Meta Van Lanschot

Of high-risk adenomas than the currently used phenotypical features size, grade of dysplasia and villous histology that define advanced adenomas. Also in other cancer types, insights in the genomic basis of progression have enhanced early detection strategies.
In chapter 3 a novel method for the detection of SCNAs was validated. By amplifying specific genomic loci using a single primer pair according to the FAST-SeqS method and analysing the data with the associated tool ‘conliga’, we were able to generate SCNA profiles that maintained high-resolution SCNA information. With the FAST-SeqS approach, the complicated and expensive library preparation steps associated with whole genome sequencing (WGS) were replaced by a two round polymerase chain reaction (PCR). This resulted in reduced preparation time and lower costs compared to WGS techniques. FAST-SeqS was able to detect SCNAs also in samples with low tumour purity. FAST-SeqS could have potential to be applied in the clinic, for example in patient screening and surveillance for cancer.
In chapter 4 the natural behaviour of colorectal lesions, including adenomas and serrated polyps, was studied. For this, we had access to a unique set of small polyps with longitudinal follow up data over a period of approximately 3 years. On these polyps a comprehensive analysis of DNA alterations was performed. We found that higher polyp growth rates were related to presence of non-random SCNAs associated with adenoma-to-carcinoma progression (cancer associated events or CAEs), as well as to increased mutation burden. Regressed lesions did not have CAEs, but did harbour some mutations, which concerned mostly APC mutations, an early event in adenoma genesis. Altogether this longitudinal study provides in vivo support in the human setting for the functional role of these molecular alterations, that mostly have been identified by cross sectional observations in tissue samples of colorectal adenomas and cancers.

Part II

Chapter 5 provided an overview of the effectiveness of primary screening colonoscopy in reducing CRC incidence and mortality. Studies have shown that the effect of colonoscopy screening is more pronounced for distal cancers than proximal cancers. Indeed, interval cancers are predominantly located in the proximal colon. This may be due to unfavourable tumour characteristics, but also to potentially avoidable factors relating to colonoscopy quality. Surveillance programs are in place to enhance the protective effect of screening. The so named ‘high-detection paradox’ refers to the detection of more diminutive and small polyps due to improved colonoscopy quality and imaging techniques, leading to an increasing number of surveillance colonoscopies. This trend emphasises that the optimal interval between surveillance examinations for different patient categories requires re-examination.
In chapter 6 we outlined a large cross-sectional study which started in 2015 and is still ongoing in multiple centres throughout the Netherlands. The aim of the MOCCAS (Molecular stool testing for Colorectal CAncer Surveillance) study is to evaluate whether stool testing could be used in the surveillance setting as a triage method to select patients with advanced neoplasia for therapeutic colonoscopy. For this, surveillance patients are asked to collect stool specimens for the multitarget stool DNA (mt-sDNA) test and two FITs (OC Sensor and FOB Gold), prior to their scheduled colonoscopy.
The interim results of the MOCCAS study were presented in chapter 7. We found that mt-sDNA had a higher sensitivity for the detection of advanced neoplasia than FIT. Lowering the screening cut-offs to reach 50% positivity rate, increased the sensitivity of the mt-sDNA test and FITs considerably, thereby reducing the risk of missing advanced neoplasia. In this scenario, a single round of mt-sDNA testing missed 24% of advanced neoplasia, while FIT missed a significantly higher proportion of 40%. Mathematical modelling approaches will be applied to the final study data in order to determine an optimal stool-based strategy for surveillance.
Despite the higher sensitivity of the mt-sDNA test, the test is also more expensive, more complicated to perform and comes with logistic challenges as it requires whole stool samples. Therefore, in chapter 8 we tested whether the performance of FIT could be improved by complementation with clinicopathological risk factors. In addition to performing stool collection, all patients included in the MOCCAS study filled in a questionnaire with risk factors for CRC. The questionnaire data were combined with the FIT result and historical colonoscopy findings to update a model that was previously developed in the screening setting to predict the risk of advanced neoplasia. The updated model included FIT result, age, calcium intake, smoking habits, (tubulo)villous adenoma in previous colonoscopy and large lesions in previous colonoscopy. Application of the updated model improved the performance of the FIT in the detection of advanced neoplasia significantly. At equal positivity rate of 50%, the sensitivity of FIT only was 68% compared to 75% when applying the model.
In chapter 9 the patient category undergoing surveillance after curative surgical CRC resection was studied more closely. These patients are currently recommended to have their first surveillance colonoscopy already one year after CRC resection. However, there is controversy whether the one-year interval between pre- and postoperative colonoscopy is indeed justified, due to improved colonoscopy quality standards. Despite confirmed high quality colonoscopies, we found that the yield of CRC was 1.7% one year after resection of the first CRC. The detected cancers included metachronous, as well as recurrent tumours. Considering the late stage of apparently metachronous cancers, these might actually represent CRCs that were missed during the previous colonoscopy. The high yield justifies the recommendation of a one-year surveillance interval after surgical CRC resection.

