|Home | About | Journals | Submit | Contact Us | Français|
A UK multicentre study compared the performance of contrast enhanced magnetic resonance imaging with X-Ray Mammography in women at high-risk of breast cancer commencing in 1997. Selection criteria were used to identify women with at least 0.9% annual risk of breast cancer
Women at high breast cancer risk, with a strong family history and/or high probability of a BRCA1/BRCA2/TP53 mutation were recruited from 22 centres. Those not known as gene carriers were asked to give a blood sample, which was tested anonymously for mutations. Women aged 35-49 years were offered annual screening for 2-7 years. Study eligibility at entry was assessed retrospectively by detailed examination of pedigrees and overall eligibility accounting for computer risk assessment and mutation results.
Seventy-eight of 837 (9%) women entered for screening were ineligible using the strict entry criteria. Thirty-nine cancers were detected in 1869 women-years in study (incidence 21 per 1000). Including 3561 further years follow up 28 more breast cancers were identified (12/1000). Incidence rates for 759 eligible women were 22/1000 in study and 13/1000 in total follow up, compared with 9/1000 and 4/1000 respectively in 78 ineligible women. Breast cancer rates were higher for BRCA2 than BRCA1 after testing anonymised samples in this selected population at 65/1000 in study and 36/1000 in total follow-up for BRCA2 compared with 44/1000 and 27/1000 for BRCA1
Strict enforcement of study criteria would have minimally improved the power of the study, whilst testing for BRCA1/2 in advance would have substantially increased the detection rates
Imaging surveillance for women at high risk of breast cancer requires a strong evidence base of demonstrated effectiveness to guide practice. Five studies of contrast enhanced breast MRI (CE MRI) have now evaluated this effectiveness in high risk women showing very high sensitivity for cancer detection compared with X-Ray Mammography (XRM) [1-5]. Furthermore two studies of cost effectiveness have shown that CE MRI is likely to be cost effective in women at high risk who are <50 years of age [6,7]. However, for CE MRI to be cost effective women need to be selected on the basis of a risk approaching 1% per year of developing breast cancer [6-8]. National programmes for MRI screening have now been introduced in a number of countries and in the UK the screening recommended by The National Institutes of Health and Clinical Excellence (NICE)  will become part of the NHS breast screening programme in 2009. We have evaluated the effect of enforcement of the original strict entry criteria of the UK MARIBS study on breast cancer incidence rates. The study involved recruitment by a multicentre collaboration of radiology and genetics units in 22 centres. It was a prospective study of asymptomatic, high risk women, comparing yearly CE MRI with XRM in women aged 35-49 years at entry (25-49 years using MRI alone for Li Fraumeni syndrome). [1,9]
Women were selected for eligibility using the criteria in table 1. All participating women gave written informed consent and the protocol and documentation were approved by the London Multicentre Research Ethics Committee (MREC) and all the relevant Local committees (LREC). The criteria were focussed on women with a high likelihood (at least 30%) of carrying a BRCA1, BRCA2 or TP53 mutation . The annual risk of breast cancer should therefore be at least 0.9%. An eligibility panel, composed of three members of the study advisory group (RE/DGE/DE), adjudicated on cases in doubt. Women at this level of risk in the United Kingdom (UK) have previously received annual mammography screening from age 35 years, or at a younger age if their first-degree relative developed cancer at a younger age than 35 . Since the end of the MARIBS study these women have returned to mammographic screening alone, with very few getting continuing MRI. Neither regular physical examination nor screening ultrasound has been generally applied for breast cancer screening in the UK in normal or high-risk groups.
Women whose genetic status was not known were asked to provide an anonymised blood sample at the start of the study for future mutation screening in BRCA1, BRCA2 and if appropriate TP53. At the end of the study, the women who had developed cancer, but who had not previously had predictive genetic testing, had anonymous testing by Myriad Genetic Laboratories. These results were reported in our original paper . Since that time testing of the other samples provided has been undertaken by sequencing of the three genes and screening for large genomic rearrangements of all three genes by Multiple Ligation dependant Probe Amplification (MLPA) . This testing was conducted solely for the purpose of this study, and the ethics committee required that this result should not be known to the woman or her physician, although it can be released to the centre if the woman subsequently asks for testing to aid mutation confirmation.
