AUTOSOMAL DOMINANT INHERITANCE OF PROSTATE CANCER: A CONFIRMATORY STUDY.

BAS A. J. VERHAGE, AGNES B. BAFFOE-BONNIE, LAURA BAGLIETTO, DEBORAH S. SMITH, JOAN E. BAILEY-WILSON, TERRI H. BEATY, WILLIAM J. CATALONA, AND LAMBERTUS A. KIEMENEY

Objectives. To confirm, in a study of a large, independent cohort of families with prostate cancer, the findings of three segregation analyses that have suggested the existence of an inherited form of prostate cancer with an autosomal dominant inheritance mode.

Methods. Between January 1991 and December 1993, 1199 pedigrees were ascertained through single, unrelated, prostate cancer probands who presented for radical prostatectomy at the Division of Urologic Surgery, Washington University Medical Center in St. Louis, Missouri. Maximum likelihood segregation analysis was used to test specifically for mendelian inheritance of prostate cancer.

Results. Segregation analyses revealed that the familial aggregation of prostate cancer can be best explained by the autosomal dominant inheritance of a rare (q = 0.0037) high-risk allele. According to the best-fitting autosomal dominant model, 97% of all carriers will be affected by 85 years of age compared with 10% of noncarriers. Furthermore, the autosomal dominant model predicts that the high-risk allele accounts for a large proportion (65%) of all patients diagnosed with prostate cancer before 56 years of age. However, of all prostate cancer cases, a relatively small proportion is inherited (8% by 85 years old).

Conclusions. These results are in agreement with earlier reports of segregation analyses of prostate cancer and strengthen the evidence that prostate cancer is inherited in a mendelian fashion within a subset of families. UROLOGY 57: 97-101, 2001. © 2001, Elsevier Science Inc.

This work was supported by grant 97-1490 from the Dutch Cancer Society.

After non-melanoma skin cancer, prostate cancer (PCa) is the most common malignant cancer among men living in the United States. Other than age, race, and family history, no clear risk factors for the disease are evident. Numerous studies have reported on the familial aggregation of PCa.¹ Overall, first-degree relatives of patients with PCa have a two to three times increased risk of PCa. According to three segregation analyses, the aggregation of PCa in families can be best explained by mendelian autosomal dominant inheritance^2-4 In 1992, Carter and colleagues² from Johns Hopkins University in Baltimore were the first to report that familial PCa may be caused by a rare (q = 0.003), highly penetrant, dominant gene (or genes). Their model predicted that 88% of the gene carriers will develop PCa by 85 years of age. Furthermore, it was suggested that this inherited form of PCa may account for a significant proportion of early onset PCa (43% of patients 55 years old or younger at diagnosis), but only a small proportion of PCa cases overall (9% by age 85)} In 1997, Grönberg et al.³ from Umeå University in Sweden and Schaid et al.⁴ from the Mayo Clinic in Rochester, Minnesota confirmed these results; Grönberg and colleagues reported a high-risk allele inherited in a dominant fashion, with a higher population frequency (q = 0.0167) and a moderate lifetime penetrance (63%). Schaid et al. reported a population frequency for the autosomal dominant gene of q = 0.006 and a risk of PCa of 89% among carriers by 85 years. Other investigators, however, suggested that an X-linked or recessive genetic model might be more accurate. This suggestion was based on the finding that brothers have a higher risk of PCa than do fathers and sons.^1,5 The purpose of the present study was to determine whether the results from the previous complex segregation analyses could be confirmed in a large, independent cohort of families with PCa.

