The Andromeda Study.Predictive Value of Combined Criteria to Tailor Breast Cancer Screening.

Breast cancer (BC) represents the most frequent neoplasm in women worldwide. In Italy about 1 out of 3 malignant cancers in women (29%) is a BC (leaving out cutaneous tumours) as reported by the Italian cancer registries between 2006 and 2009. Results from randomized trials, summarized in an updated review and meta-analysis for the U.S. Preventive Task Force by Nelson et al., showed that participating to a mammography screening program reduces BC mortality of 14-32%, depending on the age range. In Italy, mammography screening for early diagnosis has been implemented on a regional basis in several Italian areas. In compliance to national and international guidelines, most programmes invite women aged 50 to 69 years to undergo mammography every two years.

Although mammography has become standard of care in BC screening, its limitations are well recognized (overdiagnosis, the use of ionizing radiation, poor accuracy in women with dense breast tissue, a relatively high rate of false positives, and personal discomfort). In addition, the current screening protocols are based on woman's age only, ignoring other risk factors and defining a single screening periodicity within the prespecified age range.

Current challenges in the management of BC include searching for sensitive and specific minimally invasive biomarkers associated with early neoplastic changes. A number of circulating tumour markers (such as carcino-embryonic antigen, CEA or carbohydrate antigen 15-3, CA 15-3) are widely used in BC handling, but their sensitivity is very low. MicroRNAs (miRNAs) are small, noncoding RNAs (about 18-25 nucleotide long), that regulate gene expression inside cells by degrading mRNA or inhibiting protein synthesis. Since they may have hundreds of targets, they are able to control several biological processes inside cells. There is growing evidence that miRNAs may also be released outside cells, and increased levels of miRNAs released by tumour and its microenvironment as well as decrease in physiologic circulating miRNA levels have already been detected in serum/plasma of cancer patients. The potential use of such molecules for diagnostic/prognostic purposes in regard to breast cancer has been extensively evaluated and their stability in body fluids has opened new opportunities for anticipating BC diagnosis, with minimally invasive intervention especially for women at higher risk.

Hypothesis: Risk-based screening programmes according to different BC risk criteria (i.e. breast density, model based estimates of absolute risk, life-styles) may maximize the impact of screening in groups of women characterized by different risks. Gathering data for risk stratification in population based screening programmes requires large amount of resources. Non independent criteria which identify the same women at higher risk are redundant and should be considered as mutually exclusive. Validated highly accurate blood molecular biomarkers associated with BC risk may represent a complementary tool to mammography in the primary screening setting.

The first objective of the study is to evaluate and compare, in a large cohort of women, the predictive value of available criteria to define BC risk in order to identify appropriate risk-based stratifications for personalised screening. The criteria considered are:

1. general risk factors collected through a standardised risk questionnaire (e.g. reproductive factors, first-degree family history, biopsy history, physical activity, BMI and alcohol consumption). For this category the absolute risk will be evaluated using the risk prediction model of Petracci et al. that has been developed for the Italian population and validated on an independent cohort of Italian women. ii. breast density; iii.woman's life styles (through a detailed questionnaire on dietary habits and physical activity).

The study is designed for achieving the following secondary aims related to the first objective:

To measure age- and risk-specific screening performance indicators (i.e. detection rate, recall rate and benign/malignant surgical biopsy ratio).

To quantify the impact of potential interventions of tailored screening on health outcomes and costs through the knowledge of risk factors distribution in the target population; To evaluate the economic and organizational feasibility of potential tailored intervention and to design stratified breast cancer screening strategies for future pilot screening programme implementation; The extension of the follow up beyond the three-year duration of the project will allow us to determine even more consistent estimates of the predictive values abovementioned and to evaluate the criteria/risk factors considered in this study in a multivariate analyses.

The second main objective is to investigate whether selected circulating miRNAs previously found associated to BC risk are significantly altered in the plasma of cancer patients compared to matched healthy controls and if they satisfy pre-specified true- and false-positive rates that are considered minimally acceptable in the screening setting.

Related sub-objectives:

• assess the association between factors such as age, breast density, reproductive factors and family history as well as life-style and miRNA level in the cancer-free population. If such factors affect the biomarker, the threshold for screening positivity may need to be defined separately for screening subpopulations, in order to keep the false-positive rate at low level for each. - To assess factors associated with miRNA level in cancer cases - in particular, disease characteristics such as stage, grade, and available prognostic factors. Understanding the nature of cancer that is detected with a biomarker is a key issue. A biomarker that detects cancer at an early stage is more valuable for preventive purposes than one that detects only late-stage cancers.
• evaluate the presence of 18 currently established breast cancer risk SNPs obtained by genome-wide association studies (GWAS) that have identified common low penetrance alleles associated with breast cancer risk at multiple genetic loci, and test their association with circulating miRNAs, breast density and screening outcome.

