPY Volume 73, No. 3 : 2011
pist’s ability to detect polyps (particularly small polyps) will likely have the strongest impact on the APDRQ. In all probability, these factors are endoscopist-related rather than patient related, with the exception of preparation quality. We included colonoscopies performed with involvement of gastroenterology fellows for a variety of reasons. First, the supervising endoscopist is responsible for the whole of the procedure, and because the ADR is an established benchmark, we assert that the quality of the colonoscopy performed with a fellow should be at least as good as that performed without a fellow. If there is a difference in the ADR with a fellow performing the procedure, that difference should be reflected in an endoscopist’s ADR as a more accurate assessment of the quality of his or her procedures. Finally, the one study that evaluated fellow participation with regard to the ADR showed that the ADR is higher when fellows are performing colonoscopy,19 so if there is any influence on the ADR, we expect that it is a positive one. It is interesting that there was such variability in the APDRQ among endoscopists (range 0.46-1.00, standard deviation 0.13). We believe that this variability is likely attributed to endoscopist characteristics, but using the standard variables thought to influence the ADR (endoscopist experience, withdrawal time, colonoscopy volume, and cecal intubation rate), we could not find one factor that strongly predicted this quotient. There are other modalities that may improve adenoma detection or differentiation of adenomas from other types of polyps that were not accounted for in our analysis: the quality of an individual endoscopists’ vision, endoscopist visual gaze patterns, or the use of narrow-band imaging, which is available on all of the colonoscopes used in our practice. This underscores an important and relatively unexplored fact in our field: endoscopists have highly variable practice patterns in the procedures that they perform. The root of this www.giejournal.org
Francis et al
Estimation of adenoma detection rate
variability is poorly understood. Further studies are required to better define what endoscopist characteristics are important in predicting their average APDRQ and to determine whether such characteristics need to be taken into account when estimating the ADR from the PDR. One of the most important roles of a gastroenterologist is to prevent CRC. The prevention of CRC with colonoscopy requires that the colonoscopy be complete, the colonic mucosa is adequately visualized, and lesions that are identified are properly removed. In recent years, the limitations of colonoscopy to identify all neoplastic lesions and to prevent all CRC have been underscored. It is the gastroenterologist’s responsibility to ensure that the colonoscopies performed are of the highest quality possible. We believe that to improve quality, a practice and an individual must first know the quality of the procedures that they perform20 and to be able to track that quality after improvement interventions. The development of a conversion multiplier to estimate the ADR from the PDR may be a useful tool to allow patient providers and practices who do not otherwise have the time or infrastructure to determine the ADR to estimate this quality benchmark.
7.
8.
9.
10.
11.
12.
13.
14.
15.
REFERENCES 1. Ransohoff DF. How much does colonoscopy reduce colon cancer mortality? Ann Intern Med 2009;150:50-2. 2. Bressler B, Paszat LF, Chen Z, et al. Rates of new or missed colorectal cancers after colonoscopy and their risk factors: a population-based analysis. Gastroenterology 2007;132:96-102. 3. Farrar WD, Sawhney MS, Nelson DB, et al. Colorectal cancers found after a complete colonoscopy. Clin Gastroenterol Hepatol 2006;4:1259-64. 4. Pickhardt PJ, Nugent PA, Mysliwiec PA, et al. Location of adenomas missed by optical colonoscopy. Ann Intern Med 2004;141:352-9. 5. Van Gelder RE, Nio CY, Florie J, et al. Computed tomographic colonography compared with colonoscopy in patients at increased risk for colorectal cancer. Gastroenterology 2004;127:41-8. 6. Bretagne JF, Hamonic S, Piette C, et al. Variations between endoscopists in rates of detection of colorectal neoplasia and their impact on a re-
16. 17. 18.
19.
