Cancer incidence in relatives of British Fanconi Anaemia patients

It has long been hypothesised that heterozygotes for autosomal recessive DNA repair disorders may have reduced efficiency of their DNA repair systems which could cause an increased risk of cancer [1]. This hypothesis has been upheld by molecular and epidemiological studies of cancer incidence in Ataxia Telangiectasia which have confirmed a two-fold increased risk of breast cancer in female ATM heterozygotes [2, 3]. Fanconi anemia (FA) is an autosomal recessive bone marrow failure syndrome due to mutations in at least 13 different genes (FANCA, B, C, D1/BRCA2, D2, E, F, G, I, J/BRIP1, L, M and N/PALB2) [4, 5] and the incidences of acute myeloid leukaemia and certain solid tumours are all greatly increased in homozygotes [6]. Epidemiological studies of cancer risk in FA heterozygotes are conflicting: the initial study by Swift et al. analysed 102 deaths in the relatives of eight FA families and found a higher rate of leukaemia and gastric, colorectal and tongue cancer [7]. However, none of these was statistically significant and when the study was expanded to 25 families no overall or specific excess of cancers could be demonstrated, and indeed there were fewer leukaemias than expected [8]. A separate study of 125 relatives in nine FA families also failed to show a difference [9]. While molecular studies investigating cancer risk in heterozygotes have mostly been inconclusive (reviewed in [10]), a recent molecular study of 944 relatives of FA probands did suggest an increased risk of breast cancer in FANCC mutation carriers [11].

In view of these conflicting data we carried out a study of British FA families to investigate this question further, since increases in cancer risk in FA heterozygotes would have implications for counselling FA family members, and possibly also for the implementation of preventative screening measures in FA carriers.

The observed versus expected incidence, p-value (Poisson 1-sided) and 95% confidence intervals of the most frequent cancers in the relatives of FA patients are given in Table 1. Head and neck cancer and leukaemias are also listed, as these are common in FA homozygotes. There were no observed cases of other cancers that commonly occur in FA homozygotes (oesophageal, vulval, anal), and other cancers had an observed incidence of 2 cases or less. The results show no significantly different relative risks for any of the cancers and, given the small numbers studied, there is a wide range for many of the associated confidence intervals. The overall observed number of cancers is very similar to the expected number, with a relative risk of 0.97 (0.71–1.23) for FA family members.

The relative risk of all cancer in relation to carrier probability is given in Table 2. There were no significant differences in relative risk across the groups, and the rates of all cancers were similar in each group, suggesting that there is little bias in the data. As expected, most of the cancers occurred in the 0.25 carrier probability group, which mainly consisted of great-grandparents and great uncles and aunts. The overall relative risk of all cancers in FA gene carriers was calculated to be 0.946 (95% confidence limits 0.61 – 1.46; Χ² = 0.0279, p = 0.867).

Analysis of relative risk for individual cancer types in each carrier probability group did not reveal any significant differences apart from an increased incidence of prostate cancer in obligate male carriers; 2 cases were observed in 33 individuals when the expected incidence was 0.17 (O/E = 11.5, 95% C.I. = 1.7 – 78.9, p = 0.013). When this was combined with prostate case incidences in other carrier groups there were a total of five cases and the overall relative risk of prostate cancer in carriers was calculated to be 3.089 (95% confidence limits 1.09 – 8.78; Χ² = 4.767, p = 0.029).

This is the first study of cancer predisposition in UK FA families and our data suggest that there is no overall increased cancer risk in these families. Sixty-five percent of the reported cancers could be confirmed independently, and the cancer-type misreporting rate by relatives was 5.5%. It was not known how many family cancers had been missed by not being reported by relatives, and this could be a potential source of under-ascertainment. Another source of under-ascertainment could result from a situation where a family could not be contacted because one or both of the parents had died from cancer, resulting in a family with a potentially higher cancer incidence not being included in the study. Conversely, over-ascertainment could result from selection bias where relatives who perceived an increased number of cancer cases in their families were more willing to take part. The fact that rates of all cancers are similar in each carrier probability group suggests that the overall level of bias is small.

Three cancers occurred in 64 obligate carriers: a case of follicular thyroid cancer in a female (confirmed by her clinician, diagnosed age 52) and two cases of prostate cancer in males (one confirmed, diagnosed age 71, one unconfirmed, diagnosed age 69). If these observations were extrapolated to calculate relative risk, it would result in significantly increased risks for both cancers in FA mutation carriers. It is quite possible that these are chance occurrences, particularly in the case of the thyroid cancer. The increased incidence of prostate cancer is interesting in view of involvement of BRCA2/FANCD1 in the FA gene pathway; male BRCA2 carriers have an increased risk of prostate cancer [14] so there is a theoretical molecular link. In the study by Berwick et al. [11], there was an increased incidence of prostate cancer in a group that included BRCA2 carriers and others of unknown complementation status. Taken together, these two studies suggest that there could be a modest increased risk in prostate cancer in some FA heterozygotes and which merits further investigation.

Our study had a number of limitations that are important to the interpretation of our findings. Complementation data was not available on the majority of the FA families in the study so analysis according to FA gene groups could not be carried out. Given the manner in which these families were ascertained and the ethnic heterogeneity of the UK population, it is likely that the complementation group frequencies would be similar to those previously determined, and that the majority (around two-thirds) of affected families would have FANCA mutations. Families where FA is due to biallelic BRCA2 mutations will clearly have an increased cancer risk, but these cases are rare (<2% of all FA) and none of the families in this study had a pedigree typical of autosomal dominant breast/ovarian cancer predisposition. It has recently been established that mutations in two other FA genes may confer an increased risk of breast cancer; the increased risk for FANCJ/BRIP1 is around 2-fold [16], and that for FANCN/PALB2 is probably 2–4 fold [17, 18], but may be higher for certain mutations [19]. Like FANCD1/BRCA2, both these complementation groups are rare (<1% of FA cases) and interestingly all three of these FA genes act downstream of FANCD2 [20]. In their recently published study Berwick et al. found a possible increased incidence of breast cancer in carrier females, with the highest risk in FANCC mutation carriers (SIR, 2.4; 95% CI, 1.1–5.2) [1]. No overall increased risk of breast cancer was seen in our study but the number of families with FANCC mutations would have been small as this group accounts for less than 10% of non-Ashkenazi Jewish FA cases. A previous study of familial breast cancer cases did not identify mutations in FANCA, C, D2, E, F, G [21], so at present it remains unclear whether FANCC, which acts upstream of FANCD2, is a breast cancer predisposing gene.

This study did not find any significant difference in cancer incidence in FA families compared to population estimates, with the possible exception of prostate cancer. It therefore seems that only in those rare families with FA cases due to BRCA2 or FANCJ/FANCN mutations is cancer risk in heterozygotes a salient factor.