SUBJECTS
Between January 1991 and December 1993, 1345 men underwent radical prostatectomy for Stage T1-T2 PCa at the Division of Urologic Surgery, Washington University Medical Center, St. Louis, Missouri. Detailed family history and demographic data were collected in 1994 by telephone interview or postal questionnaire. For nonaffected, first-degree relatives, their current age or age at death was also recorded. Confirmation that all consecutive patients between January 1991 and December 1993 were identified was obtained by cross-referencing the Division of Urologic Surgery billing records for this period against the patient list. Patients for whom a family history questionnaire was not available were contacted by phone or mailed questionnaire. A more detailed description of the data collection has been reported elsewhere.⁶

In an earlier report from this study, it was shown that men with a family history of PCa are at a significantly increased risk. The relative risk for PCa was 3.5 (95% confidence interval 2.4 to 5.0) in the case of an affected father and 4.7 (95% confidence interval 3.0 to 7.5) in the case of an affected brother.⁶

SEGREGATION ANALYSIS
To test specifically for mendelian inheritance of PCa in the collected pedigrees, maximum likelihood segregation analyses were performed on the age at diagnosis, expressed as a censored trait, using the Statistical Analysis for Genetic Epidemiology (SAGE) program REGTL, third release.⁷ Using this program, one can either assume that the presence (or absence) of the putative disease allele influences the age at onset (model 1) or influences the susceptibility (model 2). The age at diagnosis was used as a proxy for the age at onset.

For this analysis, model 1 was chosen, using the class A regressive model.⁸ Age at onset for PCa is assumed to follow a logistic distribution described by two parameters, a and ß, with the probability distribution function according to Elston et al.⁹ Thus,

PCa was defined as a dichotomous variable (Y), where y = 1 if affected and y = 0 if unaffected (censored). The parameters estimated in the analysis included qA, the frequency of the putative high-risk allele A; ß_i, baseline parameters, where i represents an individual's type (AA, AB, BB); a, the age coefficient; and y, the susceptibility coefficient. The susceptibility parameter y describes the cumulative probability that a man develops PCa if he lives long enough.^1O Susceptibility y was fixed at 0 for females.

The probability (or penetrance) that an individual will be affected by a certain age was calculated for each type as follows:

The likelihood ratio test was used to test each of the five models against the general, unrestricted model and was computed as minus twice the natural log likelihood [-2In(L)] of the general model subtracted from that for a restricted submodel. This difference is asymptotically distributed as a chi-square distribution, with degrees of freedom equal to the difference in the number of parameters estimated in the two models. To correct for ascertainment bias, the likelihood of each pedigree was conditioned on the proband's affection status using his age at diagnosis as recorded.^13,14

POPULATION CHARACTERISTICS
The 1199 pedigrees contained 3965 males (of whom 1567 were brothers and 1199 were fathers) and 2526 females. Age (or the age at diagnosis in the case of an affected nonproband) was missing for 712 subjects (11% of the total cohort). Of these 712 subjects, 334 (47%) were female and 378 were male (53%). Three percent of the 378 males (n = 11) were affected nonprobands.

Among all participants, 93% were white, 5% were African American, and 2% were Asian or Hispanic. Because of the relatively small number of nonwhite families, we decided not to conduct any stratified analyses.

Of the 1199 probands used in the analyses, 218 (18%) had a positive family history: 113 probands had an affected father only; 75 probands had 1 affected brother only; 15 probands had an affected father and an affected brother; 11 probands had 2 affected brothers; 2 probands had an affected father and 2 affected brothers; and 2 probands had 3 affected brothers.

As expected, the mean age at PCa diagnosis in the affected nonprobands was significantly higher than the mean age at diagnosis of the radical prostatectomy probands (69.4 + 8.9 years versus 65.0 + 6.3 years, P <0.001). The geometric mean age at diagnosis for affected nonprobands was also 69.4 years.

SEGREGATION ANALYSIS
The parameter estimates and test statistics from the segregation analyses are given in Table I. The two nonmendelian models did not fit the data. The "no major gene" model, in which q was fixed at 1.0, was clearly rejected when compared with the general unrestricted model (chi-square test="29.18," P <0.001). The environmental model, which had equal transmission parameters for all three types of individuals (AA, AB, and BB), also differed significantly from the general model (chi-square test="8.34," P="0.015)." The major gene models were tested with fixed values of 1.0, 0.5, and 0.0 for the transmission parameters AA, AB, and BB, respectively. Of these mendelian models, the recessive model, defined by ßAA ßAB = ßBB," was clearly rejected (chi-square test = 24.92, P= 0.001). The codominant model, allowing for three genotype-specific age-at-onset distributions, fit the data well (chi-square test = 6.74, P = 0.23). The autosomal dominant model, assuming ßAA = ßAB ßBB, proved to fit the data best (chi-square test = 3.74, P = 0.43). The codominant model did not give an improvement in fit over the autosomal dominant model (chi-square test = 3.00, P >0.05).

