Background

Individuals with a familial predisposition to cancer pose an increasing challenge for healthcare systems hoping to provide state-of-the-art care. Optimal strategies for recognizing them, performing (and interpreting) genetic testing, and preventing cancers in at-risk family members have not been well established when considering overall benefits, harms, and costs. The challenges involved will likely become increasingly complicated with advances in understanding of the molecular genetics underlying cancer risk. In this report, we attempt to clarify many of these issues for one form of hereditary cancer (hereditary nonpolyposis colorectal cancer).

This evidence review is based upon a systematic review of the literature. The Key Questions that it addresses were proposed by the Agency for Healthcare Research and Quality (AHRQ) on behalf of the Centers for Disease Control and Prevention (CDC) Evaluation of Genomic Applications in Practice and Prevention (EGAPP) Project. EGAPP is a three-year model project developed by CDC's Office of Genomics and Disease Prevention to address the increasingly urgent need for timely and objective information that will allow health care providers, consumers, policy makers, and payers to distinguish tests that are safe and useful, and to guide their appropriate use in practice.

Hereditary Nonpolyposis Colorectal Cancer (HNPCC) can be defined clinically or genetically as will be described below. As a genetic disease it is inherited as an autosomal dominant disorder (with variable penetrance) and is caused by mutations in DNA mismatch repair genes. HNPCC is associated with a substantially increased risk for several forms of malignancy but particularly colorectal and endometrial cancer^1–3 (see Table 1).

Table 1

Lifetime cancer risk in HNPCC.

The disorder has also been referred to as the “Lynch syndrome” in recognition of Henry Lynch, who in 1966 first described familial aggregation of colorectal cancer with gastric and endometrial cancer in two large kindreds (although it was first reported by the eminent pathologist Aldred Warthin in 1913).⁴ Lynch I syndrome (also referred to as HNPCC I) refers to kindreds in which colorectal cancer predominates, while Lynch II syndrome (HNPCC II) refers to kindreds who also have extracolonic tumors. The Muir-Torre syndrome (sebaceous gland tumors with or without keratoacanthomas associated with visceral malignancy) and Turcot syndrome (HNPCC-related tumors associated with glioblastoma multiforme) describe additional subsets of HNPCC with tumor types that appear to cluster in affected families. These distinctions have become less clear as more families have been studied, and as a result, these clinical classifications are being supplanted by genetic classification.^5–7 The most common mismatch repair mutations associated with HNPCC are MLH1 and MSH2 (which together are believed to account for about 80 percent of cases) while MSH6 and PMS2 are less common.⁷

The precise cancer burden due to HNPCC has not been well defined. However, a germline mismatch repair mutation associated with HNPCC has been described in 1–5% of patients diagnosed with colorectal cancer in various reports.^8–12 Thus, HNPCC accounted for approximately 1400 to 7300 cases of colorectal cancer in 2005 in the United States based upon the overall estimate of 145,290 new cases of colorectal cancer.¹³ Similarly, approximately 0.5–2% of patients with endometrial cancer has a history compatible with HNPCC.^{14, 15} More than 1 in 3100 people between the ages of 15 and 74 are estimated to carry a defective DNA mismatch repair gene associated with HNPCC and thus are at risk for developing an HNPCC-related cancer.¹⁶

The recognition of a heightened cancer risk in carriers of mismatch repair mutations provides hope for offering screening, intensive surveillance or other measures (such as colectomy or hysterectomy) aimed at reducing the risk that cancer will develop. Furthermore, identification of a specific gene defect allows for testing of family members potentially sparing them the worry, bother, and expense associated with lifelong cancer surveillance if they do not carry the mutation. Notably, HNPCC has been excluded in 97 living members of the original family described by Henry Lynch based upon genetic testing.¹⁷

However, many uncertainties remain on the sequence of events that should occur in selecting patients with cancer to undergo testing for HNPCC and the spectrum of implications that screening, surveillance, and other management strategies have on affected patients and their families. The following summarizes many of these issues, while outlining those that are the subject of this report.

The framework for this report begins with patients with colorectal cancer, and explores the issues of screening and testing patients with colorectal cancer, identifying, counseling and testing at-risk family members, and the benefits and harms of subsequent management options for the probands and family members. We evaluate these issues from the perspectives of the patient, caregiver, family member, and policy-maker. The analytic framework is described in further detail in Chapters 2 and 3.

Who Should be Screened?

