|
Advertisement |
|
|
||
 |
|
|||||
|
|||||
|
 |
||||||
|
||||||
|
||||||
|
Journal of Clinical Oncology, Vol 22, No 24 (December 15), 2004: pp. 4934-4943 © 2004 American Society of Clinical Oncology. DOI: 10.1200/JCO.2004.11.084 Refining the Amsterdam Criteria and Bethesda Guidelines: Testing Algorithms for the Prediction of Mismatch Repair Mutation Status in the Familial Cancer ClinicFrom the Molecular and Population Genetics Laboratory, London Institute, Cancer Research UK; Department of Clinical Genetics, Guy's Hospital; Family Cancer Clinic, Colorectal Cancer Unit, Cancer Research UK, and Academic Department of Histopathology, St Mark's Hospital; Cancer Research UK, Department of Epidemiology, Mathematics and Statistics, Wolfson Institute of Preventive Medicine, St Bartholomew's, and the London Medical School, London; Department of Clinical Genetics, St George's Hospital Medical School, Tooting, London; Kennedy Galton Centre for Human Genetics, Northwick Park Hospital, Harrow; Oxford Regional Genetics Service and NHS Regional Genetics Laboratory, Churchill Hospital, Oxford, UK Address reprint requests to I.P.M. Tomlinson, Molecular and Population Genetics Laboratory, London Institute, Cancer Research UK, 44, Lincoln's Inn Fields, London WC2A 3PX, United Kingdom; e-mail: ian.tomlinson{at}cancer.org.uk
PURPOSE: Hereditary nonpolyposis colon cancer (HNPCC) is a Mendelian dominant syndrome of bowel, endometrial, and other cancers and results from germline mutations in mismatch repair (MMR) genes. HNPCC is now best diagnosed on molecular grounds using MMR mutation screening, aided by microsatellite instability (MSI) and immunohistochemistry in tumors. Selection of families for molecular investigation of HNPCC is usually based on suboptimal methods (Amsterdam Criteria or Bethesda Guidelines), but these can be improved using additional clinical data (mean ages of affected persons and presence of endometrial cancer) in a quantitative model. METHODS: We have verified the performance of the Wijnen model and have shown that it remains valid when HNPCC is diagnosed using mutation screening, MSI, and immunohistochemistry. We have also set up and verified our own models (Amsterdam-plus and Alternative), which perform at least as well as the Wijnen model. RESULTS: The Amsterdam-plus model improves on the Amsterdam Criteria by using five extra variables (numbers of colorectal and endometrial cancers in the family, number of patients with five or more adenomas, number with more than one primary cancer of the colorectum or endometrium, and mean age of presentation) and performs better than the Wijnen model. The Alternative model avoids the need to evaluate the Amsterdam Criteria and performs nearly as well as the other models. CONCLUSION: We believe that a quantitative model, such as the Amsterdam-plus model, should be the first choice for selecting families or patients for evaluation of HNPCC using molecular tests. We present an algorithm for this process.
