## Abstract

Background: Homozygous or compound heterozygous mutations in the ATM gene are the principal cause of ataxia telangiectasia (A-T). Several studies have suggested that heterozygous carriers of ATM mutations are at increased risk of breast cancer and perhaps of other cancers, but the precise risk is uncertain. Methods: Cancer incidence and mortality information for 1160 relatives of 169 UK A-T patients (including 247 obligate carriers) was obtained through the National Health Service Central Registry. Relative risks (RRs) of cancer in carriers, allowing for genotype uncertainty, were estimated with a maximum-likelihood approach that used the EM algorithm. Maximum-likelihood estimates of cancer risks associated with three groups of mutations were calculated using the pedigree analysis program MENDEL. All statistical tests were two-sided. Results: The overall relative risk of breast cancer in carriers was 2.23 (95% confidence interval [CI] = 1.16 to 4.28) compared with the general population but was 4.94 (95% CI = 1.90 to 12.9) in those younger than age 50 years. The relative risk for all cancers other than breast cancer was 2.05 (95% CI = 1.09 to 3.84) in female carriers and 1.23 (95% CI = 0.76 to 2.00) in male carriers. Breast cancer was the only site for which a clear risk increase was seen, although there was some evidence of excess risks of colorectal cancer (RR = 2.54, 95% CI = 1.06 to 6.09) and stomach cancer (RR = 3.39, 95% CI = 0.86 to 13.4). Carriers of mutations predicted to encode a full-length ATM protein had cancer risks similar to those of people carrying truncating mutations. Conclusion: These results confirm a moderate risk of breast cancer in A-T heterozygotes and give some evidence of an excess risk of other cancers but provide no support for large mutation-specific differences in risk.

Ataxia telangiectasia (A-T) is a rare autosomal recessive neurologic disorder, characterized by progressive cerebellar degeneration and oculocutaneous telangiectasia. A-T appears to be completely penetrant and is typically diagnosed in early childhood, although the precise clinical phenotype varies from patient to patient. Most cancers in A-T patients are childhood lymphoid leukemias and lymphomas, but there is also a substantial risk of epithelial tumors later in life ( 1 ) . Almost all cases of A-T have been shown to be associated with mutations in the ATM gene, the product of which plays a central role in the recognition and repair of double-strand DNA breaks and in the activation of cell cycle checkpoints ( 2 ) . Most A-T patients are compound heterozygotes; homozygous carriers are uncommon, except in consanguineous families or in the case of a few population-specific founder mutations.

It has frequently been suggested that the blood relatives of A-T patients (i.e., obligate or potential heterozygous ATM mutation carriers) have an increased risk of cancer, primarily breast cancer. Clearly, it is important to reliably establish the cancer risks in heterozygous carriers to provide appropriate advice to the relatives of A-T patients. However, the question may also have wider public-health relevance. Some estimates of the frequency of ATM mutation carriers in Western populations are as high as 1% ( 3 , 4 ) , so that a relatively modest increase in breast cancer risk could equate to a substantial population attributable risk.

Studies assessing the risk of breast cancer in heterozygous ATM mutation carriers fall in two broad categories. First, several groups have compared breast cancer incidence and/or mortality in relatives of A-T patients with that in the general population or in married-in family members ( 510 ) . A review of four such studies estimated the breast cancer relative risk (RR) to be 3.9 (95% confidence interval [CI] = 2.1 to 7.2) ( 11 ) . Subsequent studies have found slightly more modest results, with relative risks between 2.4 and 3.4; most studies report that relative risks are higher among younger women ( 5 , 9 , 11 , 12 ) .

An alternative approach is to compare the frequency of ATM mutations in breast cancer case patients with that in control subjects. Case–control studies have almost uniformly failed to find an increased frequency of pathogenic ATM mutations in case patients, even when restricted to early-onset cancers ( 4 , 1316 ) . A review of 10 studies showed that ATM mutations are statistically significantly more frequent in breast cancer case patients selected on the basis of a family history of breast cancer than in unselected case patients ( 17 ) , although other studies have not replicated this result ( 18 ) .

The findings from the family studies and the case–control studies are not necessarily incompatible, given the widths of the confidence intervals; the sample sizes in many studies are too small to detect a modest increase in risk. Moreover, some studies have suggested that certain missense ATM mutations, notably 7271T>G, may be associated with higher risks of breast cancer ( 14 , 17 , 1922 ) , whereas most of the earlier population-based studies used mutation detection techniques that are biased in favor of detecting truncating mutations.

In addition to the potential association between ATM and breast cancer, several studies have reported an increase in the overall risk of cancer in relatives of A-T patients. One review found that the risk of non-breast cancers in carriers was almost double that expected in the general population ( 11 ) . Several cancer sites have been mentioned in this context, but no statistically significant associations with particular cancers have been reported to date ( 6 , 9 , 12 , 23 , 24 ) . If the risks of any other specific tumor types are genuinely increased in heterozygous ATM carriers, no study has yet had sufficient power to demonstrate this.

This study aimed to provide more precise estimates of the risks of cancer in heterozygous ATM mutation carriers by examining the cancer incidence and mortality experienced by the relatives of 169 A-T patients from 139 families living in the UK. This is by far the largest group of A-T families outside the US to have been studied to date and represents the large majority of A-T case patients diagnosed in the UK during the last 20 years. A second aim was to investigate potential differences in cancer risks associated with different types of ATM mutations.

## S UBJECTS AND M ETHODS

### Data Collection

Families were ascertained on the basis of at least one family member having been given a clinical diagnosis of A-T. The majority of the families (121 families) were ascertained via contact with the A-T Society, a UK support group for people with A-T and their families, or after referral by their pediatric neurologist to the Cancer Research UK Institute for Cancer Studies for diagnosis, genetic testing, and research purposes. In addition, to avoid biasing the cohort towards relatives of living A-T patients, a list of all death certificates since 1979 that mentioned A-T was obtained from the Office of National Statistics, leading to the inclusion of a further 18 families. Forty-four of the families were included in a previous study ( 7 ) , but the data used here include a larger number of relatives, 7.5 years of additional follow-up, and information about cancer incidence and mortality.

After we sought permission to contact the parents of each A-T patient from his or her general practitioner, the parent or parents who had agreed to participate in the study were sent a questionnaire requesting basic information about themselves and their children, siblings, parents, and grandparents (i.e., the siblings, aunts, uncles, grandparents, and great-grandparents of the A-T patient). All parents returning questionnaires gave written informed consent. The requested information for each relative comprised name, date and place of birth, vital status, and date of death, where applicable, whether he or she had ever had a cancer, and if so, the type of cancer, age at diagnosis, and place of treatment. Dates of birth were confirmed from national birth registers, and birth, death, and marriage registers were used to trace relatives in families for which the questionnaire was incomplete. Data were also obtained in this way for families for whom no questionnaire was available and for families ascertained via death certificate. An attempt was made to “flag” each of the relatives listed above through the National Health Service Central Register (NHSCR). The NHSCR receives notification of all deaths in the UK and all cancer registrations from cancer registries covering the UK, and the study coordinator was informed of these events in study subjects. Individuals were excluded from the study if tracing was not possible. Cancer diagnoses were included in the analysis only if they had been confirmed by the NHSCR, to allow valid comparison with population-based incidence rates.

Ethical approval was obtained from the South Birmingham Research Ethics Committee and the Birmingham and the Black Country Health Authority. Approval for use of the NHSCR for tracing was given by the Patient Information Advisory Group.

### Description of Cohort

A total of 169 A-T patients from 139 separate families were included in the study. Three families each contained three siblings with A-T, and 23 families each included a pair of siblings with A-T. One pair of cousins with A-T occurred in a consanguineous family. The number of relatives per family for whom information was available ranged from two to 28 (median = 17), giving a total of 2102 blood relatives (excluding 15 stepparents of A-T patients or of their parents). We excluded 510 relatives with unknown dates of birth, 152 who were born prior to 1891, and an additional 153 relatives who could not be traced by the NHSCR. Follow-up for parents was defined as starting at the birth of their first child with A-T, and follow-up for maternal and paternal grandparents began at the birth of the A-T patient's mother and father, respectively. Follow-up for maternal and paternal great-grandparents started 28 years before the birth of the A-T patient's mother and father respectively, to approximate the date of the relevant grandparent's birth (28 years was the average age of parents at the birth of a child with A-T in the cohort). This left-truncation of the follow-up period was performed to avoid biasing the cohort toward individuals who had, by definition, still been alive at the time that the A-T patient (or his or her parent or grandparent, respectively) was born. Follow-up for all other relatives began at their own dates of birth, because the A-T patient's birth was not dependent on their being alive at any particular point in time. One father was excluded because his last follow-up (when he joined the armed forces) was before the birth of his first child. The cohort included a total of 1286 relatives.

