Editor—A report was recently published on the localisation of a chromosome segment at 1p36 which appeared to be linked (two point lod score=4.74) to a large number of families with multiple cases of early onset (mean age at diagnosis of <66 years) prostate cancer (PC) in which a brain tumour had been reported in a first or second degree relative of a PC case.1 This result is consistent with epidemiological evidence suggesting a familial relationship between brain cancer and PC as well as numerous studies of LOH at 1p36 in brain tumours.1 As part of the ACTANE (Anglo/Canadian/Texan/Australian/Norwegian/EU Biomed) familial PC Consortium, we have genotyped 207 multiple case PC families for five 1p36 tetranucleotide repeat polymorphisms, which were used by Gibbset al,1 and performed linkage analysis using GENEHUNTER2 with the following genetic map3: D1S1160 - 3.835 cM - D1S1597 - 4.339 cM - D1S407-2.611 cM - GATA29A05 (=D1S3669) - 6.520 cM - D1S552. In addition to our interest in assessing our families for evidence of linkage of prostate and primary brain cancer to this region, we also wanted to determine if any other cancer site(s) might be associated with a susceptibility locus in this region. To this end, the family histories of all cancers were abstracted from the databases of several Consortium members and included in the analysis.

Table 1 presents the characteristics of the families; details on how they were ascertained are included in the footnotes. The criteria used for the prostate cancer familial clusters in this study are more relaxed than those suggested by the Hopkins group (referred to as the “Hopkins criteria” in the field).4 The linkage analysis results of nine prostate-brain cancer families were partitioned according to mean age at diagnosis of PC in the family; these results are shown in table 2. (Only two of these brain cancers have been confirmed by pathology reports; a glioma diagnosed at 67 years in a man who had three brothers with PC, mean age of diagnosis 66.7 years, and a glioblastoma diagnosed at 69 years in a family with five cases of PC, mean age at diagnosis of 68.2 years). In the total set of nine families, all lod, NPL, and hlod scores maximised at D1S1160. Overall, we found no evidence of linkage of prostate-brain cancer to this region by either parametric (maximum lod score=−0.06) or non-parametric (NPL=0.25, p=0.39) analysis. The maximum hlod was 0.07 at an alpha (proportion of families linked) of 48%. This latter estimate is consistent with the observation that four families had positive lod scores and four had negative scores (one family was uninformative, lod=0.0). Partitioning the families by mean age at diagnosis of PC resulted in suggestive, but not significant differential linkage with five early onset families (mean age at diagnosis <66 years) providing a maximum lod score of 0.48, whereas the remainder of the families appeared unlinked according to both lod score (−0.54) and NPL score (−0.20). These results are consistent with the lod scores of Gibbs et al 1 for families with a history of brain tumours when partitioned for average age at onset of prostate cancer, 3.65 (<66 years) and –1.84 (⩾66 years).

We then generated lod and NPL scores for all 207 families and sorted them according to these linkage results as well as by their mean age at PC diagnosis. In no case did a family history of any cancer at another site appear to cluster or be associated with linkage to 1p36 or with mean age at PC diagnosis. This analysis included examination by number and mean age at diagnosis of breast cancer in first degree relatives and all relatives, by number and mean age at diagnosis of ovarian cancers, number and mean age at diagnosis of colorectal cancer, stomach and pancreatic cancer, malignant melanoma, or uterine cancer. Although none of the family history groupings gave a positive lod score generally, they gave a positive score when the mean age at onset of PC was early. We therefore hypothesised that it might not be the family history of brain cancer that was responsible for positive linkage to 1p36 but, instead, the family history of early age at PC diagnosis per se.

Table 3 presents the linkage analysis results for the entire ACTANE pedigree set subdivided according to mean age at diagnosis of affected men in the family. The four age groups presented in table 3 were chosen to give approximately equal representation and are listed from the youngest (⩽59 years) to the oldest (⩾80 years). The large negative lod score for all the families, −9.66, rejects overall linkage to this region under the presumed genetic model of Carteret al.4 However, NPL score (1.02, p=0.15) and maximum hlod (0.93, alpha=0.24) are consistent with the possibilities that either PC in a proportion of these families is the result of an autosomal dominant gene located at 1p36, or that alternative genetic models better explain the excess allele sharing among men with PC. Recessive and X linked models have not yet been tested in our family set. Over the three subgroups of families with the earliest mean age at diagnosis, the sequential maximum lods (0.49, −2.89, −5.78), NPL scores (1.60, 1.02, 0.13), and heterogeneity lod scores (1.17, 0.59, 0.00) all indicate a greater possibility of linkage the earlier the mean age at PC diagnosis. The NPL score in the group with the earliest mean age at diagnosis is close to being nominally significant (p=0.06), whereas for the other groups this score is not significant (p=0.15, 0.44, and 0.37, respectively). Also, as mean age of PC diagnosis increased, the estimated proportion of linked families decreased from 62% to 30% to 0%, also consistent with a general effect related to age at diagnosis. The group with the latest age at diagnosis does not appear to follow this pattern, but there are only 48 families in this class, and it might be expected that there would be a higher proportion of phenocopies in the upper liability class and this would decrease the power to detect linkage in this group. Although the reasons for this inconsistency are at present unclear, based on the results in the three youngest age groups, we feel that early age at diagnosis remains associated with an increased probability of linkage.

Our families provided no evidence that the putative familial prostate cancer locus, CAPB, at 1p361 is linked with primary brain tumours, or indeed cancer at any site other than the prostate. Although the five brain-prostate families with early mean age at diagnosis of PC did partition overall lod and NPL scores, we consider that because of the small number of families and the appearance of an age linkage score association when the families were subdivided by cancer at any site it remains unwarranted to postulate that susceptibility to brain tumours is increased by the inheritance of an altered gene in the 1p36 region. We found a possible overall association of linkage scores with mean age at diagnosis since families with earlier onset disease gave higher scores. Although not statistically significant, this result is similar to that found for HPC15 and PCAP6 as well as with other familial cancers such as breast and ovarian cancer.7 We note that table 6 of Gibbs et al 1 suggested the possibility of linkage in 63 early onset PC families in that the lod scores were positive when the recombination fraction was 20% or more. However, linkage was tested only at D1S407 and hlod scores were not reported. Our maximum lod, NPL, and hlod scores generally occurred 5 to 9 cM centromeric to this position, so further analysis in this region in their PC family set would be of interest. There are several possibilities for the failure to detect linkage with the prostate cancer/primary brain phenotype in our families. The first is that the analysis by Gibbset al 1 may have led to a false finding arising from multiple subgroup analysis, which always has the risk of showing significant association by chance alone. The other, more interesting possibility, from the gene hunting perspective, is that it is not the primary brain phenotype per se that is linked to this region, but rather early onset PC, and it is this association with early onset disease which is important.