A family history of colorectal cancer (CRC) is one of the most common reasons for referral to a genetics clinic. Regular surveillance, generally using colonoscopy, can greatly reduce the risk of disease.1 However, families with apparently similar histories may have quite different genetic origins, including a variety of single gene disorders, complex genetic inheritance, or aggregation resulting from shared environment or chance. It is important to diagnose the underlying cause of familial CRC, wherever possible, because there are implications for screening regimens and risks to relatives. One of the most common Mendelian CRC syndromes is hereditary nonpolyposis colon cancer (HNPCC). HNPCC results from germline mutations in DNA mismatch repair (MMR) genes, principally MSH2, MLH1, and MSH6.2 HNPCC is characterized by early-onset carcinomas of the large bowel and endometrium, with more modestly increased risks of cancers of other sites. Almost all colorectal and endometrial cancers and most colorectal adenomas in HNPCC show microsatellite instability (MSI) as a result of defective MMR.3,4 HNPCC tumors also usually show loss of the protein derived from the mutated gene.5 Approximately 10% to 15% of sporadic CRCs are MSI-positive because of silenced MLH1 expression,6 although sporadic adenomas rarely show these features.7,8 Most Mendelian CRC syndromes have distinct clinical features, such as profuse polyposis, but the tumors in HNPCC cannot readily be distinguished from their sporadic counterparts. It is increasingly recognized, therefore, that HNPCC should be a molecular diagnosis on the basis of, for example: pathogenic MSH2, MLH1, or MSH6 germline mutation; MSI or loss of MLH1 expression in multiple cancers (or a single colorectal adenoma) from a family; or loss of MSH2 and/or MSH6 expression in one or more of a family's tumors. Even if the causative germline mutation is cryptic, making or excluding the molecular diagnosis of HNPCC is important not only for determining risks to relatives, but also for deciding the screening protocol for those at risk.9,10 It is impractical to screen every occurrence of familial bowel cancer for molecular changes suggestive of HNPCC; screening must be targeted to certain families using the clinicopathologic features of the pedigree. The so-called Amsterdam Criteria11 were originally used to diagnose HNPCC if at least three family members in two or more generations had CRC, one affected person was a first-degree relative of the other two, and at least one individual was diagnosed before the age of 50 years. The Amsterdam Criteria have been highly successful, with estimated sensitivity and specificity of 60% and 70%, respectively.12,13 Certain deficiencies, however, have become increasingly recognized. For example, the original Amsterdam Criteria did not take into account extracolonic cancers, patients with new MMR mutations were not addressed, and some families with multiple polyps but without profuse polyposis were erroneously classified as having putative HNPCC. The Amsterdam Criteria II14 (Table 1) were introduced in 1999 and Syngal et al13 estimated that they have a sensitivity of 78% and specificity of 61%.
The Bethesda Guidelines15 were introduced to indicate which families should proceed to MSI testing before screening for MMR mutations (Table 2). The Bethesda Guidelines were deliberately much less restrictive than the Amsterdam Criteria. Using known MMR mutation carriers, the sensitivity of the Bethesda Guidelines has been estimated as 94%, but with specificity of only 25%.13 Recently, modified Bethesda Guidelines were produced (Table 3). These criteria will probably have impressive sensitivity with low specificity.
Wijnen et al16 set out to provide a quantitative improvement on the Amsterdam Criteria and Bethesda Guidelines by setting up a logistic regression model to predict the probability of HNPCC mutation using the following predictive variables: mean age at diagnosis of CRC; Amsterdam Criteria–positive or –negative status; number of family members with CRC; the number of family members with endometrial cancer; presence of a patient with other cancers related to HNPCC; and presence of a patient with multiple synchronous or metachronous cancers. In the multivariable analysis, mean age at diagnosis of CRC, fulfillment of the Amsterdam Criteria, and the presence of endometrial cancer were the only independent risk factors. The work of Wijnen et al16 established the principle of a practical, quantitative improvement on the Amsterdam Criteria and arguably the model has been underused in clinical practice. Three potential problems remain. First, the model of Wijnen et al has not yet been verified in an independent data set. Second, there may be good historical reasons for including a composite variable such as the Amsterdam Criteria in the regression analysis, but this may lead to unnecessarily complex input data. Third, up to 30% of HNPCC patients have germline MMR mutations that are not detected using standard techniques, and a model needs to take into account17 HNPCC diagnosed by MSI testing and/or immunohistochemistry. We have therefore set out to test both the Amsterdam Criteria II and the Wijnen model as predictors of HNPCC that has been diagnosed or excluded using a combination of mutation screening, MSI analysis, and immunohistochemistry. We have also set up new models of our own. We have then verified all of the models in an independent set of families. Our models have potential for targeting molecular testing for HNPCC.
Selection of Families Two hundred fifty families were recruited from the Cancer Research UK Family Cancer Clinic, St Mark's Hospital, Harrow, and from the Family Cancer Clinic, Guy's Hospital, London. Our inclusion criteria (Table 4) were similar to those in the Bethesda Guidelines. Families from whom archival tumor tissue specimens could not be retrieved were excluded; families without a living affected person were not specifically excluded. Pedigree information was obtained from family members and the presence of cancer confirmed, wherever possible, from hospital records, pathology reports, cancer registry records, or death certificates. Apart from distinguishing between colorectal adenomas and carcinomas, histologic tumor subtypes were not taken into account when deciding whether to include a family. Each included family was given a score18 for each of 11 variables (Table 5).