20.
gional screening program based on colonoscopy after fecal occult blood testing. Gastrointest Endosc 2010;71:335-41. Chen SC, Rex DK. Endoscopist can be more powerful than age and male gender in predicting adenoma detection at colonoscopy. Am J Gastroenterol 2007;102:856-61. Barclay RL, Vicari JJ, Doughty AS, et al. Colonoscopic withdrawal times and adenoma detection during screening colonoscopy. N Engl J Med 2006;355:2533-41. Simmons DT, Harewood GC, Baron TH, et al. Impact of endoscopist withdrawal speed on polyp yield: implications for optimal colonoscopy withdrawal time. Aliment Pharmacol Ther 2006;24:965-71. Froehlich F, Wietlisbach V, Gonvers JJ, et al. Impact of colonic cleansing on quality and diagnostic yield of colonoscopy: the European Panel of Appropriateness of Gastrointestinal Endoscopy European multicenter study. Gastrointest Endosc 2005;61:378-84. Harewood GC, Sharma VK, de Garmo P. Impact of colonoscopy preparation quality on detection of suspected colonic neoplasia. Gastrointest Endosc 2003;58:76-9. Strul H, Kariv R, Leshno M, et al. The prevalence rate and anatomic location of colorectal adenoma and cancer detected by colonoscopy in average-risk individuals aged 40-80 years. Am J Gastroenterol 2006;101: 255-62. Sanaka MR, Deepinder F, Thota PN, et al. Adenomas are detected more often in morning than in afternoon colonoscopy. Am J Gastroenterol 2009;104:1659-64; quiz 1665. Sanchez W, Harewood GC, Petersen BT. Evaluation of polyp detection in relation to procedure time of screening or surveillance colonoscopy. Am J Gastroenterol 2004;99:1941-5. Rex DK, Bond JH, Winawer S, et al. Quality in the technical performance of colonoscopy and the continuous quality improvement process for colonoscopy: recommendations of the U.S. Multi-Society Task Force on Colorectal Cancer. Am J Gastroenterol 2002;97:1296-308. Rex DK, Petrini JL, Baron TH, et al. Quality indicators for colonoscopy. Am J Gastroenterol 2006;101:873-85. Kaminski MF, Regula J, Kraszewska E, et al. Quality indicators for colonoscopy and the risk of interval cancer. N Engl J Med 2010;362:1795-803. Chen SC, Rex DK. Variable detection of nonadenomatous polyps by individual endoscopists at colonoscopy and correlation with adenoma detection. J Clin Gastroenterol 2008;42:704-7. Rogart JN, Siddiqui UD, Jamidar PA, et al. Fellow involvement may increase adenoma detection rates during colonoscopy. Am J Gastroenterol 2008;103:2841-6. Harewood GC, Petersen BT, Ott BJ. Prospective assessment of the impact of feedback on colonoscopy performance. Aliment Pharmacol Ther 2006;24:313-8.
Registration of Human Clinical Trials Gastrointestinal Endoscopy follows the International Committee of Medical Journal Editors (ICMJE)’s Uniform Requirements for Manuscripts Submitted to Biomedical Journals. All prospective human clinical trials eventually submitted in GIE must have been registered through one of the registries approved by the ICMJE, and proof of that registration must be submitted to GIE along with the article. For further details and explanation of which trials need to be registered as well as a list of ICMJE-acceptable registries, please go to http://www.icmje.org.
www.giejournal.org
Volume 73, No. 3 : 2011 GASTROINTESTINAL ENDOSCOPY 497
Application of a conversion factor to estimate the adenoma detection rate from the polyp detection rate Dawn L. Francis, MD, MHS, Daniel T. Rodriguez-Correa, PhD, Anna Buchner, MD, PhD, Gavin C. Harewood, MD, MBA, MSc, Michael Wallace, MD, MPH Rochester, Minnesota; Philadelphia, Pennsylvania; Jacksonville, Florida, USA; Dublin, Ireland
Background: The adenoma detection rate (ADR) is a quality benchmark for colonoscopy. Many practices find it difficult to determine the ADR because it requires a combination of endoscopic and histologic findings. It may be possible to apply a conversion factor to estimate the ADR from the polyp detection rate (PDR). Objective: To create a conversion factor that can be used to accurately estimate the ADR from the PDR. Design: This was a retrospective study of colonoscopies performed by board-certified gastroenterologists to determine the average adenoma to polyp detection rate quotient (APDRQ) for all endoscopists, individually and as a group. Setting: Academic group practice. Intervention: The group average APDRQ was used as a conversion factor for the endoscopist’s PDR to estimate the ADR. Main Outcome Measurements: The strength of the relationship between the estimated ADR and the actual ADR determined by Pearson’s correlation coefficient. Results: A total of 3367 colonoscopies performed by 20 staff gastroenterologists were included. The average ADR for all indications, all patient age groups, and both sexes was 0.17 (range 0.09-0.27, standard deviation 0.05). The average APDRQ was 0.64 (range 0.46-1.00, standard deviation 0.13). The correlation between the estimated ADR and the actual ADR was 0.85 (95% CI, 0.65-0.93, P ⫽ .000001). Limitations: Retrospective study in 1 practice setting with all patient types. Conclusions: The use of a conversion factor can accurately estimate the ADR from the PDR. Further study is needed to determine whether such a conversion factor can be applied to different practice settings and patient groups. (Gastrointest Endosc 2011;73:493-7.)
