To the Editor:

Section:

ChooseTop of pageTo the Editor: <<REFERENCES

In a prospective cohort study, Kuhl et al¹ reported on the outcome of a multimodal screening program for women with elevated risk for breast cancer. They analyzed 1,679 screening rounds in 687 women that led to the detection of 27 breast cancers. We want to point out that these women were recruited within the German Consortium for Hereditary Breast and Ovarian Cancer (GC-HBOC), which comprises 12 centers, and that they were subjected to a screening program developed within the GC-HBOC. By February 2010, the GC-HBOC had accumulated comprehensive screening data on 3,002 women who were at risk and underwent 12,042 screening rounds. The results presented by Kuhl et al¹ are, however, restricted to only four centers and a limited time span. In our opinion, they do not justify the conclusions drawn by the authors. The following points describe the details.

1. As has already been pointed out by the accompanying editorial by Jan G. M. Klijn,² Kuhl et al reported on a heterogeneous group of patients with a wide range of risk. Only 10x0025; of the women were carriers of a deleterious BRCA1 or BRCA2 mutation, whereas 39% had a remaining lifetime risk of only 30% or less, which represents a moderate risk. A high rate of preinvasive lesions was mainly seen in patients without a BRCA mutation, where ductal carcinoma in situ accounted for 45% of the detected tumors. Evidence that tumors in the moderate-risk group may have a more favorable prognosis, thereby raising the problem of overdiagnosis, comes from a Dutch study³ that was also referred to in the editorial.² In our own study based on an overlapping cohort of women, we found that tumor incidence rates were only marginally elevated in women at moderate risk and did not increase before the age of 50 years in this subgroup.⁴ Therefore, we would have been interested in seeing data from the study by Kuhl et al¹ plotting the tumor detection rate and false-positive rate against risk status. At present, the GC-HBOC recommends annual surveillance only for mutation carriers and women with a high risk of 30% or greater.

2. Screening examinations did not start in October 2002, as indicated by Kuhl et al;¹ most of the women who were participating had already been included in the screening program of the GC-HBOC before the study. This has previously been stated by the authors themselves⁵ and is perfectly reflected by the fact that incidence rates in the first screening round (prevalent screens) are not higher than in the following rounds (incident screens).

3. Given that the women had already undergone screening rounds previously, it is not surprising that only two breast cancers were diagnosed by suspicious calcifications in mammography, because these slowly emerging lesions were most likely removed in previous screening rounds. Therefore, in our opinion, it is misleading to conclude that calcifications are rather rare in this patient cohort and that mammography does not add to the detection of preinvasive carcinomas.

4. It is confusing that the previous publication of first results from the EVA study⁵ do not match with current results (ie, the 10 cancers detected in the first screening round in the previous publication [Table 2 in the article⁵] are not congruent with the 10 tumors of the first screening round that are described now with respect to size and nodal stage [Table 3 in the article⁵]). In addition, all three cancers detected in women younger than 40 years who were presented in the earlier publication in Tables 2 and 4 do not appear in Appendix Table A2 of the current article¹ with regard to age of onset, tumor stage, and diagnostic accuracy of the different imaging modalities. Furthermore, the two tumors detected solely by mammography in the previous article⁵ are different from the two tumors described in the current article.¹

5. Breast cancer incidence rates per 1,000 screening rounds are depicted incorrectly as percentages instead of per thousands in Results and Appendix Figure A1 (ie, 15.5% is not 15.5 per 1,000 screening rounds). Moreover, we would like to emphasize that the number of women-years equals the number of screening rounds only under the condition that the mean interval was actually 1 year. It would be interesting to see data on the adherence to the planned 12-month intervals as that has an influence on the estimated cancer incidence rates.

6. Kuhl et al⁶ as well as others^7,8 previously published on specific imaging phenotypes in hereditary breast cancers that exhibited benign kinetics and morphologic features such as nonmass-like enhancements and smooth mass margins. BRCA-associated breast cancers bear resemblance to fibroadenomas, which is especially true for sonographic appearance. However, implications of these findings (ie, the interpretation of Breast Imaging Reporting and Data System-3–like lesions) are not discussed in the article. Rather, receiver operating characteristics provided in Figure 2 of the article¹ indicate a negligible increase in ultrasound sensitivity by including Breast Imaging Reporting and Data System-2/3 lesions.¹ This is surprising as, in our hands, results-oriented training of readers led to an increase in ultrasound sensitivity from 35% to 75% in an overlapping patient cohort.⁸

7. The primary message of the article is that MRI shifts the distribution of screen-detected tumors toward preinvasive stages and that this is correlated with mortality reduction. However, it is well accepted that hereditary breast cancers, especially BRCA1-associated breast cancers, exhibit a severe spectrum bias not only with respect to different imaging phenotypes^6-8 but also with respect to different disease course. This comprises an impaired prognostic value of node-negative tumor stages^9,10 and a higher tumor doubling rate,¹¹ as is also outlined in the editorial.² On the other hand, women with a moderate risk may preferentially develop slowly growing tumors, which harbor the risk of overdiagnosis by the applied surveillance program, as was outlined. Therefore, it is mandatory to await the exclusion of a potential spectrum bias in BRCA1/2-negative familial breast cancers as well. The mere downward shift of tumor stages does not allow the conclusion of a survival benefit in these subgroups.

Taking together, with regard to the heterogeneous study cohort, the disregard of the results of previous screening rounds as well as specific imaging criteria of BRCA-associated tumors and a spectrum bias of familial breast cancers, Kuhl et al¹ are too rash in their conclusion that “the detection of high-grade [ductal carcinoma in situ] or of small, node-negative breast cancers is closely related to a reduction in breast cancer mortality” by extrapolating from data on sporadic breast cancers. In fact, it is mandatory to await data on mortality rates in the targeted risk groups before recommendations on screening modalities can be announced. Large consortia such as the GC-HBOC are dedicated to gain such data to improve life expectancy of women burdened with familial breast cancer risk.

The author(s) indicated no potential conflicts of interest.