Future perspectives

Once metastasised to distant organs, cancer survival rates decrease dramatically and treatment burden and costs increase. When thinking about reducing CRC mortality, the focus should therefore not be on curing advanced cases, but rather on early disease detection. As CRC develops over a course of 10-15 years, a wide window of opportunity exists for doing so. The aim of surveillance for CRC is to reduce disease incidence and mortality and to do so with a sustainable amount of medical and economical resources. At present, approximately 25% of the colonoscopy capacity is consumed by surveillance. Due to screening, the burden of surveillance is likely to further increase in the near future. This results in high costs and may lead to longer waiting times for colonoscopies for this as well as other indications. This situation is problematic, especially because strong evidence for the effectiveness of surveillance for all different subgroups is lacking. Therefore, the focus of this thesis was to improve surveillance strategies.
Surveillance starts with the performance of a high quality baseline colonoscopy with complete removal of all detected lesions. Only if quality parameters are met, patients are entered into the surveillance program. To avoid overuse of colonoscopy but still prevent CRC, the timing of the surveillance examination is essential. Currently, colonoscopy is used as a diagnostic as well as a therapeutic intervention for polyps. Stool tests have the potential to replace colonoscopy as diagnostic procedure and rather select those patients with relevant lesions for subsequent treatment with colonoscopy. Ideally, surveillance would not detect tumours when they have already become invasive, but just before, in a premalignant stage. Molecular alterations could be used to more precisely pinpoint premalignant lesions that are at high risk of progression. Below, all of these issues are considered in more detail.

Quality of baseline colonoscopy
The effectiveness of colonoscopy for the prevention of CRC depends, amongst others, on the quality of the performance. There is no evidence that an initial poor examination can be compensated by overuse of endoscopic surveillance. For this reason, surveillance guidelines only apply to patients that have had a high quality baseline colonoscopy.
Multiple studies have shown that missed or incompletely removed lesions contribute to the development of post-colonoscopy colorectal cancers (PCCRCs), defined as cancers diagnosed after a colonoscopy during which no cancer was found, but before the next due surveillance colonoscopy. Despite colonoscopy being the reference standard for polyp detection and removal, it misses an estimated 20% of polyps and 0.8% of cancers (in the setting of synchronous CRC), as demonstrated in chapter 9. Reasons for missed lesions include inadequate bowel preparation and endoscopist-related factors, such as completeness of colonoscopy, withdrawal time and adenoma detection rate (ADR). Therefore, these parameters are incorporated in colonoscopy quality standards. Because serrated polyps are notorious for being easily missed, also the proximal serrated polyp detection rate has been proposed as autonomous colonoscopy quality parameter. More research is needed to determine the association between endoscopists’ proximal serrated polyp detection rates and the risk of interval cancer. Another proposed performance indicator of colonoscopy is the Performance Indicator of Colonic Intubation (PICI). This measure combines cecal intubation rate, comfort and use of sedation during colonoscopy. Since less skilled endoscopists might more forcefully intubate the colon and cause more pain, which especially is remembered by those patients who are less sedated, the PICI may give a good reflection of the skills of the endoscopist. PICI could be used to identify and support low-performers and for benchmarking.
An estimated 10% of PCCRCs is caused by incompletely resected polyps. Especially large polyps resected in a piecemeal fashion are associated with inadequate polypectomy and relatively low rate of radical resection. For this reason, the Dutch and European surveillance guidelines advice endoscopic follow-up within 4-6 months after a piecemeal resection. Endoscopic submucosal dissection (ESD) enables en-bloc resection also in large polyps. Compared to piecemeal endoscopic mucosal resections (EMR), ESD however is more difficult, resulting in a longer procedure time and more complications. Future cost-effectiveness studies comparing piecemeal EMR with ESD should evaluate whether the reduced recurrence rate and higher number of radical resection after ESD outweigh these drawbacks.