The allele frequency in the autosomal dominant model was estimated to be 0.0037. The estimated cumulative risk (penetrance) of PCa for types AA/AB ("carriers") and BB ("noncarriers"), as predicted by the autosomal dominant model, is given in Figure 1. The dotted lines show the predicted cumulative risks of PCa in the Johns Hopkins study.

The results of this report show that familial aggregation of PCa can best be explained by an autosomal dominant inheritance of a rare but highly penetrant high-risk allele. The estimated gene frequency of 0.0037 implies that 0.74% of the St. Louis population would carry the putative gene. Carriers of this susceptibility gene have a much earlier age at diagnosis compared with noncarriers (Fig. 1).

These findings are consistent with previous reports^2-4 on the mode of inheritance of PCa. The methods of ascertainment (through a single proband undergoing radical prostatectomy for clinically localized PCa) and segregation analyses we used in this study were similar to those of Carter et al from the Johns Hopkins University and Schaid et al.⁴ from the Mayo Clinic.

Patients with PCa in this study were diagnosed at a similar age as the probands in the study of Schaid et al. (65.3 versus 65.6 years, respectively), but were older than the probands in the study by Carter et al. (59.3 years). The frequency of the high-risk allele reported here was equal to the frequency Carter et al. found but lower than that found by Schaid et al. (q = 0.0060).

The autosomal dominant model predicted that the high-risk allele is responsible for a large proportion (65%) of all patients diagnosed with PCa before 56 years of age. Of all PCa cases, a relatively small proportion is due to the inherited form (8% by age 85). Although the standard errors of parameter estimates from segregation analysis were large, these proportions seem to be higher than those reported by Carter et al. for early onset PCa (65% versus 43%) and approximately similar for total PCa occurrence (8% versus 9%).

In Figure 1, the predicted cumulative risks of PCa are given for carriers and non-carriers (based on the autosomal dominant model). Carriers of the high-risk allele had a cumulative probability of being affected by the age of 55, 70, and 85 years of 12.2%, 73.4%, and 96.8%, respectively. The penetrance of the high-risk allele in the Johns Hopkins study for the age of 55, 70, and 85 years was 4%, 41 %, and 88%, respectively. Thus, the estimated age-specific penetrances for carriers appear somewhat higher compared with the findings in the Johns Hopkins study. Noncarriers had a cumulative probability of being affected by the age of 55, 70, and 85 years of 0.01 %, 0.2%, and 9.6%, respectively. The corresponding percentages in the Johns Hopkins study were 0.03%, 0.5%, and 6.8%. The somewhat higher risk estimates in our study compared with those from the Johns Hopkins study may be because the data in the Johns Hopkins study were collected before the prostate-specific antigen era.

The estimated lifetime penetrance among non-carriers, who constitute the vast majority of all patients with PCa, is somewhat lower than the population-based PCa risk reported by the National Cancer Institute SEER program.¹⁵ The most likely explanation for this discrepancy is that older brothers and fathers of probands died before the widespread use of prostate-specific antigen screening in the United States.

Linkage studies of highly selected PCa pedigrees have shown that several loci, on chromosomes 1^16-20 and X,²¹ are candidates for the first PCa susceptibility gene. Recent linkage analyses suggest that chromosomes 16²² and 20²³ may harbor susceptibility loci as well. Although none of these PCa susceptibility genes have been cloned yet, it is estimated that in total only 5% to 10% of all PCa is accounted for by these hereditary factors.²⁴ This implies that among multiplex families with PCa, it is likely that phenocopies (ie, affected individuals not carrying the high-risk allele) occur. In linkage analysis, this constitutes a problem, as the discrimination between genetic and sporadic cases is based on the age at diagnosis alone.

