Simplified and more sensitive criteria for identifying individuals with pathogenic CDH1 variants

Abstract

Background

Hereditary diffuse gastric cancer (HDGC) is an autosomal-dominant syndrome most often caused by pathogenic variants in CDH1. The International Gastric Cancer Linkage Consortium (IGCLC) recently updated its criteria for genetic testing. The purpose of this study was to estimate the sensitivity of IGCLC’s 2020 criteria for identifying carriers of CDH1 pathogenic variants and to formulate a new set of criteria that is simpler and more sensitive.

Methods

Medical histories of 112 CDH1 mutation carriers, identified predominantly by multigene panel testing, and their 649 family members were reviewed. The percentage of subjects fulfilling the IGCLC 2015 and 2020 criteria was calculated, once without making any assumptions about unavailable pathology, and once assuming gastric cancer to be diffuse when pathology was unavailable. For comparison, we calculated the percentage of subjects who fulfilled our proposed criteria.

Results

When making no assumptions about missing pathology, a small (19%) and equal percentage of CDH1 mutation carriers fulfilled the IGCLC 2015 and 2020 criteria. When assuming unspecified gastric cancer to be diffuse, 45 out of 112 (40%) subjects met the 2015 criteria and 53 out of 112 (47%) met the 2020 criteria. Eighty-seven per cent (97/112) fulfilled our proposed criteria.

Conclusion

In consecutive cases, mostly unselected for clinical criteria of HDGC, the IGCLC 2020 criteria are, at best, marginally more sensitive than previous iterations, but they are also more cumbersome. Unavailable cancer pathology reports are a real-world obstacle to their proper application. Our proposed Yale criteria both address this issue and offer significantly greater sensitivity than the IGCLC 2020 criteria.

INTRODUCTION

Gastric cancer is the fifth leading cause of cancer and the third most common cause of cancer deaths worldwide, with an estimated 819 000 deaths annually.¹ While 90% of gastric cancers appear to be sporadic, 10% exhibit familial clustering, and 1%–3% can be linked to specific hereditary cancer syndromes.² Hereditary diffuse gastric cancer (HDGC) is an autosomal-dominant syndrome characterised by early-onset diffuse gastric cancer (DGC) as well as lobular breast cancer (LBC). Pathogenic germline variants in the CDH1 gene, which codes for the important cell–cell adhesion protein E-cadherin, have been identified as a major cause of this syndrome.^{3 4} DGC has a poor prognosis, largely because patients often present with vague symptoms, leading to delays in diagnosis.

To identify families carrying CDH1 and other pathogenic variants, the International Gastric Cancer Linkage Consortium (IGCLC) published a set of clinical criteria for genetic testing in 1999 and then updated it in 2010, 2015 and 2020.^{3 5–7} The major differences between the latest version and its immediate predecessor are expansion of the total number of criteria from six to nine and relaxation of previous age restrictions. In a prior study of CDH1 pathogenic variant probands mostly not selected for clinical criteria of HDGC, we noticed that a large percentage of their families did not fulfil IGCLC clinical criteria for testing and thus they would have been missed.⁸ Despite their recent modification, we hypothesised that the IGCLC criteria remain overly restrictive, leaving many CDH1 pathogenic variant carriers undiagnosed. To test this hypothesis, we used a large cohort of consecutive CDH1 mutation carriers identified mostly by multigene panel testing (MGPT), independent of HDGC clinical criteria, to calculate the sensitivity of the IGCLC 2020 criteria for detecting CDH1 pathogenic variants. For comparison, we used the same cohort to calculate the sensitivities of the IGCLC 2015 criteria, as well as our own proposed criteria for identifying CDH1 mutation carriers.

METHODS

Ethics approval

This study was approved by the Yale University Institutional Review Board. Use of deidentified data from the Ambry Genetics cohort was deemed exempt from review by Western Institutional Review Board.

