Breast cancer incidence and survival in Scotland by socio-economic deprivation and tumour subtype

Breast cancer (BC) survival has improved markedly over the last 30 years due to the introduction of mammographic screening and improvements in treatment, including targeted therapies for hormone-sensitive tumours [1]. However, socio-economic inequalities in BC survival persist in Scotland [2] and many other countries [3‐6]. It is well established that BC incidence and survival differ significantly by molecular subtype [7‐12]. Examining whether there are differences by deprivation for different subtypes could inform approaches to reducing inequalities through primary and secondary prevention.

Disparities by socio-economic status (SES) in BC incidence are complex and involve risk factor differences including race/ethnicity, access to healthcare and differences in the predisposition to different tumour types [13‐16]. Data support risk differences by SES for different subtypes [17, 18]. The prognostic disparity by SES has been attributed to patient and clinical factors, including differences in the incidence of tumours characterized by pathologically and biologically aggressive phenotypes, the prevalence of obesity and other comorbid conditions, health-risk behaviours, access to treatment, and quality of care received [19‐21].

Several studies have shown that women living in more deprived areas are more likely than those living in less deprived areas to be diagnosed with oestrogen receptor negative (ER−) and triple-negative (ER−, progesterone receptor negative (PR-), and human epidermal growth factor receptor-2 negative (HER2−)) breast cancers (TNBC) [18, 22, 23]. Race/ethnic differences in incidence of hormone negative and more aggressive BC subtypes have been observed in the US where it can be difficult to separate racial and socio-economic disparities [24‐26]. TNBC tumours are associated with early recurrence and poor survival due to lack of specific targets for commonly used adjuvant therapies [27]. It remains unclear whether differences in TNBC incidence by SES explain the observed worse prognosis of BC patients living in areas with greater socio-economic deprivation.

Greater understanding of the role of socio-economic deprivation on the incidence and survival of different subtypes of BC could inform the development of interventions aiming to reduce disparities and improve BC prognosis. Within the high-quality Scottish cancer registry, we previously showed distinct temporal trends in cancer incidence by ER status [28]. Here, we aimed to determine whether incidence (time trends) and survival by ER status and immunohistochemical (IHC) surrogate molecular BC subtypes differed by an area-based measure of SES. As a secondary aim, we investigated the effect of screening (mode of detection) on BC time trends for each SES group.

We previously reported increasing incidence of ER+ tumours by an average of 0.4%/year and decreasing incidence of ER− tumours by 2.5%/year (95% CI: − 3.9 to − 1.1%) across Scotland between 1997 and 2016 and identified that screening was a major contributor to rising incidence of ER+ tumours [28]. Here, we observed that although trends over time were similar to those previously reported regardless of deprivation, incidence increased mainly amongst women living in least deprived areas of Scotland with screen-detected ER+ tumours (AAPC of 2.9% compared to 1.6% previously reported overall) [28]. Absolute incidence for ER+ screen-detected tumours was also higher amongst the least deprived compared to the most deprived (with approximately 50 more cases per 100,000 women at the peak in 2011). Screening uptake might partially account for the differences in BC incidence observed between most and least deprived areas. This is supported by data showing uptake of BC screening in the most deprived areas of Scotland was 59.5% in 2016–2019 and 79.7% in the least deprived areas [44].

We found lower point estimates and no statistically significant association between deprivation and BC survival amongst women with the rarer subtypes of TNBC or luminal B tumours. Previous studies from Scotland and other countries have found an association between SES (at both individual and neighbourhood level) and BC mortality, with women with low SES having a higher BC mortality [22, 45, 46]. Some data show women with low SES are more likely to be diagnosed with more aggressive BC subtypes, particularly ER− and TNBC subtypes [18, 22, 23]. However, evidence of whether survival rates for subtypes differ by SES has not been investigated previously. In our multivariable analysis, deprivation was associated with statistically significantly higher BC mortality for luminal A and HER2-enriched tumour subtypes but not TNBC and luminal B tumours, for which the association was attenuated and no longer statistically significant after adjusting for screening, treatment and the Charlson index for comorbidities. Risk of BC death for HER2-enriched tumours appeared particularly high, albeit with limited power and wide confidence intervals for the most deprived areas compared to the least deprived areas, and this finding will require confirmation in other datasets to determine if it replicates. Cumulatively, our findings support associations with socio-economic deprivation in survival differ by subtypes.

