Correspondence
Familial Melanoma and Pancreatic Cancer
N Engl J Med 1996; 334:469-472February 15, 1996
- Article
To the Editor:
The studies by Goldstein and Whelan and their colleagues (Oct. 12 issue)1,2 indicate that p16 mutations are required for the development of pancreatic cancer in 10 melanoma-prone kindreds in the United States, Australia, and the Netherlands, which were previously characterized with respect to chromosome 9p21 by linkage analysis. We have found evidence relating the risk of pancreatic adenocarcinoma to a p16 mutation in bloodline members of melanoma-prone Italian families.
So far, within a small geographic area of Italy (possibly because of a founder effect), we have detected the same Gly93Trp mutation in seven apparently unrelated families and in none of 50 control persons. Nineteen cases of melanoma and three of dysplastic nevi (all histologically confirmed) were diagnosed at ages ranging from 21 to 70 years (median, 37) in the kindreds with the p16 Gly93Trp mutation. In addition, 15 cancers at other sites have been found in these kindreds, including 3 pancreatic cancers but no gastric cancers. The pancreatic tumors developed in members of three different families at the ages of 48, 51, and 60 years. The pedigree of one family with p16M is shown in Figure 1Figure 1
Pedigree of an Italian Kindred with the p16M Mutation.. In comparison, 18 cases of melanoma and 4 of dysplastic nevi (diagnosed at a median age of 42.5 years) and 9 cancers at other sites, including 4 stomach cancers but no pancreatic cancers, were detected in seven families with p16W. In two of the latter kindreds, there is strong evidence of linkage to 9p21. Each family with p16M or p16W was referred to our genetic counseling center because of suspected familial melanoma. The data support an association between the risk of pancreatic cancer and impairment of p16 function and suggest a possible role of the cyclin-dependent–kinase inhibitor 2 (CDKN2) region in the organ specificity of tumorigenesis.2 ReferencesPaola Ciotti, Ph.D.
Institute of Biology and GeneticsPaolo Strigini, M.D., M.P.H.
National Institute of Cancer ResearchGiovanna Bianchi-Scarrà, Ph.D.
Institute of Biology and Genetics, 16132 Genoa, Italyfor the Ligurian Skin Tumor Study Group
1
Goldstein AM ,Fraser MC ,Struewing JP , et al. Increased risk of pancreatic cancer in melanoma-prone kindreds with p16INK4 mutations. N Engl J Med 1995;333:970-974
Full Text | Web of Science | Medline2
Whelan AJ, Bartsch D, Goodfellow PJ. A familial syndrome of pancreatic cancer and melanoma with a mutation in the CDKN2 tumor-suppressor gene. N Engl J Med 1995;333:975-7
To the Editor:
Goldstein et al. state that there is a “striking difference” in the risk of pancreatic tumors in 10 melanoma-prone kindreds with protein-impairing p16 INK4 mutations (the group of kindreds with p16M alleles) and 9 with no impairment (the group with p16W alleles),1 and that this difference appears in both prospective and combined prospective and retrospective periods of observation. However, the methods they used raise some serious questions. First, several types of tumors were investigated for an association with the p16 mutation, without clear prior hypotheses. This raises the issue of multiple testing, since no corrections of the reported P values are discussed.
Even with the assumption of a prior hypothesis about pancreatic cancer, simple conditional tests indicate that the evidence fails to meet even modest standards of statistical significance. Second, we believe that an inappropriate argument was used to compare the risk in the prospective period. Third, the reported statistical analyses assume the independence of individual family members without regard to family structure. We believe the analysis should reflect the fact that the independently sampled units are kindreds, not individuals.