Study population

Probands with CDH1 pathogenic and likely pathogenic variants were identified from Ambry Genetics and the Smilow Cancer Genetics and Prevention Program at the Yale Cancer Center. The cohorts included 113 carriers of pathogenic or likely pathogenic CDH1 variants; one proband was excluded from analysis because her family history was not available. Ambry’s proband cohort included all CDH1 pathogenic and likely pathogenic variant carriers identified through MGPT from March 2012 through December 2016. The Yale cohort included all pathogenic or likely pathogenic variant carriers identified in the programme from June 2008 to July 2018, either through MGPT or CDH1 testing. Pathogenic variants were identified per Ambry’s classification scheme, which is based on guidelines published by the American College of Medical Genetics and Genomics and by the ClinGen CDH1 expert panel.^9–11 This series was previously described elsewhere.⁸

Sensitivity of established genetic testing criteria

Each proband’s personal and family history was examined and compared with the HDGC genetic testing criteria laid out by the IGCLC in 2015 and 2020. Histories were also compared with our own proposed criteria, which includes the National Comprehensive Cancer Network (NCCN) testing criteria for high-penetrance breast and/or ovarian cancer susceptibility genes (see below).¹² With very few exceptions, information about family members included relation to the proband (eg, mother, brother, etc), degree of relationship (eg, first, second, etc), side of family (ie, maternal or paternal), and all cancer diagnoses and the ages (exact or approximate) at which they were made. Laterality of breast cancers, relationships between family members and other clinical information (hormone receptor status, surgeries, etc) were available in some cases. When age was given as a decade, the middle of the decade was used for application of clinical criteria and calculations (eg, ‘45’ in place of ‘40s’). If a cancer was said to have been diagnosed before a certain age, the year immediately prior to that was used (eg, ‘74’ in place of ‘<75’). As information about cancer pathology of family members was often unavailable, and the IGCLC 2015 and 2020 criteria rely heavily on that information, we calculated the sensitivity of those criteria under two conditions: (1) no assumptions about unavailable pathology and (2) assuming all unspecified gastric cancers were DGC. For application of the IGCLC 2015 criteria, we considered a proband to meet criteria if he/she met either the ‘established’ criteria or the criteria for whom ‘testing could be considered.’³ Similarly, in applying the NCCN hereditary breast and ovarian cancer (HBOC) criteria, we considered a proband to meet criteria in cases in which testing was ‘clinically indicated,’ as well as in cases in which testing ‘may be considered’ (online supplemental table 1).¹²

Proposed Yale genetic testing criteria

In addition to including CDH1 in MGPT for all cases in which NCCN HBOC criteria are met,¹² we propose to complete the set of criteria for testing for CDH1 mutations in cases of ‘significant DGC history,’ which we define as (1) patients with DGC at any age and (2) patients with family history of two or more cases of gastric cancer in first-degree or second-degree relatives, when at least one is confirmed to be DGC or diagnosed at age 50 or younger.

Statistical analysis

McNemar’s test, a non-parametric test for paired nominal data, was used to determine whether the difference in sensitivity between the IGCLC 2020 criteria and our own was statistically different. This was applied using Minitab V.19 (Minitab, LLC, State College, Pennsylvania, USA).

RESULTS

Proband characteristics

Characteristics of the probands included in the study are summarised in table 1. The majority of probands were female (83%), non-Hispanic white (68%) and had a personal history of cancer (78%). Nearly half of the probands had a history of breast cancer (46%) and approximately a quarter had gastric cancer (27%).

Table 1.

Characteristics of probands

Characteristics	Total (N=112)
Sex (n (%))
Female	93 (83.0)
Male	19 (17.0)
Race or ethnicity (n (%))
Non-Hispanic white	76 (67.9)
Black	11 (9.8)
Hispanic	11 (9.8)
Asian	2 (1.8)
Ashkenazi Jewish	5 (4.5)
Other/unknown	7 (6.2)
Personal cancer history (n (%))
Yes	87 (77.7)
No	25 (22.3)
Cancer type (n (%))
Stomach	30 (26.8)
Diffuse/signet ring cells	19
Adenocarcinoma	2
Not specified	9
Breast	53 (47.3)
Lobular	25
Ductal	18
Mixed (lobular and ductal)	3
Not specified	7
Other
Colon/rectum	4 (3.6)
Thyroid	2 (1.8)
Ovaries	1 (0.9)
Thyroid	1 (0.9)
Pancreas	1 (0.9)
Cervix	1 (0.9)
Age at first cancer diagnosis (mean (SD))	46.3 (14.1)

Note that a few probands had more than one cancer, so the total number of cancer types is greater than the number of probands with a personal cancer history. Also, two gastric cancers were identified as ‘adenocarcinoma’ without clarification as to whether they were intestinal or diffuse type.

