Setting a benchmark for serrated polyp detection rate: defining the target and terminology comes first

Some experts recommend the term “postcolonoscopy colorectal cancer” (PCCRC) be used to define a colorectal cancer (CRC) that occurs after a colonoscopy in which no cancer was diagnosed. PCCRC can be subtyped into “interval” cancer (diagnosed before the next recommended screening or surveillance colonoscopy) and “noninterval” cancer (identified at or after the next recommended screening or surveillance interval, or when no repeated colonoscopy was recommended, up to 10 years). In a review of 17 studies of nearly 150,000 cases of CRC, 5% were found to be interval PCCRCs, and these have been estimated to occur in 1 out of 3174 colonoscopies. A significant proportion of PCCRCs are deemed attributable to missed lesions or incompletely resected polyps and proven to be reduced by the performance of high-quality colonoscopy and implementation of quality improvement programs. PCCRCs may arise through conventional adenomas or the serrated pathway of carcinogenesis. The adenoma detection rate (ADR), a surrogate marker of a systematic and vigilant inspection of the colon, is one of the strongest colonoscopy quality metrics. ADR is inversely associated with PCCRC including advanced-stage and fatal cancer. Whereas the current benchmark for ADR in a mixed-gender screening population is ≥25%, recent data showed that the risk of PCCRC is significantly diminished when the ADR is ≥34%, and an aspirational ADR target of ≥35% has been suggested.^{,  ,} 

Sessile serrated polyps (SSPs), also known as sessile serrated adenomas or sessile serrated lesions, are precursors to nearly one third of CRCs and a cause of interval PCCRC. Wider variability in the detection and complete resection of SSPs than adenomas has been demonstrated in numerous studies. Despite variations in the features (histologic characteristics, location, and size) of serrated polyps used to create a serrated polyp detection metric, studies have repeatedly shown a considerable difference in serrated polyp detection rates (SPDRs) among endoscopists (Table 1), including a difference of 4-fold amid high-performing endoscopists.^{,  ,  ,}  Unlike ADR, data on the effect of SPDR on PCCRC rate are very scarce. In this issue of Gastrointestinal Endoscopy, Anderson et al report their findings examining the association between the endoscopists’ SPDR, defined by the authors as clinically significant serrated polyp detection rates (CSSDRs), and PCCRC rates using data from a large colonoscopy registry in New Hampshire. CSSDR was calculated by dividing the number of complete and adequate screening colonoscopies with ≥1 CSSP (traditional serrated adenoma or SSP of any size, hyperplastic polyp[HP] ≥1 cm, or HP >5 mm proximal to the sigmoid colon) by the total number of complete screening examinations with adequate preparation. Of the 19,532 patients included in the study, 128 (0.6%) PCCRCs were diagnosed, including 40 (31%) between 6 and 36 months, 36 (28%) between 36 and 60 months, and 52 (41%) >60 months after an index examination. Slightly more than half of the cancers (72/128) were in the right side of the colon, but neither microsatellite instability status, which may infer the biologic origin of the precursor lesion, nor stage was available. Colonoscopies were performed by 142 endoscopists. The authors do not report the number of colonoscopies per endoscopist. Information on withdrawal time, another quality measure associated with higher ADR and CSSDR, was not available or examined. Gender, body mass index, and family history of CRC did not differ between the 2 cohorts of patients with or without PCCRCs. Older age, lower CSSDR and ADR at index colonoscopy, and index examination by nongastroenterologist endoscopists were associated with higher PCCRC rates. Other studies have demonstrated significantly higher non-HP serrated polyp detection rates and proximal SSP detection rates among gastroenterologists than among other endoscopists and is in line with the findings by Anderson et al. Studies have also demonstrated that procedures performed by endoscopists with more recent completion of training and higher procedure volume are independently associated with higher SPDRs.^, Anderson et al categorized CSSDRs into approximate terciles, using the cutoff points of 3% and 9%, and intervals were <3%, 3% to <9%, and ≥9%. The absolute risk and hazard of PCCRCs was significantly lower when stratified by CSSDR of 3% to <9% and ≥9% versus <3% in the 3 time periods (6-36 months, 6-60 months, or ≥6 months) examined after the index colonoscopy. The absolute risk of both right-sided and left-sided PCCRC was reduced when the CSSDR was both between 3% to <9% and ≥9%.