Collection of Blood and Tumor Samples Informed consent was obtained from patients or their next of kin according to local review board guidelines, and 10 mL of blood was sampled where possible. Samples of all colorectal tumors and endometrial cancers from each family were requested. A maximum of three tumors from any one family was obtained and for most families, we were able to obtain at least two tumors. Tumor histology was taken from pathology reports or was assessed by L.L. and A.E.S. when no reports were provided. DNA was extracted from blood by standard methods. Neoplastic and normal areas were microdissected from archival tissue and DNA was extracted using a simple proteinase K digestion.
Mutation Analysis
MSI Testing
Immunohistochemistry
Determining HNPCC Status
Second Family Set To verify each model, we tested it against an independent set of 94 kindreds from the Family Cancer Clinic, Oxford Regional Genetics Service. All Oxford families had undergone MMR gene mutation analysis only, as is typical for many departments of clinical genetics in the United Kingdom. We considered for the study all families who had undergone mutation testing, and we selected families using the same criteria as used for the St Mark's/Guy's families (Table 6).
Statistical Methods
All 250 families from St Mark's/Guy's met the Bethesda Guidelines. One hundred ten families (group 1) met the Amsterdam Criteria II, and 28 (group 2) met the Criteria but had no affected person younger than 50 years of age (Table 7). Thirty-eight families (group 3) had a single affected family member at younger than 45 years (Table 7). The 74 patients in group 4 (other pedigrees) included sibships and parent-child pairs affected by CRC with one member younger than 45, and persons with two CRCs and/or endometrial cancers older than 45 years. Using our criteria (Table 6), HNPCC was diagnosed in 67 of 250 families (27%). Of these, 34 had pathogenic mutations (25 in MLH1, eight in MSH2, and one in MSH6); of the 33 mutation-negative families with HNPCC, 21 had loss of MSH2 expression and MSI-positive cancers, five had absent MLH1 expression in two MSI-positive cancers, two had two MSI-positive cancers with loss of MLH1 expression confirmed in one of these, two had two MSI-positive cancers but no immunohistochemistry data, two had one MSI-positive cancer with no loss of MLH1 or MSH2 expression, and one had adenomas that were MSI-positive. The features of patients diagnosed as having HNPCC on the basis of germline mutations did not, in general, differ significantly (details not shown) from those diagnosed by MSI and/or immunohistochemistry (suggesting that cryptic MMR mutations have similar effects to detectable changes). The only exception was that families diagnosed by MSI and/or immunohistochemistry tended to have more individuals affected (  21 = 9.4; P = .002), perhaps because a greater supply of tumors facilitated molecular diagnosis.
Single-variable analysis (Table 5) showed that those in pedigree group 1 (Amsterdam Criteria II, AMSPOS = 1) were more likely to have HNPCC, as were families with apparently dominant inheritance (DOMREC = 1); low youngest age at diagnosis of CRC or endometrial cancer (YOUNGAGE); more persons with two primary colorectal or endometrial cancers (NTWOPRIM); and more women with endometrial cancer (NENDOCA, ANYENDO). Families with five or more adenomas (NFIVEADS) were less likely to have HNPCC. The number of persons affected with other cancers (NOTHERCA) and the right- or left-sided location of CRCs (CRCRL) were not associated with HNPCC. Overall, families with more persons affected by CRC (NCRC) and a younger mean age at diagnosis of CRC or endometrial cancer (MEANAGE) were significantly more likely to have HNPCC, although patients in group 3 (patients younger than 45 years with no family history of CRC) were less likely to have HNPCC than slightly older patients who had a positive family history (reflected in