Characteristics of family members

Medical histories from 649 family members were available for our analysis. Probands reported histories from a median of five family members (IQR 3–7). A large majority of family members (86%) carried at least one cancer diagnosis. Only one proband (1%) reported no family history of cancer. The most common cancers were breast (30%), gastric (19%), colorectal (8%), lung (4%), prostate (4%) and pancreatic (3%). Pathology data were available for a small minority of breast cancers (8%) and gastric cancers (9%). Characteristics of family members are summarised in online supplemental table 2.

IGCLC 2015 and 2020 criteria

When making no assumptions about unavailable pathology, a small (19%) and equal percentage of CDH1 mutation carriers met the IGCLC 2015 and 2020 genetic testing criteria (tables 2 and 3). When assuming unspecified gastric cancer in a relative to be diffuse, 45 out of 112 (40%) subjects met the 2015 criteria and 53 out of 112 (47%) met the 2020 criteria. Seven out of eight (88%) of the additional subjects meeting the updated criteria did so because of the increase in the age cut-off for an individual with DGC from 40 to 50. Specific criteria with the highest sensitivity for CDH1 pathogenic or likely pathogenic variants were (1) personal history of DGC before age 40 or 50 and (2) two or more cases of gastric cancer, at least one of which confirmed to be DGC.

Table 2.

Sensitivity of the IGCLC 2015 criteria

IGCLC 2015 criteria	No assumptions about missing pathology (n (%))	Assuming all unspecified GC is DGC (n (%))
Family with ≥2 cases GC, at least one confirmed DGC	11 (9.8)	31 (27.7)
DGC at age <40	11 (9.8)	30 (26.8)
Personal or family history of DGC and LBC, at least one at age <50	2 (1.8)	10 (8.9)
Bilateral LBC or 2 family members with LBC at age <50	2 (1.8)	2 (1.8)
DGC with personal or family history of cleft lip/palate	0	0
Gastric in situ or pagetoid spread of signet ring cells	0	0
Any of above criteria met	21 (18.8)	45 (40.2)

DGC, diffuse gastric cancer; GC, gastric cancer; IGCLC, International Gastric Cancer Linkage Consortium; LBC, lobular breast cancer.

Table 3.

Sensitivity of the IGCLC 2020 criteria

IGCLC 2020 criteria	No assumptions about missing pathology (n (%))	Assuming all unspecified GC is DGC (n (%))
Family with ≥2 cases GC, at least one confirmed DGC	11 (9.8)	31 (27.7)
Family with ≥1 case DGC at any age and ≥1 case LBC at age <70	1 (0.9)	11 (9.8)
Family with ≥2 cases LBC at age <50	1 (0.9)	1 (0.9)
DGC at age <50	13 (11.6)	44 (39.3)
DGC with Māori ethnicity	0	0
DGC with personal or family history of cleft lip/palate	0	0
DGC and LBC in same individual, both at age <70	1 (0.9)	1 (0.9)
Bilateral LBC at age <70	1 (0.9)	1 (0.9)
Gastric in situ or pagetoid spread of signet ring cells at age <50	0	0
Any of above criteria met	21 (18.8)	53 (47.3)

DGC, diffuse gastric cancer; GC, gastric cancer; IGCLC, International Gastric Cancer Linkage Consortium; LBC, lobular breast cancer.

There were no reports of Māori ethnicity, cleft lip/palate, or gastric in situ or pagetoid spread of signet ring cells in our probands or their family members, so those criteria provided no additional sensitivity.

Yale criteria

When we analysed fulfilment of the NCCN HBOC genetic testing criteria, the sensitivity for detecting CDH1 pathogenic or likely pathogenic variants was 71% if MGPT includes this gene. However, since these criteria do not include history of gastric cancer, several probands with dramatic gastric cancer history did not fulfil the criteria. Our proposed criteria, which are met when NCCN HBOC criteria are met or when there is a ‘significant DGC history,’ had a sensitivity of 87% (table 4). These criteria were fulfilled in all 21 cases of CDH1 pathogenic or likely pathogenic variants that fulfilled the IGCLC 2020 criteria, as well as an additional 76 cases; there were 15 cases that fulfilled neither set of criteria (online supplemental table 3).

Table 4.