For the analysis of cancer incidence, follow-up prior to 1971 was excluded because cancer registry information was not complete before then. Only first cancers were considered, non-melanoma skin cancers were excluded, and only cancers reported by the cancer registry were counted. Follow-up was assumed to cease at the earliest of July 1, 2002, the date of death, the 80th birthday, or when the individual was last reported as being alive and cancer-free. Of the 1286 relatives, 126 contributed no person-years to the cancer incidence analysis because their dates of last follow-up or death were before 1971. The proportion of relatives excluded for any of the above reasons did not differ between families with and without questionnaires ( P = 0.4).

According to the definitions above, the cohort for the cancer incidence analysis consisted of 1160 relatives of A-T patients from 132 families, who contributed a total of 26 220 person-years to the analysis (median = 9 relatives per family, 27.2 years per relative). The distribution between different types of relative is shown in Table 1 . The number of male and female relatives was approximately equal; 573 (49.4%) males contributed 12 664 person-years (48.3%), and 587 (50.6%) females contributed 13 557 person-years (51.7%). The median year of birth was 1942 (interquartile range [IQR] = 1924 to 1958). During the follow-up period, there were 355 deaths, with a median age at death of 71 years (IQR = 62 to 81 years). The remaining 805 relatives were still alive when they exited the cohort, at a median age of 50 years (IQR = 39 to 63 years).

Table 1.

Cancer incidence in 1160 relatives of A-T patients from 132 families *

No. eligible Pyears Obs Exp SIR (95% CI)
All cancer incidence, excluding breast cancer
Relationship to A-T patient
Parent 280 247 5025 10.6 0.85 (0.39 to 1.62)
Sibling 105 90 1776 0.44 2.28 (0.06 to 12.7)
Half-sibling 11 189 0.05 0.00
Aunt/uncle 437 352 10 344 22 12.6 1.75 (1.10 to 2.64)
Grandparent 454 325 7054 49 39.7 1.23 (0.92 to 1.63)
Great-grandparent 802 131 1622 14 18.4 0.76 (0.41 to 1.28)
Parent's half-sibling 13 210 0.27 0.00
Approximate carrier probability
1 280 247 5025 10.6 0.85 (0.39 to 1.62)
0.67 105 90 1776 0.44 2.28 (0.06 to 12.7)
0.5 902 685 17 587 71 52.4 1.36 (1.06 to 1.72)
0.25 815 138 1832 14 18.7 0.75 (0.41 to 1.26)
All 2102 1160 26 220 95 82.1 1.16 (0.95 to 1.41)
Breast cancer incidence (female relatives)
Relationship to A-T patient
Mother  127 2640 2.67 1.87 (0.61 to 4.36)
Sister  45 968 0.09 0.00
Half-sister  81 0.00 0.00
Aunt  174 5047 3.11 2.90 (1.33 to 5.50)
Grandmother  173 3950 6.89 1.16 (0.50 to 2.29)
Great-grandmother  62 810 1.70 0.59 (0.01 to 3.27)
Parent's half-sister  62 0.01 0.00
Approximate carrier probability
1  127 2640 2.67 1.87 (0.61 to 4.36)
0.67  45 968 0.09 0.00
0.5  351 9077 17 10.0 1.70 (0.99 to 2.72)
0.25  64 872 1.72 0.58 (0.01 to 3.24)
All  587 13 557 23 14.5 1.59 (1.01 to 2.38)
No. eligible Pyears Obs Exp SIR (95% CI)
All cancer incidence, excluding breast cancer
Relationship to A-T patient
Parent 280 247 5025 10.6 0.85 (0.39 to 1.62)
Sibling 105 90 1776 0.44 2.28 (0.06 to 12.7)
Half-sibling 11 189 0.05 0.00
Aunt/uncle 437 352 10 344 22 12.6 1.75 (1.10 to 2.64)
Grandparent 454 325 7054 49 39.7 1.23 (0.92 to 1.63)
Great-grandparent 802 131 1622 14 18.4 0.76 (0.41 to 1.28)
Parent's half-sibling 13 210 0.27 0.00
Approximate carrier probability
1 280 247 5025 10.6 0.85 (0.39 to 1.62)
0.67 105 90 1776 0.44 2.28 (0.06 to 12.7)
0.5 902 685 17 587 71 52.4 1.36 (1.06 to 1.72)
0.25 815 138 1832 14 18.7 0.75 (0.41 to 1.26)
All 2102 1160 26 220 95 82.1 1.16 (0.95 to 1.41)
Breast cancer incidence (female relatives)
Relationship to A-T patient
Mother  127 2640 2.67 1.87 (0.61 to 4.36)
Sister  45 968 0.09 0.00
Half-sister  81 0.00 0.00
Aunt  174 5047 3.11 2.90 (1.33 to 5.50)
Grandmother  173 3950 6.89 1.16 (0.50 to 2.29)
Great-grandmother  62 810 1.70 0.59 (0.01 to 3.27)
Parent's half-sister  62 0.01 0.00
Approximate carrier probability
1  127 2640 2.67 1.87 (0.61 to 4.36)
0.67  45 968 0.09 0.00
0.5  351 9077 17 10.0 1.70 (0.99 to 2.72)
0.25  64 872 1.72 0.58 (0.01 to 3.24)
All  587 13 557 23 14.5 1.59 (1.01 to 2.38)
*

N = total number of relatives in the cohort, Pyears = person-years at risk, Obs = observed cancers, Exp = expected cancers, SIR = standardized incidence ratios, CI = confidence interval.

The follow-up period for mortality was defined in the same way as for cancer incidence, except that follow-up commenced on January 1, 1950. In this cohort, 644 male and 625 females contributed a total of 41 276 person-years (median = 34.2 years per relative, IQR = 22.0 to 43.7).

### Genotyping

To identify the ATM mutations present in the families of A-T patients, mutation screening of the ATM gene was performed at the Cancer Research UK Institute for Cancer Studies using lymphoblastoid cell lines derived from blood samples of A-T patients. In all these A-T patients, including those for whom mutations have not yet been found, loss of ATM protein was confirmed by Western blotting of protein extracts from the lymphoblastoid cell lines. Proteins were separated by sodium dodecyl sulfate–polyacrylamide gel electrophoresis on 6% gels and transferred electrophoretically to nitrocellulose membranes that were incubated with a monoclonal mouse–anti-human ATM antibody (11G12) ( 39 ) . Formerly, screening for ATM mutations had been carried out using restriction enzyme fingerprinting of PCR-amplified cDNA ( 22 ) . More recently, ATM mutations in A-T patients were identified by denaturing high-performance liquid chromatography analysis of PCR-amplified exons, followed by sequencing. For those mutations identified by exon sequencing that potentially altered splicing of the RNA transcript, cDNA sequencing was also performed to confirm sequence deletion or insertion. At least one pathogenic ATM mutation was identified in 118 A-T patients from 95 families (79% of the families ascertained via the A-T Society). In eight families the A-T patients have been shown to be homozygous for different ATM mutations, and a further 40 families have been shown to carry two distinct ATM mutations. No mutation has yet been identified in 12 families, and samples are not currently available for a further 30 families. Mutations were found in both parents from 33 families, in the mother but not the father in eight families, and in the father but not the mother in 10 families.

Subsequent to the initial data collection, A-T patients in two families have been shown to carry mutations in the MRE11 gene (including the consanguineous family containing a pair of cousins with A-T), rather than in ATM, and so should more properly be described as having A-T–Like Disorder (ATLD) ( 25 ) . MRE11-associated ATLD is difficult to distinguish clinically from A-T, although the characteristic telangiectasia features are absent in ATLD patients. These families were, however, included in the main analysis, because study entry was defined on the basis of a clinical, rather than a genetic, diagnosis of A-T.

### Statistical Analysis

Standardized incidence ratios (SIR) were used to compare the cancer incidence in relatives with that expected in the general population. Expected numbers of cancers in each individual were based on the age, sex, and calendar-period specific incidence rates given for England and Wales in Cancer in Five Continents Volumes III to VIII ( 2631 ) using the PYEARS program ( 32 ) . The 95% confidence intervals (CIs) were derived as exact confidence limits for a Poisson mean ( 33 ) . For the mortality analysis, mortality rates were taken from data provided by the UK Office of National Statistics, and standardized mortality ratios (SMR) were computed.