The prevention of colorectal cancer (CRC) by colonoscopy depends on the identification and removal of precancerous polyps, the most common of which are adenomas. Unfortunately, colonoscopy does not completely prevent CRC, with recent studies suggesting that it reduces mortality by 60% to 70%1 rather than the touted 90%, as previously reported.2,3 Although the rea-
sons for this are likely multifactorial, one reason is missed adenomas. The current information regarding the adenoma miss rate is disappointing, with miss rates for adenomas larger than 10 mm in diameter as high as 12% to 17%.4,5 It is well recognized that there is variability among endoscopists in their ability to detect and remove adeno-
Abbreviations: ADR, adenoma detection rate; APDRQ, adenoma to polyp detection rate quotient; CRC, colorectal cancer; PDR, polyp detection rate.
Current affiliations: Division of Gastroenterology and Hepatology (D.L.F.), Mayo Clinic, Rochester, Minnesota, USA, Mayo Medical School (D.T.R.-C.), Rochester, Minnesota, USA, Division of Gastroenterology and Hepatology (A.B.), University of Pennsylvania, Philadelphia, Pennsylvania, USA, Division of Gastroenterology and Hepatology (G.C.H.), Beaumont Hospital, Dublin, Ireland, Division of Gastroenterology and Hepatology (M.W.), Mayo Clinic, Jacksonville, Florida, USA.
DISCLOSURE: All authors disclosed no financial relationships relevant to this publication. Copyright © 2011 by the American Society for Gastrointestinal Endoscopy 0016-5107/$36.00 doi:10.1016/j.gie.2011.01.005 Received September 16, 2010. Accepted January 3, 2011.
www.giejournal.org
Reprint requests: Dawn L. Francis, MD, MHS, Mayo Clinic Rochester, 200 First Street SW, Rochester, MN 55905.
Volume 73, No. 3 : 2011 GASTROINTESTINAL ENDOSCOPY 493
Estimation of adenoma detection rate
mas during colonoscopy.6,7 Several different variables have been identified that are associated with adenoma detection rate: withdrawal time,8,9 quality of colon preparation,10,11 patient age and sex,12 and the time of day that the colonoscopy is performed.13,14 As such, an individual endoscopist’s adenoma detection rate (ADR) has been promoted as a quality benchmark by both the U.S. MultiSociety Task Force on Colorectal Cancer15 and a joint American College of Gastroenterology/American Society for Gastrointestinal Endoscopy task force.16 The calculation of the ADR for an individual endoscopist is not an easy task, especially for large-volume endoscopy practices and practices that do not automatically link histologic and endoscopic findings. The calculation requires the number of polyps found at colonoscopy and the histology of those polyps as determined by a pathologist. For many practices, this information is not easily combined because endoscopy information and histology are often in different locations within the medical record. As a result, the calculation of the ADR often requires a “hand” audit, a time-consuming and labor-intensive undertaking. It is unknown whether there is a conversion factor that could be applied to the polyp detection rate (PDR) to estimate the ADR for individual endoscopists. If such a factor exists, it would allow easy conversion from the PDR, a number readily accessible by endoscopic data only, to the ADR. The ADR is always a proportion of the PDR; that is, adenomas are always a subset of polyps that are removed. With this in mind, a possible conversion factor is the average proportion of the PDR that is the ADR among a diverse group of endoscopists. The aim of this study was to determine whether the average ADR to PDR quotient, ie, the proportion of colonoscopies in which at least 1 adenoma is detected (ADR) divided by the proportion of colonoscopies in which at least 1 polyp is detected (PDR), could be used to estimate accurately the ADR from the PDR.