Women with previous breast cancer were excluded and women with any other cancer such that their expected prognosis was less than five years were ineligible. Subjects who underwent predictive genetic testing for a mutation known to be present in their family, during the course of the study and were found to be negative and women who developed cancer were excluded from further participation in the study.
Recruitment began in August 1997 and finished in March 2003. Screening ceased in May 2004, by which time all women had had an opportunity for at least two annual MRI scans. Eight hundred and thirty seven women were recruited to the study. In some centres logistic problems and a time lapse resulted in women who had consented never actually participating. Seven hundred and thirty two underwent screening resulting in 2065 CE MRI examinations and 1973 XRM studies. During the study, 184 examinations in 85 women were performed using CE MRI alone. These included 114 examinations on 36 women who were either known carriers of a TP53 mutation or at 50% risk of being mutation carriers on the basis of family history consistent with Li-Fraumeni syndrome. Ninety-two mammograms were undertaken in 81 women who did not have CE MRI.
Recruitment occurred over a five-year period resulting in a wide variation in the number of screening episodes for individuals (1-7 annual screening events). Twenty-two centres participated (see Appendix).
The main objective of the original study was to compare the sensitivity and specificity for malignancy of XRM and CE MRI in women at high genetic risk of breast cancer and this has been reported . Ascertainment of interval cases was undertaken by sending a follow up questionnaire to study participants, and by contacting each of the study centres. Women were also flagged at the Office of National Statistics (ONS) to ascertain subsequent cancer incidence and mortality. Two time periods were defined: “in study” refers to follow up from first scan to last scan prior to May 2004; “during extended follow up” refers to cancers occurring up to final follow up. Final follow up including at least 52 months from when the last MRI scan was censored at 31st July 2007.
Although study centres undertook to check eligibility for entry into the screening study, an anonymised pedigree was also sent to the study centre to allow for an independent retrospective assessment of eligibility. This was assessed by the panel to evaluate eligibility based on the strict criteria given in table 1, but also incorporating a computer assessment of genetic risk (60% likelihood that a family harboured a mutation) and annual incidence rate of breast cancer of at least 0.9% using the BOADICEA (Breast and Ovarian Analysis of Disease Incidence and Carrier Estimation Algorithm) programme . If individuals clearly fulfilled the criteria listed in table 1  they were deemed eligible. Borderline cases were then assessed by a combination of clinical judgement (DGE/RE/DE) and by BOADICEA (DT & DE). The eligible population was, therefore, assumed to include both BRCA1 and BRCA2 mutation carriers (with an approximate predicted cancer incidence rate of 3% per annum) and women with a strong family history of breast cancer (approximate predicted cancer incidence rate of 0.9% per annum), with an overall target incidence of 1.4% per annum. (In the published analyses, the observed incidence rate was 1.9% per annum ). To achieve the required power, the study was designed to recruit 500 women annually for 3 years, with follow-up over a five-year period (i.e. 2-4 follow-up scans), such that approximately 6000 scans would be performed .
In 2001, sample size calculations were revised, assuming a 90% sensitivity for MRI (i.e. a 20% difference) in line with more recent data [1,13]. The revised targets were for 3300 scans in 950 women, with 46 cancers predicted. The timeline of the study can be seen in Figure 1. Final follow up (after the extended period) was at a mean of 7 years, therefore Kaplan Meier curves were utilised to assess breast cancer incidence at this time point.
