Familial Cancer

, Volume 8, Issue 4, pp 483–487

Are prediction models for Lynch syndrome valid for probands with endometrial cancer?

Authors

  • Floor J. Backes
    • The Division of Gynecologic Oncology, Department of Obstetrics and GynecologyThe Ohio State University College of Medicine
  • Heather Hampel
    • Division of Human Genetics, Department of Internal MedicineThe Ohio State University College of Medicine
  • Katherine A. Backes
    • The Division of Gynecologic Oncology, Department of Obstetrics and GynecologyThe Ohio State University College of Medicine
  • Luis Vaccarello
    • Gynecologic Oncology and Pelvic Surgery AssociatesMount Carmel Health System
  • George Lewandowski
    • Gynecologic Oncology and Pelvic Surgery AssociatesMount Carmel Health System
  • Jeffrey A. Bell
    • Central Ohio Gynecologic OncologyRiverside Methodist Hospital
  • Gary C. Reid
    • Central Ohio Gynecologic OncologyRiverside Methodist Hospital
  • Larry J. Copeland
    • The Division of Gynecologic Oncology, Department of Obstetrics and GynecologyThe Ohio State University College of Medicine
  • Jeffrey M. Fowler
    • The Division of Gynecologic Oncology, Department of Obstetrics and GynecologyThe Ohio State University College of Medicine
    • The Division of Gynecologic Oncology, Department of Obstetrics and GynecologyThe Ohio State University College of Medicine
Article

DOI: 10.1007/s10689-009-9273-5

Cite this article as:
Backes, F.J., Hampel, H., Backes, K.A. et al. Familial Cancer (2009) 8: 483. doi:10.1007/s10689-009-9273-5

Abstract

Currently, three prediction models are used to predict a patient’s risk of having Lynch syndrome (LS). These models have been validated in probands with colorectal cancer (CRC), but not in probands presenting with endometrial cancer (EMC). Thus, the aim was to determine the performance of these prediction models in women with LS presenting with EMC. Probands with EMC and LS were identified. Personal and family history was entered into three prediction models, PREMM(1,2), MMRpro, and MMRpredict. Probabilities of mutations in the mismatch repair genes were recorded. Accurate prediction was defined as a model predicting at least a 5% chance of a proband carrying a mutation. From 562 patients prospectively enrolled in a clinical trial of patients with EMC, 13 (2.2%) were shown to have LS. Nine patients had a mutation in MSH6, three in MSH2, and one in MLH1. MMRpro predicted that 3 of 9 patients with an MSH6, 3 of 3 with an MSH2, and 1 of 1 patient with an MLH1 mutation could have LS. For MMRpredict, EMC coded as “proximal CRC” predicted 5 of 5, and as “distal CRC” three of five. PREMM(1,2) predicted that 4 of 4 with an MLH1 or MSH2 could have LS. Prediction of LS in probands presenting with EMC using current models for probands with CRC works reasonably well. Further studies are needed to develop models that include questions specific to patients with EMC with a greater age range, as well as placing increased emphasis on prediction of LS in probands with MSH6 mutations.

Keywords

Endometrial cancerHereditary nonpolyposis colon cancerLynch syndromeMismatch repair gene mutationsPrediction models

Abbreviations

HNPCC

Hereditary non-polyposis colorectal cancer

CRC

Colorectal cancer

MSI

Microsatellite instability

EMC

Endometrial cancer

MMR

Mismatch repair

Introduction

Lynch syndrome, also known as hereditary non-polyposis colorectal cancer (HNPCC), is the most common colorectal cancer (CRC) syndrome in western countries, accounting for 2–5% of all CRCs [1, 2]. Approximately 2% of newly diagnosed endometrial cancer (EMC) patients have Lynch syndrome [3]. Lynch syndrome is an autosomal dominant syndrome, caused by inactivating mutations of the DNA mismatch repair (MMR) genes, mostly MLH1, MSH2, MSH6 and PMS2. Patients with Lynch syndrome are at increased risk of developing colorectal, endometrial, ovarian, gastric, urinary tract, and small bowel cancers. Women with Lynch syndrome have a 50–85% lifetime risk of developing CRC and more importantly a 40–60% lifetime risk of developing EMC [4, 5].

Different methods exist to predict the chance of having Lynch syndrome. Traditionally, the Amsterdam clinical criteria have been used. These require at least three family members in two or more generations with CRC, one affected person is a first-degree relative of the two, and at least one individual was diagnosed before age 50. The Amsterdam II criteria were introduced in 1999 and also took into account extra-colonic tumors, including EMC [6]. This increased the sensitivity of the detection of Lynch syndrome to 78% [7]. However, a recent study showed that only 39% of patients with Lynch syndrome fulfilled the Amsterdam II criteria, and only 25% of patients with MSH6 mutations meet Amsterdam II criteria [8].