The parents of the A-T patients are all obligate ATM mutation carriers. No other relatives have been tested for mutations, so their carrier probabilities were estimated on the basis of their position within the pedigree, using the program MENDEL ( 34 ) , assuming that A-T is a fully penetrant recessive disorder, with mutant ATM alleles segregating according to standard Mendelian inheritance rules. The frequency of mutant alleles within the UK population was taken to be 0.3%, equivalent to approximately five new A-T cases per year. The results were not sensitive to small variations in this value.

These estimated carrier probabilities ( w i , for the ith individual) were used to obtain estimates of the relative risk of cancer associated with carrying one ATM mutation, with the observed and expected numbers of cancers in each relative ( O i and E i respectively) weighted by their estimated carrier probability; i.e., if the relative risk is denoted λ, then

$\mathrm{{\hat{{\lambda}}}}{=}\frac{{{\sum}_{i}}w_{i}O_{i}}{{{\sum}_{i}}w_{i}E_{i}}$

The relative risk of cancer for the noncarriers in the cohort, φ, was computed in the same way but with the Oi and Ei weighted instead by the estimated probability of not carrying a mutation, 1− wi . Estimates of λ and φ were obtained using the EM algorithm to iteratively update the individual carrier probability estimates and the relative risks ( 35 ) . Confidence intervals were derived from the estimated covariance matrix for λ and φ ( 36 ) . For almost all individual cancer sites, there was insufficient information to give stable simultaneous estimates of λ and φ. Simultaneously estimating λ and φ for all sites combined gave no evidence of an overall excess of cancer incidence, cancer mortality, or non-cancer mortality in noncarrier relatives; therefore, all estimates of λ presented are those estimated under the constraint that φ = 1 (i.e., noncarrier incidence rates assumed to equal general population rates).

Relative risks were also estimated separately for carriers who were younger than 50 years of age and for those aged 50 years or older. The cutpoint of 50 years was chosen to distinguish approximately between pre- and postmenopausal breast cancers. For consistency, the same cutpoint was also used for other cancers. Cumulative risks of cancer in carriers were estimated by applying the estimated carrier relative risks (younger than 50 years of age and 50 years or older) to the population rates for England and Wales (1992–1997) ( 29 ) .

Strictly, the relative risk estimates are not maximum-likelihood estimates because the dependence between the carrier probabilities of relatives from the same family is ignored in the iteration. However, the resulting estimates are consistent, whereas a full-likelihood analysis would theoretically require adjustment for familial aggregation of cancer, which is problematic to specify. In practice, the differences between the estimates presented here and the hypothetical full-likelihood estimates are likely to be negligible because there was rarely more than one cancer of the same type per family (i.e., no family had multiple cases of stomach or lung cancer; two families had two cases of breast cancer, and three families had two cases of colorectal cancer).

### Genotype–Phenotype Correlation

Given the large number of distinct pathologic ATM mutations recorded in A-T patients (81 distinct mutations in this cohort), it is impossible to evaluate risks associated with individual mutations. Because it had been previously hypothesized that the cancer risk might be related to the residual expression of the mutant ATM protein ( 21 ) , we classified mutations into three groups, according to whether any ATM protein was likely to be expressed from a mutant allele and, if so, whether the protein was likely to have kinase activity: A) frameshift mutations and substitutions leading to premature termination codons, resulting in no expression of the ATM protein from that allele; B) large (exon) or small (codon) in-frame deletions allowing some expression of a mutant ATM protein ( 37 ) that lacks kinase activity; and C) missense mutations allowing expression of mutant ATM with reduced kinase activity ( 37 ) . We have also included in this group the IVS40–1050A>G (5672ins137) “leaky” splicing mutation that can express a low level of normal ATM protein with kinase activity ( 3739 ) .

The full list of observed mutations assigned to each group is given in the Supplementary Table (available at http://jncicancerspectrum.oupjournals.org/jnci/content/vol97/issue11 ). Western blotting is routinely performed on lymphoblastoid cell lines derived from A-T patients to check for loss of ATM protein as part of the confirmation of diagnosis. The presence of some ATM protein was confirmed in A-T cells carrying all group B and C mutations (Supplementary Table, available at http://jncicancerspectrum.oupjournals.org/jnci/content/vol97/issue11 ). If ATM protein is expressed, its kinase activity can be assayed by in vitro phosphorylation of p53 ( 39 ) or detected with phosphospecific antibodies to in vivo targets (e.g., p53ser15) ( 37 ) . The ATM protein associated with the 7636del9 mutation (group B) has no detectable kinase activity ( 37 ) , although the carriers of both the 7271T>G and 5672ins137 ATM mutations (group C) express ATM protein with kinase activity ( 37 , 39 ) , as do the carriers of the other three mutations in group C. Absence of detectable kinase activity was examined and confirmed in nine patients with group B mutations (data not shown).

The pedigree analysis program MENDEL ( 34 ) was used to obtain maximum-likelihood estimates of the cancer risks associated with the three groups of mutations, assuming that all mutations must belong to one of these groups (even if there is currently insufficient evidence to say which). An iterative maximum-likelihood approach was necessary because of the incomplete genotype information available. This is an extension of the EM algorithm approach described earlier that allows for the nonindependence of genotypes within the same family. Along with relative risk parameters for breast cancer and all non-breast cancers in heterozygous mutation-carrying relatives, parameters for the relative risks of lymphoid tumors in A-T patients [C81–C96 inclusive, ICD revision 10 ( 40 ) ] were included in the models. A single relative risk parameter was used to model the risk of lymphoid tumors in A-T patients with no group C mutation (i.e., no kinase activity), whereas the relative risk parameter for patients with at least one group C mutation (i.e., some kinase activity) was fixed at 1.0. The inclusion of these parameters should improve the ability of the program to correctly predict the carrier status of untested individuals and hence give more precise relative risk estimates. In this analysis, 12 relative risk parameters were estimated for heterozygous carriers: three breast cancer relative risk parameters for women younger than 50 years of age (one for each mutation group), three for women older than 50 years of age, and three relative risk parameters each for male and female non-breast cancers.

Two families segregating the 7271T>G mutation were excluded from the genotype–phenotype analysis, because the identification of these families had prompted the hypothesis that the 7271T>G missense mutation (a group C mutation) was associated with a particularly elevated breast cancer risk ( 22 ) . One further family was excluded due to uncertainty about the function of its one identified mutation. The two families carrying mutations in the MRE11 gene were also excluded from this analysis (although they had been included in the main cohort analysis). This analysis was therefore based on 134 families, i.e., 268 mutant alleles. One hundred thirty-eight mutations have been identified (45 families have either two known mutations or two copies of the same mutation, and 48 families have one known mutation). Of these mutations, 86 were from group A, 34 were from group B, 18 were from group C, and one was of uncertain function (3403del174). The ATM mutation frequency (0.3%) was divided among the three groups of mutations according to these proportions. Estimating the allele frequencies as parameters within the model gave essentially the same results.

To improve the statistical power, the analysis was repeated with 16 additional breast cancers that were not eligible for the main analysis, because they either occurred before 1971 or after age 80 years, or were not confirmed by the NHSCR. Although including these cases might bias the overall relative risk estimate, there is no reason to believe that they would be biased toward any particular mutation group. Model selection was carried out using a conventional likelihood ratio test approach. All P values are two-sided; in the text, “statistically significant” is used to denote a P of <.05.

## R ESULTS

### Overall Results for Cohort

After the exclusions described above, the cohort consisted of 1160 relatives of A-T patients from 132 families (26 220 person-years). A total of 118 first cancers were reported by the NHSCR, compared with the 96.7 expected (SIR = 1.22, 95% CI = 1.02 to 1.46). Fifty-four of the cases were in men (50.3 expected), and 64 were in women (46.3 expected) (SIR = 1.07, 95% CI = 0.82 to 1.40, and SIR = 1.38, 95% CI = 1.08 to 1.77, in men and women, respectively). The median age was 50 years.

### Analysis by Type of Relative

The distribution of individuals, person-years, and cancer cases among relatives of each type is shown in Table 1 . Over all types of relative, the incidence of all cancers other than breast cancer was similar to that of the general population (SIR = 1.16, 95% CI = 0.95 to 1.41). The excess was attributable largely to excess risks in aunts/uncles (SIR = 1.75, 95% CI = 1.10 to 2.64) and grandparents (SIR = 1.23, 95% CI = 0.92 to 1.63). No statistically significant excess was observed in parents or great-grandparents. The overall number of breast cancers in relatives was slightly higher than expected (SIR = 1.59, 95% CI = 1.01 to 2.38, Table 1 ). Five of the 23 eligible breast cancers were in mothers, nine in aunts, eight in grandmothers, and one in a great-grandmother.

