Breast cancer risk assessment and risk distribution in 3,491 Slovenian women invited for screening at the age of 50; a population-based cross-sectional study

Women turning 50 years are invited with a personal letter to the Slovenian BC screening programme to participate in mammography screening organized according to the EU guidelines.^17,20 All eligible women aged 50 (with no previous BC diagnosis) in 2021, were sent a self-administrated structured 5-page questionnaire via postal mail together with a screening invitation and explanatory text to sign informed consent for participating in this study. The family history questionnaire was adopted from the one used at the Department of Clinical Cancer Genetics, encompassing questions about anthropometric, reproductive and hormonal anamnesis and family history (Table 1).²¹ In addition, just for the purpose of this study, breast density was assessed by the radiologist using the mediolateral oblique view of the screening mammograms, in accordance with the BI-RADS 5^th edition reporting system that classifies breast density into four levels.²²

For menopausal status, the time interval between the date of filling in the questionnaire and the date of women's last menstrual period was considered, 31 and 365 days being cut-offs for the groups (premenopausal, perimenopausal and postmenopausal).²³ For participants’ description, the characteristics of study participants were grouped into categories according to the relative risk caused by the risk factors incorporated in the Tyrer-Cuzick model.^14,24 The denominator for calculating the percentage of frequencies was the sum of participants (3,491).

In the Slovenian national health system, Slovenian IBIS is ready for use allowing an evidence-based assignment of an asymptomatic individual to a group of population, moderately increased and high BC risk. The S-IBIS software was developed at the Institute of Oncology Ljubljana in 2018 within a research project.¹⁹ It is an adjustment of the IBIS software with the Tyrer-Cuzick algorithm, where Slovenian generation-specific population BC risks were applied and it is specifically designed to calculate the individual risk of BC in Slovenian women. BC incidence and mortality rates between 2006 and 2010 were obtained from the population-based Slovenian Cancer Registry.²⁵ The Tyrer-Cuzick algorithm is recognized as one of the most consistent models and validated on several populations. It calculates both, a 10-year risk and a lifetime risk of BC based on the women's personal, reproductive, and family characteristics.¹⁴ Moreover, the latest version of the model (version 8) incorporates mammographic density.²⁶

After calculating the risk for each individual (first with breast density included and then without breast density), study participants were grouped into risk categories according to relative risk that was calculated with the Tyrer-Cuzick model.^14,24 The cut-offs for the distribution of individuals into the low, population, moderately increased and high risk categories for BC are shown in Table 2. Lifetime risk is defined as the risk of developing BC by the age of 85.^19,25 In case of missing or unknown data, the population reference relative risk was considered.

The number of screening mammographies was estimated for the same group of women (N = 3,491, considering all would attend the screening regularly by the age of 69) for two different risk-based scenarios/protocols according to the Slovenian BC diagnostic and treatment guidelines and English Predicting Risk of Breast Cancer at Screening (PROCAS) cohort study (Table 2).^18,25,27

The tool SPSS, version 24 (IBM Corp., Armonk, NY, USA), R Project for Statistical Computing (v4.3.2) and RStudio (2023.03.0, R Core Team 2023) were used for the statistical analyses and risk calculations. Statistical significance was determined using the Clopper-Pearson 95% confidence intervals.

Table 4 shows the frequencies of BC risk in the study population; 3.4% of participating women were assessed with moderately increased 10-year BC risk, and 2.2% with moderately increased lifetime risk. Only a small proportion of women had high 10-year and lifetime risk (0.4% and 0.03%, respectively). The mean of the 10-year risk was found to be 1.7% with a standard deviation of 1.0, and the mean of the lifetime risk was 6.3% with a standard deviation of 3.2. Among 3,491 analysed study participants, 27 were diagnosed with breast cancer after first screening mammography in 2021; one of them was assessed as high risk (10 years BC risk), none as moderately increased, nine as low risk and 17 as population risk.

Adding breast density information to the Tyrer-Cuzick model (10-year BC risk) significantly changed the distribution of women in our study. This information moved the majority of women to lower-risk groups. The proportion of women in high, moderate and upper-population risk groups (2.6%–3.9%) was also decreased, shifting women to lower–population (1.3%–1.7%) and low risk (below 1.3%) groups (Figure 1).

Table 5 shows the change of the number of screening mammographies in the screening programme when considering different screening protocols (age-based and risk-based using the S-IBIS model including breast density [S-IBIS+BD]) applying two different risk-based protocols, described in the Slovenian guidelines and in the PROCAS study.^18,27

When considering applying the current Slovenian risk-based screening guidelines for lifetime BC risk, 2.2% (|35,690–34,910|/34,910) more mammographies compared to the current screening strategy would be performed in 20-year screening period (aged 50 to 69 years).