Screening for HNPCC has been most widely advocated in patients with colorectal cancer while there is much less information about screening in patients with other forms of HNPCC-related cancers. This report focuses only on colorectal cancer. However, recognition of HNPCC may also be possible in patients presenting with other HNPCC-related cancers. In one study, for example, one-half of women with HNPCC (defined by the Amsterdam criteria) presented with endometrial cancer.¹⁸ It may also be feasible to identify an HNPCC kindred from individuals without a personal family history of an HNPCC-related cancer by obtaining a detailed family history.

Several studies have described estimates of the proportion of patients with colorectal cancer who have a mismatch repair mutation, fulfill clinical criteria for a familial cancer predisposition, or have clinical (e.g., tumor location or histology) or laboratory (e.g., abnormal staining for mismatch repair proteins [IHC] or microsatellite instability [MSI]) characteristics of their tumor tissue that suggest the diagnosis. In this report, we evaluate such studies in detail to produce estimates of these parameters and attempt to explain variability across studies to an extent possible. These parameters are important for defining cost-effective pathways for evaluating patients with colorectal cancer for HNPCC and caring for family members.¹⁹ However, this report does not include a formal cost-effectiveness analysis.

How Should HNPCC be Defined?

The definition of HNPCC continues to evolve. Some authorities consider HNPCC to represent a superset of individuals with a mismatch repair mutation of whom a subset is considered to have the familial cancer clustering described by Henry Lynch. Such an approach recognizes the uncertainty that remains regarding the penetrance of cancer in carriers of mismatch repair mutations; the risk of cancer in a family with Lynch Syndrome may be substantially different compared with the risk associated with a mismatch repair mutation alone without such a history.²⁰

A different view is that Lynch Syndrome represents a description of familial cancer clustering, a subset of which is caused by mismatch repair mutations. Patients with familial clustering of HNPCC-related cancers but without a mutation may have a different form of hereditary cancer (e.g., cancers caused by mutations in the MYH gene)²¹ while others may have HNPCC caused by a mismatch repair mutation that was not sought, or from a false negative result of testing. These distinctions are important because they identify subgroups of patients with variable cancer risk, and they may influence how family members are screened and subsequent surveillance strategies. For example, at least one report suggested that families who fulfill the Amsterdam criteria for Lynch syndrome (described below), but do not have an identifiable mismatch repair mutation, might be at lower cancer risk compared with those in whom a mutation is identified.²²

An advantage of this view is that it establishes a reference standard for Lynch syndrome that is relatively more objective than one based upon a clinical definition. For example, it may be possible to establish the diagnosis of HNPCC in a proband with colorectal cancer who does not fulfill classical criteria for Lynch syndrome. In addition, a genetically based reference standard provides a framework for calculating sensitivity, specificity and predictive values of various predictors of mismatch repair mutation that might be useful in selecting patients for genetic studies (see Table 2).

Table 2

2×2 table for test characteristics when considering the presence of a mismatch repair mutation as the reference standard.

As will be described further in the methods section, in this report we take both views on HNPCC depending upon which Key Question is being addressed. For questions related to the test characteristics of predictor tests (such as clinical criteria or laboratory tests of tumor tissue described below), HNPCC can be defined as the presence of a pathogenic mismatch repair mutation in a patient with an HNPCC-related mutation (see Table 2). In this model, test characteristics depend upon how comprehensively mutation testing (e.g., the number of mismatch repair mutations tested and methods of testing, and the accuracy of the specific laboratory methods used), and predictor tests on tumor tissue (e.g., the quality of IHC analysis and the specific methods used to determine MSI status) were carried out and whether a mutation was known to be pathogenic. We describe the type of testing performed in all studies included in this report and provide a general description of how they are used in Chapters 2 and 3.

Not all mismatch repair mutations that can be identified using modern genetic testing are known to be pathogenic, leaving some uncertainty as to whether a mismatch repair mutation found in a patient with an HNPCC-related cancer is responsible for the increased cancer risk. The strongest evidence that a mutation is pathogenic is when its presence correlates strongly with the clinical expression of the disease. A mutation may also be considered pathogenic when its predicted protein sequence is expected to lead to a dysfunctional protein. In this report, the methods by which the authors attempted to define pathogenic mutations (if at all) were recorded for all eligible studies.