Timing of colonoscopy
Timely colonoscopy should be offered to those patients with a substantial residual risk. So far, no randomised controlled trials have examined the effect of different surveillance intervals per risk group on long-term outcomes. It has been hypothesised that due to the improved colonoscopy quality over the last decade, current intervals recommended in the guidelines are too strict. A large-scale ongoing European trial, the EpoS-study, randomises patients from different risk categories between shorter and longer surveillance intervals. After a follow-up period of 10 years, the incidence of CRC in the various randomisation arms will be compared to identify the optimal interval between colonoscopies.

Methods of surveillance
To reduce the number of patients referred for surveillance colonoscopies, stool testing could be used as a diagnostic tool to select patients at high risk for advanced neoplasia for subsequent treatment with colonoscopy. Such a triage strategy is currently applied in the Dutch FIT screening program. In surveillance it is important that test sensitivity is high, while a lower specificity might be acceptable because all these patients currently get colonoscopy. The mt-sDNA test was previously shown to have a higher sensitivity in the screening setting than FIT. Therefore we hypothesised that the mt-sDNA test could be an appropriate triage test for surveillance.
Indeed, in chapter 7 we described that the mt-sDNA test detected more advanced neoplasia than FIT in a surveillance setting. Yet, the mt-sDNA test missed 3/10 CRCs even after lowering the test cut-off, compared to colonoscopy. For perspective, the negative predictive value (NPV) of the mt-sDNA test for CRC was 98.8% and the positive predictive value (PPV) 1.4%. A previous study conducted in the surveillance setting reported that after three rounds of annual FIT, using the low cut-off of 10 µg haemoglobin (Hb)/g faeces, 28% of CRCs were missed. In that study, the NPV of the FIT for CRC after three rounds was 99.8% and PPV 1.4%. While triage with stool tests could potentially reduce the number of colonoscopies, the clinical acceptability of such miss rates remains to be resolved.
Repeated testing could reduce the number of missed lesions within a certain period. This comes however with several caveats. First, missed cancers could be detected with later tests, but then the disease then may be more progressed. Second, stool test return could drop following a negative test and lead to false reassurance. Third, the program costs would rise. Because a larger number of patients would have a positive test and subsequent colonoscopy, the test costs need to be low to allow the strategy to be cost-effective. At the moment, costs of the mt-sDNA test are around 600 euro’s, while FIT is around 20 euro. Therefore it is unlikely that repeat mt-sDNA testing would be a cost-effective strategy. Fourth, the relieve on the colonoscopy capacity would be less pronounced. To quantify these considerations, modelling studies are required to assess the long term effect of stool-based surveillance on CRC mortality and colonoscopy burden.
