Roel Bogie

Chapter 9

CRC were excluded, as well as patients with inflammatory bowel disease or patients younger than 18 years of age. All newly diagnosed CRC cases from January 2004 to October 2014 were retrieved using the Nationwide Pathology Database (PALGA). We cross-linked detailed patient information from the hospital registries to endoscopic and pathological reports (see Figure 9.1 ). In that way we were able to identify post-colonoscopy colorectal cancers (PCCRCs). In addition, all CRC cases found in the prospective database were also screened for PCCRCs. Year of the last negative colonoscopy was used to compare PCCRC incidence pre- and post-training.

Training

End of follow-up

Post-training PCCRC indow

Post-training PCCRC window

Pre- training PCCRC window Pre- trai i

RC window

A

PALGA

Prospective cohort P ctive cohort

Incomplete follow-up

Pr

ining

Pre-training

2004

2008

2012

2016

Figure 9.1: Overview of study and the data sources used.

Definition and statistical analysis PCCRCs were defined as a CRC diagnosed 6 to 60 months after a colonoscopy that was negative for CRC. PCCRC rate after training was determined by the number of PCCRCs divided by the total number of colonoscopies (including surveillance colonoscopies), as described by Morris et al. 15 Patient years of follow-up (PYFU) were calculated using the time between index colonoscopy and end of follow-up (October 2014). Prior to training, only data on the number of colonoscopies per year were available. To estimate the PYFU we assumed that prior to training, the subset of index colonoscopies and mean follow-up was the same as in 2008 (i.e. both pre- and post-training 48.5 months). As sensitivity analysis we assumed complete follow-up (i.e. 60 months) for colonoscopies prior to training instead of the estimated 48.5 months. A Z-test was performed to compare PCCRC rates before and after training. A P value <0.05 was considered statistically significant. The other outcome parameter was the etiology of the PCCRCs, together with clinical and pathological characteristics (i.e. size, location, macroscopic appearance and histopathology). Location was subdivided in proximal colon (cecum to splenic flexure) and distal colon (descending colon to rectum). The PCCRCs were classified using the Pabby algorithm 5 modified by Huang et al. 16 ( Figure 9.2 ). Based on findings at the index-examination, time between colonoscopy and CRC diagnosis, location and stage of the tumor at diagnosis, the PCCRCs were categorized on most likely etiology: non adherence to surveillance intervals, inadequate examination/surveillance, incomplete resection, missed lesions or newly developed CRC.

176