Trends in incidence, mortality and survival in women with breast cancer from 1985 to 2012 in Granada, Spain: a population-based study

Breast cancer is the most frequent tumor in women worldwide, particularly in countries with a high Human Development Index [1]. Moreover, it is one of the leading causes of cancer mortality in females. In 2015 there were 2.4 million estimated new cases and 523,000 estimated deaths worldwide in women, which correspond to about 29% of the total incident cancer cases and 14% of all cancer deaths [2]. There is huge variability in the incidence among countries, from 27 cases per 100,000 women in Asia to 97 per 100,000 white women living in the USA [3]. In Spain, the 2015 age-standardized incidence rate referred to the world population (ASR-W) was 65.2 per 100,000 women and the age-standardized incidence rate referred to the European population (ASR-E) was 88.3 per 100,000 women, placing this country in an intermediate position worldwide [4].

Several industrialized countries including Spain have shown a 30 to 40% increase in breast cancer incidence since the 1970s [3, 5]. This rise has been related to the spread of environmental and lifestyle risk factors, and to changes in diagnostic patterns [3, 6, 7]. A trend change has been observed since the beginning of the twenty-first century, mainly among women older than 50 years [8‐10]. The main factors related to this change are the implementation of population-based screening programs at a country-wide level, and (albeit with a relatively low impact in Spain), the evolution of prescribing practices for hormonal replacement therapy [8, 11]. Analysis of breast cancer incidence trends in young women vary widely among countries, but in general show a steady increase since the early 1980s even in countries where the incidence in older age groups has decreased [12‐14]. Studies in European countries and in the US show an increase in the incidence of early-stage tumors and a parallel reduction in late-stage tumors, although this reduction seems to be smaller than expected and the incidence of metastatic breast cancer has remained stable [15‐18]. In Spain, there are no available population-based data on breast cancer incidence trends by stage.

Breast cancer mortality in Europe showed an increasing trend until the 1990s [3]. Between 1989 and 2006, breast cancer mortality (ASR-E) in European countries reportedly declined by a median of 19% [19]. The world-standardized mortality rate in Europe decreased from 21.3 in 1990 [20] to 16.7 deaths per 100,000 women in 2007 [21]. Finally, in Spain, the mortality rate (ASR-W) dropped from 17.3 per 100,000 women in 1995 to 10.8 per 100,000 in 2014 [22]. This reduction in mortality has been consistently smaller in women older than 70 years [5, 19], and correlates with the development of adjuvant treatments and, to a lesser extent, with the introduction of screening [23, 24].

Survival rates for breast cancer have generally increased since the 1980s. This trend has been related with a higher proportion of cases diagnosed at earlier stages as well as therapeutical improvements [25]. Currently, the 5-year net survival rate is higher than 85% in seventeen countries worldwide. In Europe the median survival rate ranges from 81 to 84%, with the exception of Eastern countries, where the survival rate is around 69% [26, 27]. However, no relevant increase in overall survival has been observed for metastatic tumors, or in the group of women older than 70 years [26, 28]. Spain had a 5-years survival rate of 78.4% for women diagnosed between 1997 and 1999, and this rate increased to 82.8% for those diagnosed between 2000 and 2007 [29]. Increasing trends in survival are related to early diagnosis and improvements in surgical and adjuvant treatments. Several recent studies have improved our understanding about the role played by screening, the spread of adjuvant treatments and their adverse effects, but there is still considerable controversy on this issue [30‐32].

Since 1985 the Granada Cancer Registry (southern Spain) has systematically compiled data on breast cancer incidence, mortality, and crude and net survival trends. We were able to use the data collected for a period of more than 28 years from 1985 to 2012. In addition, we analyzed a subset of the data for the years 2000 to 2012, after the implementation of a screening program in 1998. For this period, we analyzed breast cancer incidence trends according to disease stage, to shed light on the impact of screening on stage distribution and its association to mortality and survival trends. To date no such analysis has been undertaken in Spain, as far as we are aware.