Although PCCRC is a “true” patient outcome, interval PCCRC is the most important direct colonoscopy quality indicator. Establishing either PCCRC or interval PCCRC as a quality indicator may not be feasible in most clinical practices. The outcome is infrequent, requires systems to track large numbers of patients (who move in and out of individual practices or healthcare systems over time) and colonoscopies over many years, and is hampered by the lack of a national cancer registry in many countries. Additionally, it does not allow for timely feedback to endoscopists and implementation of quality improvement. For those reasons, a potentially easier to measure “surrogate” parameter. beyond ADR and associated with PCCRC, like SPDR, would be more meaningful for the endoscopist and endoscopy centers.

This is the first study to prove an association between SPDRs and PCCRC rates and to define a minimum cutoff point of competency represented by a CSSDR of 3%, using PCCRC rates as the primary outcome of interest. For that, the authors are commended on their effort. There are, however, some limitations to consider. It is not clear over what period of time the index colonoscopies were performed, given that many aspects of endoscopy and technology have changed, including the introduction of high-definition endoscopes and nuances in electronic image enhancement, lesion recognition, and polypectomy methods to ensure completeness of resection. Given their subtle appearance, serrated lesions were probably less likely to be detected by the use of standard definition colonoscopes, which could have introduced bias into the results. Although the absolute rate of PCCRC is low, with nearly two-thirds of PCCRC occurring between 6 months and 5 years, one has to question the quality of colonoscopy, including the endoscopists’ inspection technique, withdrawal time, and polyp resection methods, which were not available. A breakdown of the baseline polyp features, including conventional adenoma and serrated polyp histologic findings, would inform metachronous lesion risk and enable an understanding of the risk of an HP >5 mm proximal to the sigmoid colon on PCCRC. Another drawback, discussed by the authors, is the lack of the molecular characteristics of the PCCRCs, which makes it difficult to ascertain how many cancers occurred through the serrated pathway and therefore could have been prevented through accurate detection and resection of serrated polyps. Assuming that approximately 30% of CRCs arise through the serrated pathway, we estimate a smaller number of the outcome (PCCRC)—only approximately 40 out of the 128 PCCRCs reported—to be directly affected by the exposure factor (CSSDR), with possibly other residual confounding factors being responsible for the association.

An interesting observation in the study by Anderson et al is that regardless of endoscopist ADR <25% or ≥25%, a higher CSSDR was associated with lower PCCRC risk, implying that ADR alone may not be insightful of SPDR. Several studies have reported moderate correlation between ADR and SPDR, which suggests some but not a broad overlap between these 2 metrics, and a recent recommendation has been made for periodic monitoring of the endoscopists’ SPDRs even when their ADRs are satisfactory.^,

There is no consensus over the optimal SPDR metric to use if one is to be established. Ideally, the parameter should be easy to calculate while accurately reflecting the endoscopist’s ability to detect serrated polyps of clinical significance and should be associated with an important patient outcome, such as PCCRC. Histology-based metrics, such as SSP detection rate, relies largely on the pathologist’s interpretation of the tissue, which is not optimal, given the lack of agreement among pathologists in their diagnosis. The CSSDR proposed by Anderson et al is difficult to calculate because it requires the size, histologic features, and location of serrated polyps but has the advantage of showing the association with PCCRC. Although not validated for the clinical endpoint of PCCRC, proximal SPDR, defined as the detection rate of ≥1 proximal serrated polyps of any size and pathologic features in a screening cohort, is an easier metric to calculate and strongly correlates with CSSDR, making it more appealing for routine SPDR monitoring.

The take-home message from the study by Anderson et al is that serrated polyp detection, exemplified by the CSSDR metric, is inversely associated with PCCRC risk and may not be well accounted for when ADR alone is monitored. More research is needed to establish the ideal SPDR metric and to define a target for endoscopists to meet. Nonetheless the primary metric to focus on now is achievement of an aspirational ADR of ≥35% for all endoscopists who perform colonoscopy. Establishing a valid, clinically important, and easy to calculate serrated polyp detection metric will offer additional benefit to the quality of colonoscopy. For this to do so, consensus must be reached on the terminology and target serrated lesion(s) of interest. Future collaborative research is needed to reproduce the findings by Anderson et al outside of a single registry and must overcome the limitations cited above.