The Prospective Lynch Syndrome Database (PLSD)

In 2012 the European Hereditary Tumor Group (www.ehtg.org), at that time denoted the Mallorca Group, [1] decided to compile information on follow-up of path_MMR carriers across multiple specialist centres to answer three questions:

1.

To what degree does colonoscopy surveillance reduce CRC incidence in path_MMR carriers?

2.

What is the penetrance and expressivity of pathogenic variants in each of the four LS-associated genes?

3.

What is the survival of carriers when followed-up as recommended, to facilitate early diagnosis and treatment?

The initial results were published by Møller et al. [11,12,13]. A more extensive and detailed report confirming the results in the three first reports was recently published [14]. Below we suggest an interpretation of the findings from these and further studies that were triggered by the initial results, and of concomitant tumor biological, prevention and treatment studies and mathematic modelling of the carcinogenetic paradigms that may help to explain what we observe.

The PLSD methods

The PLSD database, it’s structure and the methods for producing the results have been described in detail elsewhere [15,16,17,18]. A randomized controlled trial including a control group of path_MMR carriers who would be denied recommended medical interventions was considered impossible. Therefore, we performed an open, prospective observational study. Independently, a complementary retrospective segregation analysis in LS families was performed by the International Mismatch Repair Consortium (IMRC) which enabled estimation of the CRC incidence before surveillance colonoscopy had been widely implemented [19]. The results confirmed that cancer incidence in path_MMR carriers was not increased significantly before 25 years of age.

Our intentions were to examine the accepted paradigms of carcinogenesis and the effects of interventions. In all studies, the results obtained reflect the parameters used for ascertainment and/or the assumptions made when considering the results. To avoid these biases, when designing the PLSD neither the ascertainment model nor the methods used to compile the results included any assumption on carcinogenesis or the effects of interventions. Instead, the methods involved an assumption-free description of the empirical information observed. Compliant with the reporting methods of cancer registries, cancers were scored as discrete events by organ and age, allowing events to be considered as the result of stochastic probabilities in a time dimension.

The results obtained from the PLSD data were therefore empirically observed cancer incidences and subsequent overall survival in path_MMR carriers who were subjected to follow-up including colonoscopy surveillance in expert hereditary cancer centres world-wide. There is no indication that methodological problems could have substantially confounded the main results. Here we provide a brief summary and an interpretation of the results published to date by the PLSD.

The PLSD results Colonoscopy and incidence of CRC

Compared with published estimates of CRC incidence in former generations before colonoscopy was widely instituted, CRC incidence in LS patients subjected to regular colonoscopy surveillance was increased for path_MLH1 and path_MSH2 carriers, not reduced for path_MSH6 carriers and possibly (but not significantly) reduced in path_PMS2 carriers < 50 years of age [20].

Penetrance and expressivity

Cumulative cancer incidences stratified by MMR gene, sex, and carriers' age (50 and 75 y.o.a.) when subjected to follow-up, including colonoscopy to achieve early cancer diagnosis, are reported [14] and illustrated in Figs. 1 and 2. Cancers of the endometrium, colon and ovary started to appear in early adult life. Cancers of other organs were diagnosed later and mainly in survivors of earlier cancers [21]. Penetrance and expressivities were specific to each gene.

Fig. 1

Cumulative incidences of cancers in male and female carriers subjected to colonoscopy stratified by gene, and sex and age (50 and 75 y.o.a.), ordered by incidence in path_MLH1 carriers. The graphs are based on figures given in [14]

Fig. 2

Median ages of onset of cancers in male and female carriers subjected to colonoscopy, by gene and sex, ordered by median ages in path_MMR carriers. The graphs are based on figures given in [14]

Survival after cancers detected during follow-up

Most early onset cancers in the endometrium, colon and ovaries detected during follow up were cured [14] (Fig. 3). However, later in life, the survivors often developed cancers in other organs most of which were associated with lower overall survival [14] (Fig. 4).

Fig. 3

10-year survival following cancer in different organs in path_MMR carriers subjected to early diagnosis and treatment including colonoscopy. The graph is based on figures given in [14]. While there was no difference in survival between carriers of path_MMR variants by gene, cancer in path_MSH6 and path_PMS2 carriers were not frequent enough to measure survival apart from after endometrial cancer in path_MSH6 carriers

Fig. 4

Cumulative incidence of death at 75 years of age following cancer in male and female carriers subjected to colonoscopy, by gene. The graphs are based on figures given in [14]

These results indicated that both penetrance and expressivities of the genes in question have been affected by a substantial time-trend as most cases of CRC are now cured, which was not the case historically. In patients who received colonoscopy surveillance and relatively recent treatments (which in most individuals did not include immunotherapy), the results suggest that there are four different inherited MSI cancer syndromes:

The four Lynch syndromes The MSH2 Syndrome

The MSH2 Syndrome is autosomal dominantly inherited. Penetrance is high.

