Not all false positive diagnoses are equal: On the prognostic implications of false-positive diagnoses made in breast MRI versus in mammography / digital tomosynthesis screening

Breast magnetic resonance imaging (MRI) is used not only for diagnostic purposes and screening women at high risk of breast cancer but also is increasingly considered as a supplemental screening tool for women at average risk who have dense breast tissue [1‐4]. In all these applications, MRI has yielded consistently higher sensitivity and increased cancer detection rates compared with digital mammography (DM).

Yet, breast MRI has frequently been reported to result in a large number of false-positive diagnoses. Such false-positive diagnoses add to the overall cost of screening because they necessitate additional workup by imaging or biopsy, may cause physical harm because of additional morbidity associated with biopsy procedures, and may cause emotional harm because they may generate breast cancer anxiety in the patient [5‐9]. Accordingly, the reported high number of false-positive diagnoses has been a major reason for limiting the acceptance of breast MRI as a screening tool [10‐12].

False-positive diagnoses in breast imaging are caused by a quite heterogeneous group of tissue changes. In principle, all positive diagnoses that prompt biopsy but yield nonmalignant tissues, including lobular carcinoma in situ/lobular intraepithelial neoplasia (LIN), are equally considered “false-positives.” Current guidelines distinguish between nonproliferative tissue changes (e.g., fibrocystic disease), proliferative tissue changes without atypias (e.g., usual ductal hyperplasia [UDH]), and proliferative changes with atypias (e.g., atypical ductal hyperplasia [ADH], atypical lobular hyperplasia [ALH], LIN) [13‐16]. This classification of nonmalignant tissues has important clinical and prognostic implications. Whereas the presence of regressive, fibrocystic changes is associated with a slightly reduced risk of breast cancer, with a relative risk of 0.65 to 1.01, the relative risk increases by the 2-fold with the presence of proliferative tissue changes. It increases by 2.8-fold if there are multiple coexisting proliferative changes in the same biopsy [15‐21] and even by 4- to 13-fold for proliferative changes that contain atypia [22‐25]. Accordingly, for an average-risk woman undergoing screening for breast cancer, a (false-positive) diagnosis of atypical tissue changes does have important prognostic implications and will impact her further management [26‐29]. This, however, is not reflected by the usual parameters that are used to benchmark the utility of screening programs (i.e., the cancer detection rate and positive predictive value [PPV]) [30].

The objective of this study was therefore to investigate the biological or prognostic significance of tissue changes that cause false-positive imaging diagnoses in MRI vs. radiographic breast imaging. Our aim was to provide a more differentiated, nuanced analysis of the clinical implications of so-called false-positive diagnoses. Because the prevalence of atypical or precancerous tissue changes depend on a woman’s risk of breast cancer, we took care to include only women without breast cancer-associated risk factors. To avoid bias secondary to underestimation or undersampling, the final analysis also included the results of all possible secondary excisional surgical biopsies and procedures done in the cohort.

In this study of 710 women undergoing biopsy for screen-detected DM/DBT and MRI findings, we found that the rate of true- over false-positive diagnoses (i.e., the PPV) was similar, with 43.3% for DM/DBT and 44.8% for MRI. However, the biological or prognostic significance of tissue changes that caused the respective false-positive diagnoses was clearly dissimilar. Whereas nonproliferative, regressive changes were the single most important cause of false-positive diagnoses based on radiographic breast imaging (97 of 195; 49.7%) (i.e., tissues that do not modulate a woman’s risk of subsequent breast cancer), the single most important cause of false-positive MRI diagnoses (81 of 202; 40.1%) was atypical proliferation. Nonproliferative tissue changes were as rarely seen to cause false-positive diagnoses based on MRI as were atypical proliferative tissue changes seen to cause false-positive diagnoses based on DM/DBT (18.8% [38 of 202] vs. 18.0% [35 of 195]).

The benchmarks used for auditing breast cancer screening efficacy are the respective programs’ cancer detection rates and the PPVs [30]. The PPV is a measure that describes the power with which an imaging method is able to classify findings. To calculate the PPV for breast cancer screening programs, all positive imaging findings that are caused by invasive cancer or DCIS are considered “true-positives,” whereas all positive imaging findings that are proven to be caused by anything but cancer or DCIS are considered “false-positives.”

For breast MRI screening, although the cancer detection rate is consistently higher than that achievable with mammographic or ultrasound screening, the high number of false-positive diagnoses (i.e., the low PPVs) that have been reported limit the perceived net benefit of this screening method [36‐39]. However, in keeping with more recent studies on contemporary breast MRI protocols [1, 40‐42], our study demonstrates that screening with MRI can be done with PPVs that are indeed comparable to those achieved with DBT. Moreover, our results suggest that simply comparing numeric PPVs might yield a misleading risk-benefit assessment of different screening methods. A significant fraction of false-positive diagnoses made on the basis of breast MRI were due to high-risk lesions. Diagnosis of such tissue atypias in women who, on the basis of their personal and family histories, were considered to carry an average risk yields important information on the respective women’s future management.