data for NCRC and MEANAGE; Table 5). The Amsterdam Criteria II (group 1 families) yielded 78% sensitivity, 69% specificity, 48% positive predictive value (PPV), and 71% families correctly classified in the St Mark's/Guy's families. In general, pedigrees with HNPCC but not fulfilling the Amsterdam Criteria II were either isolated, early-onset occurrences of colorectal cancer or families having a predominantly early-onset colorectal cancer phenotype without an extensive family history. In comparison with the correctly predicted HNPCC families, the Amsterdam-positive pedigrees who did not have HNPCC tended to have multiple adenomas, fewer multiple primaries, older age at presentation, and lower frequency of endometrial cancer (all P < .01, Fisher's exact or Wilcoxon test). We then attempted to improve on the Amsterdam Criteria II using multivariable analysis. Our first observation was that no individual in group 2 (fulfilling Amsterdam Criteria II except that all affected persons were > 50 years of age; Table 7) had HNPCC. This was an important finding for genetic testing purposes, but it inevitably led to this group being dropped from the logistic regression analysis and therefore being uninformative in relation to other variables. Groups 2, 3, and 4 were therefore combined for additional analysis to focus on the distinction between Amsterdam Criteria II and other families. We found that the Amsterdam Criteria II (AMSPOS) were improved as a predictor of HNPCC by the incorporation into the model of the following variables (Table 5): NTWOPRIM (OR, 2.51; 95% CI, 1.27 to 4.92; P = .008), MEANAGE (OR, 0.92; 95% CI, 0.88 to 0.95; P < .001), NENDOCA (OR, 3.17; 95% CI, 1.55 to 6.47; P = .002), NFIVEADS (OR, 0.32; 95% CI, 0.15 to 0.66; P = .002), and NCRC (OR, 1.41; 95% CI, 1.06 to 1.89). The corresponding logit equation was ln[p/(1 + p)] = 0.97 + 0.34 NCRC + 1.30 AMSPOS + 0.92 NTWOPRIM + 1.15 NENDOCA –0.085 MEANAGE –1.14 NFIVEADS.
This Amsterdam-plus model was associated with a pseudo-R2 value of 0.38. The Hosmer-Lemeshow test showed nonsignificant residual variation, demonstrating a good fit of the model ( We then determined whether a simpler Alternative model, which did not include the Amsterdam Criteria as a variable, might perform as well or better than the other models. The best-fitting alternative model resulted in use of variables NCRC (OR, 1.75; 95% CI, 1.35 to 2.26; P < .001), MEANAGE (OR, 0.92; 95% CI, 0.89 to 0.95; P < .001), NFIVEADS (OR, 0.33; 95% CI, 0.16 to 0.69; P = .003), NTWOPRIM (OR, 2.94; 95% CI, 1.51 to 5.75; P = .002), and NENDOCA (OR, 3.95; 95% CI, 2.01 to 7.75; P < .001). The logit equation was ln[p/(1 + p)] = 1.07 + 1.08 NTWOPRIM + 1.37 NENDOCA + 0.56 NCRC – 0.07 MEANAGE – 1.11 NFIVEADS.
This model was associated with pseudo-R2 of 0.35. The Hosmer-Lemeshow goodness of fit test showed nonsignificant residual variation ( Figure 1A shows the ROCs that demonstrate the trade-off between sensitivity and specificity in our two models and, for comparison, the Wijnen model as applied to the St Mark's/Guy's data set. All three models performed well and similarly, with an area under the curve of about 0.87 (compared with the ideal of 1.0). For the Amsterdam-plus model, we suggest that a probability cutoff of pr(HNPCC) more than 0.14 could be used to select families with suspected HNPCC for additional molecular investigation; for the Alternative model, we suggest a threshold of pr(HNPCC) more than 0.13 (Fig 1A). For both models, these thresholds produced more than 95% sensitivity, approximately 65% specificity, and 45% PPV. Using a cutoff of pr(HNPCC) more than .05, the Wijnen model had slightly lower sensitivity (92%) and specificity (63%) than our two models. Overall, these sensitivity, specificity, and PPV values seem pragmatic for use in a clinical setting, and the higher sensitivity compared with the Amsterdam Criteria II was gained at the expense of only modestly decreased specificity.