Determinants of breast cancer trends have been identified in previous studies, but unresolved controversies remain about their role. Trends studies provide an excellent opportunity to explore the specific weight of each factor. Studies at regional or national level frequently only consider either incidence or mortality [8, 15, 33, 34]. However, we present a comprehensive population-based analysis of breast cancer epidemiology, including every indicator and age group – an approach which facilitates an integral interpretation of the factors that may influence trends. Moreover, our analysis of tumor stages at diagnosis, together with the long observation period, hold the potential to provide a better understanding of trend determinants and especially the influence of breast cancer screening.

From 1985 to 2012, the incidence of breast cancer in our population has increased, as documented by the APC of 2.5%. A similar increasing trend was observed in Europe, with APCs ranging between 0.8 and 3% [44]. The introduction of the screening program may have played an important role in this trend, as has been suggested in previous European studies [8, 10, 15‐17, 36, 45, 46].

However, this trend started in our analysis before screening introduction and could also be found in age groups not invited to the program. These findings have been previously interpreted as indicators of the role played by environmental, lifestyle and behavioral exposures [3, 5, 8, 16, 23, 44, 45, 47‐50]. Several breast cancer reproductive risk factors such as parity, advanced age at first birth or breast feeding have been highlighted before [51, 52], as well as lifestyle risk factors including alcohol consumption, post-menopausal obesity and sedentarism [53, 54]. Moreover, these finding have also been connected to changes in diagnostic practices, that have increased detection rates [10, 12, 16, 24, 45, 48], like the increasing use of opportunistic screening [55‐57].

Finally, trying to disentangle the role played by screening program implementation in the increasing incidence from 1985 to 2012 shown in our population, we have performed a comparative analysis of overall and 50–69 year age group incidence trends before and after screening program introduction in our population in 1998. Incidence after 1998 showed a tendency to stabilization in overall analysis, even though trend showed a significant increase for both periods (1985–98 APC 2.9%, 1998–2012 APC 1.1%). Analysis of age group 50–69 years showed a positive, though non-significant trend after 1998 (1985–98 APC 2.8%, 1998–2012 APC 1.3%). These results suggest the influence of other determinants besides the screening program on the incidence trend shown from 1985 to 2012.

At the beginning of the twenty-first century, a change in this rising trend was seen in many European countries and in the USA. Screening programs initially led to a temporary increase in the incidence, followed by a decrease to pre-screening levels. This phenomenon is related to the diagnosis of silent prevalent cases in the first round and the need to wait until new incident cases occur in the screened population [44]. Moreover, a drop in breast cancer incidence correlated temporarily with the drastic reduction in menopausal hormone therapy in many countries [8, 15, 58] after the results of the WHI study were published [59]. These changes were observed in overall analyses and in postmenopausal women [8, 15]. However, none of these changes was found in our analysis of overall incidence, or in the 50–69 year age group.

A screening program in our population was introduced in 1998, and the whole target population was invited for the fifth round in 2002. In Granada the screening participation rate has been higher than 70% since 1999, and the median detection rate for the entire study period was 3.5‰. These surrogate indicators confirm the good performance of screening in our setting, according to the European Guidelines for Quality Assurance in Breast Cancer Screening and Diagnosis [60]. In light of this finding, the absence of changes in incidence trends seems not to be related to a reduced or delayed implementation of the program.

A previous study of the population analyzed here showed a temporary rise in incidence until 2004, similar to reports from other regions in Spain and consistent with the diagnosis of prevalent cases [9]. The longer follow-up period after screening introduction presented in our paper reduced the likelihood of finding smaller temporal trend changes in the Joinpoint analysis and could explain the absence of changes in incidence trend in our population. However, differences in age groups definition between both analyses, due to changes in the age range included in the Andalusian screening program, could also have played a role.