Comparison of genetic testing criteria

	Sensitivity	P value
IGCLC 2020 criteria	18.8%	<0.0001
Yale criteria	86.6%
NCCN HBOC criteria	70.5%
‘Significant DGC history’	31.3%

‘Significant DGC history’ is defined as a proband with DGC at any age or two or more cases of gastric cancer in first-degree or second-degree relatives, with at least one confirmed to be DGC or diagnosed at age 50 or younger. Yale criteria are met when either NCCN HBOC criteria are met or when there is ‘significant DGC history.’ IGCLC, International Gastric Linkage Consortium; NCCN, National Comprehensive Cancer Network; DGC, diffuse gastric cancer; HBOC, hereditary breast and ovarian cancer.

DISCUSSION

The identification of CDH1 mutation carriers is extremely important given the devastating effects of the cancers this defect predisposes to in many families and the particularly poor prognosis of DGC. Thus, it is of foremost importance to be able to determine who is at risk for this syndrome so testing can be performed. While the yield of CDH1 germline pathogenic variants is relatively high in families with multiple and young onset DGC and LBC, we and others have shown that the majority of individuals with these variants do not have the type of enriched family history that makes HDGC highly suspicious.^{8 13} This seems to be the limitation of the testing criteria endorsed by the IGCLC guidelines, including the recently published version.⁷ The criteria are so strict that if we were to test only individuals fulfilling that set of criteria, we would be missing most of the CDH1 mutation carriers. In fact, in our cohort, the sensitivity of the criteria was only 19% for both the 2015 and 2020 criteria, as the newer version did not afford better sensitivity in this cohort. Older studies, such as the ones by Benusiglio et al and van der Post et al, reported significantly higher sensitivity for the IGCLC 2010 criteria for detecting CDH1 mutations: 61% and 89%, respectively.^{14 15} However, the former study from 2013 only included 18 individuals with a pathogenic CDH1 variant diagnosed among 165 individuals tested. Many of the ones that tested negative (57 out of 78) had gastric/breast cancers but neither diffuse nor lobular histology, or even no pathology information. On the other hand, the study by van der Post et al reported on 499 tested families, of which only 18 were positive for a CDH1 pathogenic variant. Selection criteria for testing varied and, for instance, only 22 families (4.4% of the series) included in the study had LBC in the absence of any family member with gastric cancer, yet we⁸ and others¹³ have reported that a very significant percentage (20%–36%) of CDH1 pathogenic variant carrier families only present with LBC. The differences in patient selection between the analysed series of the mentioned studies and ours are obvious: the former analysed families that were tested due to suspicion according to contemporary understanding of CDH1 phenotypes at that time, yet our analysis included all families who tested positive for a CDH1 pathogenic variant in a clinical laboratory regardless of the reason for testing. That means that the original suspicion that prompted testing may have been for a different cancer syndrome, particularly an inherited breast cancer syndrome, and that highlighted the significant number of families with only breast cancers that were present in the cohort we analysed. We believe that this information is very valuable in helping better understand the CDH1 phenotype and a very strong argument to include CDH1 as part of the panels for testing families suspicious for an inherited breast cancer syndrome.

A significant problem encountered in cancer genetics assessment clinics is that pathology information from family members is often not available. Thus, consideration of these types of real-life limitations should be encouraged when proposing guidelines for genetic testing. In fact, when we dispensed of the requirement for histological confirmation of DGC among family members, the sensitivity in our cohort went up to 40% and 47% for the 2015 and 2020 IGCLC criteria, respectively. In a similar manner, foregoing the histological confirmation requirement yielded a sensitivity of 57% in the other large study of CDH1 mutation carriers unselected by clinical criteria.¹³ In summary, part of the IGCLC criteria’s insensitivity comes from the requirement for histological confirmation, an issue of real-world application; the remainder relates to the preconception that CDH1 pathogenic variants are only found in families with multiple, early onset DGC and LBC.

When developing genetic testing criteria, it is critical to be aware of the recommendations for genetic testing generated by other guidelines, particularly within the new framework of MGPT, the most used approach nowadays in the USA. In fact, application of current NCCN testing recommendations would have picked up 71% of the families in our cohort if the CDH1 gene is included in the MGPT. The addition of testing in cases of (1) proband with DGC at any age, or (2) two or more cases of gastric cancer in first-degree or second-degree relatives, with at least one confirmed to be DGC or diagnosed at age 50 or younger, as we propose, resulted in a sensitivity of 87% in our cohort, much higher than the sensitivity of the IGCLC 2020 criteria, even when dispensing of the requirement for histological confirmation. Another advantage of our suggested approach is that it simplifies the recommendations for testing, making it likely easier to remember, which potentially could result in better testing uptake and thus more individuals identified with pathogenic CDH1 variants.

