Colorectal cancer (CRC) is associated with a highly penetrant dominant or recessive inherited syndrome in about 5 % of cases. The most common of these is Lynch syndrome (LS), otherwise known as hereditary non-polyposis colorectal cancer (HNPCC) [1]. This syndrome is an autosomal dominant disorder caused by a mutation in one of the mismatch repair (MMR) genes. Mutation carriers have a life-time risk for CRC development of approximately 25–75 %, with a greater risk in MLH1 and MSH2 and a lower risk in MSH6 and PMS2 mutation carriers. In addition, they have an increased risk of various other malignancies, especially endometrial cancer [2]. The most frequent cause of cancer-related death in LS families, in both men and women, is CRC. In a Dutch study of mortality in LS families, the three most frequent causes of cancer-related deaths in men were CRC (56.3 %), lung cancer (7.5 %) and brain tumours (7.5 %), and in women CRC (43.4 %), endometrial cancer (14.6 %) and brain tumours (5.9 %) [3]. The first evidence that colonoscopic surveillance is useful in families with a high risk of CRC was published almost 30 years ago [4]. Since then, a number of studies of colonoscopic surveillance in LS families have been published, briefly summarized here in chronological order.

Overview of surveillance studies in Lynch syndrome (HNPCC)

In 1987, the results of surveillance of 22 “Finnish cancer family syndrome” (CFS) kindreds were reported. More than half of the family members (58 %) invited to participate in an organized surveillance program accepted the invitation. The results of this study showed that colorectal surveillance led to the detection of CRC at an earlier stage, compared to the stage in subjects from the same families not attending screening. These results also established the importance of a centralized organisation of surveillance [5]. The Finnish data were confirmed in 1989 by the surveillance results of 22 Lynch families in the Netherlands. In this study, Dukes C and D tumours were diagnosed in more than a third of the subjects of Lynch families who were referred because they were symptomatic whereas in their family members who were found to have a colorectal neoplasm by screening only Dukes A and B tumours were diagnosed [6]. By this time, the centralized organisation of surveillance for these families had been implemented in both Finland and the Netherlands, and follow-up results were published in 1995. Colonoscopic surveillance at 3 year intervals in 22 Finnish Lynch families showed a 62 % reduction in incidence of CRC and no deaths due to CRC [7]. Follow-up results on 50 Lynch families in the Netherlands showed that surveillance led to the detection of advanced CRC (Dukes C and D) in 25 % of patients while under surveillance—compared to 47 % in a control group (relatives with symptomatic CRC)—resulting in an improved 5-year survival rate in the surveillance group [8]. As the studies described above all took place prior to the availability of predictive DNA testing, all persons participating in the surveillance programs at that time were at a 50 % risk of being mutation carriers. The majority of surviving family members then participating in the surveillance programs were probably non-carriers, as many of the true carriers had died of cancer before surveillance was organized in the families. In 1995, the Dutch HNPCC Group reported an unexpectedly high occurrence of cancers detected within 2–5 years of a negative surveillance examination. This resulted in the recommendation of surveillance examinations in the Netherlands at 1–2 year intervals from then on [9]. Only a few years later, a 65–70 % decrease in mortality due to surveillance was published by both the Finnish and Dutch registries [3, 10]. However, despite surveillance, the risk of CRC for proven mutation carriers was calculated by the Finnish registry in 2000 as being 13 % every 10 years (3 year interval) and a comparable risk of almost 11 % every 10 years (1–2 year interval) was published by the Dutch HNPCC group in 2002 [10, 11].

The results of colonoscopic surveillance in 117 Finnish Lynch families were published in 2007 [12]. The recommended surveillance interval had been once every 5 years in the 1980s, every 3 years between 1989 and 1994, and when predictive testing became available the interval was reduced to every 2 years for proven mutation carriers from the age of 35. Results on 420 mutation carriers (median follow-up 7 years), with no history of CRC prior to regular colonoscopic surveillance, were published. The risk of an incident adenoma increased steadily with age, starting from age 20. The estimated risk of an adenoma by the age of 60 was 69 % in men and 48 % in women, and nearly 50 % of the adenomas were located proximal to the splenic flexure. The risk of an incident carcinoma from the age of 20–60 years was 25 % in men and 22 % in women. Carcinomas were situated mainly in the proximal colon, and 80 % were tumours at Duke stages A and B. Four of the carcinomas diagnosed within the regular surveillance protocol were at Dukes stage C. The combined risk of an incident adenoma or carcinoma by 60 years of age was estimated to be 82 % in men and 63 % in women. Twenty-six individuals died during surveillance, five of them due to CRC. In four of the cases this was owing to patient-related delays in diagnosis.