The Bethesda guidelines, using clinical criteria, were introduced in 2004 to predict which patients should have their tumor tested for microsatellite instability (MSI) testing (the phenotypic alteration associated with a MMR gene mutation, such as Lynch syndrome) [9]. The sensitivity of the Bethesda guidelines has been estimated to be 82–94% [7, 10]. Despite these common clinical criteria used to predict or test for Lynch syndrome, they were generally developed with a focus on probands or pedigrees with CRC; thus, patients and families with EMC do not meet clinical criteria for the diagnosis of Lynch syndrome.

Currently, three prediction models; PREMM(1,2) [11], MMRpro [12], and MMRpredict [13], are used clinically to predict a probands risk of having Lynch syndrome. These models use age at diagnosis, personal and family history, location of the tumor, and sometimes also information on MSI to predict the likelihood of having mutations in the MMR genes and Lynch syndrome. These models have been validated in probands with CRC, and predict between 80 and 100% of mutation carriers (where a positive screen is defined as a proband having at least a 5% probability of having Lynch syndrome). MMRpro, which estimates the probability of having MLH1, MSH2, or MSH6 mutations, has a reported concordance index of 0.79, and a ratio observed to predicted cases of 0.97 [12]. PREMM(1,2), which estimates the probability of having MLH1 and MSH2 mutations, has a concordance index of 0.80 with sensitivity at 5% cutoff for screening of 94% but specificity of 29% [11]. MMR predict stage 1 (which incorporates MLH1, MSH2, and MSH6) has a 68% sensitivity, 86% specificity and 19% PPV at a cutoff of 5% [13]. However, these models have never been validated in probands presenting with EMC. Thus, the aim of this study was to determine the performance characteristics of these prediction models in women with Lynch syndrome presenting with EMC.

Methods

Probands with EMC and mutations in DNA MMR genes defining Lynch Syndrome were identified prospectively from 1,154 eligible patients diagnosed with EMC at one of three hospital systems in Columbus, Ohio metropolitan area between January 1999 and December 2003. 562 patients completed molecular analysis: all patients underwent MSI testing, and a subset of MSI negative tumors underwent immunohistochemistry of all four MMR proteins. Probands who were found to be MSI positive, or MSI negative but with abnormal immunohistochemical staining, underwent DNA testing for mutations in MLH1, MSH2, MSH6, and PMS2. The 13 patients (2.3% of the entire population of probands with EMC) with confirmed Lynch syndrome were used for our study.

Personal and family history of cancers was entered into the prediction models MMRpro, PREMM(1,2), and MMRpredict. The first prediction model, MMRpro requires family history of colorectal and EMC with age at diagnosis, as well as the current ages of affected and unaffected family members with or without the information available from MSI testing. MSI information was not used, since our aim was to evaluate and compare the models based on the same clinical information, e.g. family history and age. Probabilities for mutations in the MMR genes MLH1, MSH2, and MSH6 using MMRpro, were recorded for each patient. This model also adds the probabilities for mutations in the three genes to give an overall probability (“total”) of having Lynch syndrome, independent of the specific gene mutation. The second model, PREMM(1,2), incorporates multiple questions regarding family history of colon and EMC, and age at diagnosis. The PREMM(1,2) model was used to estimate the probability for a mutation in MLH1 or MSH2. The third model, MMRpredict, incorporates family history of colon cancer or EMC, age at diagnosis, and location of the tumor in the first stage of the model, and adds IHC and MSI in the second stage of testing. In this report, we used only the first stage of MMRpredict; no information about IHC and MSI was included. MMRpredict was used to estimate the probability of having a mutation in MLH1, MSH2, or MSH6. This model gives one overall probability, and does not report separate probabilities for each gene. Importantly, this prediction model could only be used if the patient had EMC under age 55, as the model was developed from a population of CRC patients diagnosed under age 55 and validated on a population less than age 45. In this model, the location of CRC (“proximal” or “distal”) is a mandatory question of MMRpredict. Therefore, the model was run twice, once answering the question as “proximal”, and once as “distal”.

The sensitivities of these models were compared for the ability to correctly predict Lynch Syndrome in probands with EMC. Accurate prediction of MMR gene mutation was defined as a model predicting at least a 5% probability of a proband carrying a mutation. For comparison we also report accurate prediction of MMR gene mutation using a cutoff of 10% probability. Statistical comparisons were made using the Student’s t-test with P values < 0.05 deemed significant.

