Adverse events of colonoscopy in a colorectal cancer screening programme with faecal immunochemical test: a population-based community-based observational study

INTRODUCTION

Colorectal cancer (CRC) is the second most common cancer in Europe (500,000 estimated cases in 2018) and the second leading cause of death from malignant disease, resulting in 243,000 estimated deaths in 2018.¹ Most are preventable. Several randomised controlled trials (RCTs) have demonstrated the efficacy of screening with guaiac-based faecal occult blood test (gFOBT) and flexible sigmoidoscopy on CRC mortality.² Many countries have thus launched gFOBT CRC screening programmes.³ The efficacy of screening with a faecal immunochemical test (FIT) on CRC mortality is not demonstrated through RCTs but the superiority of FIT over gFOBT is now well established, so that many countries, including France, are now organising FIT CRC screening programmes instead of gFOBT programmes.³ However, the benefit-risk balance of these programmes is poorly evaluated.⁴ On one hand, none of the population-based screening programmes with gFOBT has been able to reproduce the CRC mortality reduction promised by the RCTs, and the expected benefit of population-based screening programmes with FIT has not been quantified yet. On the other hand, the risk of screening trials and programmes with gFOBT and FIT is poorly evaluated.⁵ We demonstrated that the adverse events (AEs) of colonoscopy were underestimated in all RCTs with gFOBT and are greater in a real-world programme.⁶ The rate of AEs requiring hospitalisation was estimated at 7.0% and 10.0% in two FIT CRC screening pilots, notably higher than the upper limit of 5% recommended by the European Society of Gastrointestinal Endoscopy (ESGE).^7–9 Owing to the established risk factors (polypectomy, polyp size…) and since the neoplasia yield is significantly higher, AEs should be more prevalent in FIT-positive colonoscopies than in gFOBT-positive colonoscopies and, a fortiori, than in direct screening colonoscopies.^10,11 Valid real-world data is needed to inform target populations, caregivers, policy-makers and funders.⁴

Our aim was to assess colonoscopy-related AEs in the CRC screening programme organised in France using an FIT and to compare them with those of the previous gFOBT programme.

METHODS

We retrospectively analysed the AEs of all colonoscopies performed in residents undergoing colonoscopy for a positive FOBT within the population-based CRC screening programme organised in Alsace, a region in eastern France, 1.8 million inhabitants, part of the French national programme. The FOBT was a gFOBT from September 2003 to July 2014 and a quantitative FIT from May 2015 to January 2018.

RESULTS

DISCUSSION

In our population-based community-based CRC screening programme with FIT, the rate of AEs requiring hospitalisation was 11.9‰ overall, 1.9‰ for perforation and 3.2‰ for bleeding. The rate of AEs requiring hospitalisation >24hrs was 5.7‰, 1.6‰ for perforation and 2.2‰ for bleeding. The rate of AEs necessitating surgery was 0.8‰, 0.7‰ for perforation and 0‰ for bleeding. Gastrointestinal AEs requiring hospitalisation and hospitalisation >24hrs remained stable during the gFOBT and FIT periods despite a significant increase in several risk factors for complications: the rate of therapeutic colonoscopies increased from 48.7% to 64.7%, the mean number of polypectomies per therapeutic colonoscopy from 2.0 to 2.4, and the rate of proximal polyps from 28.9% to 35.9%. Compared with the gFOBT period, the need for surgical management of AEs decreased significantly during the FIT period, from 70.0% to 38.9% for perforation and from 4.3% to 0% for bleeding. During the whole gFOBT and FIT period, we observed one death (30-day mortality 1 / 27,000 colonoscopies) and three splenic injuries (1 / 9000 colonoscopies).

The main strengths of our study are the population-based community-based setting along with the direct patient surveys. Indeed, this study is one of the first to assess the harms of colonoscopy in a population-based FIT CRC screening programme. It is also noteworthy that the details of all AEs were confirmed through direct access to hospital charts, colonoscopy reports, and, whenever necessary, phone calls to the gastroenterologist and/or the general practitioner and/or the patient. Albeit, this study is not without weaknesses. The response rate to the patients’ survey was 51.3%, low in comparison with the 83.7% of the NHS BCSP,¹⁰ and hospitalisation claims for the period could not be analysed, so that some AEs may have been missed, leading to a potential underestimation of the rate of AEs. However, the low response rate induced an underestimation of incidents and AEs without hospitalisation only, without consequence for AEs requiring hospitalisation, hospitalisation >24hrs and surgery. Therefore, underestimation of clinically relevant AEs is probably less of an issue. Another limitation is that the postal surveys were performed every other year, so that the delay between colonoscopy and survey could be long for some patients. Likewise, this delay may have induced an underestimation of incidents and AEs without hospitalisation as patients may have forgotten some minor complications.

