Effectiveness of MRI in screening women for breast cancer: a systematic review

Abstract

Artificial intelligence techniques for the diagnosis of disease continue to develop with rapid pace. This review article systematically determines incremental accuracy and other parameters of current methods, including sensitivity, specificity, positive predictive value and negative predictive value with regard to breast MRI as a screening tool for women under 50 years. Articles were included from the databases of health technology assessment agencies from 2000 to 2019, using various medical subject heading terms. A total of 23 eligible studies were included incorporating a total of 11,688 patients out of which two were multicentered, four were accuracy studies, seven were prospective studies and four were retrospective studies. MRI screening showed an adequate detection of invasive cancers, premalignant lesions and pre-invasive cancers, suggesting that MRI is a powerful surveillance tool to detect cancer in high-risk populations. These findings have indicated that MRI has particular sensitivity and specificity for the diagnosis of breast cancer.

PROSPERO Registration Number: CRD42020158372.

Keywords: :

• artificial intelligence
• breast cancer
• MRI
• screening

Practice points

• MRI has particular sensitivity and specificity for the diagnosis of breast cancer.

• This review has systematically determined incremental accuracy and other parameters of current methods (sensitivity, specificity, positive predictive value, and negative predictive value) with regard to breast MRI as a screening tool for women.

• The results showed MRI to be more precise and sensitive than mammography for annual surveillance of patients with hereditary risk.

• MRI screening is beneficial specifically for BRCA2 carriers, while screening parameters are comparable to population screening data, with more favorable results in the youngest patients (aged less than 40 years) who are BRCA1/2 carriers.

• MRI has the potential to give false negative results in the screening of patients with high risk of breast cancer, although it is cost effective and has high sensitivity.

• An MRI session in breast cancer patients would likely be a cost-effective strategy for higher risk patients.

The prevalence rate of breast cancer in women ranges from 19.3 per 100,000 women in Africa to 89.9 per 100,000 women in Europe, Asia and America [1]. There are many risk factors that are increasing the incidence of breast cancer in specific populations, in particular, genetic factors [2]. A predisposing gene mutation or a strong family history of breast cancer among women has an increasing lifetime risk of developing breast cancer [3]. The morbidity and mortality of this cancer can be ameliorated using prophylactic oophorectomy, chemoprevention and mastectomy with a selective estrogen receptor modulator [4]. Early detection in high-risk women by means of breast imaging is highly recommended since most risk factors are not preventable [5].

Mammography is less sensitive in younger women and those with a genetic predisposition to breast cancer, although it is broadly necessary to screen postmenopausal women [6,7]. The effectiveness of ultrasound is becoming increasingly evident for the effective detection of breast cancer in younger women [8,9]. Emerging evidence, on the potential patient-related advantages of MRI has led to its increased use in the detection of breast cancer [10]. Evidence for the effectiveness of MRI has been increasingly reported in the US and UK, although a number of institutions in various countries question whether there is yet enough evidence to integrate recommendations based on this research [11]. Nevertheless, a review of the evidence in this regard guides screening decisions, especially considering the current uncertainty of possible harms and benefits of MRI [12]. For particularly high-risk groups, MRI has found to have greater sensitivity in comparison to ultrasound and mammography in the detection of invasive cancer. The smallest lesions can now be detected with the assistance of high-powered MRI machines. Thus, this modality is becoming crucial for early detection in such patient groups [13].

At present, as far as the present author is aware, there are no systematic reviews that examine the incremental value of MRI over imaging using ultrasound and mammography in young high-risk women. A systematic review may help in determining the incremental accuracy of breast MRI as a screening tool for women under 50 years, and to examine other parameters, such as sensitivity, specificity, positive predictive value and negative predictive value. Moreover, this study evaluates the reliability of evidence for early detection of breast cancer in this population.