Keetch et al.,²⁵ using in part the same population as in the present report, found that the mean Gleason score ( + standard deviation) in 50 men with sporadic cancer was 6.2 + 1 compared with 5.6 + 0.9 in those with "hereditary disease" (P = 0.008). No other substantial clinical or pathologic differences were evident.²⁵ Others²⁶ did not find any clear distinct clinical features of hereditary cases either, which further hampers the search for PCa susceptibility genes in linkage analysis. Grönberg et al.²⁷ reported that men from families potentially linked to the HPCl gene had a significantly lower mean age at diagnosis (63.7 versus 65.9 years), more higher grade (grade 3) cancer (39% versus 29%), and more often had advanced disease (41% versus 31 %) than men from unlinked families. However, others challenged these results and suggested that ascertainment bias occurred, leading to a more advanced stage.^28,29

Although family members in, for example, hereditary breast cancer families and hereditary non-polyposis colon cancer families have an increased risk of several types of cancer, including PCa,^3O other types of cancer do not seem to cluster in HPC families.^6,31 Only modest increases in the occurrence of cancer of the breast, colon, and brain were seen in first-degree relatives of patients with PCa.^31,32 Isaacs et al.³¹ showed that members of HPC families have an increased relative risk of tumors of the central nervous system (relative risk = 3.02; 95% confidence interval 1.08 to 8.41). Gibbs et al.²⁰ recently reported a potential hereditary PCa susceptibility locus on chromosome 1p36. They suggest that the excess of brain cancer within a subset of families with a high risk of PCa, as well as the frequent loss of heterozygosity at 1p36 in brain cancer, is a result of the same genetic cause.

In the present study, nuclear families were ascertained through a single patient with PCa who presented at the Division of Urologic Surgery, Washington University, for radical prostatectomy. The eligibility criteria for this surgery are age, confined disease, and good general health. Because of this ascertainment, our results may not apply to all ages and stages of disease. Furthermore, this study was unable to address the issue of race heterogeneity because of the relatively small strata of nonwhites (5% African American and 2% Asian or Hispanic).

In summary, our findings strengthen the statistical evidence of an autosomal dominant mode of inheritance of PCa.

1. Bishop DT, and Kiemeney LALM: Family studies and the evidence for genetic susceptibility to prostate cancer. Semin Cancer Biol 8: 45-51, 1997.

2. Carter BS, Beaty TH, Steinberg GD, et al: Mendelian inheritance of familial prostate cancer. Proc Natl Acad Sci USA 89: 3367-3371, 1992.

3. Grönberg H, Damber L, Damber JE, et al: Segregation analysis of prostate cancer in Sweden: support for dominant inheritance. Am J Epidemiol 147: 552-557, 1997.

4. Schaid JS, McDonnell SK, Blute ML, et al: Evidence for autosomal dominant inheritance of prostate cancer. Am J Hum Genet 62: 1425-1438,1998.

5. Monroe KR, Yu MC, Kolonel LN, et al: Evidence of an X-linked or recessive genetic component to prostate cancer risk. Nat Med 1: 827-829, 1995.

6. Keetch DW, Rice JP, Suarez BK, et al: Familial aspects of prostate cancer: a case control study.J Urol 154: 2100-2102, 1995.

7. Statistical Analysis for Genetic Epidemiology, Release 3.0. Computer package available from the Department of Epidemiology and Biostatistics, Case Western Reserve University, Cleveland, Ohio, 1997.

8. Bonney GE: Regressive logistic models for familial diseases and other binary traits. Biometrics 42: 611-625, 1986.

9. Elston RC, and George VT: Age at onset, age at examination,and other covariates in the analysis of family data. Genet Epidemiol 6: 217-220, 1989.

