The value of MRI for downgrading of breast suspicious lesions detected on ultrasound

According to global cancer statistics in 2018, breast cancer is the most common cancer in women and is the leading cause of death associated with cancer [1]. However, mortality rate of breast cancer has reduced over decades, one of the reasons is detection of breast cancer at an early stage by imaging tools [2]. Ultrasound (US) is the commonly used imaging technique for detection of breast cancer due to its convenience and cost-effectiveness [3, 4].

According to breast imaging reporting and data system (BI-RADS) lexicon, biopsy is recommended for suspicious lesions classified as BI-RADS 4 category. However, positive predictive value (PPV) of BI-RADS 4A and 4B categories on US is relatively low [5‐7], implying that numerous benign biopsies are unnecessary. Although US-guided biopsy has advantages of convenience, low cost and high accuracy, it is challenging to perform biopsy in some situations, for instance, lesions are multiple, patients have contraindication for biopsy. In such clinical settings, additional noninvasive diagnostic imaging test with a high sensitivity would be most welcome.

MRI is a noninvasive modality in breast imaging for plenty of indications, for example, screening of high-risk patients, preoperative staging for newly diagnosed breast cancer [8]. Compared with mammography and US, MRI has a high sensitivity of 90%-100% for breast cancer detection [8, 9]. Moreover, negative findings on MRI are reliable to exclude malignancy due to its high negative predictive value (NPV) [9, 10]. Several studies have found MRI has great values in evaluation of inconclusive findings on mammography [11‐14]. However, few studies investigated the value of MRI exclusively on suspicious findings on US [15].

Therefore, the purpose of our study is to assess the value of MRI for suspicious lesions classified as US BI-RADS 4A and 4B categories.

BI-RADS category has been widely used in clinical practice. According to BI-RADS atlas, lesions classified as 4 category warrant biopsy rather than follow-up. However, nearly 90% biopsies prompted by US yielded benign results in a large prospective multicenter ACRIN 6666 trials [17], resulting in increases of medical cost and psychological burden for patients. Although US-guided biopsy is accessible with high accuracy and less harm, it is invasive and may be challenging to perform in some situations. Thus, the purpose of our study is to investigate the value of MRI for suspicious lesions with low-intermediate risk of malignancy detected on US.

Our study indicated MRI was useful in both mass and non-mass lesions detected on US. In our study, no malignancies were misdiagnosed and most of benign lesions could be downgraded. MRI increased PPV from 34.9% (65/186) to 87.8% (65/74) and reduced false-positive biopsies by 92.6% (112/121). Furthermore, 3 of 4 additional suspicious lesions were proven malignant detected on MRI.

There were some controversies on the value of MRI as a problem-solving modality for suspicious clinical and radiological findings. In 2004, a multicenter study by Bluemke et al. [18] involved 821 suspicious mammography or clinical findings revealed a sensitivity of 88.1%, specificity of 67.7%, NPV of 85% of breast MRI, indicating that MRI could not eliminate the need of biopsy due to its moderate specificity. Subsequently, several meta-analyses suggested MRI might not be conclusive as a problem-solving tool for equivocal findings [19‐21]. Due to lack of evidence in this setting, European Society of Breast Imaging recommends MRI may not be an alternative to biopsy for equivocal findings at conventional imaging, and can be used only when biopsy can not be performed [8]. However, several recent studies have suggested that MRI can be used as a problem-solving tool for plenty of clinical indication and reduces unnecessary benign biopsy. Strobel et al. [22] obtained a sensitivity of 95.5%, specificity of 92% and NPV of 98.9% of MRI on the suspected mammography and US findings. In a study by Spick et al. [10], negative findings on MRI for lesions with BI-RADS 0 category were reliable to exclude malignancy. Additionally, MRI can also be helpful to decrease unnecessary benign biopsy for suspicious calcification and architectural distortion on mammography [11, 12]. Our study provided additional empirical data and supported the aforementioned results. However, in a retrospective study by Sarica et al. [15], the results suggested MRI might not be effective in the assessment of BI-RADS 4 lesions on US, in discordance with ours. In that study, 5 of 79 benign MRI findings were false negative and 60 of 110 MRI suspicious findings were false positive for US BI-RADS 4 lesions, resulting in a low-moderate specificity of 56.7%, NPV of 46.4%. The reasons for the discrepancy between Sarica's study and ours may be as follows: MR imaging interpretations were evaluated solely on the basis of morphological criteria and kinetics in the study of Sarica et al. [15], and DWI was not incorporated into the evaluation. However, there was some overlap on morphology and kinetics between breast benign and malignant lesions, resulting in a relatively low specificity [23]. Several studies have suggested DWI can be used as a valuable adjunct to improve diagnostic accuracy and alleviate false-positive findings [23‐25]. In our study, DWI combined morphological features and kinetics was used to assess breast lesions. Therefore, our study may be more convincing as DWI is widely used in imaging evaluation in clinical practice.