Some specific characteristics of our population may partially explain the absence of changes in temporal trends. During the study period, hormonal replacement therapy was prescribed to a lesser extent in Spain compared to other European countries [11, 61], so the increase in incidence during the 1990s and later decrease during the beginning of the 2000s due to the usual prescribing patterns were probably not as large as in other countries. Our setting (southern Spain) is at a relatively low socioeconomic level within the European Union, and this factor is known to be associated with a lower incidence [1]. This circumstance may mean a smaller number of silent prevalent cases at the beginning of the screening program, and hence a less dramatic fall after screening began. Finally, we should consider the effect of opportunistic screening as a source of potential bias, as previously described by international organisms [62]. This diagnostic practice shows high detection rates, especially for early stage and in-situ cancer [63], so it could have reduced the amount of prevalent cases that otherwise would have been detected in the first screening round [10]. No information is available regarding the extent of opportunistic screening in Granada province before or during our study period. However, this practice has been proved to be common in other countries [55], as well as in other regions of Spain before screening program introduction [56, 57], and its effect over incidence trends have been considered in previous studies [45, 64].

To better understand the effects of population-based screening, we undertook an analysis by tumor stage at the time of diagnosis for the period from 2000 to 2012. As expected, a statistically significant increase was observed in stage I tumors at diagnosis. The age distribution confirmed that this increase occurred mainly in the age group targeted for screening (50–69 years) – a trend consistent with earlier diagnosis due to screening. However, the absence of a parallel decrease in advanced-stage tumors in our distribution, has been attributed to the non-progressive nature of a large proportion of tumors potentially detectable by the program, and does not support this earlier diagnosis [17]. A favorable stage distribution due to screening is suggested by the lower proportion of stage III tumors in the screened age group (50–69 years old), but no decreasing trend was seen for this group in the Joinpoint analysis.

The decrease we observed in breast cancer mortality was noted throughout the whole period analyzed here. In Spain there has been a generalized decrease in mortality since 1992, although there is some variability among geographical regions [65]. This downward trend started in our cohort before the screening program was implemented, as in almost every region in Spain [65] and in other European countries [23]. Hormonal treatments and new polychemotherapy schemes were also introduced during the 1990s, and together with the increased use of effective radiotherapy regimens, probably played an important role in this trend [24, 36, 45, 66‐68]. Metanalysis of the effectiveness of clinical trials with adjuvant treatments showed a marked reduction in breast cancer mortality, and in some cases, in all-cause mortality [31, 32].

The favorable evolution of survival trends is consistent with findings reported for other European countries [29] and the USA [27]; these trends correlate with tumor stage at diagnosis [69]. In our analysis, we found an increase in stage I tumors during the 2000–2012 period. Despite this favorable trend, survival did not increase in the 70–84 year group or in the subgroup with metastatic tumors at diagnosis. Adjuvant treatment, one of the factors responsible for this trend, is less effective for this stage and age group [70]. Women older than 70 years also have more comorbidities, and breast-conserving surgery plus adjuvant therapy are used to a lesser extent; both of these factors are related to decreased survival [71]. In the 85–99 year age group survival increased markedly from 23% in 1985–1989 to 62% in 2010–2012. However, the small number of deaths in this age group precludes any conclusions regarding this particular subgroup.

Mortality in women older than 70 years in Europe has shown an increasing trend or a smaller decrease than in younger age groups [19]. In our results, mortality increased in this age group (data not shown). This trend was also found for women older than 84 years in a separate analysis. In the 70–84 year and > 84 year age groups the proportion of metastatic tumors was larger than in other age groups (Table 2). Both older age and a greater proportion of metastatic tumors are important factors in the response to treatment. Moreover, women older than 70 years are less likely to receive standard treatment [72].

In our analysis of women younger than 40 years, the incidence trend (APC 3.6%) was larger than the trend reported for this age group in other European countries: the European median APC is 1.2% [13]. There appear to be no clear correlations between trends in this age group and known risk factors [13]. In younger women at least one earlier study found that factors related with tumor biology were associated with a greater risk of death and a worse prognosis [14].