Considering the PROCAS protocol for 10-year BC risk, the number of screening mammographies in the risk-based screening would drop by 38.6% (|21,418–34,910|/34,910) in the 20 years compared to the current screening. It considers enhancing the screening intervals among above-average BC-risk women, but also less frequent screening intervals in the below-average risk group. An increase of 100.0% (|280–140|/140) in the number of mammographies would be observed in the high risk group and a decrease of 40.6% (|19,948–33,580|/33,580) in the low and population risk groups. In fact, even in case of risk stratification using S-IBIS with no breast density information less mammographies would be performed considering PROCAS protocols for 10-year BC risk, namely 30.5% (|24,254–34,910|/34,910) less.

In our study, we experienced a satisfactory up-take (57%). Our results confirm the feasibility of determining BC risk at the entry in the Slovenian BC screening programme and this result is in accordance with literature reports. The 1971 birth cohort of women eligible for screening was reached and women were offered preventive mammography screening and voluntary study participation. Therefore, no additional interventions were anticipated. The invitation to the study was part of a regular screening programme. Furthermore, the women recognized as high risk for BC will be offered genetic counselling, where BC risk factors will be reverified following the clinical pathway and BC risk will be recalculated by using verified data and genetic data where applicable.²¹ As reported in the PROCAS study, the majority of women (95%) indicated they wished to receive risk information.²⁹ Also, the DECIDO study showed a positive attitude and a high understanding of risk-based screening.³⁰

According to our results, 3.4% of Slovenian women invited for mammographic screening at the age of 50 and consented to participate in our study belong to the moderately increased 10-year BC risk group and 0.4% to the high risk group and would have to be screened more often according to our guidelines. So far, these data were unavailable, since no population-based cross-sectional study has been performed to assess the BC risk in Slovenian women. Some regional and hospital/breast centres-based research have been conducted to assess women's BC risk.^19,25,31,32 Due to different subpopulations assessed and low sample volumes, the results of these studies cannot be directly compared to our study. They were all conducted among women referred to breast centres, where women with a positive family history or previous biopsies are normally assessed. Therefore, we may predict their BC risk could be higher when compared to the general population.

In 2016, a prospective cohort study among 100 asymptomatic women (aged 20–49) from one regional breast centre was conducted, testing the S-IBIS calculator (version 8) for lifetime BC risk. 18% of women were identified as moderately increased risk and none at high risk (above 30% risk). 86% of women referred for another mammography in 12 months would not need annual screening mammography. This study proved the S-IBIS is effective in decreasing the number of referrals for annual mammography.³¹ In 2018, a study recruiting women from regional breast units proved that 148 (75.1%) out of 197 interviewed and examined women were assigned to the population risk category, 49 (24.9%) to the moderately increased risk category and none to the high or low risk categories.²⁵ For that study, the S-IBIS tool (version 8) was used and tested.¹⁸ Another Slovenian study from 2020 was assessing the proportion of women with above average 10-year risk of BC (more than 2%) using the S-IBIS calculator version 8 (breast density not considered). All assessed women in the study were already at higher BC risk at the baseline. They were either healthy with some breast symptoms or already diagnosed with BC (for the latter, the data prior to BC diagnosis was considered); 48.7% and 39.2% of women were recognised with above average BC risk, respectively. The study concluded, that inclusion of additional risk factors into the S-IBIS is needed to reliably stratify women into the BC risk groups.³²

As shown above, the advantage of BC risk assessment is the use of S-IBIS, a BC risk calculator in the Slovenian breast centres, which could reduce the number of unnecessary preventive mammographies for the majority of women assessed with the population risk, giving room for symptomatic women and women at moderately increased and high risk. This was proved reasonable in all aforementioned Slovenian risk studies.^25,31,32