Path_MSH2 carriers are at high risk of cancers in all organs that are affected across the Lynch syndromes, with early onset of cancer in endometrium/ovaries and colon. Most carriers survive these first cancers following early detection and treatment. Cancers in other organs are most often diagnosed in survivors of the early onset cancers and include rectal, upper urinary tract, prostate and brain cancers.

Few founder variants – fitness is low

Most cancer deaths are associated with non-CRC cancers, particularly those of the endometrium, rectum, upper urinary tract, prostate and brain

Colonoscopy overdiagnoses colon cancer.

The MLH1 syndrome

The MLH1 Syndrome is autosomal dominantly inherited. Penetrance is high.

Early onset and high incidence of cancer in the colon, endometrium and ovaries. Most carriers survive their first cancer following early detection and treatment. Cancers in other organs are most often diagnosed in survivors of the early onset cancers, and most often include rectal, stomach, small intestine, bile duct and pancreatic cancers.

Few founder variants – fitness is low

Most cancer deaths are associated with non-CRC cancers, particularly those of the endometrium, bile duct and pancreas.

Colonoscopy overdiagnoses colon cancer.

The MSH6 syndrome

The MSH6 Syndrome is autosomal dominantly inherited with sex limitation. Penetrance is high in females but low in males.

There is a high incidence of endometrial and ovarian cancer that occurs at older ages than in path_MSH2/MLH1 carriers. There is an increased incidence of CRC in both sexes that is nonetheless much lower than in path_MSH2/MLH1 carriers. Incidences of cancers in other organs are low.

Few founder variants – fitness is low.

Detection rate by family history is low because of the sex-limited inheritance.

The inconclusive effect of colonoscopy is due to the low number of carriers reported to the PLSD.

The PMS2 syndrome

The PMS2 Syndrome is autosomal dominantly inherited.

Path_PMS2 carriers have a slightly increased incidence of CRC and endometrial cancer in young adults, with higher incidence in older ages. Increased cancer incidence in other organs is not demonstrated.

Estimates of cancer incidences are difficult because of ascertainment biases.

Founder variants present, fitness seems good.

In contrast to the three other syndromes, the low incidence of CRC in young adults receiving colonoscopy surveillance may suggest that colonoscopy reduces CRC incidence.

Because of the very much lower cancer risks in the PMS2 syndrome and potentially different carcinogenetic mechanisms associated with CRC development, the discussion below may not be pertinent to path_PMS2 carriers.

The PLSD and the InSiGHT variant database

When using the international InSIGHT database that indicates whether variants of the MMR genes are disease-associated or not [2], the PLSD database that indicates the penetrance and expressivity of pathogenic or likely pathogenic variants is displayed by selecting the MMR CANCER RISK tab. The PLSD is also available directly at www.plsd.eu. It interactively displays the remaining life-time risk for cancer in each organ when the user indicates the carrier’s age, sex and genetic variant. This feature of PLSD may be helpful to carriers and health care workers.

Inherited congenital mismatch repair deficiency (CMMRD)

Inherited, biallelic path_PMS2, path_MLH1, path_MSH2 and path_MSH6 variants are recognized to cause the recessively inherited congenital mismatch repair deficiency (CMMRD) syndrome [22]. CMMRD is characterized by a high incidence of MSI non-endodermal malignancies in early life. Path_PMS2 variants are the most frequent cause of CMMRD. The relative rarity of path_MLH1 and path_MSH2 causing CMMRD has caused speculation that most homozygotes or compound heterozygotes may not survive fetal development. A more detailed discussion of CMMRD is outside the scope of this paper.

Relations in time between cancers and consequences of early diagnosis and treatment

Most of the frequent and early onset cancers in the four Lynch syndromes are cured following early diagnosis and treatment, which may be achieved in many cases through colonoscopy and gynaecologic examinations [13] and by promotion of cancer awareness with early consultation in relation to “red flag” symptoms. Most cancer-associated deaths in carriers who are subject to follow-up are now associated with non-CRC cancers [14].

Follow up studies suggest that the occurrence of a non-CRC first cancer was not associated with CRC incidence [12] and neither stage at CRC diagnosis, nor survival following CRC was associated with the interval between colonoscopies [23,24,25,26]. These findings were not as expected from the adenoma-carcinoma sequence paradigm of CRC. Instead, they were in keeping with the notion that the observed cancers are the result of stochastic probabilities in time, in conflict with the notion of linear progression from acquisition of an initial somatic pathogenic variant that causes increased mitotic activity, to a dysplastic adenoma and eventually an invasive and metastasizing cancer [27, 28]. If the stochastic model is correct for most LS cancers, the average incidences we have published are valid for groups but may have limited predictive value for an individual. No person has an “average sex “or a pathogenic variant in an “average Lynch syndrome gene” and results that are not stratified by gene and sex will be valid for no one. Modifiers of penetrance may be important in determining which cancers occur and when. More detailed analyses investigating whether there are associations between cancers, or whether they do indeed reflect stochastic chances, would be interesting.

