In total this study is comprised of 221 EC patients used to compare the detection rate of Lynch syndrome using different strategies. The use of Amsterdam II criteria had a detection rate of 0.9%, Bethesda criteria 0.9%, and universal testing 2.3%, which might be close to the LS frequency of LS in Sweden, even if the study used a relatively small number of cases.

A defect mismatch repair system or MSI phenotype can be screened for in tumor cells using PCR, next-generation sequencing techniques as well as IHC [20] to detect possible MMR deficiency. MSI and IHC usually display a high concordance [21], but both methods have their advantages as well as obstacles. MSI is a fast and non-user dependent technique, but cannot distinguish which specific MMR protein is involved in the MSI phenotype. It has also been shown that standard MSI-testing has a sensitivity of 80–90% for MLH1/PMS2 and MSH2 loss, but a reduced sensitivity of 55–77% to find deficient MMR caused by MSH6 loss [22]. IHC is dependent on a pathological evaluation, but can indicate which MMR-protein has been lost, and therefore helps select which gene should be screened in the germline DNA to confirm the LS diagnosis [23]. Furthermore, IHC has fairly equal sensitivity to both MLH1/PMS2 or MSH2/MSH6 loss with 83% sensitivity and 89% specificity regardless of gene involved (14).

However, one of our IHC findings were unexpected. One case with a MLH1/PMS1 loss on IHC did not have a MLH1 mutation, but instead had a pathogenic germline variant in MSH2. The hypermethylation test later conducted on this patient’s tumor DNA turned out to be positive which indicates that her EC had a sporadic origin and was not caused by her inherited MSH2 pathogenic variant. It should be noted that this patient fulfilled the Amsterdam II criteria, and other individuals in the family with the same pathogenic germline MSH2 variant displayed loss of MSH2 and MSH6 upon IHC on their tumor tissue indicating that this variant does lead to protein loss in LS-associated tumors. Sporadic tumors can of course also occur in LS patients with increasing incidence in older ages. In fact, this patient was affected at 65 years of age and was the oldest in our LS cohort. If we would have used a study protocol with hypermethylation test on all MLH1/PMS2 losses before germline testing this patient would have been missed. However, with our strategy of performing germline screening of all four MMR genes on all tumors with IHC loss, we could still identify hereditary LS in this patient. Thus, our small study might suggest a benefit to always screen all four MMR genes when there is a MMR loss instead of only the gene suggested from the IHC result in cases that fulfill the Amsterdam II criteria.

44% of tumors with loss of MSH2 and MSH6 have a germline mutation in MSH2, but we only had available genomic DNA from one such patient with a normal germline test. Seven other cases had a loss of only MSH6 with three of them having germline mutations which is in line with other studies (23).

When using universal IHC on EC tumors as the first line screening tool our study found three previously unknown MSH6 mutation carriers. None of these three patients fulfilled the Bethesda criteria, which is in line with previous literature on MSH6 pathogenic variant carriers, who are known to have negative clinical criteria to a higher degree compared to MLH1 and MSH2 carriers, making them more difficult to find using clinical criteria as the primary screening method [24]. It is also in line with previous results that indicate that inherited MSH6 mutations also make up a substantial part of all LS associated ECs [25]. One of these patients had a first degree relative with EC before 50 years of age which would increase suspicion of an inherited risk, but it was not sufficient to fulfill the revised Bethesda guidelines in their current design.

Our results are in concordance with other similar studies which have shown that universal IHC screening on endometrial and colorectal tumors would lead to approximately a doubled increase in detection of LS in randomly selected EC tumors or CRCs [26]. In November 2018 the American Cancer Society therefore updated their recommendations, and now suggest screening of all new diagnosed ECs with IHC to find MMR loss and which patients to continue investigation for LS [27].

Another important finding in our data is that Bethesda criteria did not show a high concordance with MMR protein loss (13.0%) and the patients with positive Bethesda were also evenly distributed between the group with loss of MMR and intact MMR (13.0% vs. 15.0%) (Table 3). This shows that Bethesda had a low sensitivity to detect MMR loss in EC and would probably perform poorly to detect LS in EC as the primary screening method. This can of course be explained by the fact that the Bethesda is designed primarily to detect LS in CRCs.