In the prospective period, the incidence of pancreatic cancers was examined in relation to the incidence in the general population, with two cases occurring in different kindreds with p16M alleles (P<0.05) and none in the kindreds with p16W alleles (P not significant). The authors conclude from this finding that there was a striking difference in risk, although the reported confidence intervals overlap considerably. The two risk estimates should be compared directly, as was done in the combined prospective–retrospective analysis. The evidence can be assessed as follows: with 10 kindreds with p16M alleles, 9 kindreds with p16W alleles, and 2 kindreds with pancreatic cancer (regardless of the group), what is the chance that the affected kindreds will belong to the same group? The result is obviously not significant, or even suggestive. An application of Fisher's exact test to the hypergeometric data yields a P value of 0.4737. This argument in fact underestimates the true P value, since the average number of person-years of observation was higher in the kindreds with p16M alleles.
The authors also report seven pancreatic cancers in the kindreds with p16M alleles and none in the kindreds with p16W alleles when the retrospective data are considered. Note that these cancers occurred in only 4 of 19 kindreds, an event with a probability of 0.038 under an independent assignment of cancers to kindreds of equal sizes. Fisher's exact test applied to the four kindreds yields a P value of 0.0867, as compared with the reported P value of 0.02.
1 ReferencesFred A. Wright, Ph.D.
Ronald G. Thomas, Ph.D.
University of California, San Diego, La Jolla, CA 920931
Hussussian CJ ,Struewing JP ,Goldstein AM , et al. Germline p16 mutations in familial melanoma. Nat Genet 1994;8:15-21
CrossRef | Web of Science | Medline
To the Editor:
The articles by Goldstein et al. and Whelan et al. address the association of pancreatic cancer and cutaneous malignant melanoma in certain families with mutations of the CDKN2 gene.
In 1990 we reported clinical evidence of this association in nine large Dutch kindreds with the familial atypical multiple-mole melanoma syndrome.1 Besides pancreatic carcinoma (standardized incidence ratio, 13.4), other types of gastrointestinal cancers were found in excess (standardized incidence ratio, 3.3). These tumors were not distributed in the gastrointestinal tract in the pattern that would be predicted from Dutch incidence data, a pattern in which about half the gastrointestinal cancers are colorectal and approximately 10 percent are pancreatic. In these families 9 of 18 gastrointestinal cancers were pancreatic and none were colorectal.
Recent data show that all families in which pancreatic carcinomas were found have a 19-bp germ-line deletion in the CDKN2 gene, impairing the function of the p16 protein.2 Analysis of primary pancreatic tumor tissue from one of the family members who was a carrier of the 19-bp CDKN2 deletion showed the retention of the mutated allele and the loss of the wild-type allele. This finding suggests the involvement of the CDKN2 mutation in the development of the tumor.
However, three of the nine families with familial atypical multiple-mole melanoma, which are large enough to express the gastrointestinal cancer trait, have the same 19-bp deletion3 without any excess incidence of systemic cancers. In two of the nine families, no CDKN2 mutations have been found so far, whereas linkage data are consistent with the linkage of the disease trait to chromosome 9p21. The results in these Dutch families suggest that the inconsistent occurrence of other cancers cannot be explained by the kind of CDKN2 mutation, but instead must be due to the influence of other factors or genes.
These families are under surveillance for early detection of melanoma. But knowing the increased risk of gastrointestinal cancer in such families presents us with the problem of not being able to detect pancreatic carcinomas early. We wonder how Goldstein and Whelan and their colleagues are going to inform such families and manage their care with regard to pancreatic and other carcinomas of the digestive tract.
3 ReferencesWilma Bergman, M.D., Ph.D.
Nelleke Gruis, Ph.D.