The extra sensitivity of the quantitative models was shown when the 15 true HNPCC families who did not fulfill the Amsterdam Criteria II were considered. When the pr(HNPCC) more than .14 and pr(HNPCC) more than .13 cutoffs were used, respectively, the Amsterdam-plus and Alternative models both correctly predicted nine and 10 of these families to have HNPCC, respectively. To detect these patients with HNPCC, 46 extra families would have to have undergone molecular analysis under the Amsterdam-plus model and 68 extra families would have to have undergone molecular analysis under the Alternative model. The small number of true HNPCC families misclassified by the Amsterdam-plus and Alternative models using the above cutoffs had no specific distinguishing features, except that no occurrence of endometrial cancer was present in any of these pedigrees. Sixty-two families were incorrectly predicted to have HNPCC by the Amsterdam-plus model and 71 families were incorrectly predicted to have HNPCC by the Alternative model. When compared with those correctly predicted as having HNPCC, the incorrectly predicted pedigrees were more likely to have nondominant inheritance, left-sided cancers, a lower frequency of endometrial cancer, fewer members affected by CRC, and fewer patients with more than one primary cancer (all P < .01, Fisher's exact or Wilcoxon test). The Wijnen, Amsterdam-plus, and Alternative models were then verified using the patient set from Oxford, which comprised 20 mutation carriers (21%). One of the Oxford families in group 2 (none affected younger than age 50, but otherwise Amsterdam Criteria II-positive) had a germline MMR mutation, although no patient with five or more adenomas had a germline MMR mutation. The Amsterdam Criteria II performed slightly better in the Oxford data set than in the St Mark's/Guy's families, with 85% sensitivity, 62% specificity, and 38% PPV. The ROCs for the three quantitative models are shown in Fig 1B and are similar to those obtained using the St Mark's/Guy's data set, with the exception that the Wijnen model performed slightly worse. With the same cutoffs as used above, the Wijnen model yielded 85% sensitivity and 40% specificity, the Amsterdam-plus model yielded sensitivity of 90% and specificity of 30%, and the Alternative model provided sensitivity of 95% but only 15% specificity.
The aims of all clinicopathologic methods for predicting HNPCC are to improve diagnosis and avoid unnecessary laboratory investigations. These aims are often in conflict and the result is that no optimal model has emerged, owing to different opinions about the correct balance between sensitivity and specificity. The result is that historical or consensus criteria have predominated over analytic models such as that of Wijnen et al.16 In our opinion, a model should have approximately 90% sensitivity and then aim to maximize specificity within that constraint. We believe that the Bethesda Guidelines, for example, have overemphasized sensitivity. Conversely, the Amsterdam Criteria II are too specific, detecting only 78% of HNPCC families in our series. We have tested the Wijnen HNPCC prediction model and two new models in two data sets comprising family cancer clinic patients who fall into categories that are frequently investigated as possibly having HNPCC. Our conclusions are four-fold. First, we have confirmed the general finding of Wijnen et al16 that the Amsterdam Criteria II can be improved using quantitative approaches. This conclusion holds even though the quantitative models have to cope with suboptimal data, such as incomplete reporting, and therefore have to use relatively crude inputs, such as number of affected individuals uncorrected for family size. Second, we have shown that the Wijnen model using a cutoff of pr(HNPCC) of more than .05 (rather than .20 originally suggested) is an improvement on the Amsterdam Criteria II and Bethesda Guidelines (when both of our data sets are considered). The Wijnen model also holds when HNPCC is detected using MSI analysis and immunohistochemistry in addition to mutation screening. Third, our models slightly outperform the Wijnen model. The incorporation of new variables—in particular, presence of multiple colorectal adenomas and number of individuals with more than one primary colorectal or endometrial cancer—can improve the prediction of HNPCC. Fourth, we have shown that some aspects of the Amsterdam Criteria, such as the requirement for an individual affected at younger than 50 years of age, were particularly well judged. However, the importance of other variables that have been supposed to be consistent, independent predictors of HNPCC may need reassessment: there was, for example, no predominance of right- over left-sided colorectal cancers in our data set. Moreover, the presence of cancers apart from those of the colorectum and endometrium was not associated with HNPCC. This finding probably results from the relatively low risks of other cancers in HNPCC mutation carriers,23 although there may also have been some referral bias in our data set toward bowel cancer. Any of the three quantitative models could be substituted for the Amsterdam Criteria II in the family cancer clinic. Although the optimal cutoffs require refinement, all three models can provide improved sensitivity (85% to 95%). Specificity (20% to 65%) and PPV (30% to 45%) are lower, but are tolerable in a diagnostic setting and are better than those produced by the Bethesda Guidelines. We suggest that all three quantitative models undergo additional evaluation to assess their relative merits. Evaluation should be extended, if possible, from the types of families that we have analyzed to pedigrees with less evidence of a genetic predisposition and, perhaps, even to a population-based series. If one model is shown to be superior, it should be used, but if all are shown to have similar performance, then the Alternative model should be used because its inputs are simplest (Table 8) .