The 14 cancers diagnosed in parents of A-T patients are listed in Table 2 . No cancer site showed a statistically significant excess. Overall, the cancer incidence in parents was similar to that predicted using general population rates.

Table 2.

Cancer incidence in 247 parents of A-T patients from 132 families *

Cancer site Obs Exp SIR (95% CI)
Esophagus 0.27 3.65 (0.09 to 20.4)
Colorectal 1.45 0.69 (0.02 to 3.83)
Lung 1.89 1.59 (0.33 to 4.63)
Breast (female) 2.67 1.87 (0.61 to 4.36)
Prostate 0.55 1.83 (0.05 to 10.2)
Bladder 0.63 3.17 (0.38 to 11.4)
Brain 0.34 2.98 (0.08 to 16.6)
All sites 14 13.3 1.06 (0.58 to 1.77)
All except breast 10.6 0.85 (0.39 to 1.62)
Cancer site Obs Exp SIR (95% CI)
Esophagus 0.27 3.65 (0.09 to 20.4)
Colorectal 1.45 0.69 (0.02 to 3.83)
Lung 1.89 1.59 (0.33 to 4.63)
Breast (female) 2.67 1.87 (0.61 to 4.36)
Prostate 0.55 1.83 (0.05 to 10.2)
Bladder 0.63 3.17 (0.38 to 11.4)
Brain 0.34 2.98 (0.08 to 16.6)
All sites 14 13.3 1.06 (0.58 to 1.77)
All except breast 10.6 0.85 (0.39 to 1.62)
*

Obs = observed cancers, Exp = expected cancers, SIR = standardized incidence ratio, CI = confidence interval.

### Weighted Relative Risk Estimation

Consistent with previous observations, a statistically significant excess of female breast cancer in heterozygous ATM mutation carriers was seen (RR = 2.23, 95% CI = 1.16 to 4.28, Table 3 ) compared with the general population. Excluding breast cancer, there remained some evidence of an overall increased cancer risk to ATM carriers compared with that of the general population (RR = 1.47, 95% CI = 1.00 to 2.16), which was slightly greater in female carriers (RR = 2.05, 95% CI = 1.09 to 3.84) than in male carriers (RR = 1.23, 95% CI = 0.76 to 2.00). In addition, a statistically significant excess risk was observed for colorectal cancer (RR = 2.54, 95% CI = 1.06 to 6.09), and there was some suggestion of an excess of stomach cancer (RR = 3.39, 95% CI = 0.86 to 13.4).

Table 3.

Cancer incidence in 1160 relatives of A-T patients from 132 families, with estimated relative risks (RRs) and 95% confidence intervals (CIs) to heterozygous ATM carriers estimated using the EM algorithm *

Cancer site ICD 9 Obs Exp RR (95% CI)
Buccal cavity and pharynx 140–149 1.78 1.59 (0.15 to 16.8)
Esophagus 150 2.17 2.34 (0.47 to 11.6)
Stomach 151 10 4.74 3.39 (0.86 to 13.4)
Colorectal 152–154 20 12.1 2.54 (1.06 to 6.09)
Gallbladder 156 0.53 12.2 (1.26 to 118)
Pancreas 157 2.63 2.41 (0.34 to 17.1)
Lung 162 21 18.2 1.38 (0.64 to 2.97)
Breast (female) 174 23 14.6 2.23 (1.16 to 4.28)
Uterus 179 2.15 1.38 (0.09 to 22.4)
Ovary 183 2.67 1.90 (0.20 to 18.2)
Prostate 185 5.34 1.29 (0.30 to 5.48)
Bladder 188 5.22 1.41 (0.41 to 4.82)
Brain 191 1.93 0.06 (0.01 to 0.33)
Unknown 199 5.19 0.70 (0.10 to 4.92)
Myeloma 203 1.09 4.49 (0.32 to 62.2)
Other female genital 184 0.43 10.2 (0.30 to 345)
All sites except breast  95 82.1 1.47 (1.00 to 2.16)
Male: all sites  54 50.4 1.23 (0.76 to 2.00)
Female: all sites except breast  41 31.8 2.05 (1.09 to 3.84)
Cancer site ICD 9 Obs Exp RR (95% CI)
Buccal cavity and pharynx 140–149 1.78 1.59 (0.15 to 16.8)
Esophagus 150 2.17 2.34 (0.47 to 11.6)
Stomach 151 10 4.74 3.39 (0.86 to 13.4)
Colorectal 152–154 20 12.1 2.54 (1.06 to 6.09)
Gallbladder 156 0.53 12.2 (1.26 to 118)
Pancreas 157 2.63 2.41 (0.34 to 17.1)
Lung 162 21 18.2 1.38 (0.64 to 2.97)
Breast (female) 174 23 14.6 2.23 (1.16 to 4.28)
Uterus 179 2.15 1.38 (0.09 to 22.4)
Ovary 183 2.67 1.90 (0.20 to 18.2)
Prostate 185 5.34 1.29 (0.30 to 5.48)
Bladder 188 5.22 1.41 (0.41 to 4.82)
Brain 191 1.93 0.06 (0.01 to 0.33)
Unknown 199 5.19 0.70 (0.10 to 4.92)
Myeloma 203 1.09 4.49 (0.32 to 62.2)
Other female genital 184 0.43 10.2 (0.30 to 345)
All sites except breast  95 82.1 1.47 (1.00 to 2.16)
Male: all sites  54 50.4 1.23 (0.76 to 2.00)
Female: all sites except breast  41 31.8 2.05 (1.09 to 3.84)
*

The cancer sites shown are those for which at least two cases were observed. In addition, there was a single observed case of each of the following cancers: melanoma, cervix, testis, kidney, and thyroid. ICD = International Classification of Disease, Obs = observed cancers, Exp = expected cancers.

### Age Groups

The estimated relative risks for carriers younger than 50 years of age and 50 years of age or older are summarized in Table 4 . The overall relative risk of cancer was greater for both male and female carriers younger than 50 years of age, with little evidence of an excess risk for carriers aged 50 years and older (RR = 1.04, 95% CI = 0.59 to 1.83 in males; RR = 1.64, 95% CI = 0.81 to 3.30 in females, excluding breast cancer). The estimated relative risk of breast cancer in carriers younger than 50 years of age was close to 5 (RR = 4.94, 95% CI = 1.90 to 12.9), but there was no statistically significant risk for women 50 years of age and older. The overall excess cancer risk in carriers younger than 50 years of age appeared to be due to several different cancer types (for myeloma, RR = 43.3, 95% CI = 2.70 to 694; for stomach cancer, RR = 15.8, 95% CI = 1.63 to 153). One of the two buccal cavity cancers was a nasopharyngeal cancer in the 6-year-old brother of an A-T patient; this was the only juvenile cancer in a relative.

Table 4.

Cancer incidence, by age group, in 1160 relatives of A-T patients from 132 families, with estimated relative risks (RRs) and 95% confidence intervals (CIs) to heterozygous ATM carriers estimated using the EM algorithm *

Less than 50 years old

50 years or older

Site Obs Exp RR (95% CI) Obs Exp RR (95% CI)
Stomach 0.33 15.8 (1.63 to 153) 4.51 2.16 (0.40 to 11.6)
Colorectal 1.10 3.20 (0.55 to 18.3) 18 11.0 2.45 (0.90 to 6.69)
Gallbladder 0.04  0.49 13.5 (1.39 to 132)
Lung 1.05 0.78 (0.02 to 39.0) 20 17.2 1.42 (0.65 to 3.11)
Breast 11 4.34 4.94 (1.90 to 12.9) 12 10.1 1.14 (0.48 to 2.72)
Prostate 0.04  5.30 1.31 (0.31 to 5.57)
Female genital 0.07  0.36 12.3 (0.36 to 423)
All sites 30 15.4 3.16 (1.77 to 5.65) 88 81.2 1.20 (0.81 to 1.78)
Male: all sites 5.33 2.14 (0.86 to 5.30) 45 45.1 1.04 (0.59 to 1.83)
Female: all sites except breast 10 5.78 3.81 (1.09 to 13.4) 31 26.0 1.64 (0.81 to 3.30)
Less than 50 years old

50 years or older

Site Obs Exp RR (95% CI) Obs Exp RR (95% CI)
Stomach 0.33 15.8 (1.63 to 153) 4.51 2.16 (0.40 to 11.6)
Colorectal 1.10 3.20 (0.55 to 18.3) 18 11.0 2.45 (0.90 to 6.69)
Gallbladder 0.04  0.49 13.5 (1.39 to 132)
Lung 1.05 0.78 (0.02 to 39.0) 20 17.2 1.42 (0.65 to 3.11)
Breast 11 4.34 4.94 (1.90 to 12.9) 12 10.1 1.14 (0.48 to 2.72)
Prostate 0.04  5.30 1.31 (0.31 to 5.57)
Female genital 0.07  0.36 12.3 (0.36 to 423)
All sites 30 15.4 3.16 (1.77 to 5.65) 88 81.2 1.20 (0.81 to 1.78)
Male: all sites 5.33 2.14 (0.86 to 5.30) 45 45.1 1.04 (0.59 to 1.83)
Female: all sites except breast 10 5.78 3.81 (1.09 to 13.4) 31 26.0 1.64 (0.81 to 3.30)
*

The cancer sites shown are those for which either the overall carrier RR was statistically significantly greater than 1 or for which there were 10 or more cases. In addition, there were two or more cases in the younger age group of buccal cavity and pharynx cancer (two cases), uterus cancer (two cases), and myeloma (two cases). Obs = observed cancers, Exp = expected cancers.