Instead of using stool tests with one cut-off for all, also a more tailored approach could be considered. One option is to use clinicopathological risk factors, such as age, life-style factors or previous colonoscopy findings, in addition to the FIT value, as described in chapter 8. The diagnostic prediction model we created included FIT result, age, calcium intake, smoking habits, (tubulo)villous adenoma in previous colonoscopy and large lesions in previous colonoscopy and was able to improve advanced neoplasia detection compared to FIT only. The model should be externally validated in a cohort of surveillance patients, before implementing this strategy in the clinic. In addition, similar efforts could be made to improve mt-sDNA test performance by developing a new prediction model. A second approach is to use the quantitative FIT value to personalise the frequency of surveillance colonoscopy. It has been demonstrated that FIT values below the cut-off value are predictive for advanced neoplasia detection at follow up, in screening as well as surveillance cohorts. This suggests that colonoscopy intervals could be lengthened relative to FIT Hb concentrations. Further modelling studies are needed to assess the risk of advanced neoplasia based on Hb and optimal time intervals. Yet another approach comes from fruits of the genomic era. Based on our increasing knowledge of DNA variants associated with CRC risk and the ease with which these now can be determined, the concept of polygenic risk scores (PRS) has been developed. The potential of these PRS for personalising screening, as well as strategies require further clinical validation.
Successful implementation of stool testing in surveillance also depends on several practical matters. The test has to be user-friendly and easy to apply at large scale. Therefore, participation rates and number of analytical drop-outs should be incorporated in modelling studies in order to identify the optimal stool-based surveillance strategy.
In general, the probability of the disease not only depends on the result of the test, but also on the probability of the disease before the test was performed. For proper implementation of the stool testing in surveillance, it is essential to take the a priori chance of advanced neoplasia, and especially CRC, into account. In chapter 7 we found that patients with a previous diagnosis of CRC, had a higher risk of subsequent CRC (1.7%) compared to post-polypectomy (0.4%) and familial risk patients (zero). As outlined in chapter 9, this may be explained by CRCs being missed due to quicker withdrawal once a CRC is detected or maybe suboptimal bowel preparation due to a stenosing cancer. Also the background risk of the patients (high-risk genetic make-up/ life-style) may play a role. Because of the pronounced risk of the post-CRC surveillance patients, colonoscopy likely remains warranted for this population especially in the first few years after resection. For the post-polypectomy and familial risk population with a low a priori chance of CRC, stool testing seems more appropriate.
Lastly patient attitude towards stool tests replacing routine colonoscopy should be considered. Surveillance-experienced patients have previously expressed concerns about the sensitivity of FIT and did therefore not endorse the idea of FIT as alternative for colonoscopic surveillance. Whether patient preference would be different for the mt-sDNA test or is dependent on historical colonoscopy findings, needs to be evaluated.