Results

Thirteen patients with EMC and Lynch syndrome confirmed by genetic testing were identified (Table 1). Nine patients had a mutation in MSH6, three in MSH2, and one in MLH1. Median age at cancer diagnosis was 58; four patients were diagnosed with EMC under age 50. Patients with MSH6 gene mutations had a median age of 60, whereas patients with MLH1 or MSH2 mutations had a mean age of 44.5 (P = 0.01).
Table 1

Patient characteristics of those diagnosed with Lynch syndrome and endometrial cancer

Study ID

Age at diagnosis

Diagnosis

Mutation

Amsterdama

Bethesdab

59092

39

EMC

MLH1

N

N

1020

44/46

EMC/CRC

MSH2

Y

Y

862

55

EMC

MSH2

Y

Y

537

45

EMC

MSH2

N

N

18

58

EMC

MSH6

N

Y

345

60/65

EMC/ureter

MSH6

N

N

475

40/69

CRC/EMC

MSH6

Y

Y

501

60

EMC

MSH6

N

N

1079

56

EMC

MSH6

N

N

140

47

EMC

MSH6

N

N

55251

68

EMC

MSH6

N

N

553

61

EMC

MSH6

N

N

1493

59

EMC

MSH6

N

N

EMC endometrial cancer, CRC colorectal cancer

aWhether or not the patient met Amsterdam II criteria

bWhether or not the patient met revised Bethesda guidelines

Two patients had metachronous EMC and CRC (one at age 44 and 46, and one at age 69 and 40, for EMC and colon cancer, respectively). One patient, diagnosed at age 60, developed transitional cell cancer of the ureter at age 65. Three patients met Amsterdam II criteria (and, by definition also met the Bethesda guidelines), and one patient met only the revised Bethesda guidelines. Nine (69%) patients did not meet either Amsterdam II criteria or the revised Bethesda guidelines.

The first model, MMRpro, correctly predicted a mutation in the MSH6 gene in 3 of 9 probands (more than 5% probability for the individual mutation), in the MSH2 gene in 3 of 3 probands, and a mutation in the MLH1 gene in 1 of 1 proband (Table 2).
Table 2

Efficiency of the existing prediction models for Lynch syndrome in endometrial cancer patients

Prediction model

Correct predictiona

Median estimated probability (%)

5% Cutoff

10% Cutoff

MMRpro

    MLH1

1/1

0/1

5.1

    MSH2

3/3

3/3

46.7

    MSH6

3/9

1/9

1.4

    Total

11/13

8/13

15

PREMM(1,2)

    MLH1/MSH2

4/4

3/4

22.5

    Total

13/13

11/13

11

MMRpredict (MLH1/MSH2/MSH6)

    Classified as “Proximal”

5/5

4/5

66

    Classified as “Distal”

3/5

3/5

30

aThe number of times the model is correct in predicting Lynch syndrome by using a cutoff of either a 5 or 10% probability of having Lynch syndrome as reported by the different models

More importantly, when combining the probabilities for all MMR gene mutations (MLH1, MSH2, and MSH6), MMRpro predicted that eleven of thirteen would have Lynch syndrome. The second model, MMR predict, was only used for patients diagnosed at less than 55 years of age. Therefore, we could only include five patients (four presenting with endometrial and one with colon cancer at age <55). MMRpredict was evaluated twice, using the designation of both “proximal” and “distal” for the location of the CRC. When entered as “proximal”, MMRpredict predicted 5 of 5 patients to have more than a 5% chance of having Lynch syndrome; when entered as “distal”, MMRpredict predicted 3 of 5 to have Lynch Syndrome. The third model, PREMM(1,2), was developed based on a group of patients with either MLH1 or MSH2 mutations. In our patients with EMC, PREMM(1,2) predicted that 4 of 4 probands with confirmed MLH1 or MSH2 mutations had Lynch syndrome. When this model was applied to our entire population with Lynch syndrome (including the nine patients with an MSH6 mutation), PREMM(1,2) correctly predicted 13 of 13 MMR patients had Lynch syndrome (Table 3).
Table 3

Number of correct predictions per mutation and model using the 5% probability as a cutoff

 

MMRpro

MMRpredict

PREMM(1,2)

MLH1 (n = 1)

1

1

1

MSH2 (n = 3)

3

2

3

MSH6 (n = 9)

3

2

9

Overall, using a 5% probability of having a DNA MMR mutation as a cut off for defining a positive screening test for Lynch syndrome, MMRpro predicted 84% of cases correctly, MMRpredict 60 or 100% (depending on whether the tumor was entered as “distal” or “proximal” colon cancer, respectively), and PREMM(1,2) 100%. As expected, the performance of all of these prediction models was worse when a threshold of a 10% probability of Lynch syndrome was used to classify a positive screening test (Table 2).