In 2010, the results of colonoscopic surveillance in 205 LS families in the Netherlands were published [13]. Screening recommendations included colonoscopy starting from age 20–25 years, whereas the recommended interval was 2–3 years until 1995, when it was further reduced to 1–2 years. Follow-up results on 745 mutation carriers were presented, and mutation carriers with a previous subtotal colectomy, known to have an advanced CRC (Dukes C or D) or with a prevalent CRC were excluded. The median follow-up time was 7 years, and 33 patients developed CRC. The cumulative risk of an incident carcinoma at 10 years follow-up was 6 %, with 62 % of the carcinomas situated in the proximal colon. Thirteen of the 33 patients (39 %) showed intervals of >2 years between the previous colonoscopy and the detection of cancer. Tumours diagnosed at or within the regular (<2 years) surveillance protocol (61 %) were localized cancers (Dukes A and B) in 90 % of cases and Dukes stage C in 6 %. None of the patients died as a result of CRC during surveillance.

The year 2010 also saw the publication of surveillance results from the German Consortium for Hereditary Non-Polyposis Colorectal Cancer, initiated in 1999 [14]. Screening recommendations included annual colonoscopy from the age of 25, or 5 years prior to the earliest disease manifestation within a given family. Proven and obligate gene carriers were included, in addition to individuals with a tumour related to the HNPCC spectrum. First-degree relatives of these individuals were also eligible, but individuals from families in whom a known familial mutation could be excluded were not included in the analysis. The study comprised patients with and without a history of CRC before inclusion in the study. In individuals lacking a history of CRC prior to inclusion in the study, the cumulative CRC incidence was 23 % at the age of 60 years. Five percent of the CRCs detected by follow-up endoscopies were at stage III, and no metastatic CRCs were diagnosed during follow-up colonoscopies. In eight of nine patients with an incident symptomatic CRC, the time since their last colonoscopy exceeded the recommended 12-month interval by more than 3 months (median time 39 months).

The most recent surveillance results originate from Canada, with surveillance results on 18 Lynch syndrome families with an MSH2 mutation published in 2011 [15]. Screening recommendations included colonoscopy every 1–2 years from the age of 20–25 years. Proven, obligate and presumed (LS tumour <50 years of age) gene carriers were included. Median follow-up was 9 years in men and 11 years in women. Screening delayed the age of onset of CRC by more than 10 years for both male and female mutation carriers. Twenty-one out of 109 individuals developed interval CRC, and 13 (62 %) of these CRCs were detected within 2 years of a previous colonoscopy. The stage distribution of these 13 CRCs was 62 % in stages I and II, 23 % in stage III and unknown in 15 %. Only one patient died as a result of CRC during surveillance.

Role of Lynch syndrome registries

What can be learned of colonoscopic surveillance in LS families from the studies described above?

First, the enormous effort underlying registries in a range of countries worldwide should be emphasized, including the detection and recruitment of a large number of Lynch families over the past 25 years (Fig. 1). The most significant surveillance results in a given population can be achieved in countries with good surveillance coverage of LS families on a national level and of all mutation carriers within each diagnosed family. These issues in particular are the most important aims of the registries: they promote surveillance in LS families and they guarantee the continuity of surveillance. The exact methods and approaches of the different registries have been described elsewhere [6, 12, 14]. By the end of the 1980s, results on only 44 LS families were published. Today (2013), the total number of LS families with a known mutation registered in Finland, Germany, Canada and the Netherlands has increased to almost 2000 (confirmed by email correspondence with the different registries) and the number of tested mutation carriers is reaching 10,000, in these four countries alone. The distribution of the different known gene mutations in these families are: MLH1 (n = 780), MSH2 (n = 828), MSH6 (n = 271), PMS2 (n = 67) and EPCAM (n = 17). How many families remain to be identified? Assuming for instance that the frequency of LS is 1 in 1,000 in Finland, there should be about 5,000 mutation carriers in the country. To date, around 1,400 mutation carriers have been diagnosed in Finland and if these mutation carriers have about 2,000 growing children who will later take genetic tests as adults, 50 % of them will prove to be mutation carriers. Thus, 2,500 mutation carriers have already been identified, which may represent 50 % of all LS mutation carriers in Finland. Surveillance programs that offer regular endoscopic surveillance to these gene carriers have led to a significant decrease in CRC-related mortality in such families and the results of these programs have led to the adoption of various clinical practice guidelines that have been published [16].

Fig. 1
figure 1

Number of registered Lynch families published in the literature. The numbers registered in 2013 were obtained by email correspondence with the different registries

Full size image

An optimal surveillance interval?

