Socioeconomic aspect of breast cancer incidence and mortality in women in Lower Silesia (Poland) in 2005–2014

The importance of socioeconomic factors as health determinants is well known. As the impact of social factors on health is supported by strong and commonly observed relationships between a wide range of health indicators and measures of socioeconomic resources or social position of individuals, such as employment status, income, education level, or position in the occupational hierarchy, it is important to systematically analyze the relationship these parameters have with the territorially observed incidence of civilization diseases, which include neoplasms, including breast cancer. Unemployment translates into many adverse factors that increase cancer morbidity and mortality. Loss of employment worsens access to health care (loss of insurance, lack of financial resources to use in the private sector, lack of resources for commuting and medications). Increased unemployment in the population translates into deterioration of health-seeking habits (increased alcohol consumption, smoking, promotion of unhealthy dietary patterns, and discontinuation of participation in preventive screenings). Forced unemployment causes serious psychological implications (trauma, depression, chronic anxiety, and stress) which translate into increased biological viability for cancer [1, 2, 3, 4, 5, 6].

Many current studies showed significant findings regarding the association between being unemployed and breast cancer (BC) incidence and mortality [7, 8, 9, 10, 11]. However, on the other hand, in some articles the relative risk among areas with high and low unemployment was not found for BC, concurrently [5, 12]. Convincing evidence of the influence of socioeconomic factors has not been found in the context of geographical location and breast cancer. This aspect is so important because compared with urban women, non-urban women were more likely to live in disadvantaged areas. They had differing clinical management and patterns of care [13]. In the recent extensive systematic review by Williams and colleagues [14], half of the studies reported higher proportions of women diagnosed with breast cancer who had resided in urban rather than rural counties, whereas the remaining half showed no statistically significant association between place of residence and the BC diagnosis.

The aim of this study is to update the results of research on changes and correlations in breast cancer incidence and mortality rates in the Dolnośląskie (Lower Silesia) province, Poland (≈ 20 K km², 30 counties, ≈ 2.9 M ppl., 2019), based on the latest economic, spatiotemporal, and epidemiological statistics. The obtained results can be used for the socioeconomic assessment of risk factors and improvement of health care organization (distribution of full- profile cancer centers, organization of local preventive and educational programs) in local populations, and translationally in global terms.

Two data sources were used for this article. Breast cancer (C50, after International Classification of Diseases, Tenth Revision) incidence and mortality data were derived from the Lower Silesia Cancer Registry in Wroclaw, Poland. Within the decade of 2005–2014, 14,158 newly diagnosed cases and 4096 deaths from BC were registered in the province. Women's population and annual unemployed numbers at the county level were obtained from the Statistical Office (unemployment rate has been calculated from the absolute number of unemployed persons to the total number of females).

The average 2005–2014 female population and unemployment rates in the region in the analyzed administrative units is presented on the thematic maps in Figure 1.

Fig. 1

The standardized incidence and mortality ratios (SIRs and SMRs) were estimated based on the conditional autoregressive (CAR) model, providing spatial patterns of the disease [15]. SIR is a ratio between the observed numbers of incidence in the population and the expected number of cases (if the ratio of observed/expected incidences is greater than 1.0 = 100%, there is said to be “excess risk,” whereas when SIR<1.0, then we are dealing with “moderate risk”). The SIR/SMR is estimated by the ratio of observed (O) and expected cases (E). E is calculated as E=∑_i,jpy_ijIR_ij where py_ij are the observed person-years in the full cohort for age group i and period j, and IR_ij the corresponding incidence rates in the reference population. In general, based on the Bayesian approach, the resulting estimators represent a weighted compromise between the overall rate and a local mean of the relative rate in nearby areas.

Cancer incidence and mortality growth rates (GRs) in each county were estimated following the spatiotemporal modeling by Congdon [16]. Spatiotemporal modeling relates to problems where disease varies over space and time. The computation was performed in the WinBUGS platform [17], following the WinBUGS code provided by Bailey [18] and based on the initial burn-in of 1000 samples and the following 10,000 production run cycles of the Gibbs sampler (an equilibrium state of streams of values was established via an examination of within chain statistics available in the WinBUGS software).