Target for surveillance
Ideally, surveillance would target those lesions that are not yet malignant, but would have likely progressed to cancer when left in situ. Yet, leaving colorectal polyps in place an following them over time until they progress, is considered unethical. The standard endoscopic removal of polyps disrupts the natural behaviour of polyps and hinders the identification of the exact characteristics of these high-risk lesions. Most of the knowledge on the adenoma-to-carcinoma progression is based on cross-sectional data, comparing molecular profiles of premalignant with malignant tissue. From these studies it can be concluded that in adenomas, chromosomal instability (CIN) occurs at a late stage and this is a critical step in progression to cancer. The role of CIN in malignant transformation has also been confirmed in mouse and human intestinal organoid models. Instead of randomly, these chromosomal alterations arise in specific patterns. When leading to amplification of oncogenes or deletion of tumour suppression genes, this may confer a growth advantage, as has been shown for CDK8 on the 13q ampicon and AURKA on the 20q amplicon. Altogether, these findings suggest that SCNAs could be used to distinguish adenomas that are likely to progress from the ones that are not, i.e. the low-risk ones.
A previous study has attempted to more specifically define the regions of copy number gains and losses that could be used as adenoma progression biomarkers. In that study, the presence of two or more out of seven frequently occurring DNA copy number alterations (CAEs) distinguished adenomas with and without a malignant component with high accuracy (78% sensitivity and 78% specificity). In chapter 4 we found that these CAEs were present in adenomas that grew over time, as well as those that remained stable, but were absent in polyps that had regressed. This provides evidence for the functional role of these alterations in a clinical setting. Furthermore, when comparing to the traditionally used advanced adenoma criterion, a much smaller proportion of polyps was classified as molecular high-risk adenomas, suggesting an overestimation of the number of high-risk lesions with the current definition. Therefore, molecular high-risk adenomas could potentially be a useful measure in clinical practice. The implementation of the concept of molecular high-risk adenomas is envisioned in several ways.
First, instead of measuring sensitivity and specificity of novel diagnostic stool tests against advanced adenomas, molecular high-risk adenomas could be applied as a more precise intermediate endpoint for CRC. Clinical use of tests that are in particular sensitive for molecular high-risk adenomas, could then reduce the number of patients being referred for colonoscopy, while still effectively reducing CRC mortality. In such a scenario, the polyp tissue resected during colonoscopy would undergo copy number profiling to identify high-risk and low-risk adenomas. The FAST-SeqS method described in chapter 3 could provide an assay for realising copy number profiling of tissues in large cohorts of patients, as it is a simple and low-cost technique which can easily automated in a high throughput platform.
Second, molecular high-risk adenomas could be used to predict future CRC risk. A research collaboration between the Netherlands and Norway called IntEnd has recently been initiated for this purpose. In this study, DNA copy number profiling will be performed on a large retrospective series of advanced and non-advanced adenomas. Molecular high- and low-risk lesions will be related to the risk of metachronous CRC during 10 year follow-up. If indeed molecularly-defined intermediate endpoints appear to be more accurate in predicting future risk than the currently used concept of advanced adenomas, this could eventually lead to revision of surveillance guidelines.
Lastly, the changes that define molecular high-risk adenomas, i.e. the seven CAEs, could be used as diagnostic markers to design new diagnostic stool tests. For example, the mt-sDNA test is currently based on the detection of methylation and mutation markers, but could in the future be replaced or complemented by markers detecting CAEs. An important consideration in this respect is that the test should reliably detect very small quantities of marker analytes against large amounts of background DNA. Until recently, no reliable assay for the detection of SCNAs was available. According to the results presented in chapter 3, FAST-SeqS is able to detect copy number alterations in low purity samples. So, possibly, FAST-SeqS could also detect CAEs in stool samples. The analytical sensitivity however, will depend largely on the occurrence of identical LINE-1 sequences in bacterial DNA, the main constituent of stool DNA. LINE-1 sequences have been shown to be human specific and our first explorations show poor alignment of the FAST-SeqS primers to the genomes of bacterial phyla most prevalent in human faeces. Therefore, efforts investigating further CAE markers as basis for stool tests could be a meaningful next step.
As opposed to adenoma progression, the serrated pathway is not characterised by CIN, but by MSI. MSI might prove an appropriate progression marker, although it likely coincides with dysplasia in serrated polyps, from when it takes very little time to progress into cancer. This indicates that the windows of opportunity to use genomic instability markers, i.e. CIN or MSI, are likely to differ for the traditional adenomas and serrated polyps, respectively. At the moment, research is conducted in which the whole exome, as well as the whole methylome of progressed and non-progressed serrated polyps is studied. This could help to identify appropriate progression biomarkers in serrated polyps.

The complexity of surveillance becomes clear when considering all the different aspects that need to be optimised and aligned. In this thesis, the focus has been on finding ways to reduce the colonoscopy burden of surveillance, by applying stool testing as diagnostic medium. In addition, I have sought to better understand which premalignant lesions should be the target for surveillance to prevent CRC, while avoiding overdiagnosis. Hopefully, the results of this thesis can in this way contribute to the development of new, more efficient surveillance strategies.