A second challenge is that the most optimal surveillance interval for gene carriers is still unknown. This is important for the efficacy of surveillance on a personal level. In the surveillance studies described above, different surveillance intervals were recommended by the various registries and the recommended time intervals changed over the years, making the comparison of the results of these studies difficult. The risk of an incident carcinoma in Finland (2007) between the ages of 20 and 60 years was 25 % in men and 22 % in women (surveillance interval 2–3 years), and a mean age of 44 years (range 29–75) at occurrence of an interval CRC was observed. In the Netherlands (2010), a calculated cumulative risk of 6 % (surveillance interval 1–2 years) for an incident carcinoma at a 10-year follow-up and a mean age at diagnosis of 51 years (range 34–71 years) were observed. In Germany (2010), the cumulative CRC incidence was 23 % at the age of 60 years (annual surveillance interval) and the median age at detection of CRC was 50 years [12–14]. With respect to the number of advanced CRCs diagnosed while enrolled in a program and death due to CRC, the following is a summary of these results: Finland (2007)—four of the carcinomas diagnosed within the regular surveillance protocol were Dukes C stage and five patients died from CRC during surveillance, four of them due to a delayed diagnosis owing to patient-related factors. In the Netherlands (2010)—6 % of the tumours diagnosed at or within the regular (<2 years) surveillance protocol were at Dukes stage C and none of the patients died of CRC. In Germany (2010)—no significant differences in tumour stage distribution and the time since the last colonoscopy were detected, and no metastatic CRCs were diagnosed at follow-up colonoscopies. In eight of nine patients with an incident symptomatic CRC, the time since their last colonoscopy exceeded the recommended 12-month interval by more than 3 months (median time 39 months). Järvinen reported that surveillance led to a 62 % reduction in CRC incidence, most likely based on the detection and removal of adenomas during 10- and 15-year controlled trials [6, 9]. The above described surveillance results from the last decade suggest that further reductions in intervals between examinations will not further reduce the cumulative risk of CRC, but might lead to the detection of CRC at an earlier stage and increase the average age at detection of CRC [12–14]. Most authors now recommend surveillance with a 1–2 year interval for proven mutation carriers.

Causes of the high frequency of interval cancers?

Why is the risk of CRC still substantial despite enrolment in a surveillance program? Poor quality surveillance endoscopies (such as incomplete endoscopies or endoscopies with poor bowel preparation) are a known cause of a substantial part of the published interval cancers [11]. Compliance also plays an important role. In the Finnish series, 4 of the 5 patients who died of CRC had compliance issues [12]. In the German series, with annual surveillance interval, the median time interval between two successive colonoscopies was 12 months. As only 81 % of colonoscopies were performed within 15 months, this means that 19 % of the subjects were not compliant to the recommended interval [14]. In patients with no history of CRC and who belonged to families in whom no MMR mutation could be identified, the adherence was somewhat worse but there were no deaths due to CRC [14]. In the Canadian series, subjects were considered compliant if all screening colonoscopies were performed within 2 years of the previous examination; 44 % of the males and 41 % of the females were compliant [15]. Although one patient died of metastatic CRC, the previous endoscopy was within the recommended interval. However, the majority of interval cancers detected within recommended surveillance intervals are at a local stage and therefore compliance probably mainly reduces the risk of death due to CRC. Another possible explanation of high frequency interval cancers could be the so-called “field effect in cancer”; molecular abnormalities in tissues that appear histologically normal. This would mean that we are simply not able to identify (and remove) the colonic tissue at risk for cancer development [17].

Clinical implications

How can we further improve the outcome of colonic surveillance in LS? A high yield from surveillance programs requires the utmost effort of both patient and doctor. First of all, patient and doctor delay should be minimized. This is achieved by education of patients on cancer risk and the importance of surveillance examinations. Furthermore, reminders sent by the registry to specialists play an important role, and patients should therefore be encouraged to enrol in registers themselves. For those that decide not to register, hospital-based call systems can offer a solution. Secondly, cleansing of the colon should be optimal if optimal adenoma detection is to be achieved. Split dose polyethylene glycol is the recommended bowel preparation and clear patient information leaflets are required [18]. The quality of bowel preparation should be documented using standardized scoring systems such as the Boston Bowel Preparation Scale (BBPS), as higher BBPS scores are associated with a higher polyp detection rate [19]. In cases with a poor quality of documented preparation, repeating the endoscopy at a shorter interval should be considered. Thirdly, there should be no doubt as to the completion of colonoscopy. As colorectal cancer develops in the most proximal part of the colon in at least two-thirds of CRC cases in Lynch syndrome [11, 20], photographic evidence showing two of the three landmarks (ileum, ileocecal valve and appendicle orifice) should be documented. In cases of incomplete colonoscopy, the examination should be repeated as soon as possible following specific preparations (i.e. general anaesthesia, double balloon enteroscopy, more experienced endoscopist). Furthermore, as it is well known that particularly flat adenomas are easily missed with conventional colonoscopy, better visualization of the colonic mucosa leads to an increased adenoma detection rate. It is also relevant to note that in two studies chromoendoscopy detects significantly more adenomas than standard white light colonoscopy or Narrow Band Imaging (NBI) [21, 22]. In another study chromoendoscopy did not detect more missed adenomas than intensive inspection (lasting >20 min) [23]. Clearly, withdrawal time should be sufficient and in cases where adenomas are detected, they should be removed radically and without leaving residual adenomatous tissue. In the near future, improvements in the quality of colonoscopy and visualization techniques will hopefully further reduce the CRC risk of those enrolled in a surveillance program.