METHODS After receiving approval from our institutional review board, we retrospectively evaluated colonoscopies at the Mayo Clinic Rochester over a 6-month period. We included colonoscopies performed by staff gastroenterologists, and the colonoscopies were performed with or without fellows and for all indications. We excluded procedures in which the cecum was not reached or the colonic preparation was considered inadequate to detect polyps that were smaller than 5 mm. We excluded endoscopists who had performed fewer than 200 colonoscopies in the preceding year. The Mayo Clinic Rochester has a proprietary electronic database that is used to enter procedural information in real-time called the Mayo Electronic Record for Gastrointestinal Endoscopy. Some of the items that are recorded in 494 GASTROINTESTINAL ENDOSCOPY Volume 73, No. 3 : 2011
Francis et al
Take-home Message ● ●
A conversion factor can accurately estimate the adenoma detection rate (ADR) from the polyp detection rate. An estimated ADR cannot take the place of an actual ADR calculation for determining compliance with established quality benchmarks, but may allow some providers and practices, who are not otherwise able to calculate an actual ADR, an accurate estimation.
this database include patient identifiers, the indication for the procedure, the performing endoscopist, the presence of a gastroenterology fellow, insertion and withdrawal times during colonoscopy, and the presence, size, and number of polyps. Polyp histology is not entered into this database. Therefore, histologic information was obtained from the electronic records separately. The PDR was considered to be number of colonoscopies in which at least 1 polyp was detected divided by the total number of colonoscopies and was calculated for each endoscopist. Similarly, the ADR is the number of colonoscopies in which at least 1 adenomatous polyp is detected divided by the total number of colonoscopies for each endoscopist.
Statistical analysis We determined the adenoma to polyp detection rate quotient (APDRQ) for all individual endoscopists and the weighted average for the group as a whole. Endoscopist A. ADR (proportion of colonoscopies in which at least 1 adenoma is detected) ⫼ PDR ⫽ APDRQ. ADR (0.25) ⫼ PDR (0.40) ⫽ 0.625 We then took the weighted group average (APDRQ) and used it as a conversion factor for each endoscopist’s PDR. Group average APDRQ ⫽ 0.64 Endoscopist A PDR ⫽ 0.40 PDR (0.40) ⫻ group average APDRQ (0.64) ⫽ 0.256 (estimated ADR) We used Pearson’s correlation coefficient to evaluate the strength of the relationship between the estimated and actual ADR. We used linear regression analysis to determine whether there were any endoscopist characteristics that we could identify that would predict the variability of the APDRQ among endoscopists. The following endoscopist characteristics were evaluated as independent (predictor) variables in the regression model: experience (⬍ or ⬎5 years since completion of fellowship training), average withdrawal time, colonoscopy completion rate, annual colonoscopy volume (preceding year), and average patient age. www.giejournal.org
Francis et al
Estimation of adenoma detection rate
Endoscopist
PDR
Actual ADR
Estimated ADR
1
0.40
0.25
0.26
2
0.23
0.08
0.15
3
0.29
0.19
0.19
4
0.31
0.19
0.20
5
0.15
0.13
0.09
6
0.28
0.18
0.18
We used linear regression analysis to evaluate whether an endoscopist or patient characteristic could be identified that would predict the variability of the APDRQ. We tested endoscopist experience (⬍5 years out of fellowship vs ⬎5 years out of fellowship), the total number of colonoscopies performed by that endoscopist in the preceding year, the average withdrawal time for average risk, screening colonoscopies for patients in whom no polyp was found, the cecal intubation rate for the preceding year, and the average age of the patient. None of these factors were found to significantly predict an individual endoscopists’ average APDRQ (Table 2). Only average patient age approached significance (P ⫽ .08).
