A comparison of breast cancer incidence and cancer stages before and after the introduction of the Austrian national breast cancer screening program in the federal state of Salzburg

In January 2014 a national quality-assured breast cancer screening program was introduced in Austria. Women between 45 and 69 years of age are invited biannually to undertake a mammography and in the case of dense breast tissue additional ultrasound with double reading of the mammograms by a second radiologist. The goal of the screening program is to detect breast cancer at an early stage and to subsequently reduce breast cancer mortality. Of the invited women in the federal state of Salzburg, a constant rate of 45% participated in the screening program between 2014 and 2021 [1]. When diagnostic mammograms are considered as well, the mammography supply rate for women aged 45–69 years is calculated as 51% [1], which is not very different from the estimated rate of 55% during the unorganized, opportunistic breast cancer screening which was performed before the start of the national screening program [2]. Irrespective of the participation rate, breast cancer screening is controversial. A systematic review from 2021, conducted on behalf of the European Commission Initiative on Breast Cancer, evaluated 10 randomized controlled trials conducted in the 1960s–1990s and including more than 600,000 women aged 38–75 years [3]. This review found a 20–23% relative and a 0.05–0.2% absolute risk reduction in breast cancer mortality over 10 years, similar to a Cochrane Review from 2013 [4]; however, for several studies included in this review the risk for bias was rated as high. When those potentially biased studies were excluded, no statistically significant reduction in breast cancer and all-cause mortality could be detected anymore [4]. In addition, screening is associated with a substantial risk of overdiagnosis and false positive results [5, 6], leading to anxiety and unnecessary interventions [7]. Furthermore, several population-based studies showed that mammography screening did not reduce the incidence of advanced tumor stages or only to a small extent [8‐19] and had no detectable effect on breast cancer mortality [20‐23].

A successful screening program should not only reduce breast cancer mortality but also the incidence of advanced breast cancer stages. Therefore, we investigated the data from the Tumor Registry Salzburg (TRS), which collects the epidemiological data on all tumor types diagnosed in the federal state of Salzburg and asked the question if the introduction of the national breast screening program had an influence on the tumor stage at diagnosis.

Given that younger women have a lower risk of breast cancer than older women, the absolute benefit from mammography screening is likely to vary between different age groups [24]. Therefore, we investigated the influence of age on tumor stage before and after the introduction of the screening program. Furthermore, we investigated if the place of residence (urban vs. rural) played a role in this context. Previous data have shown that women living in rural areas tend to have higher breast cancer stages at diagnosis compared to women from urban areas, probably because they face more barriers to access mammography screening and further diagnostic procedures [25].

Out of approximately 95,000 women aged 45–69 years living in Salzburg between 2010 and 2022, 2793 breast cancer cases, including 292 cases of DCIS (stage 0), were registered in the Tumor Registry Salzburg (potential screening cohort; Fig. 1). This corresponds to an average age-standardized incidence rate of annually 228.5 cases per 100,000 standard population (95% CI 198.6–260.8).

Since the introduction of the national breast cancer screening program, the average age-standardized incidence rate in this cohort continuously decreased over time from 245.7 (95% CI 213.8–280.4) per 100,000 standard population in the 2010–2013 period to 205.3 (95% CI 177.5–235.6) in the 2020–2022 period. The difference between 2010–2013 and 2016–2019, however, was not statistically significant (P = 0.483; Fig. 2). When the incidence rates were evaluated over time without categorization in time periods, a short incidence peak was recorded in the first half year of 2015 (Figure S1 in the supplement).

According to the World Health Organization (WHO), the purpose of cancer screening is to detect precursor lesions or early stages in asymptomatic individuals in order to enable early treatment with better outcome and ultimately to reduce mortality and morbidity [34]. An efficient breast cancer screening program consequently should lead to an increasing incidence of early stages and correspondingly to a decreasing incidence of advanced stages of breast cancer. In several population-based trials; however, this effect was not observed [8‐19]. Additionally, the prospectively shown positive effect of mammography screening on breast cancer mortality in the 1960s–1990s is thought to be “diluted” by significantly more efficient treatment options improving the prognosis of breast cancer also in advanced tumor stages [9, 35]. A recent simulation study from the USA suggested that the reduction of breast cancer mortality by 58% in the USA from 1975 to 2019 is mainly driven by improved treatments for stages I–III breast cancer (47%), while screening contributed to a much lower extent (25%) [35]. On the other side, radiology applied to the breast has advanced as well to more modern applications such as contrast-enhanced mammography (CEM), artificial intelligence, and radiomics [36], thus potentially leading to a stage shift effect with higher numbers of advanced and lower numbers of localized stages just by technological improvement.