University of Leiden, 2300 RC Leiden, the Netherlands1
Bergman W ,Watson P ,de Jonge J ,Lynch HT ,Fusaro RM . Systemic cancer and the FAMMM syndrome. Br J Cancer 1990;61:932-936
CrossRef | Web of Science | Medline2
Gruis NA ,Sandkuijl LA ,van der Velden PA ,Bergman W ,Frants RR . CDKN2 explains part of the clinical phenotype in Dutch familial atypical multiple-mole melanoma (FAMMM) syndrome families. Melanoma Res 1995;5:169-177
CrossRef | Web of Science | Medline3
Gruis NA ,van der Velden PA ,Sandkuijl LA , et al. Homozygotes for the CDKN2 (p16) germline mutation in Dutch familial melanoma kindreds. Nat Genet 1995;10:351-353
CrossRef | Web of Science | Medline
Author/Editor Response
The authors reply:
To the Editor: Ciotti et al. and Bergman and Gruis present data from several Italian and Dutch melanoma-prone kindreds that provide further evidence of an association between p16 mutations and pancreatic cancer and familial melanoma. The data, however, also reveal the complexity of melanoma: a number of putative 9p-linked kindreds do not carry detectable p16 mutations, only a subgroup of melanoma-prone kindreds with p16 mutations have pancreatic cancer, and in some families with identical p16 mutations, there are no cases of pancreatic cancer. Additional research will be required to resolve these questions and to determine the precise associations among p16 mutations, melanoma, pancreatic cancer, and possibly other systemic tumors.
Wright and Thomas are concerned about the methods we used and the conclusions we drew from our analyses. In the overall period, the absolute difference between the incidence rates in the kindreds with p16M alleles (7 cases of pancreatic cancer observed; 0.32 expected) and the kindreds with p16W alleles (0 cases observed; 0.27 expected) was 6.4 per 10,000 person-years (95 percent confidence interval, 0.9 to 10). In the more limited prospective period, the difference between the incidence rates in the kindreds with p16M alleles (2 cases observed; 0.15 expected) and the kindreds with p16W alleles (0 cases observed; 0.09 expected) was 8.9 per 10,000 person-years (95 percent confidence interval, -6.5 to 20).1,2
We cannot rule out chance as a reason for the reported association. We believe, however, that observing multiple cases of pancreatic cancer when less than one case is expected is itself too striking an observation not to report. As we mentioned in our discussion, the analyses are based on small numbers and require corroboration from studies of other melanoma-prone kindreds. Ciotti et al. and Bergman and Gruis provide further support for the reported relation.
Although a subgroup of melanoma-prone kindreds has an increased risk of pancreatic cancer, at this time we cannot identify the genotype (or genotypes) or the phenotype (or phenotypes) that are predictive of this excess risk. In addition, a surveillance or screening program is useful only when an intervention would have an effect on the outcome by either preventing the disease or detecting it at a curable stage. Neither of these options is currently applicable to pancreatic cancer. As additional information or new techniques for screening become available, we will modify our strategy for identifying persons at risk. Melanoma remains, however, the principal cause of morbidity and mortality in these families.
2 ReferencesAlisa M. Goldstein, Ph.D.
Margaret A. Tucker, M.D.
National Cancer Institute, Bethesda, MD 20892Author/Editor Response
We thank Bergman and Gruis for pointing out that in 1990 their group described the association between the familial atypical multiple-mole melanoma syndrome and pancreatic carcinoma and other gastrointestinal cancers.1 The family we investigated was ascertained to be a pancreatic-cancer–prone kindred.2 Our review of the literature focused on familial pancreatic carcinoma and not familial melanoma, which explains in part why we overlooked the 1990 report by Bergman and colleagues.1 Subsequent investigations by Gruis and colleagues have revealed that families prone to melanoma and pancreatic cancer have a CDKN2 mutation.3 Bergman and Gruis note that not all kindreds with the same CDKN2 mutation have an excess of cancers of the gastrointestinal tract. We agree with their suggestion that other environmental or genetic factors contribute to the development of gastrointestinal cancers in kindreds with CDKN2 mutations. This is even more clear when one considers the results of our study2 along with those of Hussussian and colleagues4 and Goldstein et al.5 In the family we investigated, pancreatic cancers developed in three of four carriers of the Gly93Trp mutation.2 Hussussian and colleagues4 described the same mutation in three melanoma-prone kindreds. Only one of these three kindreds had members in whom pancreatic cancers developed.5 In the other two families with the Gly93Trp mutation, there were a total of 10 carriers of the disease allele, none of whom had pancreatic cancer. This variable expressivity of the Gly93Trp mutation is largely consistent with that described by Bergman and Gruis in the case of another CDKN2 mutation.