However, in some cases, the observed genotype has an unclear relationship to the clinical expression of the disease. Thus, even under the best of circumstances, genetic testing may produce an ambiguous result, making it unclear how the patient and their family should be counseled.²

Test characteristics are also vulnerable to several other features of study design, particularly selection bias, spectrum effects and verification bias, potentially helping to explain the variable results that have been described in the literature. In this report, we attempt to define these issues clearly to permit valid comparisons among studies. For example, a study applying clinical criteria to an unselected, consecutive population of patients with colorectal cancer may produce substantially different estimates of the accuracy of the Amsterdam criteria compared with a study enrolling patients who were referred to a tertiary care medical center because of multiple occurrences of cancer in the family. The latter group would be expected to have a higher prevalence of HNPCC and correspondingly better predictive values for the Amsterdam criteria.

In considering a clinical diagnosis of Lynch syndrome, sensitivity can also be defined as the proportion of patients with Lynch syndrome (defined by the Amsterdam criteria) with a specific predictor (see Table 3). Specificity would be the proportion of unselected patients with colorectal cancer in whom these predictor variables are present.

Table 3

2×2 table for test characteristics when considering Amsterdam criteria I as the reference standard.

However, this view permits only a limited understanding of the sensitivity or specificity of clinical criteria used to screen patients for HNPCC since many such criteria are included in the Amsterdam criteria. Nevertheless, the proportions determined using a clinical or genetic reference standard for HNPCC are complementary since they all describe relationships among predictor test, mutations, and the clinical syndrome of familial cancer clustering.

Identifying HNPCC in Patients With Colorectal Cancer

Most patients with colorectal cancer do not have a mismatch repair mutation making it impractical to consider universal genetic testing, especially when considering the cost (about $3,000 for comprehensive mutation testing).²³ As a result, several strategies have been proposed to identify patients who should undergo additional testing. As a general rule, these have been based either upon clinical criteria, predictive laboratory testing of tumor tissue, or a combination of both. Statistical models incorporating these approaches have also been proposed.¹¹ These approaches can be characterized quantitatively by determining their sensitivity, specificity and predictive values as described above.

In this report, we attempt to understand test characteristics of various approaches to identifying HNPCC in patients with colorectal cancer. We used these parameters to develop models (based upon decision-analysis) that explore different strategies for recognizing patients who carry the mutations.

Clinical Criteria That Suggest the Diagnosis

The recognition that certain types of cancers cluster in families with HNPCC and that cancer develops at relatively early ages compared with the general population provided the rationale for development of criteria that could be used to aid in the diagnosis. Two sets of criteria (the Amsterdam criteria and Bethesda guidelines) developed by a consensus of experts, have been most widely accepted and best studied, although many similar criteria have been proposed.^{24, 25} Both have been revised since their initial development.^{26, 27} These criteria have applied by healthcare providers and genetic counselors during interviews with the patient and/or their family and/or with assistance of written documents such as a survey.

The Amsterdam criteria (see Table 4) were designed to establish the diagnosis of HNPCC based upon familial clustering of HNPCC-related tumors. As described above, some authorities consider the Amsterdam criteria as the formal description of Lynch syndrome.

Table 4

Original and revised Amsterdam criteria.

By contrast, the Bethesda guidelines (see Table 5) were designed to help predict which patients with colorectal cancer are likely to have a mismatch-repair mutation and should thus undergo further testing. However, both the Amsterdam criteria and Bethesda guidelines have been studied for predicting the presence of mismatch repair mutations. The Amsterdam criteria are much stricter than the Bethesda guidelines and thus have lower sensitivity but higher specificity. The Bethesda guidelines are also more applicable in small families.

Table 5

Original and revised Bethesda guidelines.

Although the Bethesda guidelines and Amsterdam criteria continue to be used widely, several studies evaluating them (both the original and revised) have underscored the limitations of their accuracy in predicting the presence of mismatch repair mutations.^{11, 12, 20, 28, 29} A 2006 review of the literature reported that the sensitivity of the original Amsterdam criteria ranged from 54 to 91%.⁷ Such a wide range of estimates leaves substantial uncertainty as to the role of the Amsterdam criteria as a screening test for mismatch repair mutations. As described, above, there are many potential explanations for the variability across studies. In this report, we attempt to clarify important differences across studies that may help explain the variability.