In clinical practice, we suggest that the family data should be ascertained initially and then, the chosen model applied. Families would be subject to further testing if an HNPCC risk greater than the threshold was predicted. Two alternative approaches can then be used. First, if the clinic has much easier access to germline mutation screening than to tumor analysis, we suggest that mutation testing on the youngest available affected family member should be undertaken. If no mutation were detectable or no DNA from an affected individual were available, immunohistochemistry and MSI testing should be carried out and families classified as true HNPCC using the same diagnostic criteria that we have used above (Table 6). Second, if there exists ready access to local histopathology expertise, we propose using MSI and immunohistochemistry to target mutation screening (and to diagnose HNPCC where mutations are cryptic) using the flow chart shown in Fig 2. The decision-making process attempts to take account of factors such as phenocopies in HNPCC families and the possibility that HNPCC adenomas might not be MSI-positive, although high-quality analysis is assumed. If the decision tree were believed to be too complex, it could be simplified by routinely performing immunohistochemical and MSI analyses together.
In summary, we have shown that quantitative models can provide better prediction of HNPCC than the Amsterdam Criteria II or the Bethesda Guidelines. In our opinion, the Wijnen or Amsterdam-plus model should be used in clinical practice—a simple computer program is sufficient to calculate risks. Families for whom the probability of HNPCC exceeds a suitable threshold can then be thoroughly investigated using a combination of mutation screening, MSI analysis, and immunohistochemistry. Using these methods, a greater number of families can be correctly classified as having true HNPCC and screening can be undertaken with confidence and an appropriate risk-to-benefit ratio.
The authors indicated no potential conflicts of interest.
We thank colleagues at St Mark's, Guy's, and Northwick Park Hospitals; Ella Barclay; Richard Houlston and Mike Bradburn for statistical advice; family members; and histopathology departments that supplied tumor specimens.
Authors' disclosures of potential conflicts of interest are found at the end of this article.
1. Jarvinen HJ, Aarnio M, Mustonen H, et al: Controlled 15-year trial on screening for colorectal cancer in families with hereditary nonpolyposis colorectal cancer. Gastroenterology 118:829-834, 2000[CrossRef][Medline]
2. Peltomaki P: Role of DNA mismatch repair defects in the pathogenesis of human cancer. J Clin Oncol 21:1174-1179, 2003
3. Wheeler JM, Loukola A, Aaltonen LA, et al: The role of hypermethylation of the hMLH1 promoter region in HNPCC versus MSI+ sporadic colorectal cancers. J Med Genet 37:588-592, 2000
4. Loukola A, Eklin K, Laiho P, et al: Microsatellite marker analysis in screening for hereditary nonpolyposis colorectal cancer (HNPCC). Cancer Res 61:4545-4549, 2001
5. Wahlberg SS, Schmeits J, Thomas G, et al: Evaluation of microsatellite instability and immunohistochemistry for the prediction of germ-line MSH2 and MLH1 mutations in hereditary nonpolyposis colon cancer families. Cancer Res 62:3485-3492, 2002
6. Lindor NM, Burgart LJ, Leontovich O, et al: Immunohistochemistry versus microsatellite instability testing in phenotyping colorectal tumors. J Clin Oncol 20:1043-1048, 2002
7. Loukola A, Salovaara R, Kristo P, et al: Microsatellite instability in adenomas as a marker for hereditary nonpolyposis colorectal cancer. Am J Pathol 155:1849-1853, 1999