Eight hundred and thirty seven women were entered into the MARIBS study. We were able to assess eligibility in all these women (Tables (Tables22 and and3).3). Seven hundred and fifty nine (91%) women from 612 families were considered by the panel to be eligible for the study at entry. Seventy-eight women from 57 families were considered by the panel to be ineligible. The main reason for ineligibility was failure to meet the criteria of 4 breast cancers <60 years with at least two confirmed. However, a much higher proportion of those entered as Li Fraumeni Syndrome (LFS) were ineligible using the strict LFS criteria in table 1.
In the 837 women entered for screening by CE MRI, 39 breast cancers were detected in 1869 years in study (incidence 21 per 1000). When we included a further 3561 years of extended follow up, a further 28 breast cancers were identified for a combined total 12 per 1000 incidence. Incidence rates for the 759 eligible women were 22 per 1000 in study and 13 per 1000 in total follow up, compared to 9 per 1000 and 4 per 1000 in the 78 ineligible women.
All groups in table 3 met the minimum 0.9% detection rates, but the 4+ ineligible group did not meet the target 1.4% per annum overall target. If the BRCA1/2 group were taken alone this gave the highest power to the study. Excluding those who tested negative during the study and for a known family mutation after anonymised testing the overall cancer incidence in individuals still eligible rose to 26 and 16 per 1000 women years.
Of the 837 women entered into this study, 66 had tested positive for a BRCA1 mutation and 27 for a BRCA2 mutation prior to study entry (Figure 1). Five further women with a BRCA2 mutation were identified through anonymous testing, having developed a breast cancer during the study. Clinical testing during the study and anonymised testing post-study identified 30 further BRCA1 and 22 BRCA2 carriers in families with known mutations and 39 BRCA1 and 41 BRCA2 mutations in families without a known mutation. Therefore there were 135 BRCA1 and 90 BRCA2 mutation carriers in the study. Thus 81/301 (27%) of the tested eligible samples in the 4+ group had pathogenic BRCA1/2 mutations. Rates of breast cancer were higher for BRCA2 than BRCA1 after testing anonymised samples in this selected population at 65 per 1000 in study and 36 per 1000 overall for BRCA2 compared with 44 and 27 per 1000 for BRCA1 (Table 3). This is largely due to the occurrence of 10 breast cancers amongst the 41 BRCA2 carriers identified from the families with no previously known mutation compared with only six in the 40 BRCA1 carriers identified in families with no previous known mutations.
If all samples had been tested prior to commencement in study 171/837 (20%) would have been ineligible, and one cancer would have been missed (table 3). However, the rate in the 4+ eligible category who had uninformative testing was still 12 per 1000 annual rate in study, although this fell to 7 per 1000 annually in total follow up. The three breast cancers occurring in women identified by the panel as ineligible were all in women retrospectively identified as BRCA1/2 mutation carriers in anonymised testing and a further 6 women deemed ineligible by the panel were also found retrospectively to have a mutation. One cancer occurred in a woman who was initially eligible but subsequently tested negative for a family mutation. Seven ineligible women were found not to have a family BRCA1/2 mutation. Of women tested who did not meet eligibility criteria at trial entry 9/50 (18%) were found to have mutations and 28 did not supply a sample for testing. No cancers occurred amongst the 69/78 ineligible women who did not have a mutation identified. Breast cancer incidence from trial registration is shown for all groups in Figure 2. BRCA2 eligible individuals had a cumulative risk of breast cancer of 22.3% (95%CI 18.0-26.6) at 7 years, which was significantly higher than for BRCA1 at 14.7% (95%CI 11.8-17.6%). The total number of known mutation carriers and an assessment of likely total carrier numbers for the study is shown in table 4. A surrogate for penetrance estimates is the reduction in rates of those testing positive for a known family mutation. Of those still unaffected on censor day 30/82 (37%) of those originally at 50% risk for BRCA1; 22/66 (33%) of BRCA2 and 0/5 (0%) TP53 tested positive in anonymous testing.