To test correlations in SIR and SMR among nearby locations in space, the Moran's spatial autocorrelation statistic was applied, based on both feature locations and feature values simultaneously [19]. This method evaluates whether the spatial pattern expressed by SIR and SMR is clustered or random. In addition, using spatial regression coefficients, we determined the effect of the unemployed female population on the risk of developing breast cancer and risk of death from BC in the Lower Silesia region based on the mentioned CAR model [15].

The spatial clusters of the disease were detected using spatial scan statistics [20]. Routinely, the term disease cluster (“hot spot” or “cold spot”) is used if an unusually greater or smaller than expected number of cases of the disease occurs in a group of people living in close proximity and over a limited period of time. The risk of cancer within the detected hot and cold spots were expressed by the relative risk (RR>1 and RR<1, respectively) with the same interpretation within SIR and SMR. The disease clusters were computed via SatScan software [21]. All the estimates were illustrated graphically in thematic maps. Significance level was established at p<0.05.

The counties of the Lower Silesia are characterized by an uneven distribution of population density (Figure 1A) and unemployment rate (Figure 1B). Observable corresponding geographical patterns in Figures 1A and 1B can be seen. A simple Pearson's coefficient estimate (r= −0.46, p= 0.011) shows statistically significant negative linear correlation between female population numbers in Lower Silesia and unemployment rates.

Following the spatial patterns, illustrating breast cancer standardized incidence and mortality ratios in Lower Silesia, Poland (2005–2014) (Table 1), we observed a fairly dynamic spatial pattern of the incidence in Lower Silesia (Figure 2A) but a more stable mortality model from BC (Figure 2B). In the above analysis, the Oławski County appears to be the most unfavorable county in terms of SIR (>1.3) in the Lower Silesian province (Figure 2A), with the lowest SMR level (<0.9) at the same time (Figure 2B). Such an atypical relationship has not been observed in any other county. On the other hand, in the adjacent Oleśnicki County, an inverse relationship was observed: the lowest range of SIR≤ 0.8 and highest SMR>1.1 (Figure 2).

Fig. 2

The cancer incidence and mortality growth rates and provincial means (with credible 95% intervals) for the entire region are shown separately in Table 1, respectively. The highest BC incidence growth rate (GR>15%) in the analyzed time period was observed in the city of Wałbrzych (Figure 3A). Altogether the positive GRs (>0%) of BC incidence are observed in half (15) of the counties and in 13, in the GR mortality pattern. Visually, both the models in Figure 3 are not associated with each other. The Pearson's r linear correlations between SIR, SMR, and GRs are given in Table 2.

Fig. 3

Following the results reported in Table 2, the borderline statistically significant (p<0.1) inverse linear correlation (r= −0.35) was calculated between SIR and SMR, and a positive correlation (r= 0.34) between SMR and GR (incidence). At the same time, moderate positive (r= 0.60) but statistically significant (p<0.05) correlation was found between SMR and GR (mortality). Using Moran's I test, the estimated p-values are given in Table 3. Based on the Moran's I inferential statistic (Table 3), it can be established that it is quite possible that the spatial distribution of SIRs and SMRs (Figure 2) are the result of random spatial processes (if p>0.05, then SIR and SMR are spatially independent), whereas the geographical patterns of GRs in both the models (Figures 3) are more spatially clustered (if p<0.05, then GRs must be spatially dependent).

The results of modeling the statistical impact of the unemployed fraction in the entire female population on the breast cancer morbidity and mortality are summarized in Table 4. As can be seen from the results collected in Table 4, in Lower Silesia in 2005–2014, a negative impact of unemployment among women on breast cancer morbidity was found, while a positive effect on deaths of women due to BC, in parallel, and both the results are statistically significant (p<0.05). Arithmetically transforming these data, it can be inferred that the fraction of 10% of unemployed women generated epidemiologically circa (1-exp(−0.03)¹⁰)*100% ≈ 25% decrease in the risk of developing breast cancer. The same percentage of unemployed women was associated with an approximately (exp(0.04)¹⁰-1)*100% ≈ 50% increase in deaths from BC.