7
0.20
0.15
0.13
DISCUSSION
8
0.41
0.21
0.26
9
0.19
0.12
0.12
10
0.24
0.16
0.15
11
0.15
0.15
0.10
12
0.39
0.25
0.25
13
0.22
0.13
0.14
14
0.28
0.16
0.18
15
0.36
0.26
0.23
16
0.37
0.27
0.24
17
0.28
0.15
0.16
18
0.26
0.13
0.15
19
0.30
0.14
0.17
20
0.31
0.18
0.18
TABLE 1. Actual polyp detection rate, actual adenoma detection rate, and estimated adenoma detection rate using the conversion factor of 0.64
ADR, Adenoma detection rate; PDR, polyp detection rate.
RESULTS A total of 3367 colonoscopies performed by 20 staff gastroenterologists were included. The average patient age was 64, and 54% were male. The average PDR for all indications was 0.28, whereas the average ADR for all indications was 0.17 (range 0.09-0.27, standard deviation 0.05). The weighted group average APDRQ was 0.64 (range 0.46-1.00, standard deviation 0.13). We then used 0.64 as a constant multiplier for each endoscopists’ PDR to estimate the ADR: the estimated ADR for endoscopist 1: 0.64 ⫻ 0.40 ⫽ 0.26; actual ADR for endoscopist A ⫽ 0.25. The actual PDR, actual ADR, and estimated ADR of each endoscopist are listed in Table 1. We used Pearson’s correlation coefficient to evaluate the strength of the relationship between the estimated and actual ADR, and it was 0.85 (95% CI, 0.65-0.93, P ⫽ .000001) (Fig. 1). www.giejournal.org
These data suggest that it is possible to estimate accurately the ADR from PDR by using a conversion factor. Moreover, this multiplier seems to remain stable regardless of endoscopist-related variables. We believe that this is an important finding because it will allow easier evaluation of colonoscopy quality in many practices without the requirement of laborious chart review. We firmly believe in the promotion of quality benchmarks for colonoscopy and recognize that the easier it is to gather the data for evaluation, the more likely it is that practices will measure and improve the quality of colonoscopies performed. Certainly the best marker of quality in colonoscopy is the CRC detection and miss rate. Given the relative rarity of CRC and the difficulty in tracking miss rates, the ADR was developed as a surrogate marker. A recent study shows that ADR is, in fact, a good surrogate for the CRC miss rate in that endoscopists with a low ADR also had a high CRC miss rate.17 One potential criticism of this study is that we have taken the accepted surrogate (ie, ADR) and diluted it even further—in essence, a surrogate for a surrogate. Although this notion is valid, we believe that the current limitation of endoscopy databases to interface endoscopic findings with pathology records is pervasive and an obstacle for many practices to track ADR, including our own. This study provides a ready solution until pathology and endoscopic data are automatically and uniformly linked. Most institutions and database vendors recognize that the calculation of the ADR must be automated with endoscopy tracking systems, and this should be aggressively pursued, but it will likely be years before this is accomplished on a broad scale. In the meantime, this conversion factor concept may be used to generate a thumbnail estimation of the ADR for providers who would otherwise have no ability to know their own ADR. This study was not performed to evaluate what a “good” ADPRQ is, but rather to create a robust conversion factor for ADR estimation. It is beyond the scope of this study to conclude whether it is better to remove more polyps, including small ones, with a lower ADPRQ or to remove fewer polyps and having more of them be larger Volume 73, No. 3 : 2011 GASTROINTESTINAL ENDOSCOPY 495
Estimation of adenoma detection rate
Francis et al
Figure 1. Estimated versus actual ADR.