### Cumulative Cancer Risks

Cumulative risks of cancer were estimated by applying the estimated relative risks for carriers to the incidence rates in the general population. The cumulative risk of breast cancer in heterozygous ATM mutation carriers was estimated to be 8.8% (95% CI = 3.5% to 21.4%) by age 50 years and 16.6% (95% CI = 9.1% to 29.3%) by age 80 years ( Fig. 1, A ). The latter risk, that approximately one woman in six will develop breast cancer, compares with a risk of approximately one in 11 in the general population of England and Wales (1992–1997) ( 29 ) . The estimated risk of any other cancer type by age 50 years was 5.3% (95% CI = 2.2% to 12.6%) in males and 9.0% (95% CI = 2.6% to 28.1%) in females, compared with 2.5% and 2.4%, respectively in the general population ( 29 ) . The cumulative risk of any non-breast cancer by age 80 years was similar in male and female carriers (38.9%, 95% CI = 25.6% to 56.0%; and 35.1%, 95% CI = 20.9% to 55.0%, respectively), although the risk in females was more strongly elevated above the population risk ( Fig. 1, B and C ).

Fig. 1.

Cumulative risks of cancer in heterozygous ATM mutation carriers, estimated from cancer incidence in 1160 relatives of A-T patients from 132 UK families. A ) Estimated cumulative risks of breast cancer in female heterozygous ATM mutation carriers. B ) Estimated cumulative risks of all cancers in male heterozygous ATM mutation carriers. C ) Estimated cumulative risks of all cancers other than breast cancer in female heterozygous ATM mutation carriers. Estimated cumulative risks to carriers along with 95% confidence intervals ( solid lines ) and cumulative risks in the general population [England and Wales, 1992–1997 ( 29 ) hatched lines ] are shown, at each 10-year age point. Cumulative risks were obtained by applying the estimated RRs to carriers below and above age 50 (estimated using the EM algorithm) to the general population rates.

Fig. 1.

Cumulative risks of cancer in heterozygous ATM mutation carriers, estimated from cancer incidence in 1160 relatives of A-T patients from 132 UK families. A ) Estimated cumulative risks of breast cancer in female heterozygous ATM mutation carriers. B ) Estimated cumulative risks of all cancers in male heterozygous ATM mutation carriers. C ) Estimated cumulative risks of all cancers other than breast cancer in female heterozygous ATM mutation carriers. Estimated cumulative risks to carriers along with 95% confidence intervals ( solid lines ) and cumulative risks in the general population [England and Wales, 1992–1997 ( 29 ) hatched lines ] are shown, at each 10-year age point. Cumulative risks were obtained by applying the estimated RRs to carriers below and above age 50 (estimated using the EM algorithm) to the general population rates.

Based on the observed case frequency over the period 1979–1997, we estimate the heterozygous carrier frequency to be 0.4%. Therefore, our best estimate of the fraction of breast cancer cases attributable to ATM mutations is 0.5% overall, rising to 1.6% for cases diagnosed before age 50 years.

### Mortality

The overall mortality rate in males was almost identical to that expected (SMR = 1.01, 95% CI = 0.87 to 1.16). However, this rate reflected the combination of a modestly increased risk of cancer deaths (SMR = 1.35, 95% CI = 1.07 to 1.70) with a slight, statistically non-significant deficit of non-cancer deaths (SMR = 0.88, 95% CI = 0.74 to 1.05). The relative risk of non-cancer death was similar in female relatives (SMR = 0.85, 95% CI = 0.67 to 1.09), but a higher risk of cancer deaths (SMR = 1.82, 95% CI = 1.43 to 2.32) in these relatives resulted in an overall borderline statistically significantly increased mortality rate (SMR = 1.16, 95% CI = 0.98 to 1.37) as compared with the general population. The mortality in fathers was close to that expected in the general population, as was the mortality in other male relatives (data not shown). The mortality in female relatives other than mothers was also close to that expected in the general population. However, there were only two deaths in mothers (a lung cancer and a pancreatic cancer), as compared with an expected 7.90 deaths.

Seventeen deaths from breast cancer were observed in female relatives (SMR = 2.08, 95% CI = 1.21 to 3.32). Ten of these were included in the incidence analysis; the other seven were ineligible because they were reported only on death certificates and not by the NHSCR.

Statistically significant excess cancer mortality was observed in ATM carriers of both sexes (SMR = 1.88, 95% CI = 1.14 to 3.10 and SMR = 3.56, 95% CI = 1.83 to 6.93 for males and females, respectively), whereas non-cancer mortality was slightly, but not statistically significantly, lower than expected ( Table 5 ). There was no evidence of excess mortality from either vascular or respiratory disease. Statistically significant excesses in mortality in ATM carriers were estimated for breast cancer (RR = 4.18, 95% CI = 1.38 to 12.7), stomach cancer (RR = 4.19, 95% CI = 1.49 to 11.8), colorectal cancer (RR = 3.19, 95% CI = 1.24 to 8.23), and lung cancer (RR = 2.36, 95% CI = 1.24 to 4.50) as compared with the general population.

Table 5.

Mortality in 1269 relatives of A-T patients from 132 families, with estimated relative risks (RRs) and 95% confidence intervals (CIs) to heterozygous ATM carriers estimated using the EM algorithm *

Death cause ICD 9 Obs Exp RR (95% CI)
Cancer deaths
Esophagus 150 2.68 1.09 (0.08 to 14.6)
Stomach 151 15 7.14 4.19 (1.49 to 11.8)
Colorectal 152–154 18 9.87 3.19 (1.24 to 8.23)
Pancreas 157 3.61 3.21 (0.89 to 11.5)
Lung 162 39 24.9 2.36 (1.24 to 4.50)
Breast (female) 174 17 8.18 4.18 (1.38 to 12.7)
Ovary 183 2.76 1.84 (0.19 to 17.8)
Prostate 185 3.09 0.93 (0.14 to 6.29)
Bladder 188 2.53 1.87 (0.19 to 18.0)
Brain 191 2.19 1.53 (0.11 to 20.7)
Unknown 199 4.33 2.76 (0.59 to 12.9)
Myeloma 203 1.03 1.51 (0.01 to 358)
Other   1.68 4.00 (0.45 to 35.3)
Male: all cancer sites  70 51.9 1.88 (1.14 to 3.10)
Female: all cancer sites  66 36.3 3.56 (1.83 to 6.93)
Female: all cancer sites except breast  49 28.1 3.21 (1.64 to 6.27)
Circulatory disease  119 135 0.78 (0.53 to 1.17)
Respiratory disease  43 35.2 1.63 (0.81 to 3.28)
Injury and poisoning  14.5 0.17 (0.04 to 0.68)
Male: all non-cancers  128 144 0.75 (0.52 to 1.08)
Female: all non-cancers  67 78.4 0.79 (0.45 to 1.38)
Death cause ICD 9 Obs Exp RR (95% CI)
Cancer deaths
Esophagus 150 2.68 1.09 (0.08 to 14.6)
Stomach 151 15 7.14 4.19 (1.49 to 11.8)
Colorectal 152–154 18 9.87 3.19 (1.24 to 8.23)
Pancreas 157 3.61 3.21 (0.89 to 11.5)
Lung 162 39 24.9 2.36 (1.24 to 4.50)
Breast (female) 174 17 8.18 4.18 (1.38 to 12.7)
Ovary 183 2.76 1.84 (0.19 to 17.8)
Prostate 185 3.09 0.93 (0.14 to 6.29)
Bladder 188 2.53 1.87 (0.19 to 18.0)
Brain 191 2.19 1.53 (0.11 to 20.7)
Unknown 199 4.33 2.76 (0.59 to 12.9)
Myeloma 203 1.03 1.51 (0.01 to 358)
Other   1.68 4.00 (0.45 to 35.3)
Male: all cancer sites  70 51.9 1.88 (1.14 to 3.10)
Female: all cancer sites  66 36.3 3.56 (1.83 to 6.93)
Female: all cancer sites except breast  49 28.1 3.21 (1.64 to 6.27)
Circulatory disease  119 135 0.78 (0.53 to 1.17)
Respiratory disease  43 35.2 1.63 (0.81 to 3.28)
Injury and poisoning  14.5 0.17 (0.04 to 0.68)
Male: all non-cancers  128 144 0.75 (0.52 to 1.08)
Female: all non-cancers  67 78.4 0.79 (0.45 to 1.38)
*

Obs = observed cancers, Exp = expected cancers, ICD = International Classification of Disease.