In addition to the limitations regarding their predictive accuracy, there are practical problems with policies based on the implementation of these clinical criteria. Patients' report of the family history may not be accurate, particularly for cancers other than colorectal that are potentially related to HNPCC.³⁰ Issues of uncertain paternity may also be relevant in some families while some families may be too small (or have insufficient contact among family members) to obtain a clinically meaningful family history. In addition, the criteria are not always remembered or practical to obtain; as a result many caregivers (including oncologists) fail to obtain a detailed family history, and among those who do, many do not act appropriately upon it.^{30, 31}

Analytic Validity

As noted above, there are several laboratory methods used for predictive testing for mismatch repair mutations and for genetic testing itself. The accuracy of these methods can be influenced by several factors such as the definition of the reference standard, how tissue was collected and processed, and the specific method by which it was analyzed. Laboratory errors (e.g., mislabeling of a specimen, contamination, incorrect interpretation of results) all weigh into overall accuracy. These considerations have been collectively referred to as analytic validity.

A related issue is the reliability of testing (both within a laboratory and between laboratories). In some clinical areas, reliability has been assessed by a method known as “proficiency testing” in which samples of known positive and negative biological materials are submitted blindly to a laboratory.⁴⁰ Results can be used to determine sensitivity, specificity, and reliability. Proficiency testing for MSI became available in the United States in 2006.

In this report we attempt to define analytic validity and reliability of the predictive laboratory tests (i.e., IHC and MSI) and genetic testing methods. However, there are several limitations to attempting to assess these parameters using a literature-based approach:

• The literature search was based upon HNPCC, not the specific laboratory techniques, thus limiting the pool of potentially relevant studies.
• Many of the techniques described in the literature are experimental or old and thus do not reflect contemporary methods.
• There is likely to be publication bias since information regarding reliability and accuracy of the laboratory methods have been evaluated by individual laboratories or by manufacturers of the testing methods.
• Because it can be difficult to establish a clear reference-standard, many studies attempting to define analytic validity included the predictor tests in the reference standard, thereby making the results of test characteristics uninterpretable.

We attempted to clarify these points both by adhering to strict criteria for literature selection and in analyzing individual studies.

Benefits and Harms of Screening, Testing, and Subsequent Management Strategies

Genetic testing for a cancer predisposition has profound implications for the affected patients and their families. The genetic test results have the potential to prevent cancer in the affected patient and their families, and may influence how patients with cancer are managed, but they can also lead to harm from discrimination, the risks associated with surveillance strategies or other interventions (such as colectomy or hysterectomy), and the psychological impact of recognizing a cancer predisposition. These are issues that are common to all forms of genetically based diseases and represent areas of intense study. In this report we critically evaluate the literature exploring these issues in patients and their families with HNPCC from the perspective of the affected patients, their family and from the point of view of providers and policy-makers.

Enthusiasm for genetic testing is based upon the belief that knowledge of the genetic basis for a disease will allow for improved treatment and prevention. How these objectives can be best achieved in the care of patients and families with HNPCC has not been well established. Few high-quality studies have evaluated the effectiveness of screening strategies based upon the results of genetic testing for HNPCC. It is generally agreed that patients undergoing genetic testing should fully understand the implications of a positive and a negative result and the level of certainty of a positive or negative result as a predictor of disease.⁴¹ The knowledge-base used to counsel patients on these parameters is still evolving.

Correlation of genotypes with phenotypes in HNPCC is incompletely understood, as are the corresponding implications for screening for HNPCC-related tumors. At least two studies^{42, 43} and a cost-effectiveness analysis ¹⁹ suggested a reduction in colorectal cancer and mortality from colorectal cancer screening based upon results of genetic testing. However, the primary studies were not randomized and were vulnerable to selection bias. There is even less information regarding the effectives of screening for other forms of HNPCC-related cancers (particularly endometrial cancer).^{44, 45}

The uncertain benefits must be balanced against the potential for harms, which include the risks associated with screening procedures, the potential for false-positive results leading to further, possibly unnecessary testing, and the psychological, social, and economic implications from stigmatization. An observational study that included 16 HNPCC and HNPCC-like families illustrated the difficulties that may be encountered when attempting to implement a program of genetic screening and counseling.⁴⁶ Problems encountered included lack of compliance, ambiguous results of genetic tests, incomplete documentation of pathologic materials or medical history, poor cooperation among family members and/or their physicians, patient fear and anxiety and perception of insurance discrimination, and lack of knowledge among referring physicians. Thus, the realities of implementing a program for testing patients with colorectal cancer for HNPCC must be understood along with the full spectrum of implications of various strategies for establishing the diagnosis and testing family members