Furthermore, risk-based screening is already ongoing in several countries across the world, but at the moment only in study settings. The English cohort study PROCAS conducted in the UK in the period from 2009 to 2020 and recruiting 63,000 women in two large studies (aged 50–70), concluded that the Tyrer–Cuzick risk prediction model (version 6) accurately predicts BC risk.²⁷ However, some further improvements are still required. The study showed that 11% of women in the general population have moderately increased BC risk and 85% of women have average population risk or very low risk. It also indicated that adding breast density and genetic information improved risk precision and can be used to tailor screening. Using a combination of both predicts that 70% of the population with average or below-average risks have very low rates of advanced BC. Moreover, 3-yearly screening interval appeared effective in 70% of the population in the UK. Additionally, giving women their risk information and management feedback increased their next screening participation, and even more, it encouraged them to improve their lifestyles.^27,33 In numbers, 24% of participating women were found at low 10-year risk, 61% at average (population) risk, 11% at moderate and 4% at high with breast density information added to the Tyrer-Cuzick model (version 6).^27,33 Similar proportions were found in our study – 34%, 62%, 3%, and less than 1%, respectively. The difference, however, may be due to the age gap – only 50-year-olds in our study vs. women aged 50–70 in the PROCAS study. Furthermore, if we transpose the risk groups’ distribution from our study to the PROCAS potential risk-based protocol (Table 2 and Table 5), the number of screening mammographies will decrease by more than one-third (38.6%) in 20 years of screening compared to current screening programme workload. The main reason is the longer screening interval for the majority of women with population BC risk.

At the moment, two big randomized controlled trials (RCT) are trying to answer whether personalized screening is non-inferior to the standard age-based screening protocols. WISDOM is a multicentre RCT ongoing in the USA, comparing risk-based screening to annual screening in women aged 40–74 years and determining whether risk-based screening is as safe as annual mammographic screening, which is the screening policy in that country.³⁴ Similarly, an European RCT ongoing in 6 countries is called MyPeBS (My Personalized Breast Screening).^28,35 Study MyPeBS is an international randomized, multicentric study assessing the effectiveness of a risk-based BC screening strategy compared to a standard screening in detecting stage 2 or higher breast cancers. It will recruit 85,000 women from Belgium, France, Israel, Italy, the United Kingdom and Spain. Each participating country has different current national guidelines – biennial or triennial mammography screening beginning from the age of 40 to 50 years and ending from 69 to 74 years.^28,35 Both RCTs are integrating polygenic risk scores (with 313 single-nucleotide polymorphisms) in the risk calculations, for which BCSC and Tyrer-Cuzick calculators are used.^34,35 Risk score 313 provides the highest level of BC risk stratification in the population, followed by mammographic breast density and other risk factors.⁶ Those RCTs are aimed at investigating whether the personalised approach is at least equally or more appropriate than the standard one.³⁵

Our results also show that the vast majority of women (96.2%) have low (34.0%) or population (62.2) 10-year BC risk and are thus appropriately screened every 2 years, according to Slovenian and NICE guidelines.^18,36 However, 3.8% of women at the age of 50 would need more intensive BC surveillance.

Regarding the reliability of the data collected, risk feedback in PROCAS (in person or by telephone) showed that women's information on their risk factors stated in the questionnaires was not always accurate, and in some cases, women changed risk groups after consultation. The greatest proportion of changes in risk occurred in those originally assessed as having a 10-year TC risk of ≥ 8%.²⁷ With this in mind, the proportion of the low and population risk groups in our study may be overestimated (listed in study limitations), since the self-administrated questionnaire was quite long and demanding for an average user. Besides, it may have deterred some women from participating and entering the complete information about risk factors. Some women may have not enquired about the cancer history of their family members, especially distant relatives. However, we expect, that positive cancer diagnoses are well-known in families and that women with the highest risk were not missed.³⁷ Overall, risk scores should be calculated with verified data on risk factors, where more effort to obtain accurate data should be considered. The risk feedback to women (and consultation, if possible) is an example of how to improve the data accuracy for risk identification, as it is planned for our identified high risk women.

The analysed study women are representative group of the Slovenian population. In Slovenia, 10% of women at the age of 50 are nulliparous (in our study 9.6%) and 50% of women at the age of 45 to 54 are overweight or obese (body mass index higher than 25) – in our study 38.3%.4,38 Noteworthy, 27% of study participants did not provide the data on body weight or height.

In addition to the data collection, mammographic density is a strong independent risk factor for BC and it is not a part of standard screening mammography results.³⁹ For almost all women participating in our study, breast density was estimated and the majority were assessed with breast density BI-RADS b (51.6%) and BI-RADS c (40.4%). These results are in accordance with the literature reports.^40,41 However, it is known that the BI-RADS assessment method is subjective and depends on the reader, reading volume and image quality.⁴² In the PROCAS study, the percentage of women in each risk category changed when density was added to the Tyrer-Cuzick risk model. Adding density moved many women from average (population) to higher or lower risk of developing BC.^27,33,43 On the contrary, in our study, adding density mainly moved women from higher-risk groups to lower-risk groups (shift to the left) (Figure 1). We can assume that density contributes to the model to decrease the risk at the age of 50 (in the PROCAS study, women's age was 50–70, the majority of women were assessed for breast density as BIRADS b (lower density)).