TABLE 2. Linear regression analysis to evaluate influence of endoscopist and patient characteristics on the adenoma to polyp detection rate quotient Variables
Estimate
Pr > |t|
Intercept
0.459
0.621
Cecal intubation rate
⫺0.017
0.167
Withdrawal time
⫺0.016
0.396
Experience (⬎5 y)
0.061
0.517
Average patient age
0.030
0.083
None of these factors was found to significantly predict an individual endoscopist’s average adenoma to polyp detection rate quotient.
adenomas. One study suggests that the more polyps that are removed, the lower the ADPRQ but the higher number of total adenomas removed.18 The importance of removing very small adenomas remains to be defined. A distinct drawback in using the PDR to estimate the ADR is the risk that endoscopists may perform more biopsies of lesions than they would otherwise to falsely inflate their rates, so-called gaming the system. This may be especially tempting in pay-for-performance models of reimbursement. We chose to include all completed colonoscopies with adequate bowel preparations and for all indications because we believe that the APDRQ is determined primarily by endoscopist characteristics rather than patient characteristics. Further, including all these colonoscopies for analysis provides a more robust statistical model. We acknowledge, however, that the inclusion of all colonoscopy indications and patient groups (eg, young and old) may also be a limitation of this study design. It is unknown whether the APDRQ varies among different patient populations. Factors that most strongly influence an endosco496 GASTROINTESTINAL ENDOSCOPY Volume 73, No. 3 : 2011
pist’s ability to detect polyps (particularly small polyps) will likely have the strongest impact on the APDRQ. In all probability, these factors are endoscopist-related rather than patient related, with the exception of preparation quality. We included colonoscopies performed with involvement of gastroenterology fellows for a variety of reasons. First, the supervising endoscopist is responsible for the whole of the procedure, and because the ADR is an established benchmark, we assert that the quality of the colonoscopy performed with a fellow should be at least as good as that performed without a fellow. If there is a difference in the ADR with a fellow performing the procedure, that difference should be reflected in an endoscopist’s ADR as a more accurate assessment of the quality of his or her procedures. Finally, the one study that evaluated fellow participation with regard to the ADR showed that the ADR is higher when fellows are performing colonoscopy,19 so if there is any influence on the ADR, we expect that it is a positive one. It is interesting that there was such variability in the APDRQ among endoscopists (range 0.46-1.00, standard deviation 0.13). We believe that this variability is likely attributed to endoscopist characteristics, but using the standard variables thought to influence the ADR (endoscopist experience, withdrawal time, colonoscopy volume, and cecal intubation rate), we could not find one factor that strongly predicted this quotient. There are other modalities that may improve adenoma detection or differentiation of adenomas from other types of polyps that were not accounted for in our analysis: the quality of an individual endoscopists’ vision, endoscopist visual gaze patterns, or the use of narrow-band imaging, which is available on all of the colonoscopes used in our practice. This underscores an important and relatively unexplored fact in our field: endoscopists have highly variable practice patterns in the procedures that they perform. The root of this www.giejournal.org
Francis et al
Estimation of adenoma detection rate
variability is poorly understood. Further studies are required to better define what endoscopist characteristics are important in predicting their average APDRQ and to determine whether such characteristics need to be taken into account when estimating the ADR from the PDR. One of the most important roles of a gastroenterologist is to prevent CRC. The prevention of CRC with colonoscopy requires that the colonoscopy be complete, the colonic mucosa is adequately visualized, and lesions that are identified are properly removed. In recent years, the limitations of colonoscopy to identify all neoplastic lesions and to prevent all CRC have been underscored. It is the gastroenterologist’s responsibility to ensure that the colonoscopies performed are of the highest quality possible. We believe that to improve quality, a practice and an individual must first know the quality of the procedures that they perform20 and to be able to track that quality after improvement interventions. The development of a conversion multiplier to estimate the ADR from the PDR may be a useful tool to allow patient providers and practices who do not otherwise have the time or infrastructure to determine the ADR to estimate this quality benchmark.
7.
8.
9.
10.
11.
12.
13.
14.
15.
