Breast density as indicator for the use of mammography or MRI to screen women with familial risk for breast cancer (FaMRIsc): a multicentre randomized controlled trial

The FaMRIsc study is a multicentre, randomized controlled trial (RCT), with balanced randomisation (1:1), and a parallel group design conducted in the Netherlands. The study is in compliance with the Helsinki declaration and ethical approval has been granted on 8 November 2010 by the Institutional Review Board of the Erasmus University Medical Centre, Rotterdam, the Netherlands (reference-number: MEC-2010-292).

Eligible participants are women aged 30–55 years with a cumulative lifetime risk (CLTR) of >20% because of a familial predisposition according to the modified tables of Claus [1, 27] or as assessed at a Clinical Genetics Centre. BRCA1 and/or BRCA2 mutation carriers or women with a 50% likelihood of such a mutation are excluded, since MRI screening is already advised for these women by the American Cancer Society (ACS), the American College of Radiologists (ACR), the United Kingdom’s NICE guideline and the European guideline of the European Society of breast imaging (EUSOBI) [18, 28‐30]. Exclusion criteria are previous invasive cancer (potentially of influence on survival data), and a contraindication for contrast-enhanced MRI (decreased creatinin clearance, metal implants or claustrophobia).

Participants are recruited from outpatient breast or family cancer clinics at all eight academic medical centres in the Netherlands and the Netherlands Cancer Institute/Antoni van Leeuwenhoek Hospital ( Additional file 1). Women who are already in a screening programme because of an increased familial risk and meeting inclusion criteria are sent study information 2 weeks before a scheduled visit. Women who meet all criteria and visit the outpatient clinic for an initial screening are given information on site.

After informed consent is obtained participants are randomized through a computer-generated randomization sequence with stratification for centre, in one of the two groups.

Group 1 receives screening according to the 2012 Dutch guidelines [17] with yearly mammography and clinical breast examination (CBE).

Group 2 is screened with yearly MRI and CBE, and mammography biennially (Figure 1). Additional investigations are performed if deemed necessary due to findings at clinical examination, on mammography or MRI. Mammography still has a place in both arms, since DCIS is generally easier to detect with mammography [8, 10, 31‐33], although in one study MRI was found to detect more aggressive grade III DCIS than mammography [34]. In the intervention arm however, the frequency of mammography is reduced from annually to biennially. DCIS not detected by MRI will most likely be low-grade, progress slowly, and be detected by the next mammographic examination. Leaving out mammography every other year seems safe in the MRI arm and may prevent over diagnosis of low-grade DCIS. Mammographic examination is done using full field digital mammography (FFDM). All examinations are scored in a standardized way, according to the Breast Imaging Reporting and Data System (BI-RADS) mammography classification of the American College of Radiology [30]. To determine mammographic density an automated breast density measurement is done on raw data of the first FFDM of all participants [35, 36]. Dynamic breast MRI with gadolinium-containing contrast medium is performed according to standard protocol. In premenopausal women, the MRI is performed between day 5 and 20 of the menstrual cycle [37].

Primary endpoints are the number and stage of detected breast cancers, both DCIS and invasive, in each arm. Secondary endpoints are the false-positive results in the screening arms, and the sensitivity and positive predictive value of both screening strategies. Furthermore, cost-effectiveness and breast cancer mortality reduction of both strategies will be assessed. All analyzes will also be performed stratified for mammographic density.

A positive screening test is defined as a mammographic or MRI examination with a BI-RADS score of 0, 3, 4, or 5 and/or a clinical breast examination classified as ‘suspicious’. Interval cancers are defined as tumours diagnosed after a negative screening examination but before the next scheduled screening examination. Sensitivity is calculated as the number of screen-detected tumours divided by the total of screen-detected and interval tumours. The positive predictive value of a screening strategy is calculated as the proportion of women with a positive screening test, which after pathology indeed proved to be breast cancer.

Prospective assessment of mortality reduction requires a very lengthy follow-up and a large study population, which may not be feasible. To study this issue we will start with a less costly and time-consuming approach and estimate mortality reduction through a micro-simulation model: MISCAN, a well-validated micro-simulation model, originally developed to estimate the cost-effectiveness of the population-based screening program in the Netherlands [38‐40]. In the model, the natural history of breast cancer is modelled as a progression through 5 pre-clinical and invasive disease stages. At each pre-clinical stage, a tumour may either be clinically diagnosed or grow into the next pre-clinical stage. Screening may detect the tumour in a preclinical stage. Transition probabilities, stage durations and survival after clinical diagnosis or screen detection are based on data from the Dutch nation-wide screening program [41, 42]. The improvement of prognosis after detection by screening is based on the long-term effects of Swedish trials [43‐45]. A detailed description of the model has been published previously [39].

We will develop a family history risk model by using the number of women enrolled in the study, the age distribution at entry of the study, the duration of follow-up and the screening protocol, attendance and sensitivity of different screenings methods as inputs. The model will be calibrated using the number of screens, the number of screen detected cancers and interval cancers, the stage distribution and the age at diagnosis. Likelihood ratio tests will be used to compare the goodness of fit. Using the calibrated model, predictions of the number of screens, number of screen detected and interval cancers, the stage distribution, the mortality reduction and the life years gained will be made for the different screening arms in the study.

A cohort of 5 million women will be simulated. All costs and effects will be predicted for a life-time follow-up. The costs will be presented in European currency (€). Cost-effectiveness ratios will be expressed as cost per life year gained (LYG). Costs and effects will be discounted at an annual rate of 3.5%.

Our primary aim is to detect a difference in tumour stage between the intervention and the current practice group. In the Dutch MRI Screening Study (MRISC) study, conducted from 1999 to 2007, over 1500 women with familial risk were included in the 6 participating centres [11]. The incidence rate in this risk group was 7/1000 women years screened. Since the FaMRIsc study will have three more participating centres we intend to include 2000 women. We expect to detect about 50 breast cancers (both DCIS and invasive) in 4 years. With this number we are able to detect a difference in tumour size of 8 mm (SD tumour size: 9 mm) as statistically significant (two sided alpha =0.05) with a power of 80%. Eight mm is considered to be a clinically relevant difference.

The accrual will be evaluated after two years. If adequate inclusion numbers cannot be achieved, appropriate measures will be taken in the remaining two years, consisting of expansion of the number of participating centres or longer continuation of the study.

Primary outcome will be incidence and the difference in mean tumour size at diagnosis between the two arms. If normally distributed, this will be tested by means of the independent samples (unpaired) t-test. If not normally distributed, medians will be estimated and differences between distributions will be tested with the non-parametric Mann–Whitney U test.

Breast cancer incidence rates will be calculated as the total number of breast cancers detected per 1000 woman-years at risk. This will be calculated both including and excluding DCIS. Differences between these proportions will be compared using a chi-square test or Fisher’s exact test as appropriate.

All tests will also be performed stratified by mammographic density to examine the influence of density on the efficacy of MRI screening versus usual care. The influence of breast density on detection rates, tumour stage and false positive results in both arms will be analysed by means of analysis of variance (ANOVA).