We agree with Bergman and Gruis that the clinical treatment of kindreds with a CDKN2 mutation is problematic, since the penetrance of pancreatic carcinoma is not established and the efficacy of various screening approaches in the early diagnosis of pancreatic carcinoma is not known. In our kindred, the single member at risk consulted a medical geneticist and a gastroenterologist. Risk factors, including tobacco smoking and alcohol consumption, were discussed, as were surveillance options, including the use of computed tomography, magnetic resonance imaging, and endoscopic or abdominal ultrasonography. Prophylactic pancreatectomy was also discussed. The proband has elected to undergo endoscopic ultrasonography or magnetic resonance imaging every six months.
The description of the melanoma-prone kindreds followed by Ciotti and colleagues provides further evidence that there is an increased risk of pancreatic cancer in families with the CDKN2 Gly93Trp mutation and may speak to the organ specificity of tumorigenesis.
5 ReferencesAlison J. Whelan, M.D.
Washington University School of Medicine, St. Louis, MO 63110Detlef Bartsch, M.D.
Phillipps University, 35033 Marburg, GermanyPaul J. Goodfellow, Ph.D.
Washington University School of Medicine, St. Louis, MO 631101
Bergman W ,Watson P ,de Jong J ,Lynch HT ,Fusaro RM . Systemic cancer and the FAMMM syndrome. Br J Cancer 1990;61:932-936
CrossRef | Web of Science | Medline2
Whelan AJ ,Bartsch D ,Goodfellow PJ . A familial syndrome of pancreatic cancer and melanoma with a mutation in the CDKN2 tumor-suppressor gene. N Engl J Med 1995;333:975-977
Full Text | Web of Science | Medline3
Gruis NA ,Sandkuijl LA ,van der Velden PA ,Bergman W ,Frants RR . CDKN2 explains part of the clinical phenotype in Dutch familial atypical multiple-mole melanoma (FAMMM) syndrome families. Melanoma Res 1995;5:169-177
CrossRef | Web of Science | Medline4
Hussussian CJ ,Struewing JP ,Goldstein AM , et al. Germline p16 mutations in familial melanoma. Nat Genet 1994;8:15-21
CrossRef | Web of Science | Medline5
Goldstein AM ,Fraser MC ,Struewing JP , et al. Increased risk of pancreatic cancer in melanoma-prone kindreds with p16INK4 mutations. N Engl J Med 1995;333:970-974
Full Text | Web of Science | Medline
- Citing Articles (16)
Citing Articles
1
Julie Lang, Nicholas Hayward, David Goldgar, Hensin Tsao, David Hogg, Jane Palmer, Mitchell Stark, Edward S. Tobias, Rona MacKie. (2007) The M53I mutation inCDKN2A is a founder mutation that predominates in melanoma patients with Scottish ancestry. Genes, Chromosomes and Cancer 46:3, 277-287
CrossRef2
Kristin B. Niendorf, Hensin Tsao. (2006) Cutaneous melanoma: family screening and genetic testing. Dermatologic Therapy 19:1, 1-8
CrossRef3
J Andrew Carlson, Andrzej Slominski, Gerald P Linette, Martin C Mihm Jr, Jeffrey S Ross. (2003) Biomarkers in melanoma: predisposition, screening and diagnosis. Expert Review of Molecular Diagnostics 3:2, 163-184
CrossRef4