10. Go RCP, King MC, Bailey-Wilson J, et al: Genetic epidemiology of breast cancer and associated cancers in high-risk families. I. Segregation analysis. J Natl Cancer Inst 71: 455- 461, 1983.

11. Box GEP, and Cox DR: An analysis of transformations. J R Stat Soc B 26: 211-252.1964.

12. Elston RC, and Stewart J: A general model for the genetic analysis of pedigree data. Hum Hered 21: 523-542, 1971.

13. Cannings C, and Thompson EA: Ascertainment in the sequential sampling of pedigrees. Clin Genet 12: 208-212. 1977.

14. Elston RC, and Sobel E: Sampling considerations in the gathering and analyses of pedigree data. Am J Hum Genet 27: 31-45. 1979.

15. Ries LAG, Kosary CKL, Hankey BF, et al: SEER Cancer Statistics Review, 1973-1996. :Bethesda. National Cancer Institute, 1999.

16. Smith JR, Freije D, Carpten JD, et al: Major susceptibility locus for prostate cancer on chromosome 1 suggested by a genome-wide search. Science 274: 1371-1374, 1996.

17. Berthon P, Valeri A, Cohen-Akenine A, et al: Predisposing gene for early-onset prostate cancer, localized on chromosome 1q42.2-43. Am J Hum Genet 62: 1416-1424, 1998.

18. Hsieh CL, Oakley-Girvan I, Gallagher RP, et al: Re: prostate cancer susceptibility locus on chromosome 1q: a confirmatory study. J Natl Cancer Inst 89: 1893-1894. 1997.

19. Cooney KA, McCarthy JD, Lange E, et al: Prostate cancer susceptibility locus on chromosome 1q: a confirmatory study. J Natl Cancer Inst 89: 955-959, 1997.

20. Gibbs M, Stanford JL, McIndoe RA. et al: Evidence for a rare prostate cancer-susceptibility locus at chromosome Ip36. Am J Hum Genet 64: 776-787. 1999.

21. Xu J, Meyers D, Freije D, et al: Evidence for a prostate cancer susceptibility locus on the X chromosome. Nat Genet 20: 175-179, 1998.

22. Suarez BK, Lin J, Burmester JK, et al: A genome screen of multiplex sibships with prostate cancer. Am J Hum Genet 66: 933-944, 2000.

23. Berry R, Schroeder JJ, French AJ, et al: Evidence for a prostate cancer-susceptibility locus on chromosome 20. Am J Hum Genet 67: 82-91. 2000.

24. Carter BS, Bova GS, Beaty TH, et al: Hereditary prostate cancer: epidemiologic and clinical features. J Urol l50: 797- 802, 1993.

25. Keetch DW, Humphrey PA, Smith DS, et al: Clinical and pathological features of hereditary prostate cancer. J Urol 155:1841-1843,1996.

26. Bastacky SI, Wojno KJ, Walsh PC, et al: Pathological features of hereditary prostate cancer. J Urol 153: 987-992, 1995.

27. Grönberg H, Isaacs SD, Isaacs MS, et al: Characteristics of prostate cancer in families potentially linked to the hereditary prostate cancer 1 (HPCl) locus. JAMA 278: 1251-1255, 1997.

28. Laniado ME: Prostate cancer potentially linked to the HPC1 gene (letter). JAMA 279: 507, 1998.

29. McClellan MW: Prostate cancer potentially linked to the HPCl gene (letter). JAMA 279: 507-508, 1998.

30. Ford D, Easton DF, Bishop DT, et al, for the Breast Cancer Linkage Consortium: Risks of cancer in BRCAl-mutation carriers. Lancet 343: 692-695.1994.

31. Isaacs SD, Kiemeney LALM, Baffoe-Bonnie A, et al: The risk of cancer in relatives of prostate cancer probands. J Natl Cancer Inst 87: 991-996, 1995.

32. Goldgar DE, Easton DF, Cannon-Albright LA, et al: Systematic population-based assessment of cancer risk in first-degree relatives of cancer probands. J Natl Cancer Inst 86: 1600-1608,1994.