The harm caused by colonoscopies in our FIT programme is in line with that recently reported by two other programmes (table 4).^7,8 Their rates of 7.0‰ and 10.0‰, along with our 11.9‰ rate of complications requiring hospitalisation, are notably higher than the maximum standard of 5‰ for 7-day hospital admission/readmission rate recommended by the ESGE.⁹ They were higher than those usually reported, that is 0.5‰ for perforation and 2.6‰ for bleeding in a systematic review and meta-analysis of 21 studies along with those reported in another review of six studies on colonoscopies for stool-positive testing, i.e. 0.8‰ for perforation and 1.9‰ for bleeding.^19,20 However, the comparison with such recommendations and aggregated figures is not pertinent as there is considerable heterogeneity in the literature. Today, it is almost impossible to compare different series dealing with colonoscopy-related AEs because indication and yield are different; AE nomenclature and sources of information are different; rates of referral for surgery are different. Even within a single programme, AE rates vary twofold: e.g. from 7.2‰¹⁶ to 14.2‰¹⁰ in the English Bowel Cancer Screening Programme (BCSP).

Neoplasia yield is correlated with the risk of complication and increases significantly depending on the indication for colonoscopy (symptoms, direct screening, positive gFOBT, positive FIT).^6,7 The rate of therapeutic colonoscopies varies more than twofold between series, from 32.9% in the German colonoscopy screening programme²¹ to around 50% in gFOBT programmes and 55% to 70% in FIT programmes.^7,8 A way to compare different series is to analyse separately diagnostic and therapeutic colonoscopies. For diagnostic colonoscopies, our perforation and bleeding rates requiring hospitalisation were 0.9‰ and 0‰, in line with those reported in the meta-analysis (0.4‰ and 0.6‰, respectively).¹⁹ For therapeutic colonoscopies, comparison is possible only between series having similar indications and neoplasia yields. Our 5.0‰ rate of bleeding requiring hospitalisation was significantly lower than the 9.8‰ reported in the meta-analysis.¹⁹ By contrast, our perforation rate was 2.4‰, significantly higher than the 0.8‰ in the meta-analysis.¹⁹ Our high rate is probably related to the high number of polyps per therapeutic colonoscopy (2.4), their large size (5.6% of polyps ≥ 2 cm), their proximal location (35.9%) and the low rate of surgical resection of benign polyps (1.8%). This level of neoplasia yield is seldom encountered, almost exclusively in FIT-positive colonoscopies. The rate of polyp(s) ≥ 2 cm was 2.1% in the English gFOBT BCSP.²² The rate of patients with at least one polyp ≥ 2 cm was 1.1% in the Austrian colonoscopy screening programme (7.4% in our programme).¹⁸

Colonoscopy techniques have evolved continuously towards more comfort, safety and ability to remove large lesions. Recent haemostasis and suture techniques allow endoscopic management of complications. Gastrointestinal AEs can be expected to be more prevalent today with AEs related to endoscopic mucosal resection of numerous, large and right-sided polyps. In fact, during our FIT period, all bleeding episodes were managed endoscopically. Moreover, the rate of perforations managed endoscopically increased significantly from 30% to 61% between the gFOBT and FIT periods. In the latter period, 39% of perforations were probably inevitable, caused by an “expert endoscopist” during large polypectomies. Almost half (45%) were diagnosed immediately and managed endoscopically, the median length of the hospital stay being 2.5 days. These figures indicate that in recent years in Alsace community gastroenterologists have improved their performance level significantly. One might wonder whether accredited gastroenterologists could perform better. In the French and Scottish screening programmes,¹⁷ colonoscopies are performed by any certified community gastroenterologist, whereas in England and the Netherlands they are performed by a set number of accredited gastroenterologists.