The “other” cancers were three female genital cancers and a cancer of the middle ear.

ATM carrier relative risks were also estimated separately for deaths before or after age 50 years ( Table 6 ). The estimated cancer mortality relative risks were higher for carriers younger than 50 years of age than for carriers aged 50 years and older (RR = 3.59, 95% CI = 1.74 to 7.38; and RR = 2.23, 95% CI = 1.44 to 3.45, respectively). Consistent with the incidence analysis, the relative risk of breast cancer mortality was higher for carriers below age 50 years (RR = 6.08, 95% CI = 1.05 to 35.3) than for carriers aged 50 years and older (RR = 3.45, 95% CI = 0.89 to 13.4). Mortality from stomach cancer and colorectal cancer was also particularly elevated in ATM carriers below age 50 years (stomach cancer, RR = 14.0, 95% CI = 3.18 to 61.9; and colorectal cancer, RR = 11.0, 95% CI = 2.55 to 47.2).

Table 6.

Mortality, by age group, in 1269 relatives of A-T patients from 132 families, with estimated relative risks (RRs) and 95% confidence intervals (CIs) to heterozygous ATM carriers estimated using the EM algorithm *

Less than 50 years old

50 years or older

Death cause Obs Exp RR (95% CI) Obs Exp RR (95% CI)
Cancer deaths
Stomach 0.58 14.0 (3.18 to 61.9) 11 6.55 2.94 (0.75 to 11.5)
Colorectal 0.87 11.0 (2.55 to 47.2) 13 8.99 2.23 (0.67 to 7.46)
Pancreas 0.27   3.34 3.65 (1.01 to 13.2)
Lung 1.60 2.16 (0.37 to 12.5) 37 23.3 2.38 (1.19 to 4.76)
Breast 1.89 6.08 (1.05 to 35.3) 12 6.30 3.45 (0.89 to 13.4)
Male: all sites 4.65 2.55 (0.98 to 6.62) 62 47.3 1.75 (0.99 to 3.08)
Female: all sites except breast 3.98 4.45 (1.06 to 18.6) 40 24.1 2.92 (1.44 to 5.91)
Male: non-cancer deaths 17 22.3 0.61 (0.27 to 1.39) 111 122 0.79 (0.53 to 1.18)
Female: non-cancer deaths 10 12.6 0.78 (0.23 to 2.68) 57 65.7 0.79 (0.42 to 1.48)
Less than 50 years old

50 years or older

Death cause Obs Exp RR (95% CI) Obs Exp RR (95% CI)
Cancer deaths
Stomach 0.58 14.0 (3.18 to 61.9) 11 6.55 2.94 (0.75 to 11.5)
Colorectal 0.87 11.0 (2.55 to 47.2) 13 8.99 2.23 (0.67 to 7.46)
Pancreas 0.27   3.34 3.65 (1.01 to 13.2)
Lung 1.60 2.16 (0.37 to 12.5) 37 23.3 2.38 (1.19 to 4.76)
Breast 1.89 6.08 (1.05 to 35.3) 12 6.30 3.45 (0.89 to 13.4)
Male: all sites 4.65 2.55 (0.98 to 6.62) 62 47.3 1.75 (0.99 to 3.08)
Female: all sites except breast 3.98 4.45 (1.06 to 18.6) 40 24.1 2.92 (1.44 to 5.91)
Male: non-cancer deaths 17 22.3 0.61 (0.27 to 1.39) 111 122 0.79 (0.53 to 1.18)
Female: non-cancer deaths 10 12.6 0.78 (0.23 to 2.68) 57 65.7 0.79 (0.42 to 1.48)
*

Obs = observed cancers, Exp = expected cancers.

### Genotype–Phenotype Correlations

Risks of breast and non-breast cancers in relatives were estimated for the three categories of ATM mutation. There were no statistically significant differences between the mutation groups in the risks of either non-breast cancer ( P = .5) or breast cancer ( P = .8). When 16 additional breast cancer cases were included, the risk was highest for patients with mutations expressing some protein without kinase activity (group B) (comparing groups B and A, RR = 2.5, 95% CI = 0.7 to 8.9) and slightly lower for those with mutations retaining kinase activity (group C) (comparing groups C and A, RR = 0.9, 95% CI = 0.1 to 8.9), although the differences were not statistically significant ( P = .4).

## D ISCUSSION

We have studied the cancer incidence and the mortality in 1160 blood relatives of A-T patients from 132 families and have found evidence for an increased risk of breast cancer in heterozygous ATM mutation carriers, chiefly at young ages, accompanied by a more moderate increase in the risk of other cancers. The overall estimated breast cancer relative risk to heterozygous ATM carriers was 2.23 (95% CI = 1.16 to 4.28), with a relative risk of 4.94 (95% CI = 1.90 to 12.9) in carriers younger than 50 years of age. This is equivalent to a lifetime (until age 80 years) risk of approximately one woman in six, as compared with one in 11 in the general population of England and Wales.

The estimated risk of any cancer in male carriers by the age of 80 years was only slightly higher than in the general population (39% vs. 36%), whereas the risk by age of 80 years of any cancer other than breast cancer in female carriers was considerably higher than in the general population (35% vs. 21%).

Although there was little evidence for an overall excess risk of cancers other than breast cancer in ATM heterozygotes, there was some evidence for excess risks of colorectal cancer and stomach cancer. We also observed a clear excess mortality from cancer, with statistically significant excess risks of stomach, colorectal, and lung cancer deaths. The higher relative risks based on mortality might reflect some underreporting of cancers by the NHSCR but could also reflect a more aggressive behavior of cancers in ATM carriers. Two previous studies have hinted at a possible association between ATM and cancers of the gastrointestinal tract, although neither association was statistically significant ( 7 , 9 ) . In contrast to our study, neither study found any evidence of a specific excess of colorectal cancers in relatives of A-T patients.

Some limitations of this study that may lead to biased relative risk estimates include incomplete ascertainment of families, possible nonpaternity or de novo ATM mutations, and the possibility that some A-T patients may not carry ATM mutations. A further limitation is that we were able to genotype only the parents of A-T patients and not any other relatives. Although this reduced the power of the study and the precision of the relative risk estimates, it should not result in any bias, providing that ATM mutations are inherited according to Mendelian rules. The precision of the estimates was also limited by the number of available A-T families. Further precision should, however, be obtained through combined analysis of our data with those from other European studies.

We have attempted to minimize bias in this study by systematically following a defined cohort of relatives of all known A-T patients and by basing analysis only on registered cancers and deaths reported through national records, to allow direct comparability of observed and expected rates. Nevertheless, some potential biases remain. First, families in which a parent died at a young age might be less likely to have participated in the study. We attempted to minimize this bias by including additional families ascertained through a mention of A-T on a death certificate. That some bias remains is borne out by a marked deficit in mortality in mothers, with two deaths observed, compared with nearly eight expected. This bias is reflected in the slight deficit in overall mortality from nonmalignant causes and suggests that the excess mortality from and incidence of cancer may therefore have been underestimated.

Other events that would reduce the number of mutations in relatives, and hence underestimate the risks, are nonpaternity and de novo mutations. One A-T patient in the cohort is known to carry an inherited truncating mutation alongside a de novo paternal missense mutation, 8189A>C. This is generally considered to be a very rare event in A-T. There was no evidence of incompatible paternal genotypes among the genotyped parents of the A-T patients. Although there may be instances of false paternity among grandparents or great-grandparents, any such events would not affect the carrier probabilities of as many family members.