Possible additional factors that could be contributing to survival differences by deprivation are alcohol intake, obesity and smoking. In Scotland, alcohol-related hospitalisation and mortality was up to 8 times higher across people from the most deprived areas. However, men and women in the most deprived areas of Scotland are less likely to drink hazardous or harmful alcohol levels than those in the least deprived areas (10% drinking at hazardous/harmful level vs 20%) [47]. Further, heavy drinking has been also consistently linked to weight gain [48]. In Scotland, obesity prevalence in women is around 30% and 20% in the most and least deprived areas, respectively [49]. Physical activity is also a noted risk factor that is approximately 20% lower in the most deprived compared to the least deprived communities in Scotland [50]. Smoking could also be a contributing factor given that prevalence was 30% compared to 9% in women in the most and least deprived areas of Scotland in 2018 [47]. The more marked differences by deprivation amongst women diagnosed with luminal A or HER2−enriched tumours than for luminal B and TNBC tumours may also be related to differences in prognosis and/or treatment adherence [51].

This study has several strengths as to our knowledge is the first study in the UK to investigate BC incidence and survival by SIMD and molecular subtypes utilising high-quality data from the Scottish cancer registry with linkage to mortality and comorbidity records. As for any observational study, the validity of our findings must be assessed in terms of potential confounding and bias. Although our analysis controlled for some potential confounders, there was no information about other risk factors, such as obesity alcohol consumption, smoking and physical activity. Another limitation is that survival rates can be affected by lead time and length biases [52].The lead time bias refers to the additional number of years added to the survival time of all women whose tumours were detected by screening [53]. Our data support that this is likely differential between deprived groups and needs to be considered in analyses of inequities with BC. On average, lead time bias is estimated to be 3 years, hence reporting 5-year survival estimates might help reduce its impact on survival rates. Length bias relates to the tumour’s presymptomatic period when it is mammographically detectable, called the sojourn time. Screening preferentially detects tumours with longer sojourn times; therefore, tumours detected through screening are slower growing and less lethal. Although we present stratified analyses by ER status and adjustment for whether tumours were detected through screening in our analysis, the potential for residual confounding remains as women who accept invitations to screening are likely to differ from women who do not attend screening in ways that may influence survival. Another limitation is the validity of BCSS analysis depends on the accuracy of cause of death as recorded in the registry which assumes that the underlying cause of death has been accurately determined for each woman. Skyrud et al. [36] compared cause-specific and relative survival estimates and found cause-specific estimates to be as reliable as relative survival estimates, particularly for common cancers. Another possible limitation is competing risks of death with women in most deprived areas being more likely to die from other causes than BC. In order to minimise competing risks of death, we restricted survival estimates to 5 years. Finally, the SIMD is an area-based measure of deprivation so it can misclassify individuals’ SES [54]. Potential misclassification is a particular risk for rural areas where the index domains particularly the ‘access’ domain fails to capture important singularities of the rural areas, such as, frequency and cost of public transport [55].

This analysis using high-quality population-based data in Scotland shows differences in incidence and prognosis between an area-based measure of SES for different molecular subtypes of BC. Determining factors that are associated with differences in the incidence and survival for different subtypes could help identify interventions for modifiable risk factors and/or identify high risk individuals to try and detect cancers earlier through screening. Tackling inequalities in BC require more detailed analyses such as ours that report incidence and survival for disease subtype characteristics [56] including stage, ER and IHC-defined molecular subtypes. These analyses should help identify where inequalities exist (or don’t) allowing cancer control programmes to focus on inequalities where they are greatest. Few studies have been able to stratify these results by subtype as well as mode of detection due to the lack of availability of such data. More detailed data on risk factors by SES, screening participation, lifestyle behaviours (e.g. smoking, physical activity, and alcohol intake), comorbid conditions, treatment and tumour subtype are required. As recently proposed [56], future analyses using modern methods of causal mediation analysis would be important to accurately estimate the contribution of potential explanatory factors for inequalities; this would provide evidence that could translate into improvements in primary and secondary prevention of BC that would have the most impact with regard to mortality.