Although 25–30% of all EC tumors display the MSI phenotype or MMR loss, this is most often due to somatic mutations and not a result of inherited pathogenic variants (5). From the patients with MMR loss on IHC in our study the vast majority had a loss of MLH1 and PMS2 and only one showed a germlineMLH1 variant (Fig. 1). This finding is in line with a similar study who tested 500 EC tumors with IHC for MMR proteins and 132 turned out to have MMR loss on IHC with 100 of them with loss of MLH1 and/or PMS2 (75.8%). They further used MLH1 promotor hypermethylation on these 100 MLH1 deficient tumors and found 83 of them (83.0%) to have positive hypermethylation. Interestingly the group with hypermethylated tumors were significantly older than those with no hypermethylation [28]. We did not conduct routine promotor methylation test in our study, but chose to screen for germline variants in these patients.

IHC and gene sequencing results. BC ±  = Bethesda criteria positive/negative, AmII ±  = Amsterdam II positive/negative

In future investigation protocols for LS it may be feasible to have a hypermethylation test done on MLH1 deficient tumors to exclude them for germline testing as long as those patients do not have a strong family history for LS which our MSH2 carrier had. The high proportion of confirmed LS in patients with loss of MSH2 and/or MSH6 on IHC. (3/9, 33%,) is also in line with other studies on EC tumors. One study had 22 tumors in total with loss of either MSH2 and/or MSH6, twelve of which had a LS mutation (55%) and four additional cases had variants of unknown clinical significance (18%) [28]. A similar study to ours used IHC for MMR on 223 EC tumors in total and identified 60 with MMR-loss, nine of which had a MSH2 and/or MSH6 loss. In turn, seven of these nine cases underwent germline testing with four having a LS mutation (57%) [29].

It should be noted that 47 of 52 with loss of MMR on IHC that underwent genome screening did not have a germline disease causing variant in our study which indicates that the vast majority of MMR loss is explained by sporadic somatic mutations in the tumor. This could be an argument against universal IHC screening, but the MMR loss is still valuable information for the clinical management and outcome of EC. A study, that analyzed MMR loss and DNA methylation in 466 ECs found that MMR loss due to DNA methylation had a more advanced stage of cancer at diagnosis with worse clinical outcome than tumors with intact MMR or MMR loss without methylation [30]. The FDA has also approved the use of pembrolizumab as checkpoint inhibitor treatment in all tumors with MMR loss regardless of tumor origin [31].

This means that the MMR status will be important to know in EC from both a genetic and a clinical perspective, which further implies the use of universal screening with IHC on all EC tumors.

Of importance in the implementation of universal IHC screening on EC is the potential benefit of the patient, their relatives and the health care service of knowing if the person is a LS carrier in relation to the increased cost for tumor surveillance in the patient and their carrier relatives. Snowsill et al. [32] did a cost–benefit analysis on EC tumor screening from LS and compared three scenarios, where one option would be to not screen at all, one was to screen all ECs for a germline mutation directly, or to use IHC/MSI with or without methylation-test before proceeding to germline testing. They concluded that screening with IHC together with a methylation-test was the most cost- effective option of these three alternatives up to 65 years of age with a cost of 14.200 pound per quality-of-life year (QALY) gained. The study could not see any economic reason to screen patients older than 65 years due to the lower probability of LS associated cancer with increasing age, and the older age of the patient and her potential relatives. The obvious weakness with their model is that they did not compare the IHC and methylation alternative with a model based on clinical criteria for screening. Another study compared costs to screen CRCs universally or to have targeted IHC-screening based on age < 60 years and/or other Bethesda criteria. There was a 2.5 – sevenfold lower cost upon targeted screening compared to universal testing with the same detection rate of LS [33]. Although the detection rate is debatable, it can still be reasonable to use Bethesda criteria or age limits for IHC on a solely economic basis. The same group did a follow-up study with a cost benefit analysis for universal testing on all ECs. What they found was that IHC with methylation test is the most cost-effective option to diagnose LS in EC and the most cost-effective option for QALY gained if taking both the patient and relatives into account [34]. Kwon et al. showed that the most cost-effective strategy would be to screen with IHC on all ECs that have at least one first degree relative with a LS-cancer [35]. The other options were not cost effective at all, due to their low sensitivity for LS. The disadvantage with that analysis though is that they did not use methylation tests in their model, and the study is 10 years old when costs for IHC were higher. Still, there is a massive support for IHC as a primary screening tool with or without additional clinical criteria. Interestingly, their cost-effective strategy would have missed three LS-carriers we found in our study. A micro-cost effect study by Ryan et al. [36] confirmed the beneficial effect of adding methylation test together with IHC or MSI in the screening, where the costs were almost equal between using MSI or IHC.

The cost of MPS on tumor and germline DNA has dropped significantly in the last years but is not yet in clinical practice in Sweden. However, if the MPS technique continues to get cheaper, faster and easier to interpret, the next step in LS diagnostics might be paired tumor/normal MPS including MSI on all new EC cases.