Some authors have noted that changes in diagnostic patterns with the increased use of mammography and ultrasonography, along with wider access to MRI, are likely to be important factors in the reported increases in incidence among younger women [20]. In our study, more than 75% of tumors were diagnosed at stages I–II, and survival rates were similar to those in other age groups. These findings are consistent with the concurrent use of opportunistic screening in parallel with population-based screening programs. The influence of opportunistic screening was demonstrated in Barcelona, where 27.1% of women younger than 40 years received routine screening with mammography before a population-based program was introduced [57]. Moreover, 23.5% of women younger than 45 years reported having a mammography examination in 2014 [73], and 5% of this age group had visited a gynecologist for reasons other than pregnancy in the previous year [74]. Unfortunately, the lack of information regarding hormonal receptors and HER2 overexpression prevented us from analyzing these trends according to pathologic subtypes.

The decrease in overall mortality in Europe is reportedly greater in women younger than 50 years [19], and international studies confirm greater mortality with advancing age [5]. In our cohort, the 0–50 year age group showed a stable trend, in contrast to the decrease observed for women 50 to 69 years old (data not shown). In a differential analysis of the 40–49 year age group, we also found no statistically significant decrease. Previous research in Spain, however, reported a decrease in mortality among women younger than 40 years [75], so the difference between studies may reflect regional differences in incidence. However, caution should be used when interpreting these results given that the number of deaths in this age range is low.

The results of our analysis are strengthened by the inclusion of the most recent available data from the Granada population-based cancer registry, which has been in operation for 25 years and holds data for approximately 9000 registered cases of invasive breast cancer. We used appropriate statistical methods to detect trend changes and to calculated net survival rates. The high quality of the data was ensured by quality control measures, e.g. microscopic confirmation of the diagnosis in 96% of all registered cases, only 1.8% of which recorded the diagnosis based only on information from the death certificate.

Nevertheless, several limitations should be considered when interpreting our findings. The Granada Cancer Registry limits its target population to the provincial level, with a total population of approximately 1 million people. Population-specific characteristics need to be considered, along with the low number of events for some analyses. The absence of consistent information about risk factors prevalence in our population has not permitted us to present a direct interpretation of their role in time trends. Because of the mentioned factors, the external validity of our results may be limited, and the statistical power of some analyses was insufficient to reach definitive conclusions. In addition, lack of stage at diagnosis data before year 2000 refrained us from performing Joinpoint regression of breast cancer incidence according to stage for the whole period. Despite these limitations, this study provides population-based results for a long period of study, documents for the first time the evolution of breast cancer incidence according to stage at the time of diagnosis and provides some clues regarding the effect of screening on this important prognostic variable.

In conclusion, the data from a population-based cancer registry in southern Spain show an increasing trend in breast cancer incidence from 1985 to 2012. The increases were greatest in the 0 to 49 and 50 to 69 years age groups. No change points in incidence trends were found in Joinpoint regression analysis for this period. This evolution is consistent with the spread of risk factors and the rise in diagnostic pressure [3, 8, 16, 23, 36, 44, 45, 47, 55, 76]. Further analyses of incidence trends determinants should be carried out, especially in young women, for designing future prevention strategies. We did not observe the previously turning point in incidence at the beginning of the twenty-first century in our country [8‐10, 77] and, as we have already discussed, several causes may explain this difference.

Incidence trends by stage at diagnosis for the 2000–2012 period show an increase in stage I tumors. The absence of an equivalent decrease in advanced stage tumors suggests that at least a proportion of the tumors detected thanks to screening are non-progressive, raising doubts about screening effectiveness. However, the reduced proportion of stage III tumors in the 50–69 year age group points to a favorable shift in stage distribution. It should also be considered that tumors identified as metastatic at diagnosis represent a more aggressive type of breast cancer that may not benefit from mammographic screening, neither from advances in treatment. Therefore, it is possible that specific prevention strategies for metastatic breast cancer should be developed.

Mortality decreased slightly during the 1985–2012 period, although analysis by age group showed that this trend was statistically significant only in women aged 50–69 years. Although this is the age group the screening program is targeted to, the absence of any change in trend after screening was introduced, and the lack of a clear decrease in incidence during the 2000–2012 period, do not support a substantial beneficial effect.