One important limitation of our study is that, as discussed, our cohort consists only of families carrying CDH1 pathogenic or likely pathogenic variants and, as a result, we cannot determine the specificity or positive predictive value of the IGCLC criteria of our own criteria. Due to our inclusion of the NCCN HBOC criteria, which calls for testing of many people who may have other non-CDH1 pathogenic variants, our criteria are likely less specific than the IGCLC criteria. Nevertheless, our criteria do not call for testing many individuals who would not get genetic testing otherwise. Individuals meeting the NCCN’s HBOC criteria are recommended to get genetic testing anyway, and the NCCN currently endorses multigene testing and specifically lists CDH1 as one of six genes to be included on panels.¹² Our other two proposed criteria (personal history of DGC or family history of two gastric cancers in first-degree or second-degree relatives, with at least one before age 50 or confirmed DGC) do not encompass many individuals. Thus, there are about 27 000 new cases of gastric cancer annually in the USA.¹⁶ Approximately 20% (5400) of those cases are DGC.¹⁷ Roughly 10% (2700) of gastric cancers show familial aggregation.¹⁸ By random chance, few of those familial cancers are likely to be DGC or diagnosed before age 50, as DGC only makes up 20% of gastric cancers and fewer than 17.4% (4700) of gastric cancers in the USA are diagnosed before age 50.¹⁶ In sum, we estimate that instituting our criteria would amount to genetic testing for no more than 10 000 people who would not be tested otherwise. Once we extend the benefits of diagnosing a CDH1 mutation to other family members, the potential benefits very likely outweigh the costs of the additional tests.

We acknowledge the possibility that we may overestimate the sensitivity of our criteria due to ascertainment bias, as meeting the NCCN HBOC criteria is a common reason for referral to genetic testing. Using data from large-scale molecular sequencing studies performed on all-comers in oncology will be essential to overcome this potential bias and get more accurate estimates. Other limitations are that our dataset has no information about Māori ethnicity, cleft lip/palate or pagetoid spread of signet ring cells. Nonetheless, these criteria are unlikely to significantly impact the IGCLC criteria’s overall sensitivity. For instance, the Maori population in the USA is extremely small; in the USA census of 2000, only 1994 individuals indicated that they were of Māori descent.¹⁹ Finally, we did not address individuals who have CTNNA1 pathogenic variants, which have been identified in a small number of patients with clinical features of HDGC, as this was outside of the scope of this project.

One may question the clinical relevance of identifying CDH1 pathogenic variant carriers whose family histories are not striking enough to satisfy the IGCLC criteria. Our group and Roberts et al have previously shown that cancer penetrance is lower in CDH1 pathogenic variant carriers not preselected by IGCLC clinical criteria.^{8 13} Nevertheless, the cumulative incidence of gastric cancer (37%–42% for men and 24%–33% for women) and female breast cancer (43%–55%) in these individuals remains significantly higher than that of the general population. Like the IGCLC, we would not recommend prophylactic total gastrectomy or bilateral mastectomy in the absence of significant family history. A similar situation has recently been reported in Lynch syndrome where very important phenotypic differences are being unveiled even within the same gene when comparing different families.²⁰ We hope this study fosters further research and helps the scientific community appreciate even more the complexity of CDH1 germline mutations and the associated phenotypes. Realising that there is a very important subset of families that have a much milder phenotype or a breast cancer only phenotype should prompt future studies geared at understanding the reason for these differences so eventually we have a much more educated way to counsel these patients. Selection criteria that allow the identification of more mutation carriers will be essential in helping understand the differences in these distinct phenotypes. We cannot hope to better understand the full spectrum of phenotypic syndromes related to CDH1 mutations if most of these individuals remain undiagnosed.

In summary, we believe our proposed criteria could pick up a much higher number of patients than the IGCLC 2020 criteria with no significant increase in testing burden. We would like to encourage testing the set of proposed criteria in other cohorts from various regions to gain a better understanding of their effectiveness and evaluate if they fair better than the IGCLC 2020 criteria, as we have seen in our cohort.

Data availability statement

All data relevant to the study are included in the article or uploaded as supplementary information.