Discussion

Lynch syndrome is equally prevalent in patients with EMC as in patients with CRC (2%). Patients with EMC and Lynch syndrome are at risk for multiple other cancers, including CRC. Recognition of Lynch syndrome is important since effective screening methods for associated cancers like CRC are in place. In addition, family members may benefit from prophylactic hysterectomy and oophorectomy for prevention of ovarian and EMC related to Lynch syndrome [14]. Importantly, in patients with Lynch syndrome cancers, EMC has been shown to precede colon cancer in half of the cases and can serve as a ‘sentinel’ cancer [15]. To improve detection of Lynch syndrome, prediction models have been developed for patients with CRC and their families. However, these models have not been validated for the prediction of Lynch syndrome in probands presenting with EMC. When the prediction models were applied to Lynch syndrome probands with EMC, all three models performed reasonably well with sensitivities of 80–100%. However, we identified differences in ability of different models to predict mutations in specific MMR genes. All three models correctly predicted Lynch syndrome in patients with either MLH1 or MSH2 mutations; however, MMRpro predicted Lynch syndrome in less than one-third of probands with MSH6 mutations. Clinically, this is further complicated by the fact that the Amsterdam II clinical criteria also do not perform well in patients with MSH6 mutations (compared with its performance in pedigrees with mutations in MLH1 or MSH2). One possible explanation for the limitations of the models in probands with MSH6 mutations is that these patients are more commonly diagnosed with EMC than CRC [1619], and that CRC is mainly associated with MLH1 and MSH2 mutations [1921]. In our series, MSH6 mutations were more common than mutations in MLH1 and MSH2 combined. Therefore, in pedigrees with Lynch syndrome and an MSH6 mutation, CRC may be less prevalent, and a more atypical distribution of HNPCC associated malignancies may be seen. Furthermore, although the average age at diagnosis for CRC and EMC in patients with Lynch syndrome is comparable (~48 years [22]), in our population, the mean age of diagnosis of our Lynch syndrome patients with EMC was 58 years (and 9 of 13 were diagnosed after age 50), which makes it less likely to fit the clinical criteria for the diagnosis of Lynch syndrome. Our finding of a higher prevalence of MSH6 mutation combined with an older age at diagnosis is consistent with data from patients with CRC due to MSH6 mutation, in which these patients have been shown to be diagnosed on average 10 years later than patients with MLH1 or MSH2 mutations [12, 22, 23].

These differences between CRC and EMC in the setting of Lynch syndrome may affect the performance of the existing prediction models. Since MSH6 is more prevalent in EMC, consideration of new Lynch syndrome prediction models should more heavily consider the role of MSH6 in EMC, and expand the inclusion of patients with Lynch associated cancers who are older than 50.

Another interesting finding from our study was that even though PREMM(1,2) only provides estimation for MLH1 and MSH2 gene mutation and does not include prediction of MSH6 gene mutations, the model correctly predicted all cases of Lynch syndrome (even for the nine patients with MSH6 mutations). This indicates that this model, despite developed based on a cohort of 898 consecutive, unrelated probands subjected to full gene sequencing of MLH1 and MSH2, is able to estimate the probability of a gene mutation fairly well, even when the mutation is in MSH6. Nonetheless, as with all of the models tested, more information is needed regarding the specificity and positive predictive value of when applied to probands with EMCA.

While we acknowledge that our study design investigates only sensitivity, we are in the process of running all of the 562 patients through these models to investigate the specificity and positive predictive value.

In conclusion, identification of EMC probands with a high probability of Lynch syndrome using an accurate prediction model would be beneficial in enrolling these patients in screening protocols to prevent and detect CRC. Improved identification of EMC patients with Lynch syndrome could also engage at risk family members in genetic testing for Lynch syndrome. We found that these models, particularly PREMM(1,2), perform reasonably well for probands with EMC. However, limitations in all models include the moderate performance in prediction of Lynch syndrome in probands with MSH6 mutations, and the fact that the age restrictions of certain models limit their utility in patients with EMCA, especially when caused by MSH6 mutation. Future studies are needed to evaluate further performance characteristics before applying these models to EMC patients suspicious of Lynch syndrome.

Copyright information

© Springer Science+Business Media B.V. 2009