Materials & methods

Search strategy

Electronic websites and databases were used to extract information using medical terms as search criteria between the years 2019–2000. Also, various text words were used to extract relevant articles, such as MRI and breast cancer diagnosis, MRI for women at high risk of breast cancer, MRI and dense breast tissues, MRI and screening tool, breast MRI and young patients and MRI and BRCA gene (See Supplementary File 1). The scanned citation abstracts were also scanned again by two independent reviewers to identifying articles that evaluate the efficacy of MRI in asymptomatic high-risk women, and they reported adequate data for computation of the incremental sensitivity and specificity of MRI, by means of histology, as the reference standard.

Inclusion & exclusion criteria

Articles of all types were included in this review, such as cohort studies, diagnostic accuracy studies, cross-sectional studies and clinical trials. Articles examining the value of MRI in screening women with a family history of breast cancer, or other high-risk groups of developing breast cancer, are given focus. Articles were excluded if no full text was retrievable or if they were published in journals without an assigned impact factor.

Quality appraisal

The Quality Assessment of Diagnostic Accuracy Studies 2 (QUADAS-2) tool was used to assess the quality of studies by the two independent reviewers. The classification of studies was given based on the following criteria:

• First criteria (Low quality): Studies not meeting the criteria or not conducted prospectively for an appropriate reference standard or test interval.

• Second criteria (High quality): Studies conducted prospectively; studies validated the results of study tests; studies interpreted test results without the reference standard knowledge; studies explained subject withdrawals; studies reported indeterminate test results; and studies reported on the execution of study tests.

• Third criteria (Fair quality): Studies that does not meet all the criteria for a high-quality study, but it has minimal bias and meets most of the criteria and cannot be classified as low-quality.

Data analysis

The study has estimated specificity, sensitivity, relative risk reduction and cost–effectiveness from the data using a random-effects model.

Results

Study characteristics

A total of 11,688 articles were examined initially, which resulted in 23 eligible articles after reviewing the titles and abstracts found in the electronic search and also removing duplicates. The majority of these 23 studies were based in a single center, while two were multicentered studies [11,12]. A total of four studies were accuracy studies [14–17], four were retrospective studies [10,12,13,18] and seven were prospective studies [7,11,15,19–22].

MRI for women at high risks breast cancer

A total of six studies have used MRI for detecting high-risk association with breast cancer [16–18,20,23,24]. According to Stoutjesdijk et al. [16] MRI is an accurate and precise tool compared with mammography in annual breast cancer surveillance of patients with a hereditary risk (Table 1). Breast MRI was found to be effective in the screening of dense breasts in young women at high-risk [24]. Also, MRI showed to be sufficient in the detection of invasive cancers, premalignant lesions and pre-invasive cancers, which suggest that MRI is a powerful surveillance tool in the detection of breast cancer in high-risk populations [25]. The absolute density volume in MRI predicts breast cancer risk that requires additional investigation. This volumetric breast density is actually highly correlated with mammographic density, but these are not equally measured as in MRI [21]. For that, the use of MRI to assess high-risk groups has important potential in the detection of mammographically occult cancers [26].

Table 1. Quality of assessment and findings of studies related to high-risk breast cancer.

Quality assessment			Summary of findings
Author	Design	Centers	Patients	Age group	Relative reduction	Sensitivity	Specificity	Ref.
Stoutjesdijk et al.	Retrospective cohort	1	112	25–65	RR = 0.54	42%	96%	[16]
Tilanus et al.	Prospective design	1	109	25–55	RR = 0.79	45%	67%	[17]
Riedl et al.	Prospective comparison study	1	327	5–66	RR = 0.90	88.9%	19.7%	[18]
Thompson et al.	Prospective study	1	837	31–49	RR = 0.76	–	–	[23]
Brekelmans et al.	Prospective study	1	1198	21–70	RR = 23.7	–	–	[20]
Weinstein et al.	Prospective study	1	609	25–80	RR = 0.28	–	–	[24]

RR: Relative risk.

Screening parameters were comparable to population screening data, with more favorable results in the youngest age group (<40) with BRCA1/2 carriers as suggested by Hagen et al. [22] Gareth et al. [12] have indicated that MRI screening is beneficial specifically for BRCA2 carriers.