We have analysed the breast cancer incidence both in study and in extended follow up post MRI screening in a large cohort of 837 high-risk women. The incidence rates for breast cancer were high with 21 per 1000 developing breast cancer annually in study and 12 per 1000 including overall extended follow up. There is a drop off in incidence rates after the study despite the fact that the women were getting older and the rates would be expected to rise. This is likely to be because MRI has not been available to the majority of women post study, due to health service issues in the UK. The rates during study are also boosted by the initial prevalence scan. Detection rates were 26.9 per 1000 at prevalence and thereafter 12.8 per 1000 in study . Therefore women had an initial scan that produced a lead time effect and after study completion the lead time was lost resulting in an apparent decrease in breast cancer incidence despite the fact that women were getting older.
The overall breast cancer incidence in study easily exceeded the required power of 9 per 1000. Indeed even the 78 women who were entered who did not meet the strict eligibility criteria had an incidence that exactly met the 9 per 1000 incidence on study. However, this was not adjusted for the lead time bias and dropped to only 4 per 1000 overall including the extended follow-up period. If BRCA1/2 testing had been carried out on all women before study entry, the efficiency of the study would have increased as only one cancer occurred in the 171 women who were ineligible by study criteria and/or negative on gene testing in families with proven mutations. This group had a rate of breast cancer of only 0.4 per thousand per year equivalent to the risk for the average woman in the UK aged 30-39 years . The only cancer that did occur in this group was in a woman originally eligible who tested negative for a family mutation and as such was a “phenocopy” The 1 per thousand rate amongst those with “true” negative tests does not support nor disprove a possible increased risk amongst this group of women . Even among the women meeting study criteria the overall incidence dropped to only 7 per 1000 after excluding BRCA1/2 mutation families tested during or post study. Given the expense of MRI screening (~£300 annually) and the relatively low cost of a one off genetic test (£600-1000) a case could be made to test women from families with no known mutation who would otherwise qualify for MRI screening. This would obviously be a matter of patient choice, but in many families in the UK testing is not currently possible as there is no living affected family member to test . The “negative” or “uninformative” test in families with no previous testing may in fact represent a “true” negative particularly in our highly selected patient cohort. This is reflected in the low breast cancer incidence in the 4+ eligible group after “uninformative” testing. There is therefore clear cost utility in offering women, who could have 20 years of MRI screening at a cost of £6,000, a test which would be reassuring regarding their breast cancer risk (an incidence of only 5.5 per 1000 with a mean of nearly 7 years follow up) and could save the health system in excess of £5,000 per woman. A suggested new algorithm for offering testing based on MRI eligibility is presented in Figure 3. Although testing of unaffected individuals without a known family mutation is not currently covered by NICE guidance, the 20% threshold for testing would still be maintained  and there would be a clear cost utility in saving of MRI screening in those testing negative. We therefore propose that consideration is given to BRCA1/2 testing in unaffected women whose risk of having a mutation is at least 20% (40% chance of a mutation in the family).
The overall ineligibility rate in the MARIBS study of around 9% did not substantially affect the study power. However, there was a strikingly high rate of ineligibility amongst the LFS criteria group. When families with a TP53 mutation are excluded, 44% of women entered into this study were shown to be ineligible. We are unsure why the written LFS criteria proved more difficult to apply than the Breast Cancer Linkage Consortium (BCLC) criteria outlined in table 1 for breast/ovarian cancer. It is nonetheless possible that geneticists used their own judgement or other published criteria for LFS [15,16] to assess eligibility. No breast cancers occurred amongst this ineligible group and in retrospect a more strict approach to study entry for this group would have been warranted. Another issue is the accuracy of LFS family history . Schneider et al reported on 191 cancer diagnoses among relatives reported by 32 LFS and 52 breast/ovary participants in genetic testing programs. Cancer diagnoses of relatives were more accurately reported in the breast/ovary cohort (78%) than in the LFS cohort (52%). Almost all breast cancer diagnoses were accurately reported, this compares to 74% of ovarian cancer diagnoses and only 55% of other LFS-related cancers that were accurately reported .