Our findings suggest that the rate of AEs is significantly higher in a community-based population-based FIT CRC screening programme than usually reported.^19,20,23 The invited population should be informed for improved shared decision-making. Today, the leaflet accompanying the letter inviting the French population to participate in FIT CRC screening mentions in tiny characters that serious colonoscopy-related AEs are rare, estimated at 3‰. The word “serious” is not further explained. Maintaining this figure that underestimates by a factor of two the true prevalence of AEs is a kind of propaganda. It must be updated. We propose this statement “Colonoscopy-related adverse events, mainly bleeding and perforation, are rare. It is estimated that for 1000 colonoscopies performed for a positive FIT, six complications requiring hospitalisation longer than 24 hours and one complication necessitating surgery will occur”.

The ASGE lexicon is essential as it is almost the only one to propose a standardised nomenclature for AEs¹⁴ However, it is not satisfactory for common use:^6,10,16,23 it is rather complex so that it has to be detailed every time it is used; it is best suited for the United States healthcare system; and certain criteria are somewhat arbitrary and open to criticism.⁶ Moreover, it does not define any specific method for identifying AEs. The less one seeks, the less one finds. All studies that rely on voluntary reporting underestimate actual complication rates.^8,24 In our study, 15% of AEs requiring hospitalisation were not reported by gastroenterologists. In Germany, the AE rate was threefold higher in an audit than in the national colonoscopy screening registry.²⁴ The Danish CRC screening database registered only 29.4% of recognised complications.⁸ A survey directed toward all patients should be mandatory for the publication of any article dealing with AEs. Moreover, should we still be calling a perforation occurring during a large polypectomy that is immediately diagnosed, closed and under surveillance < 48hrs in hospital a complication? There is an imperative and urgent need for a worldwide consensus on AE nomenclature, collection and reporting. An international task force should be created to formulate a series of recommendations, standardise definitions and categories, develop standardised methodology to search for AEs, and establish rules for reporting endoscopy-related AEs, for the same reasons and in the same way as were established the consensus on post-colonoscopy CRC and the STROBE statement.²⁵ In anticipation, we adopted three simple, precise, self-explanatory indicators, all relevant for patient information: AEs requiring hospitalisation, as it is the most commonly used criterion to define clinically relevant complications; AEs requiring hospitalisation >24hrs, as we feel that an overnight admission for surveillance after colonoscopy (around 50% of our hospitalisations) is not a true complication; and AEs necessitating surgery.

Rate of referral for surgery must also be considered. An AE following the treatment of a large sessile polyp is attributed either to endoscopy or to surgery depending on the type of resection. Surgical resection-related morbidity-mortality is significantly higher than endoscopic polypectomy-related morbidity-mortality.^26,27 Consequently, if the rate of referral for surgery for benign polyps is high, the individual endoscopist’s or CRC screening programme’s colonoscopy-related AE rate will be low while the overall AE rate will be high. Unfortunately, the rate of referral for surgery is never specified in articles concerning colonoscopy-related AEs and results of CRC screening programmes. Yet, 1) it is far from negligible as surgery for benign colorectal polyps represents 25% of surgeries for colorectal neoplasia in the USA,²⁸ and 2) surgery-related morbidity-mortality is almost entirely preventable, provided the endoscopist who encounters a polyp he/she cannot manage personally refers the patient to an “expert endoscopist” rather than to a surgeon. “Expert endoscopists” are able to remove endoscopically more than 90% of large polyps.²⁹ Overall, all studies concerning AEs of colonoscopy series and CRC screening programmes should mention their rate of referral for surgery for benign polyps (or, failing that, mention their surgery-related AEs). The rate of referral for surgery was 1.8% in our programme, about half that reported in Brittany (4.1%).³⁰ Our low rate of surgical resection may partly explain our high rate of perforations following therapeutic colonoscopies.

The benefit-risk balance of CRC screening remains to be assessed correctly.^4,5 At a time when most countries are organising CRC screening programmes, there is an urgent need for a worldwide consensus on standardised indicators. Overall, owing to the risk factors of colonoscopy-related AEs (polyp number and size) the risk is roughly proportionate to the benefit. We propose a new indicator that is the number of persons harbouring at least one adenoma ≥ 10 mm per AE requiring hospitalisation >24hrs (that could be abbreviated A10+/AE24). Between our gFOBT and FIT programmes, A10+/AE24 increased from 39 to 52. This indicator would be the cornerstone for assessment and comparison of the benefit-risk balance of different CRC screening programmes (e.g. between different countries) and strategies (e.g. between FIT and colonoscopy screening programmes). Unfortunately, this indicator cannot be calculated in any previous study, so we presented another close indicator in table 4, that is the number of persons harbouring an advanced adenoma per AE requiring hospitalisation >24hrs.