The performed spatiotemporal disease clusters analysis of BC data discovered a statistically significant (p<0.05) 2 hot and 3 cold spots in incidence and only 1 hot disease cluster in mortality (Table 5). The hot (Clusters 1 & 2) and cold spots (Clusters 3–5) of BC incidence are presented graphically in Figures 4 A & B, respectively, whereas the hot spot of BC mortality (Cluster 6) is shown in Figure 4 C. The hot clusters (Figure 4) are localized in the Walbrzych area and neighboring counties. In all the above examples (Figure 4), the disease clusters’ locations roughly correspond with the SIR, SMR, and GRs displayed earlier (Figures 2 & 3).

Fig. 4

The importance of mortality data in health assessment is well known. The status of any population in disease control planning and health interventions is essential for the proper functioning of the healthcare system. The relationship between the socioeconomic status (SES) of populations, the distribution of health care, and cancer incidence and mortality are important prognostic factors, useful in health-care screening programs, especially oncological screening programs. On the basis of seemingly routine socioeconomic and epidemiologic data and modern geostatistical methods, a large handful of research findings can be made. In our analysis, breast cancer incidence and mortality were subject to many interesting spatial interactions. One of the indirect results is dependence between female population numbers in Lower Silesia and unemployment rates. This relationship can certainly be justified by a higher percentage of the unemployed in agricultural areas, and vice versa in urban areas, which has already been demonstrated in many socioeconomic analyses [22, 23].

A controversial observation is certainly the opposite tendency (relationship on the border of statistical significance) between the risk of developing breast cancer and mortality caused by the disease. This result could be explained by the fact that in metropolitan areas there is better access to health care facilities than in rural areas, which results in faster oncological diagnostics and a theoretically higher rate of breast cancer incidence. Moreover, closer therapeutic contact favors better treatment outcomes of the diagnosed disease in patients, resulting in lower rates of breast cancer-related deaths. Similar observations were already found in other studies [24, 25, 26] and do not require further comment.

On the other hand, in some counties, despite the similar availability of oncological care, we observed extremely different breast cancer incidence and mortality rates. Good examples are the adjacent Oławski and Oleśnicki Counties, both are characterized by similar geographic and communications accessibility to the main oncology center in the region, located in the city of Wrocław, but differing in the percentage of unemployment among women (Figure 1B). This observation indicates that here are additional very powerful factors (other than health care infrastructure) influencing the incidence and mortality rates of breast cancer [10]. In terms of the analyzed BC risk indicators, the positive relationships between SMR and the incidence and mortality GRs seem to be novel discoveries, and not only in our regional epidemiological research. These results may indicate a relationship between the introduction of population-based mammography screening in Poland in 2006, which resulted in improved breast cancer detection but not reduced mortality: due to the longterm survival of stage IV patients, mortality was not reduced in the analyzed period [27]. However, clear confirmation of the coexistence of the discussed epidemiological phenomena requires further observations. A positive correlation between the mortality rate and growth in the mortality rate also invites further discussion, but the problem seems to be difficult to interpret at this stage of research and requires further epidemiological analysis.

The spatial independence of the incidence and death rates in the studied Wrocław province is similarly vague, in light of the spatial dependence of adjacent surface units in the context of temporal BC risk increments. Unknown additional risk factors could explain the observed territorial randomness of disease incidence and the tendency to spatial aggregation of mortality. This, however, requires further epidemiological studies.

The inversely proportional relationship found between the unemployment rate and the breast cancer incidence rate, which is directly proportional to deaths from the disease, seems to be already explained by the urban-rural effect and worse health habits in societies with lower socioeconomic status [1, 2, 3, 4, 5]. All these results would not be possible without modern epidemiology and geostatistical methods that meet the social and health needs of the population.

Finally, in the light of the estimated results, we think, the reorganization of oncological care in the region, combined with the opening of new oncology centers, including a radiotherapy center in 2009 in the city of Wałbrzych, may have contributed to the above dependence. The above action improved the availability of specialist health care, which could have provided both the improvement of BC detection and the increase in the registration of deaths from it [25].

The following general conclusions can be drawn from this study:

The estimated spatiotemporal incidence rates and incidence growth rates correlate with mortality growth rates (Table 2) and show relationships within neighboring territorial units (Figures 2 and 3).