Kriege et al. [15] have screened 1909 eligible women, including 358 carriers of germ-line mutations. A total of 51 tumors and one lobular carcinoma in situ were identified throughout a median follow-up period of 2.9 years. It has been reported that the sensitivity of clinical MRI in the detection of invasive breast cancer is higher compared with mammography. Kuhl et al. [21] investigated a cohort study of asymptomatic women that were detected or verified for having a breast cancer vulnerability gene on the basis of their family history of mutational analysis. A total of 43 breast cancers were detected, which include nine ductal carcinomas and 34 invasive carcinomas in situ that showed the high sensitivity for high-risk breast cancer patients (100%) in comparison to mammography (25%). Hagen et al. [22] examined women with truncating mutation in either BRCA1 or BRCA2 genes at a counseling genetic department which were given breast MRI examinations in a traditional screening program which also comprised mammography. The study detected 25 cancers, where higher sensitivity (86%) was reported for MRI in 20% of interval cancers over mammography (50%) at the time of diagnosis.

Skaane [25] detected women with cancer at the baseline screening, corresponding cancers and interval cancers with respect to age group. The follow-up was 18 months because the program was suspended during the research period by the Norwegian government. A total of two groups of women were included in the prospective analysis attending a breast care clinic for a palpable breast lump or screening [26]. The study reported on the performed mammography in 564 women, which were categorized as BIRADS I in 40%, BIRADS II in 20%, BIRADS III in 9% BIRADS IV in 23% and BIRADS V in 8%. Sardanelli et al. [27] recruited asymptomatic women with ovarian cancer, first-degree relatives of BRCA mutation carriers, with a history of personal breast cancer. The study found higher sensitivity and specificity among 501 women, with 1592 rounds.

However, there is insufficient data to recommend or refute MRI in patients with extremely dense tissues in mammogram [10]. It is important to have clear indications for MRI in women with extremely dense breast tissues as BI-RADS clinical evaluation, they are relatively 2.3-times more likely to develop cancer than those with scattered fibroglandular density [28]. Also, women with breast cancer are at higher risk of developing cancer in contralateral breast as well, which is not detected accurately in mammogram due to dystrophic calcifications or scar formation and are mostly detected by palpation [10]. Similarly, axillary nodal carcinoma is better detected in mammogram and has a lower fraction in women where MRI is used for screening [10].

Cost–effectiveness of screening MRI

This section discusses the role of MRI in breast cancer diagnosis; seven studies were found to be eligible in this category. Two studies were multicentered, while the remaining five were single-centered studies.

According to Mann et al. [10] MRI does not appear to be cost-effective, even with the preference for paying thresholds above $100 000 per quality-adjusted life-year, where quality-adjusted life-year is an economic measure for the disease burden and evaluation of its intervention cost. Even though MRI offers health advantages for high-risk patients when compared with mammographic screening. It has also been observed that the inclusion of screening using ultrasound or MRI for high-risk breast cancer patients resulted in false-positive findings, although with higher cancer detection capabilities [11,17]. This false positive findings could be due to enhanced detection of high-risk lesion; however, it is still important to detect them for the screening purpose [10]. This can also be prevented by using ultrasound annually and comparing with the first screening [11]. Similarly, there is a matter of false negative results which could be decreased by BRCA mutation [10].

Table 2 showed the cost–effectiveness of MRI for breast cancer diagnosis. It has been highlighted that the minimal cost–effectiveness was reported at 40% by Gareth et al. [12] among 2809 patients with relative risk reduction (0.76). By contrast, the highest cost–effectiveness was reported at 88% by Hassanein et al. [14], among 120 patients with a 76% relative risk reduction.

Table 2. Cost–effectiveness of MRI for breast cancer diagnosis.

Quality assessment			Summary of findings
Author	Design	Centers	Patients	Age group	Relative risk reduction	Cost–effectiveness	Ref.
Hassanein et al.	Accuracy study	1	120	35–50	RR = 0.76	88%	[14]
Mann et al.	Accuracy study	1	79	40–50	RR = 0.85	70%	[10]
Berg et al.	Retrospective review	21	2809	40–69	RR = 0.76	58%	[11]
Gareth et al.	Prospective cohort	1	649	35–55	RR = 0.75	40%	[12]
Yuan et al.	Retrospective design	1	412	24–88	RR = 0.77	63%	[13]

RR: Relative risk.