Another striking finding in our study is the very high rate of breast cancer incidence amongst BRCA2 carriers. This is not in keeping with the very low penetrance estimates for BRCA2 of 40-49% reported in recent papers [18,19]. In one of these papers derived from a meta-analysis for all patients , clinicians were advised to use the 49% penetrance estimate by 70 years. This advice may be misleading. In MARIBS, the 3.6% annual risk over 5.6 years mean follow up for 108 BRCA2 carriers and those at 50% risk would breach the penetrance estimate from the meta analysis within 14 years. Our figure does seem high even for a highly selected group of women, but all these women were clinically unaffected at study entry and any bias in testing affected women has been overcome by testing all provided samples. The extended follow up also reflects the “loss” of the lead time bias as very few women continued to receive MRI screening. The high incidence amongst BRCA2 mutation carriers could be partially explained by the fact that the selection for genetic testing in the UK is stricter than in most countries  and therefore the great majority of BRCA2 families entered to MARIBS had a strong family history of breast cancer. There is also emerging evidence of the fact that breast cancer risk associated with BRCA2 is more modifiable than BRCA1 . However, increased mammographic density, which has a heritable component is a strong modifier of risk for both BRCA1 and BRCA2 . The very high rate of breast cancers in MARIBS among the previously untested BRCA2 families with multiple early onset cases also adds weight to the premise that selection truly affects breast cancer incidence and that many recent attempts to correct for ascertainment bias may have “overcorrected” . Penetrance estimates for BRCA1 and BRCA2 from the Breast Cancer Linkage Consortium (BCLC)  give very similar risks for BRCA1 and BRCA2. As a higher proportion of BRCA1 carriers were identified in families not fulfilling BCLC criteria, this could at least partly explain the lower risks in BRCA1 carriers. Support for the higher penetrance amongst BRCA2 carriers also comes from table 4 where it can be seen that only 33% of individuals at 50% risk of a BRCA2 mutation who were still unaffected at 31st July 2007 were mutation positive compared to 37% for BRCA1. The evidence from our prospective study would suggest that higher estimates of breast cancer risk are given to women who carry BRCA2 mutations and come from highly selected families, in line with previous estimates from these families . Use of a model such as BOADICEA  allows adjustment for this extra familial effect almost certainly due to modifier genes . This study shows that in families highly selected for breast cancer, prospective incidence is high and risks given to unaffected female mutation carriers should reflect the family pattern.
Our study has shown that in highly ascertained families the incidence of breast cancer in BRCA1 and BRCA2 carriers is very high and further justifies MRI screening for these women. In individuals tested negative for BRCA1/2 without a known family mutation, rates of breast cancer may not justify MRI screening. However, testing of unaffected females at potential high risk of a mutation in a family without previous testing may be justified to assess eligibility for MRI.
The national study was supported by a grant from the Medical Research Council (G9600413) and Cancer Research UK Project Grant G5047/A5830. The cost of the MRI studies was paid for from subvention funding for research from the United Kingdom National Health Service. The protocol is based in part on developments supported by the Cancer Research UK and the Yorkshire Cancer Research Campaign. Contributions towards training and education have been made by Schering Healthcare Ltd. and Oracle Education. DFE and DT are funded by Cancer Research UK. RE is funded by The Institute of Cancer Research and HEFCE. SJR is funded by the Mermaid component of the Eve Appeal. This work was carried out at three centres who received a proportion of funding from the Department of Health’s NIHR Biomedical Research Centres funding scheme namely: Central Manchester Foundation Trust, the Royal Marsden Foundation Trust and University College Hospital London.
The study is a UK-wide collaboration of 22 genetics centres, and their associated MRI and mammography departments. The key clinical contributors are named in the Appendix. We acknowledge the work of many others, radiographers, nurses, clerical staff, physicists, engineers, whose contribution is important but who have not been named. A special thank you is extended to the women and their surgeons and oncologists who referred them, without whom the study would not have been possible.
This work was supported by a project grant from the UK Medical Research Council (G09600413) and Cancer Research UK (G5047/A5830)