REFERENCES 1. Ransohoff DF. How much does colonoscopy reduce colon cancer mortality? Ann Intern Med 2009;150:50-2. 2. Bressler B, Paszat LF, Chen Z, et al. Rates of new or missed colorectal cancers after colonoscopy and their risk factors: a population-based analysis. Gastroenterology 2007;132:96-102. 3. Farrar WD, Sawhney MS, Nelson DB, et al. Colorectal cancers found after a complete colonoscopy. Clin Gastroenterol Hepatol 2006;4:1259-64. 4. Pickhardt PJ, Nugent PA, Mysliwiec PA, et al. Location of adenomas missed by optical colonoscopy. Ann Intern Med 2004;141:352-9. 5. Van Gelder RE, Nio CY, Florie J, et al. Computed tomographic colonography compared with colonoscopy in patients at increased risk for colorectal cancer. Gastroenterology 2004;127:41-8. 6. Bretagne JF, Hamonic S, Piette C, et al. Variations between endoscopists in rates of detection of colorectal neoplasia and their impact on a re-
16. 17. 18.
19.
20.
gional screening program based on colonoscopy after fecal occult blood testing. Gastrointest Endosc 2010;71:335-41. Chen SC, Rex DK. Endoscopist can be more powerful than age and male gender in predicting adenoma detection at colonoscopy. Am J Gastroenterol 2007;102:856-61. Barclay RL, Vicari JJ, Doughty AS, et al. Colonoscopic withdrawal times and adenoma detection during screening colonoscopy. N Engl J Med 2006;355:2533-41. Simmons DT, Harewood GC, Baron TH, et al. Impact of endoscopist withdrawal speed on polyp yield: implications for optimal colonoscopy withdrawal time. Aliment Pharmacol Ther 2006;24:965-71. Froehlich F, Wietlisbach V, Gonvers JJ, et al. Impact of colonic cleansing on quality and diagnostic yield of colonoscopy: the European Panel of Appropriateness of Gastrointestinal Endoscopy European multicenter study. Gastrointest Endosc 2005;61:378-84. Harewood GC, Sharma VK, de Garmo P. Impact of colonoscopy preparation quality on detection of suspected colonic neoplasia. Gastrointest Endosc 2003;58:76-9. Strul H, Kariv R, Leshno M, et al. The prevalence rate and anatomic location of colorectal adenoma and cancer detected by colonoscopy in average-risk individuals aged 40-80 years. Am J Gastroenterol 2006;101: 255-62. Sanaka MR, Deepinder F, Thota PN, et al. Adenomas are detected more often in morning than in afternoon colonoscopy. Am J Gastroenterol 2009;104:1659-64; quiz 1665. Sanchez W, Harewood GC, Petersen BT. Evaluation of polyp detection in relation to procedure time of screening or surveillance colonoscopy. Am J Gastroenterol 2004;99:1941-5. Rex DK, Bond JH, Winawer S, et al. Quality in the technical performance of colonoscopy and the continuous quality improvement process for colonoscopy: recommendations of the U.S. Multi-Society Task Force on Colorectal Cancer. Am J Gastroenterol 2002;97:1296-308. Rex DK, Petrini JL, Baron TH, et al. Quality indicators for colonoscopy. Am J Gastroenterol 2006;101:873-85. Kaminski MF, Regula J, Kraszewska E, et al. Quality indicators for colonoscopy and the risk of interval cancer. N Engl J Med 2010;362:1795-803. Chen SC, Rex DK. Variable detection of nonadenomatous polyps by individual endoscopists at colonoscopy and correlation with adenoma detection. J Clin Gastroenterol 2008;42:704-7. Rogart JN, Siddiqui UD, Jamidar PA, et al. Fellow involvement may increase adenoma detection rates during colonoscopy. Am J Gastroenterol 2008;103:2841-6. Harewood GC, Petersen BT, Ott BJ. Prospective assessment of the impact of feedback on colonoscopy performance. Aliment Pharmacol Ther 2006;24:313-8.
Registration of Human Clinical Trials Gastrointestinal Endoscopy follows the International Committee of Medical Journal Editors (ICMJE)’s Uniform Requirements for Manuscripts Submitted to Biomedical Journals. All prospective human clinical trials eventually submitted in GIE must have been registered through one of the registries approved by the ICMJE, and proof of that registration must be submitted to GIE along with the article. For further details and explanation of which trials need to be registered as well as a list of ICMJE-acceptable registries, please go to http://www.icmje.org.
www.giejournal.org
Volume 73, No. 3 : 2011 GASTROINTESTINAL ENDOSCOPY 497