It is well established that MRI-guided biopsies yield many high-risk tissue changes [43‐47]. However, so far, breast MRI, and thus MRI-guided biopsy, has been reserved for women who carry a high lifetime risk of breast cancer or who undergo MRI because of clinical or imaging findings suspicious of breast cancer. Yet, the distribution of histological tissue types, especially the prevalence of atypical proliferation or of high-risk lesions in general, will greatly depend on the individual woman’s risk of breast cancer [13, 15, 16, 19]. Meaningful comparisons of the natural distribution of tissue changes in false-positive MRI- and DM/DBT-guided biopsies are possible only if the respective screening cohorts exhibit similar risk profiles. This is what our study offers; both cohorts (i.e., women undergoing biopsy for positive DM/DBT screening findings or for positive MRI screening findings) were asymptomatic and had no breast cancer-associated risk factors.

On pathological or pathophysiological grounds, it is plausible that the majority of false-positive diagnoses made on the basis of DM and DBT are due to regressive tissue changes. Microcalcifications, especially those caused by benign histological changes, usually represent regressive changes and are not correlated with cell proliferation [48]. In contrast, tissue changes leading to false-positive diagnoses based on MRI are depicted because of their contrast enhancement. Dynamic contrast enhancement in MRI does indeed correlate with tissue proliferation [49].

The association of atypical tissue changes with subsequent breast cancer has been consolidating over the past years. As evidenced by common patterns of genomic additions and deletions, breast cancer progression is a biological continuum, starting from normal breast tissue via flat epithelial atypia, ADH, to DCIS, and then to invasive breast cancer [50, 51]. Tissue changes with atypias such as ADH and ALH would thus not only mark an increased risk of breast cancer but also might represent true precursors of the actual breast cancer that may follow [52, 53]. In keeping with these concepts, recent evidence on the natural history of women diagnosed with ADH or ALH suggests that both histological types are associated with similar degrees of risk, indicating that it is the presence or absence of atypical cells, regardless of their morphology or presumed cell of origin, that drives the risk of subsequent breast cancer [22‐25, 53]. All in all, it appears that the actual risk of subsequent invasive breast cancer is not substantially different for women diagnosed with ADH or ALH vs. low-grade DCIS [23, 53, 54].

Accordingly, we propose a careful discussion on what constitutes a true-positive vs. a “false-positive” diagnosis in women at average risk who undergo screening for breast cancer. If it is agreeable that a true-positive diagnosis should be one that has an impact on a woman’s further management (i.e., an “actionable” diagnosis), then diagnosis of atypical tissue changes could indeed be considered a true-positive diagnosis. Most practice guidelines recommend intensified surveillance, preventive surgery, or even chemoprevention for women with a tissue diagnosis of ADH, ALH, or LIN [26‐29]. Regardless of how we label false-positive imaging diagnoses, breast radiologists should be aware of the fact that not all false-positive diagnoses are equal in that not all are only unwanted side effects of the desire to establish a diagnosis of breast cancer early, but many, especially many false-positive MRI diagnoses, provide valuable information that is helpful for guiding further patient management. Last, but especially important, we suggest that this information be given to women with dense breasts when they are counseled about the advantages and disadvantages of different supplemental breast cancer screening methods.

A possible limitation of our study is the fact that the breast density distribution of the study population that underwent MRI was slightly shifted toward those with denser breast tissue, in accordance with the clinical use of MRI for supplemental screening of women with dense breast tissue. However, it has previously been shown that breast density and rate of proliferative breast disease are not correlated [19]. The switch from pure DM to DM/DBT screening may have influenced our false-positive biopsy rate for radiographic screening. However, the PPV observed for DM/DBT in our cohort is within the range of PPVs expected for routine screening mammography, so the impact of switching of technology, if it exists, should not be very large.

One may argue that the MRI cohort was “enriched” for high-risk lesions because the lower-risk lesions were verified by DM/DBT. However, the null hypothesis of our study was that the distribution of low- and high-risk tissue changes is identical in women with DM/DBT-positive vs. MRI-positive screening findings. If the null hypothesis was to be accepted, then no enrichment would be possible, because high- and low-risk lesions would have been biopsied with similar frequency. Yet, our null hypothesis must be rejected; the distribution is indeed quite dissimilar. Accordingly, just because DM/DBT will indeed preferably pick up low-risk findings is an enrichment at all conceivable. Such enrichment would thus be a consequence of the very findings made in this study. Furthermore, the degree of enrichment is determined by the degree of overlap (i.e., the number of women who had suspicious findings on both MRI and DM/DBT). However, this number was small, with 1.5% of DM/DBT-positive findings also yielding MRI-positive results, all representing intermediate- or high-risk lesions based on histology. Accordingly, the effect of enrichment in our cohort is likely negligible.

Our study does not merely confirm previous results on the similar PPVs (PPV3) of radiographic (DM/DBT) and breast MRI. We demonstrate that the prognostic relevance of the false-positive diagnoses that drive the respective PPVs differs depending on the imaging method, with high-risk lesions being the predominant cause of false-positive MRI diagnoses, as compared with low-risk lesions being the predominant cause of false-positive radiographic diagnoses. This should be considered when discussing the rate of false-positive findings in MRI vs. radiographic breast cancer screening and should be explained when counseling women about their choices and the risks and benefits of supplemental screening.