With risk-based screening and following the Slovenian BC guidelines, the number of mammographies increased (by 2.2%) on account of women with above population risk. Less frequent screening in women with lower BC risk is not considered at this point.¹⁸ It is clear that communicating the reduction in screening frequency in the general population is rather demanding even though more screening does not prove higher efficiency.^28,44 In reality, this is the biggest uncertainty in the risk-based screening, because it is very unlikely that less screening would be accepted in the target population. While recruiting participants for the MyPeBS study, 60% of women recognised as low risk, opted-out the intervention group due to prolonged screening intervals.⁴⁵ However, this ongoing RCT in Europe does predict less frequent screening intervals for women with population and low risk, i.e. 4 years.²⁸ Therefore, we can expect a smaller amount of annually performed screening mammographies in the national BC screening programme in case risk-based screening protocol that includes less frequent screening for lower-risk women is recommended. Nevertheless, at the time there is not enough evidence for such recommendation at the Europe level, since not many prospective RCTs are being conducted nor concluded.

For Slovenian situation as for the other countries, first, communicating clinical safety of less intensive screening can be an important obstacle, and secondly, the risk-based screening protocol should not be to complex to be feasible and to be able to follow-up the participants efficiently. To add, recalculation of 10-year BC risk should be considered after 10 years.

For the first time in Slovenia, a population-based sample of women was assessed for BC risk and it is the first time potential changes in the organised screening programme in case of introducing risk stratification have been estimated.

The mammographic density was assessed exclusively for our study (available for 99% of our participants), which made it possible to define BC risk more accurately and proved to be feasible to incorporate this information into screening data. Equally important, the IBIS software has recently been adjusted using Slovenian-specific population BC risks making it more valid.¹⁹ Women's uptake to study participation (57%) was satisfactory when compared to other studies. In the same year, the screening participation rate of women aged 50 was 74.4%.⁴⁶ In the PROCAS study (first phase of recruitment similar to ours, where all women invited for BC screening were sent a participant invitation letter), screening uptake was 68% and study uptake 37%.²⁷

The study questionnaires were self-administrated and data verification by enquiring with the women in person or using the health records was not performed. Furthermore, the family members’ names were not collected so the family cancer history was not medically confirmed; some women did not remember all the family cancer diagnoses. In addition, some may wrongly interpret the topography of cancer (e. g. ovarium cancer instead of cervical cancer). Additionally, the health literacy of women is very variable, which can affect answering the questionnaire without explanations. Thus, some data we used might be unreliable, and probably we have under or overestimated the risk scores. To improve data quality, assistance with filling out the risk questionnaires is needed. From clinical work it is known, that majority of people cannot finish the risk tool/questionnaire without assistance. In practice for risk-based population-wide screening this means, that trained radiographers or administrative personnel should administer women at their first screening visit to gather adequate information. Besides, legal framework for data verification through other databases (like cancer registry and registry of genetically tested individuals) should be legislated.

We assume that some women did not participate in the screening programme nor in this study since they have already been regularly and thoroughly surveilled at the Institute of Oncology Ljubljana in the High risk breast clinic. After the BC risk had been assessed in the Department of Clinical Cancer Genetics, which has been operating for more than 20 years, women with above-population BC risk were referred to the High risk breast clinic at the Institute of Oncology Ljubljana.²¹ For this reason, a certain number of women with moderately increased and high risk for BC may not have been considered in our study (selection bias), thus decreasing the proportion of women in higher-risk groups.

In conclusion, assessing a personalised risk score at a woman's first screening appointment is reasonable. It can improve screening benefits for low and higher-risk groups in the target population. To plan more efficiently the BC screening and BC patients’ care if the risk-based screening over the current one-size-fits-all screening strategy would be evidence-based and recommended in the future, we assessed the BC risk in a 50-year old cohort of women in Slovenia and found the majority of women belonging to the population 10-year BC risk (62.2%) and 3.4% to moderately increased BC risk group. Our evidence supports the effectiveness of the current Slovenian screening protocol for the majority of screened women. Potential future risk-based screening would change the manner of BC screening for approximately one third of Slovenian women (38% at high, moderately increased, or very low risk) with either additional screening methods at a higher frequency or with prolonged screening intervals, respectively. Risk assessment is feasible at the entry to screening. Due to the study uptake, where more than half of screened women took part, rather high risk assessment uptake among Slovenian women is expected. However, data accuracy can be improved with in-person risk assessment and risk counselling.

Still, only randomised and observational studies will answer the main question regarding personalised BC screening in the future. And this is if risk-based screening over the current one-size-fits-all strategy should be recommended.