One concern for breast MRI was additionally detected lesions due to its high sensitivity. These additional lesions needed further work-up and potentially limited the usage of diagnostic breast MRI. Previous studies showed detection rate of additional suspicious lesions on MRI was low, but a considerable proportion of these lesions were malignant. Strobel et al. [22] reported 8 of 340 patients (2.4%) had incidental findings, and 3 of 8 (37.5%) were small invasive breast cancer. Spick et al. [26] found that 5.3% (16/302) patients had incidental MRI findings, and 37.5% (6/16) were malignant. In our study, the rate of incidental findings on MRI was 2.2% (4/186), 75% (3/4) of which were malignant. These suspicious additional lesions may have a vital impact on treatment plan, especially for patients with existed breast cancer. In any case, further evaluation is required for the suspicious additional MRI finding. A second-look or MRI-directed US is beneficial in this setting [27].

Another concern for breast MRI was false-negative findings. The reported false-negative rate of MRI was from 0%-7.5% in studies [11, 12, 28]. Majority of false-negative findings were ductal carcinoma in situ, which showed absence of enhancement on MRI. Mammography could assist in reducing false-negative cases of ductal carcinoma in situ, which commonly manifested suspicious microcalcification [29, 30]. Moreover, from pathological point of view, these non-enhanced ductal carcinoma in situ may be biologically dormant [31] and are considered potential overdiagnosis and overtreatment [32]. To sum up, the benefits of MRI to solve suspicious findings and avoid substantial unnecessary benign biopsies outweighed its negative impact.

The PPV was 24.1% (95% CI: 17.4%-31.9%) for US BI-RADS 4A category and 73.2% (95% CI: 57.1%-85.8%) for 4B category in our study, higher than that of ACR benchmarks (4A: 2%-10%; 4B: 10%-50%). Previous studies showed inconsistent results of PPV of US BI-RADS 4 category. Zou et al. [33] and He et al.[7] obtained PPV of 23.7% (52/219) and 70.7% (53/75) versus 13.6% (100/733) and 50% (136/272) for BI-RADS 4A and 4B category respectively. On the one hand, the US BI-RADS subcategories are assessed predominantly depending on experience of doctors rather than objective criteria. Some doctors are very cautions and prefer a lower subcategory in order to decrease anxiety and mental burden of patients. [7, 33] On the other hand, plenty of benign findings were excluded in our study, as some clinicians opted for US-guided biopsy instead of further MRI examination. The selection bias may contribute to high PPV of US 4A and 4B categories in our study.

There are several limitations in our study. Firstly, this was a single center study and the sample size was relatively small. Secondly, 23 lesions assigned as US 4A categroy showed negative findings on MRI and thus biopsy was not performed. These lesions were considered benign by imaging follow-up with mean time of 21.6 months, which may lead to overestimating sensitivity and NPV of MRI. However, among theses 23 lesions, only 5 lesions were classified as MRI BI-RADS 3 category and the rest were classified as MRI BI-RADS 1–2 categories. In view of the low malignancy likelihood of MRI BI-RADS 3 category [34] and approximate 100% of NPV of MRI to exclude malignancy [9, 10, 22], unknown pathology of these 23 lesions may have minor impact on the results. Finally, we did not assess inter-observer variability in imaging analysis among radiologists. In our study, two radiologists evaluated the imaging separately and reached a consensus for disagreement, this double-reading pattern is consistent with daily work practice.

In conclusion, our study indicates MRI yielded high diagnostic performance in suspicious lesions classified as US BI-RADS 4A and 4B category. Breast MRI could be used as a problem-solving tool in such clinical settings to reduce unnecessary benign biopsies without missing malignancy. Furthermore, additional suspicious lesions detected on MRI needed further work-up.