Most cancers diagnosed before 50 years of age in male path_MLH1 carriers are colon cancers, and cancers occurring in other organs reflect the time trend of increased colon cancer survival. Most cancers in young male path_MSH2 carriers are also colon cancers, but rectal, urinary tract, prostate and brain cancers are more frequent later in life and with increased colon cancer survival they have become the major causes of death.

In females, however, endometrial and ovarian cancers together are the first and most frequent cancers in path_MLH1 carriers and even more so in path_MSH2 carriers. Colon cancer has a slightly lower incidence and older onset in path_MSH2 than path_MLH1 carriers. Considering LS as a syndrome of inherited CRC or gastro-intestinal cancers ignores gynaecological cancers as the main manifestation in women, and that gynaecological cancers together with urothelial, prostate and brain cancers are the leading causes of death in path_MMR carriers who receive follow-up with colonoscopy for early diagnosis and treatment.

Ascertainment biases when estimating variant frequencies, penetrance and expressivities

Identification of LS families has been and still is biased. The first clinical criteria to identify affected families were constructed for research purposes to identify the causative gene(s) and reflected the misconception that we were looking for inherited CRC, not a syndrome of inherited cancer in many organs, and they primarily identified path_MLH1 families [29]. The understanding that endometrial cancer was part of the syndrome was reflected in later criteria [30] and led to more path_MSH2 and especially path_MSH6 families being recognized. Because of sex-limited inheritance many path_MSH6 families did not fulfil these clinical criteria which assumed high cancer incidences in both sexes [31]. Path_PMS2 penetrance is so low that discrimination from normal variation by using family history is nearly impossible. A similar pattern of biases occurs when genetic testing is undertaken based on family history and when incident testing is focused on young onset CRC [32]. Only after all incidental cancers are tested for the four genes in question (in all ages and for all the genes in question) and the combined results are assembled, we will be able to fully determine the incidence and expressivities related to the genetic variants of the genes.

Contrary to the assumptions underlying the clinical Amsterdam 1 (AMS1) [29] and Amsterdam 2 (AMS2) [30] criteria, our studies in the PLSD have shown that ovarian cancer can be grouped together with endometrial and colon cancer as the early onset cancers in carriers who are subject to colonoscopy, while rectal cancer can be grouped with the other less frequent cancers that are seen mainly in survivors of the early onset cancers. Therefore, grouping colon and rectal cancer epidemiologically as one organ, as was done by the AMS1/2 criteria might now be considered a mistake. A possible interpretation of our findings is that colonoscopy may prevent rectal but not colon cancer in LS. The PLSD data also show that not including ovarian cancer in the AMS2 criteria was a mistake. And especially so when its treatment often includes hysterectomy which will prevent endometrial cancer. Including cancer in other organs will have little effect on the sensitivity of the criteria in the identification of LS families, because most of these cancers appear in survivors of the early onset cancers [21, 33].

Despite the consensus that colonoscopy reduces CRC incidence in the population as a whole, there is no recognized evidence that this statement is correct in path_MMR carriers, as there is only limited historical evidence based on three publications, all reporting observations made on a cohort of only 22 Finnish families [34,35,36,37].

The Finnish families were selected on the basis of one case of very early onset CRC with multiple CRC-affected relatives. Some of the additional methodological problems that were not discussed in those reports are as follows: the index clusters were not removed when calculating CRC incidences; it was incorrectly assumed that the family members had 50% carrier probability after the CRC-affected cases were excluded, and lead-time bias in the non-intervention group was not discussed. Therefore, the validity of the conclusions included in those reports is arguable. A later segregation analysis in 70 Finnish families of which 65 had a path_MLH1 variant [38] reported higher CRC incidence than in French families [39] with path_MLH1 variants, and much higher incidence than in a multi-national report on European families with path_MLH1 variants [20]. Additionally, these Finnish papers described findings in carriers of the local Finnish path_MLH1 founder variant which may not be representative of all path_MLH1 variants and may not be representative for carriers of pathogenic variants of the other genes. Findings in a single series should be confirmed in another before they are used to support clinical decision-making. Despite the considerable time since the Finnish reports, there is no other reported evidence that colonoscopy reduces CRC incidence in LS [34].

Ovarian MSI cancer

Ovarian cancer in LS carriers is dramatically different from ovarian cancer in path_BRCA1/2 carriers [40]. The observed risk of dying from gynaecologic cancer diagnosed before 40 years of age for carriers of path_MMR variants was 0%, leading us to conclude that prophylactic hysterectomy and/or oophorectomy before 40 years of age solely for cancer prevention reasons is unwarranted and unethical. Similarly, the observed risk of dying from ovarian cancer in path_MSH6 or path_PMS2 carriers diagnosed before 50 years of age was 0%, and in these carriers, prophylactic oophorectomy before 50 years of age solely for cancer prevention reasons is considered unwarranted and unethical [41]. Etiologic diagnosis of ovarian cancer cases is crucial to select proper treatment for the affected individual, for planning their follow-up for subsequent cancers, and for cascade testing of their relatives when hereditary cancer is demonstrated.