8. Iino H, Simms L, Young J, et al: DNA microsatellite instability and mismatch repair protein loss in adenomas presenting in hereditary non-polyposis colorectal cancer. Gut 47:37-42, 2000 9. Lynch HT, Smyrk T, Jass JR: Hereditary nonpolyposis colorectal cancer and colonic adenomas: Aggressive adenomas? Semin Surg Oncol 11:406-410, 1995[Medline]
10. Sieber OM, Lipton L, Crabtree M, et al: Multiple colorectal adenomas, classic adenomatous polyposis, and germ-line mutations in MYH. N Engl J Med 348:791-799, 2003 11. Vasen HF, Mecklin JP, Khan PM, et al: The International Collaborative Group on Hereditary Non-Polyposis Colorectal Cancer (ICG-HNPCC). Dis Colon Rectum 34:424-425, 1991[CrossRef][Medline] 12. Liu T, Wahlberg S, Burek E, et al: Microsatellite instability as a predictor of a mutation in a DNA mismatch repair gene in familial colorectal cancer. Genes Chromosomes Cancer 27:17-25, 2000[CrossRef][Medline]
13. Syngal S, Fox EA, Eng C, et al: Sensitivity and specificity of clinical criteria for hereditary non-polyposis colorectal cancer associated mutations in MSH2 and MLH1. J Med Genet 37:641-645, 2000 14. Vasen HF, Watson P, Mecklin JP, et al: New clinical criteria for hereditary nonpolyposis colorectal cancer (HNPCC, Lynch syndrome) proposed by the International Collaborative group on HNPCC. Gastroenterology 116:1453-1456, 1999[CrossRef][Medline]
15. Rodriguez-Bigas MA, Boland CR, Hamilton SR, et al: A National Cancer Institute Workshop on Hereditary Nonpolyposis Colorectal Cancer Syndrome: Meeting highlights and Bethesda Guidelines. J Natl Cancer Inst 89:1758-1762, 1997
16. Wijnen JT, Vasen HF, Khan PM, et al: Clinical findings with implications for genetic testing in families with clustering of colorectal cancer. N Engl J Med 339:511-518, 1998 17. Scott RJ, McPhillips M, Meldrum CJ, et al: Hereditary nonpolyposis colorectal cancer in 95 families: Differences and similarities between mutation-positive and mutation-negative kindreds. Am J Hum Genet 68:118-127, 2001[CrossRef][Medline]
18. Syngal S, Fox EA, Li C, et al: Interpretation of genetic test results for hereditary nonpolyposis colorectal cancer: Implications for clinical predisposition testing. JAMA 282:247-253, 1999 19. Wijnen J, Khan PM, Vasen H, et al: Majority of hMLH1 mutations responsible for hereditary nonpolyposis colorectal cancer cluster at the exonic region 15-16. Am J Hum Genet 58:300-307, 1996[Medline] 20. Liu T, Wahlberg S, Rubio C, et al: DGGE screening of mutations in mismatch repair genes (hMSH2 and hMLH1) in 34 Swedish families with colorectal cancer. Clin Genet 53:131-135, 1998[Medline]
21. Ji HP, King MC: A functional assay for mutations in tumor suppressor genes caused by mismatch repair deficiency. Hum Mol Genet 10:2737-2743, 2001 22. Cravo M, Lage P, Albuquerque C, et al: BAT-26 identifies sporadic colorectal cancers with mutator phenotype: A correlative study with clinico-pathological features and mutations in mismatch repair genes. J Pathol 188:252-257, 1999[CrossRef][Medline] 23. Watson P, Vasen H, Aarnio M, et al: Incidence of extra-colonic/endometrial cancers in hereditary non-polyposis colorectal cancer. Familial Cancer 2:23-24, 2003 Submitted November 14, 2003; accepted September 24, 2004.
This article has been cited by other articles:
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
|||||||||||
|
|||||||||||
|
|
Copyright © 2004 by the American Society of Clinical Oncology, Online ISSN: 1527-7755. Print ISSN: 0732-183X
|
TOP
ABSTRACT
INTRODUCTION
2 test for categoric variables and the Wilcoxon rank sum test for continuous variables. Multivariable analysis was performed by logistic regression using a stepwise approach to identify independent predictors of HNPCC. The 11 explanatory variables used (