Strictly speaking, the estimates are a weighted average of the risks conferred by ATM and MRE11 mutations. The apparent A-T cases in two of the families are in fact due to compound heterozygous mutations in the MRE11 gene. The relative sizes of the two genes would suggest that approximately 6% of A-T patients might in fact carry MRE11 mutations, i.e., approximately six further families ( 25 ) . MRE11 acts in the same DNA damage response pathway as does ATM, but mutations in the two genes need not predispose to cancer to the same extent; there is no evidence that homozygous Mre11 mutations are associated with tumors in mice ( 41 ) . If MRE11 mutations conferred no excess cancer risk, then the ATM excess cancer risk estimated in this study could be underestimated by approximately 6%.

In addition to the excesses of breast, colorectal, and stomach cancer noted above, a statistically significant excess of cancer of the gallbladder was also observed, but this was based on only three cases. A high relative risk was estimated for “other female genital cancers,” but this was based on just two cases and was not statistically significant. Three further female genital cancers were also reported but did not contribute to the analysis. A high relative risk (RR = 4.5) was also estimated for myeloma, based on three cases. It is noteworthy that these were the only lymphoid tumors seen in relatives and that no myelomas were observed in A-T patients.

The apparent excesses at some or all of these sites could be due to chance, given the number of cancer sites evaluated, and larger studies will be required to determine whether these effects are genuine. Conversely, moderate risks of other cancers in ATM carriers cannot be ruled out. The modest overall increase in the risk of non-breast cancer appears to be due largely to a combination of small increases at many sites; it is notable that all the relative risk estimates in Table 3 are greater than 1, with the exception of brain cancer and cancers of unknown site.

Our study is comparable in design to two recent European studies, one based in France ( 42 ) and the other in the Nordic countries ( 9 ) . The Nordic study obtained cancer incidence data for 1218 relatives of A-T patients from 50 families via record linkage to national cancer registries. Its authors estimated the breast cancer relative risk for ATM carriers to be 2.4 (95% CI = 1.3 to 4.1), which is very similar to our estimate. Breast cancer was the only individual cancer with a statistically significant excess; apart from breast cancer, they observed only 15% more cancers than expected in relatives of A-T patients.

The French study was based on 1423 relatives of A-T patients from 34 families. ATM genotyping was performed on over a quarter of the relatives, but not all cancer cases had been formally confirmed. The relative risk of breast cancer, weighted by prior carrier probability (RR = 2.43, 95% CI = 1.32 to 4.09) was also very similar to our estimate (RR = 2.23, 95% CI = 1.16 to 4.28). In the French study, the relative risk was higher for women below age 45 years, but no excess was seen in women above this age (RR = 6.32, 95% CI =1.94 to 15.2, and RR = 0.68, 95% CI = 0.08 to 2.46, respectively). There was no evidence of an increased risk of cancers other than breast cancers in carriers in this study ( 42 ) .

The results presented here are generally in line with the French and Nordic studies. Our study has the advantage of being based on a far larger number of families, and, although the number of eligible relatives in our cohort was slightly smaller, the exclusion of cousins and great-aunts/uncles meant that the cohort had a higher density of mutation carriers. Previous studies have either presented separate relative risks for each type of relative, often with large confidence intervals as a consequence of the small numbers of cases in each group, or have pooled all relatives into a single group, without taking into account their different carrier probabilities. In contrast, our use of the EM algorithm to obtain maximum-likelihood estimates of the carrier relative risks, based on weighting the information from all relatives, made more efficient use of the data. In common with the Nordic study (but not the French study), we considered only cancer cases that had been formally confirmed. Neither of the other European studies considered both cancer incidence and mortality.

We found no evidence for any difference in risk of breast or other cancer according to the type of ATM mutation. If anything, the trend was toward a lower breast cancer risk for the group C mutations, in contrast with previous reports that showed that missense mutations, in particular 7271T>G, are associated with a markedly increased risk of breast cancer ( 1922 ) . Our estimates were necessarily imprecise, because group C mutations were the least frequent in this set; after the exclusion of the two hypothesis-generating 7271T>G families ( 22 ) , there was only one breast cancer in a family branch known to carry a group C mutation. Furthermore, because the 5762ins137 mutation accounted for 14 out of 18 of the known group C family branches, the results may not be generalizable to all ATM mutations retaining kinase activity.

The mutation categories were devised in the context of A-T patients with two germline mutations in trans , whereas the analysis of cancer risks was restricted to heterozygous carriers, in whom these particular differences between mutations may be less important. For a single mutation in the presence of a wild-type allele, alternative mechanisms may become relevant to the disease process, potentially including haploinsufficiency (group A), dominant-negative effects (groups B and C), or some gain in function (groups B and C). For example, lymphoblastoid cell lines with a heterozygous missense mutation have been shown to have higher ATM mRNA expression than do cell lines with a truncating mutation and to have poorer cell survival following irradiation ( 43 ) .

A recent study of 34 French A-T families found no difference between the breast cancer risks associated with heterozygous truncating and missense/in-frame deletion ATM mutations but identified three groups of truncating mutations with particularly high breast cancer relative risks, each relating to a known binding domain ( 42 ) . However, we observed no breast cancers in the seven family branches with mutations that truncate the ATM protein in these domains.

It is important to note that our results do not exclude the possibility of more substantial heterogeneity at the mutation level. Despite all these uncertainties, the results do appear to confirm that a substantial risk of breast cancer is conferred by mutations that eliminate the ATM protein and that the risk is not restricted to a subset of missense mutations.

The breast cancer risk we have estimated would be sufficient to classify an ATM carrier as “moderate risk,” according to recent guidelines of the National Institute for Clinical Excellence (2004). These guidelines suggest that annual mammography beginning at age 40 years may be appropriate in this risk group. However, given the role of ATM in radiation-induced DNA repair, it is not clear whether mammographic screening would be beneficial in ATM carriers. Recent studies have suggested that magnetic resonance imaging may be a sensitive screening tool in women at high risk of breast cancer, such as BRCA1 and BRCA2 carriers ( 44 ) , and it may provide an alternative management approach for ATM carriers. Further research would be needed to evaluate the appropriateness of any specific screening for gastric, colorectal, or other cancers.

In conclusion, this study has confirmed an approximately twofold-increased risk of breast cancer in female carriers of ATM mutations, with a higher relative risk for those younger than 50 years. We also identified increased risk of colorectal cancer and a possible increased risk of stomach cancer. Combined analyses with similar cohorts and further follow-up will be required to provide reliable risk estimates for other cancer sites and to investigate mutation-specific effects.

We thank the A-T patients and their families for their willingness to participate in this research. This study was supported by grants from Cancer Research UK and the A-T Society. DFE is a Principal Research Fellow of Cancer Research UK. We also thank staff at the Office of National Statistics for its help.

## References

(1)
Morrell D, Cromartie E, Swift M. Mortality and cancer incidence in 263 patients with ataxia-telangiectasia.
J Natl Cancer Inst

1986
;
77
:
89
–92.
(2)
Savitsky K, Bar-Shira A, Gilad S, Rotman G, Ziv Y, Vanagaite L, et al. A single ataxia telangiectasia gene with a product similar to PI-3 kinase.
Science

1995
;
268
:
1749
–53.
(3)
Swift M, Morrell D, Cromartie E, Chamberlin AR, Skolnick MH, Bishop DT. The incidence and gene frequency of ataxia-telangiectasia in the United States.
Am J Hum Genet

1986
;
39
:
573
–83.
(4)
FitzGerald MG, Bean JM, Hegde SR, Unsal H, MacDonald DJ, Harkin DP, et al. Heterozygous ATM mutations do not contribute to early onset of breast cancer.
Nat Genet

1997
;
15
:
307
–10.
(5)
Swift M, Reitnauer PJ, Morrell D, Chase CL. Breast and other cancers in families with ataxia-telangiectasia.
N Engl J Med

1987
;
316
:
1289
–94.
(6)
Swift M, Morrell D, Massey RB, Chase CL. Incidence of cancer in 161 families affected by ataxia-telangiectasia.
N Engl J Med

1991
;
325
:
1831
–6.
(7)
Inskip HM, Kinlen LJ, Taylor AM, Woods CG, Arlett CF. Risk of breast cancer and other cancers in heterozygotes for ataxia-telangiectasia.
Br J Cancer

1999
;
79
:
1304
–7.
(8)
Janin N, Andrieu N, Ossian K, Lauge A, Croquette MF, Griscelli C, et al. Breast cancer risk in ataxia telangiectasia (AT) heterozygotes: haplotype study in French AT families.
Br J Cancer

1999
;
80
:
1042
–5.
(9)
Olsen JH, Hahnemann JM, Borresen-Dale AL, Brondum-Nielsen K, Hammarstrom L, Kleinerman R, et al. Cancer in patients with ataxia-telangiectasia and in their relatives in the nordic countries.
J Natl Cancer Inst

2001
;
93
:
121
–7.
(10)
Su Y, Swift M. Mortality rates among carriers of ataxia-telangiectasia mutant alleles.
Ann Intern Med

2000
;
133
:
770
–8.
(11)
Easton DF. Cancer risks in A-T heterozygotes.