Despite this controversy, in 2014 a national quality-assured breast cancer screening program was introduced in Austria, much later than in other European countries (e.g., Sweden 1986, United Kingdom 1989, Germany 2002) [37]. To improve the performance of the program, a supplementary breast ultrasound can be offered to women immediately after mammography at the radiologist’s discretion [38]. In order to indirectly evaluate the effectiveness of the program, we investigated the data from the Tumor Registry Salzburg to evaluate whether it reflected the anticipated changes in the patient population outlined above. We focused on disease stage at diagnosis as, in contrast to mortality, this parameter is not influenced by treatment, allowing a more direct view on screening effectiveness.

Unexpectedly, we did not find an increase in the overall incidence of invasive breast cancer and DCIS after the introduction of systematic screening. This stands in contrast to many observational trials (e.g. [9, 10]), and is most probably caused by a similar mammography supply rate during the organized national screening program (about 51% in Salzburg) [1] and the opportunistic screening before 2014 (estimated about 55%) [2]. We did observe a slight nominal shift in the empirical stage distributions, with a lower proportion (as well as incidence) of stage IV tumors; however, this could not be distinguished statistically from random fluctuations, owing to the high variance in the statistics of small numbers. Accordingly, the statistical tests performed failed to reject the hypothesis that the screening program introduction had no discernible effect. Demonstrating the absence of an impact by the introduction of the program is epistemically difficult. Our results, however, provide an upper bound on the size of any additional screening effect. This is consistent not only with the well-documented challenges of large-scale centralized screening programs but also the estimates suggesting that the rate of screening itself may not have changed much.

In a similar study, also conducted in Austria but in the federal state of Tyrol, participation in the mammography screening program was associated with a 28% risk reduction for tumors ≥ stage II and a 73% risk reduction for stage IV tumors [39]. This may appear contradictory to our results; however, substantial differences to our investigation exist both in terms of the program studied and the methodology of the investigation. Mammography screening in Tyrol started earlier (2008), included women aged 40–69 years, and invited women aged 40–59 years annually instead of biannually. Moreover, Oberaigner et al. classified patients as “exposed” or “unexposed to screening”, irrespective of the reason for breast cancer diagnosis (screening detected vs. not). As such, their results support the benefits of screening itself while our focus lies on the impact of the organized screening program.

Whether or not the trend toward lower rates of de novo metastatic breast cancer, without reduction of stages II-III tumors, can be attributed to screening cannot be answered using our dataset.

When comparing node-negative with node-positive disease (after exclusion of stage IV), more pronounced nominal differences between the two time periods were observed: while the proportion of node-negative tumors increased from 72% to 77%, node-positive tumors decreased from 23% to 29%; however, again, these effects of were within the magnitude of expected random fluctuations and thus yielded no substantial statistical evidence in favor of a benefit of the screening program.

Furthermore, we investigated if the place of residence was associated with tumor stage at diagnosis, because women living in rural areas could experience more barriers to access mammography screening than women in urban areas. Several previous datasets suggest that patients from rural areas are diagnosed at later breast cancer stages compared to women from urban areas [25]. Our results, however, did not suggest any influence of rural-urban residence or, in fact, different age categories on tumor stage at diagnosis.

One major limitation of this study is the relatively high percentage of unknown tumor stages (17%) and the uneven distribution over the different time periods, thereby hampering the comparison of incidence rates. We tried to overcome this by comparing relative distributions of tumor stages through logistic regression, alongside the incidence rate analysis. Furthermore, we performed exploratory imputation of missing information on metastatic status in otherwise early stage tumors, which did not present any different results. A potential limitation is, in addition, the fact that patients living in Salzburg could be diagnosed outside the federal state, thereby hindering the inclusion in the registry. Because of the centralized structure of the healthcare system in Salzburg, the number of affected patients is however thought to be very low.

Our observations should not be carelessly extrapolated to other countries, first and foremost due to the extent of the pre-existing opportunistic screening in Austria, a similar program in populations with less access to screening may well yield very different results. Other limitations are the lack of information on screening exposure for the patients in the registry as well as the relatively small source population of approximately 95,000 women aged 45–69 years. This reduces the precision of our estimates and limits the generalizability of our results. At the same time, this is a weakness that would be inherent to any attempt to study the net effect of the screening program on the whole breast cancer population in Salzburg and on the other side our analysis makes use of all available information on breast cancer cases in the population, including both screened and unscreened women.

In conclusion, we did not observe a reduction of locally advanced breast cancer stages (II and III) at diagnosis in the federal state of Salzburg after the introduction of the Austrian national breast cancer screening program compared to the opportunistic screening established before. These results encourage a critical evaluation of the whole Austrian breast cancer screening program.