Using artificial intelligence in screening MRI

Artificial Intelligence (AI) provides the opportunity for streamlining and integrating the diagnostic expertise of the radiologist, which include the stratification and recognition of complicated patterns in images, outcome prediction and clinical translation of tumoral phenotype to genotype as it relates to prognostic and therapeutic prediction.

According to Hassanein [14], the implemented imaging method provides the overall accuracy, and recommended that the dynamic contrast-enhanced (DCE)-MRI technique is most effective in comparison to other machine learning techniques available. Yuan et al. [13] have revealed that single-modality computer aid design with combined extracted features from Full Field Digital Mammography (FFDM) and Dynamic Contrast-Enhanced Magnetic Resonance Imaging (DEC-MRI) images are beneficial in order to differentiate between benign and malignant lesions.

Some of the artificial intelligence imaging literature has questioned the advantage of using different types of processed image data as an input to convolutional neural networks because of limited datasets available. For instance, Antropova et al. [29] have revealed that maximum intensity projection images yield enhanced deep-learning computer performance, rather than early-phase subtraction images, when using DCE-MRI. In addition, Grimm et al. [30] examined 278 breast cancer patients and revealed considerable relationships between DCE-MRI breast imaging reporting and data system (BIRADS) features and molecular subtypes. They also revealed statistically a significant correlation between Luminal A and B cancers and MRI radiomic characteristics. Data were used by Li et al. [31] from the The Cancer Genome Atlas (TCGA) to explain relationships between molecular subtypes and radiomic MRI tumor features.

Summary of results

The review question of the study was to determine incremental accuracy and other parameters of current methods, including sensitivity, specificity, relative risk reduction and cost–effectiveness with regard to breast MRI as a screening tool for women under 50 years. MRI was found to be more precise and sensitive then the mammography for annual surveillance of patients with hereditary risk. Whereas, it has same specificity as mammography. Screening parameters are comparable to population screening data, with more favorable results in the youngest age group (<40) with BRCA1/2 carriers, whereas MRI screening is beneficial specifically for BRCA2 carriers. MRI is not cost-effective. Among AI, the implemented imaging method provides the overall accuracy, and the DCE-MRI technique is most effective in comparison to other machine learning techniques available. The results of the study show that, even though MRI is cost-effective and has high sensitivity, it should not be neglected that it has a potential of giving false negative results when included in screening of patients with high risk of breast cancer. The study was limited as it did not address MRI utility in male breast cancer. Further studies should be conducted to assess and compare specificity and sensitivity of MRI in male patients with breast cancer.

Conclusion

Breast MRI is an important tool for the screening of young women at high risk of breast cancer, especially when cancers progress but avoid detection using mammography and are not yet clinically identifiable. MRI might be essential for the follow-up of breast cancer patients whose cancer is detected initially by MRI. Also, the suspicious lesions diagnosed should be refined by predicting tumor recurrence and tumor subtypes on the basis of computer-extracted MRI features. Indeed, the implementation of AI in breast imaging research may facilitate the development of imaging biomarkers incorporating patient and tumor-inherent features and; therefore, enable the risk stratification of patients with personalized imaging guidelines. The study has offered evidence that the cost–effectiveness of applying routine MRI to examine and observe breast cancer in women with increased risk depends on the level of lifetime risk of the particular patient, the cost of the MRI involved, and the combination chosen for DCE-MRI to be implemented. The present findings have shown that staggering MRI sessions in breast cancer patients would likely be a cost-effective strategy for higher risk patients. The high costs and low specificity of MRI are restricting factors for annual examination schedules.

Financial & competing interests disclosure

The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

No writing assistance was utilized in the production of this manuscript.

Ethical conduct of research

The authors state that they have obtained appropriate institutional review board approval or have followed the principles outlined in the Declaration of Helsinki for all human or animal experimental investigations. In addition, for investigations involving human subjects, informed consent has been obtained from the participants involved.

Acknowledgments

The authors acknowledge all the associated personnel who contributed in the completion of this study.