1994
;
66
:
S177
–82.
(12)
Geoffroy-Perez B, Janin N, Ossian K, Lauge A, Croquette MF, Griscelli C, et al. Cancer risk in heterozygotes for ataxia-telangiectasia.
Int J Cancer

2001
;
93
:
288
–93.
(13)
Bebb DG, Yu Z, Chen J, Telatar M, Gelmon K, Phillips N, et al. Absence of mutations in the ATM gene in forty-seven cases of sporadic breast cancer.
Br J Cancer

1999
;
80
:
1979
–81.
(14)
Buchholz TA, Weil MM, Ashorn CL, Strom EA, Sigurdson A, Bondy M, et al. A Ser49Cys variant in the ataxia telangiectasia, mutated, gene that is more common in patients with breast carcinoma compared with population controls.
Cancer

2004
;
100
:
1345
–51.
(15)
Izatt L, Greenman J, Hodgson S, Ellis D, Watts S, Scott G, et al. Identification of germline missense mutations and rare allelic variants in the ATM gene in early-onset breast cancer.
Genes Chromosomes Cancer

1999
;
26
:
286
–94.
(16)
Shafman TD, Levitz S, Nixon AJ, Gibans LA, Nichols KE, Bell DW, et al. Prevalence of germline truncating mutations in ATM in women with a second breast cancer after radiation therapy for a contralateral tumor.
Genes Chromosomes Cancer

2000
;
27
:
124
–9.
(17)
Thorstenson YR, Roxas A, Kroiss R, Jenkins MA, Yu KM, Bachrich T, et al. Contributions of ATM mutations to familial breast and ovarian cancer.
Cancer Res

2003
;
63
:
3325
–33.
(18)
Chen J, Birkholtz GG, Lindblom P, Rubio C, Lindblom A. The role of ataxia-telangiectasia heterozygotes in familial breast cancer.
Cancer Res

1998
;
58
:
1376
–9.
(19)
Chenevix-Trench G, Spurdle AB, Gatei M, Kelly H, Marsh A, Chen X, et al. Dominant negative ATM mutations in breast cancer families.
J Natl Cancer Inst

2002
;
94
:
205
–15.
(20)
Dork T, Bendix R, Bremer M, Rades D, Klopper K, Nicke M, et al. Spectrum of ATM gene mutations in a hospital-based series of unselected breast cancer patients.
Cancer Res

2001
;
61
:
7608
–15.
(21)
Gatti RA, Tward A, Concannon P. Cancer risk in ATM heterozygotes: a model of phenotypic and mechanistic differences between missense and truncating mutations.
Mol Genet Metab

1999
;
68
:
419
–23.
(22)
Stankovic T, Kidd AM, Sutcliffe A, McGuire GM, Robinson P, Weber P, et al. ATM mutations and phenotypes in ataxia-telangiectasia families in the British Isles: expression of mutant ATM and the risk of leukemia, lymphoma, and breast cancer.
Am J Hum Genet

1998
;
62
:
334
–45.
(23)
Bernstein JL, Bernstein L, Thompson WD, Lynch CF, Malone KE, Teitelbaum SL, et al. ATM variants 7271T>G and IVS10–6T>G among women with unilateral and bilateral breast cancer.
Br J Cancer

2003
;
89
:
1513
–6.
(24)
Spurdle AB, Hopper JL, Chen X, McCredie MR, Giles GG, Newman B, et al. No evidence for association of ataxia-telangiectasia mutated gene T2119C and C3161G amino acid substitution variants with risk of breast cancer.
Breast Cancer Res

2002
;
4
:
R15
.
(25)
Stewart GS, Maser RS, Stankovic T, Bressan DA, Kaplan MI, Jaspers NG, et al. The DNA double-strand break repair gene hMRE11 is mutated in individuals with an ataxia-telangiectasia-like disorder.
Cell

1999
;
99
:
577
–87.
(26)
Muir C, Waterhouse J, Mack T, Powell J, et al. Cancer incidence in five continents. Vol V. Lyon (France): IARC;
1987
.
(27)
Parkin DM, Muir CS, Whelan SL, Gao YT, et al. Cancer incidence in five continents. Vol VI. Lyon (France): IARC;
1992
.
(28)
Parkin DM, Whelan SL, Ferlay J, Raymond L, et al. Cancer incidence in five continents. Vol VII. Lyon (France): IARC;
1997
.
(29)
Parkin DM, Whelan SL, Ferlay J, Teppo L, et al. Cancer incidence in five continents. Vol VIII. Lyon (France): IARC;
2002
.
(30)
Waterhouse JAH, Muir C, Correa P, Powell J. Cancer incidence in five continents. Vol III. Lyon (France): IARC;
1976
.
(31)
Waterhouse JAH, Muir C, Shanmugaratnam K, Powell J. Cancer incidence in five continents. Vol IV. Lyon (France): IARC;
1982
.
(32)
Coleman MP, Hermon C, Douglas A. Person-years (PYRS)—a fortran program for cohort study analysis. IARC internal report 89/006. Lyon (France): IARC;
1989
.
(33)
Breslow NE, Day NE. Statistical methods in cancer research. Vol II—the design and analysis of cohort studies. Lyon (France): IARC;
1987
.
(34)
Lange K, Weeks D, Boehnke M. Programs for pedigree analysis: MENDEL, FISHER, and dGENE.
Genet Epidemiol

1988
;
5
:
471
–2.
(35)
Dempster AP, Laird NM, Rubin DB. Maximum likelihood from incomplete data via the EM algorithm (with discussion).
J Roy Stat Soc Series B

1977
;
39
:
1
–38.
(36)
Louis T. Finding the observed information matrix when using the EM algorithm.
J Roy Stat Soc Series B

1982
;
44
:
226
–33.
(37)
Stewart GS, Last JI, Stankovic T, Haites N, Kidd AM, Byrd PJ, et al. Residual ataxia telangiectasia mutated protein function in cells from ataxia telangiectasia patients, with 5762ins137 and 7271T–>G mutations, showing a less severe phenotype.
J Biol Chem

2001
;
276
:
30133
–41.
(38)
McConville CM, Stankovic T, Byrd PJ, McGuire GM, Yao QY, Lennox GG, et al. Mutations associated with variant phenotypes in ataxia-telangiectasia.
Am J Hum Genet

1996
;
59
:
320
–30.
(39)
Sutton IJ, Last JI, Ritchie SJ, Harrington HJ, Byrd PJ, Taylor AM. Adult-onset ataxia telangiectasia due to ATM 5762ins137 mutation homozygosity.
Ann Neurol

2004
;
55
:
891
–5.
(40)
World Health Organization. International statistical classification of diseases and related health problems, 1989 revision. Vol I. Geneva (Switzerland): World Health Organization;
1992
.
(41)
Theunissen JW, Kaplan MI, Hunt PA, Williams BR, Ferguson DO, Alt FW, et al. Checkpoint failure and chromosomal instability without lymphomagenesis in Mre11(ATLD1/ATLD1) mice.
Mol Cell

2003
;
12
:
1511
–23.
(42)
Cavaciuti E, Lauge A, Janin N, Ossian K, Hall J, Stoppa-Lyonnet D, et al. Cancer risk according to type and location of ATM mutation in ataxia-telangiectasia families.
Genes Chromosomes Cancer

2005
;
42
:
1
–9.
(43)
Fernet M, Moullan N, Lauge A, Stoppa-Lyonnet D, Hall J. Cellular responses to ionising radiation of AT heterozygotes: differences between missense and truncating mutation carriers.
Br J Cancer

2004
;
90
:
866
–73.
(44)
Kriege M, Brekelmans CT, Boetes C, Besnard PE, Zonderland HM, Obdeijn IM, et al. Efficacy of MRI and mammography for breast-cancer screening in women with a familial or genetic predisposition.
N Engl J Med

2004
;
351
:
427
–37.
(45)
National Institute of Clinical Excellence Guidelines, CG0143 Familial Breast Cancer. Available at: http://www.nice.org.uk/CG014NICEguideline . [Last accessed: April 28,
2005
.]