Michela Mantelli, Monica Barile, Paola Ciotti, Paola Ghiorzo, Francesca Lantieri, Lorenza Pastorino, Caterina Catrical
, Gabriella Della Torre, Ugo Folco, Paola Grammatico, Laura Padovani, Barbara Pasini, Dario Rovini, Paola Queirolo, Maria Luisa Rainero, Pier Luigi Santi, Roberto M. Sertoli, Alisa M. Goldstein, , , Giovanna Bianchi-Scarr
CrossRef5
Detlef Bartsch, Mercedes Sina-Frey, Sven Lang, Anja Wild, Berthold Gerdes, Peter Barth, Ralf Kress, Robert Grutzmann, Mario Colombo-Benkmann, Andreas Ziegler. (2002) Annals of Surgery 236:6, 730
CrossRef6
William B. Goggins, Dianne M. Finkelstein, Hensin Tsao. (2001) Evidence for an association between cutaneous melanoma and non-Hodgkin lymphoma. Cancer 91:4, 874-880
CrossRef7
Nicholas Hayward. (2000) New developments in melanoma genetics. Current Oncology Reports 2:4, 300-306
CrossRef8
Mark Harland, Elizabeth A. Holland, Paola Ghiorzo, Michela Mantelli, Giovanna Bianchi-Scarr
CrossRef9
Ramon M. Fusaro, Henry T. Lynch. (2000) The FAMMM Syndrome: Epidemiology and Surveillance Strategies. Cancer Investigation 18:7, 670-680
CrossRef10
Paola Ghiorzo, Paola Ciotti, Michela Mantelli, Abdelhamid Heouaine, Paola Queirolo, Maria Luisa Rainero, Carlo Ferrari, Pier Luigi Santi, Roberto De Marchi, Alessandro Farris, Franco Ajmar, Paolo Bruzzi, Giovanna Bianchi-Scarr
CrossRef11
Werner Hilgers, Scott E. Kern. (1999) Molecular genetic basis of pancreatic adenocarcinoma. Genes, Chromosomes and Cancer 26:1, 1-12
CrossRef12
Elizabeth A. Holland, Helen Schmid, Richard F. Kefford, Graham J. Mann. (1999) CDKN2A (P16INK4a) andCDK4 mutation analysis in 131 Australian melanoma probands: Effect of family history and multiple primary melanomas. Genes, Chromosomes and Cancer 25:4, 339-348
CrossRef13
J. Aitken, J. Welch, D. Duffy, A. Milligan, A. Green, N. Martin, N. Hayward. (1999) CDKN2A Variants in a Population-Based Sample of Queensland Families With Melanoma. JNCI Journal of the National Cancer Institute 91:5, 446-452
CrossRef14
Emanuel A. Yakobson, Avraham Zlotogorski, Rafael Shafir, Meir Cohen, Michael Icekson, Marina Landau, Sarah Brenner, Sally Usher, Hava Peretz. (1998) Screening for Tumour Suppressor p16(CDKN2A) Germline Mutations in Israeli Melanoma Families. Clinical Chemistry and Laboratory Medicine 36:8, 645-648
CrossRef15
Pamela M. Pollock, Nigel Spurr, Tim Bishop, Julia Newton-Bishop, Nelleke Gruis, Pieter A. van der Velden, Alisa M. Goldstein, Margaret A. Tucker, William D. Foulkes, Ray Barnhill, Daniel Haber, Jane Fountain, Nicholas K. Hayward. (1998) Haplotype analysis of two recurrentCDKN2A mutations in 10 melanoma families: Evidence for common founders and independent mutations. Human Mutation 11:6, 424-431
CrossRef16
Sophie Sun, Pamela M. Pollock, Ling Liu, Sepideh Karimi, Serge Jothy, Benedict J. Milner, Andrew Renwick, Norman J. Lassam, Nicholas K. Hayward, David Hogg, Steven A. Narod, William D. Foulkes. (1997) CDKN2A mutation in a non-FAMMM kindred with cancers at multiple sites results in a functionally abnormal protein. International Journal of Cancer 73:4, 531-536
CrossRef
