Statistics from Altmetric.com
Quality indicators for colonoscopy should be designed to render the outcomes of healthcare services measurable and transparent for both patients and physicians. For example, as the goal of colonoscopy is to prevent and detect colorectal cancer (CRC), improved quality should minimise the postcolonoscopy colorectal cancer (PCCRC) rate. In fact, most other indicators of procedural quality, such as adenoma detection rate and caecal intubation rate, owe their validation to a correlation with PCCRC rate. In an ideal world, rigorous monitoring of PCCRC rates can be used for benchmarking at multiple levels (regional, national, international) and would be a key driver of colonoscopy quality improvement within and outside screening programmes. It is therefore crucial to use a common language and common methodology when measuring, monitoring and reporting PCCRC.
The process of benchmarking is neither quick nor simple. It starts with the implementation of a uniform terminology for a PCCRC. The term ‘PCCRC’ refers to colonoscopy in general, performed for screening, surveillance or symptoms, whereas the term ‘interval CRC’ refers to screening and colonoscopy surveillance, when a follow-up time interval is specified (intention-to-screen).1 The next key issue is what and how to monitor for calculating PCCRC rates. Several caveats should be recalled, foremost of which are the lack of complete clinical information, hurdles in crosslinking a cancer registry to colonoscopy databases and ambiguity on how to calculate rates. Such factors hinder meaningful interpretation of PCCRC rates and defining of quality standards, as shown by Morris et al in Gut.2
The study by Morris and colleagues is the first to examine PCCRC rates across the English National Health service (NHS). In a retrospective observational population-based study, 94,648 of the 297,956 individuals with a first primary diagnosis of CRC had undergone a colonoscopy in the 3 years prior to CRC diagnosis. The authors demonstrate that previous methods used for calculating PCCRC rates3–6 could significantly affect outcomes, with PCCRC rates varying from 2.7% to 7.5%. They propose a novel method for calculating PCCRC rates, based on the year of colonoscopy in lieu of the CRC diagnosis (as used in previously published studies). Using this methodology, the NHS’ overall PCCRC rate within 3 years of colonoscopy (without applying any exclusion criteria) was 8.6% between 2001 and 2007, and gradually declined from 10.6% in 2001 to 7.3% in 2007.
Morris’ methodology models several desirable aspects of PCCRC measurement. For example, the use of England's National Cancer Data Repository (NCDR) is a valuable example of how the creation of large-scale cancer registry databases is a necessary first step in measuring the PCCRC rate. In addition, a carefully described and detailed methodology was used to link the CRC registry to the colonoscopy database. Finally, sources of bias were sought, and the importance of quality and accuracy of data sources in reporting PCCRC rates was discussed, revealing, for example, that the Hospital Episode Statistics component of the NCDR (from which colonoscopy findings have been extracted) could overestimate the PCCRC rate when used alone.
Therefore, beyond the implementation of a uniform terminology, the process of benchmarking requires the adoption of an organised approach to PCCRC measurement and management, as depicted in figure 1. Fundamentally, PCCRC calculation is derived from a comprehensive CRC registry that links with colonoscopy data. Beyond the patient's age and gender, information about the indication for colonoscopy (screening, surveillance or symptoms), and the a priori risk for CRC of the subject affected by PCCRC as compared with the background population (eg, presence of IBD or hereditary forms of CRC) would greatly enhance our understanding of the PCCRC rate.
In the second phase, the crosslinking strategy needs to be verified and validated through case ascertainment. In particular, it can be challenging to differentiate a detected cancer (eg, cancer diagnosed within 0–6 months after a colonoscopic examination) from a PCCRC (eg, cancer diagnosed 6–36 months after an examination) in certain situations, such as in patients with multiple or complex polyps requiring follow-up examinations.
The results should be carefully examined and categorised (data mining). In this step, manual chart review performed on a segment of the target population is an important aspect of case ascertainment, and cautious interpretation of the results is advised in its absence. PCCRC may be attributable to factors that are unrelated to colonoscopy quality. For example, a CRC may occur as a result of non-compliance with the recommended follow-up interval, possibly related to systemic issues, such as poor communication or poor access. In other cases, the PCCRC rates may be underestimated or overestimated by purely administrative causes. A single-centre study of PCCRCs from Pennsylvania7 revealed that nearly half of the PCCRCs diagnosed in their study were either systemic or (so-called) ‘administrative errors’, such as failure to return for repeat colonoscopy (after incomplete colonic examination or incomplete polypectomy), and delays in surgical pathology diagnosis or cancer registration.
While the optimal measure of PCCRC is yet to be defined, and should probably include all colonoscopies in the fraction denominator as proposed in the Morris paper, the study highlights, once again, that high-resolution data is mandatory for both CRC registry and colonoscopy databases, so that, ultimately, we could obtain measures of PCCRC that are age-specific, gender-specific and indication-specific. A minimum dataset for registration of PCCRC should ideally include the context in which the procedure was performed (eg, screening programme or opportunistic screening), the initial test employed (eg, faecal immunological test (FIT)/ guaiac faecal occult blood test (gFOBT)), the recommended surveillance interval (according to published guidelines), the time elapsed from the index colonoscopy to CRC diagnosis, the location of CRC, the histopathology and the CRC stage at diagnosis.1 Collection of comprehensive datasets, including high-quality photodocumentation, and subsequent data mining will help to reconstruct the most likely aetiological factors of the PCCRCs and, in particular, to disentangle PCCRCs attributable to procedural factors from those associated with a more aggressive tumour biology, or to systemic issues that impact patient compliance with recommendations.8 ,9 In turn, the correlation of the PCCRC rate with quality markers, such as adenoma detection10 ,11 and polyp resection12 rates, can be validated. Such a strategy will allow the creation of tailored interventions needed to minimise PCCRC rates. Finally, monitoring and reporting PCCRC can be complex and therefore requires a multidisciplinary framework including clinical, epidemiological and technical communication expertise.
In conclusion, the Morris paper improves our understanding of the gaps in monitoring PCCRC and provides opportunities to close them. The authors of this commentary encourage the application of a structured standard operating procedure to facilitate monitoring of PCCRC rates and progress in this field. However, such progress will depend on the comprehensive characterisation of the subjects affected by PCCRC, the context in which such cancers occurred and the tumour characteristics. Careful data mining will pave the way towards PCCRC benchmarking, and help to identify room for improvements.
Competing interests None.
Provenance and peer review Not commissioned; internally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.