Phase 1 - Enrolment About 36 600 women (age-range: 46-67 years) participating to BC screening program in selected areas, will be offered participation in the Andromeda study, of whom 22 000 are expected to provide informed consent. At enrolment information on BC risk factors and life-style will be collected, together with breast density and blood samples for serum, plasma and buffy-coat storing.

Women who will agree to participate in Andromeda will be asked to fill in a questionnaire to design their BC standard risk and a detailed questionnaire on their life styles. They will also be asked to provide a blood sample.

Furthermore, breast density will be calculated during the breast examination. Exams will be read by two expert radiologists and breast density will be classified into categories by means of the Breast Imaging Reporting and Data System (BI-RADS).

Women accepting blood sampling will receive a specific appointment where trained personnel, beyond the blood withdrawal, will collect anthropometric measures and will support them in completing the study 's questionnaires, if needed.

Given that positivity for dense breast, Gail score, lifestyle related risk and miRNA are determined on the same subjects, the most efficient estimate of the ratio between relative (positive vs. negative) positive predictive values (PPV) ratios for two different factors is given by a x b/c x d, where a = cancers positive factor 1 but not to factor 2; b= women with abnormal mammography but no cancer, positive for factor 2 but not for factor 1; c=cancers positive for factor 2 to but not for factor 1; d= women with abnormal mammography but no cancer, positive for factor 1 but not for factor 2. Assuming that 40% of 21000 enrolled women have a dense breast, 20% a certain Gail score, 50% a lifestyle risk and 20% a given miRNA, that 11% of study women carry breast cancer and that the overall PPV of mammography is 20%, some 100-120 discordant cancers and 640-740 discordant women with abnormal mammography but no cancer are expected for the considered couples of factors. Under these assumptions, the study has a 80% power to reject (alpha = 0.05) the null hypothesis that the relative PPV ratio is =1 if its true value is at least 1.8-2.0, depending on the couple of factors considered.

Phase 2 - Biomarkers analysis The Andromeda study comprises the molecular analyses to evaluate SNPs and circulating miRNAs that will be performed blindly on an appropriate case-control sample extracted from the cohort of women enrolled, as soon as a minimum number of BC cases will be reached. Blinding allows avoiding that knowledge of subject's outcome status affect the interpretation of an assay result or the care with which the specimen is handled. Genomic DNA will be extracted from buffy coat with High Pure PCR Template preparation kit. 18 SNPs will be amplified by PCR and evaluated on a qPCR apparatus with specific assays. The 18 breast cancer risk related SNPs are: rs11249433, rs1045485, rs13387042, rs4973768, rs10941679, rs889312, rs2046210, rs13281615, rs1011970, rs2981582, rs2380205, rs10995190, rs704010, rs3817198, rs614367, rs999737, rs3803662 and rs6504950.

Circulating total RNA extraction from plasma samples will be carried out with miRNeasy minikit according to Exiqon protocol, implying use of a RNA carrier (batteriofage MS2 RNA, Roche Diagnostics) to promote RNA precipitation and its purification on membranes. Cel-miR-139 will be used as spike-in control to check extraction yield and to normalize data, together with hasmiR- 1228. RNA samples will be eluted with nuclease free water and stored at -80° C. A novel method to check for haemolysis will be applied.

To calculate the sample size necessary for the pivotal evaluation of the biomarkers (expressed on a continuous scale), minimally acceptable and desirable levels of typical performance measures of interests, such as the true-positive rate (TPR) and the false-positive rate (FPR) have been defined. For general population screening, the false-positive rate must be quite low to avoid a huge number of people undergoing unnecessary costly medical procedures. Thus, we hypothesized a maximally acceptable false-positive rate (FPR0) of 4% (the minimally acceptable specificity is therefore 96%) and a minimally acceptable sensitivity of 80% (TPR0). The null hypothesis to be rejected is the following: H0: TPR0≤0.80 or FPR0≥0.04.

The sample size required with an 80% power (alpha=5%) and assuming desirable true-positive rate and false-positive rate of TPR1=0.90 and FPR1=0.05, respectively, was of 179 cases and 537 controls. Sample size computation was based theoretically on the ROC curve. Assuming an annual BC detection rate (DR) of 0.006 in the first year and 0.005 in the second year of the study (therefore the biennial DR considered was 0.011) a total of 233 case patients in the first two years of the study are expected to be observed, thus exceeding the 179 required. These estimates are based on DRs observed in the previous years in the same centers considered in this study. Thus, the molecular analyses will be performed on 233 cases and 699 matched controls. For each case, three control subjects will be selected from the cohort of women, on the basis of the following criteria: no history of cancer, similar age at enrolment (within 5 years), similar race, availability of blood sample, similar date of blood draw.