Background
Advances in breast cancer screening and treatment have enlarged the pool of breast cancer survivors [
1]. These women are at risk of breast cancer recurrence and premature death, but data on how socioeconomic position (SEP) influences this risk are scarce.
SEP has been associated with breast cancer recurrence, but most studies stem from settings of unequal healthcare access and non-uniform insurance coverage [
2‐
6]. Furthermore, most studies focused on racial, ethnic, or insurance-related disparities [
2‐
5]. Countries with tax-funded healthcare could be expected to offer equal treatment and follow-up after cancer irrespective of SEP. However, in Nordic countries, health inequality persists despite the build-up of the welfare states (also called the Nordic paradox) [
7]. Studies set in Denmark, a country with tax-funded healthcare, report increased risks of breast cancer recurrence (and other cancer types) among patients with low education and in those living alone [
8], and a higher 5-year mortality in women with lower education and income [
9,
10], even after adjusting for tumor characteristics, treatment, and comorbidities [
11,
12]. The same tendency has been observed in other European countries [
12]. Thus, SEP-related disparities in breast cancer prognosis are evident even in populations with universal healthcare access.
The potential influence of SEP on breast cancer prognosis may differ according to patients’ characteristics. Scandinavian studies show increased mortality in women with breast cancer with low SEP aged below 50 years (assumed premenopausal) [
13‐
15], whereas no such association was seen in their older counterparts [
13,
14]. Moreover, stage-specific survival has improved less over time in Norwegian women with low SEP, compared to women with high SEP [
15].
Breast tumors in premenopausal women often have poor prognostic characteristics—higher stage and grade and estrogen receptor (ER) negative [
16,
17]. Accordingly, most premenopausal women are recommended taxane-based chemotherapy, followed by up to 10 years of tamoxifen therapy for those with ER+ tumors, whereas postmenopausal women to a greater extent can forego chemotherapy [
18‐
20]. Yet, studies on breast cancer prognosis according to SEP did not distinguish pre- from postmenopausal breast cancer [
8‐
11], lacked treatment-specific information [
8,
15,
21,
22], failed to stratify by ER status [
8‐
11,
14,
15,
21,
22], and included patients with metastatic breast cancer [
11,
13,
14,
22,
23]. We therefore investigated the association between individual-level SEP and breast cancer recurrence and overall mortality in a contemporary cohort of Danish premenopausal women with non-metastatic breast cancer, treated with taxane-based chemotherapy. We evaluated interaction by ER status and associated tamoxifen therapy.
Discussion
Single women had increased recurrence risk and mortality compared with their married counterparts; this was mainly seen in those with ER+ tumors. We observed increased mortality in women who had low education, low income, were unemployed, or had health-related absenteeism, especially among those with ER+ tumors.
As outlined by Green at al [
42], SEP consists of several elements, which cover separate aspects of SEP, that overlap in a shared core dimension of SEP. One single measure may not completely cover the underlying SEP, but our similar findings across several SEP measures support an impact of SEP on breast cancer prognosis, especially mortality. Although we used administrative data assuring high validity, some misclassification of SEP was possible, and other categorizations could possibly have shed light on other aspects of SEP. Some of our estimates were imprecise, but across SEP measures, the lower bound of the 95% CI had the same direction, supporting that the core dimension of SEP may be associated with prognosis. A slightly larger sample size would have increased the precision of our estimates.
Our findings were partly modified by ER status, similar to findings reported by others [
43]. A previous study set in the ProBeCaRe cohort showed that 22% of women with ER+ tumors prematurely discontinued their tamoxifen therapy [
44]. Early discontinuation increased the recurrence risk, and lower adherence was observed in single women [
44]. Tamoxifen is sometimes associated with bothersome side effects, which may impact treatment adherence [
45]. Likewise, taxanes incite a wide range of side effects leading to treatment disruption and/or dose reduction, limiting treatment effectiveness [
46]. Social support from a partner might influence treatment compliance, particularly among women suffering from side effects. This may have impacted our findings of higher recurrence and mortality among single women. We did not have information on chemotherapy adherence. A study in postmenopausal women shows that discontinuation of chemotherapy is most often due to comorbidities or age, and rarely based on patient request [
47]. Our cohort of young patients had a low prevalence of comorbidities and a long life expectancy. We therefore expect it to be rare for them to discontinue a time-limited treatment such as chemotherapy. Still, any discontinuation of therapy could differ by SEP and could be more likely among patients with low SEP [
44,
48]. This could have influenced our findings.
Similar to our crude estimates, the aforementioned Danish study on breast cancer recurrence reported lower risks in women living alone, also after adjustment for age, education level, comorbidity, calendar period, tumor and lymph node stage, and adjuvant therapy (chemotherapy and/or radiotherapy) [
8]. However, their findings may be affected by residual confounding from other dimensions of SEP, which were not incorporated in their study (e.g., income and employment). We note that adjustment for other SEP measures attenuated our estimates. The previous Danish study used a validated algorithm capturing recurrences in Danish registries [
49]. However, their adjusted estimates for education were similar to ours, suggesting non-differentially underreporting in DBCG across SEP categories. As such, our relative estimates may not be affected by the underreported recurrences in DBCG.
Our sensitivity analyses confirmed that deaths in this young population were mainly due to breast cancer. Accordingly, the equivalent mortality, but not recurrence risk, especially across levels of education and employment, may indicate underdetection of recurrences in deprived groups. Danish women with low SEP are less likely to be referred to, and to attend, cancer rehabilitation programs [
50,
51], including mammography screening [
52]. Recurrences may therefore be underdetected in women with low SEP if they opted out of the follow-up programs. These findings may also enhance the understanding of SEP-related disparities in non-tax-funded healthcare settings, by indicating that inequality in cancer survivorship extend beyond obvious factors like insurance coverage.
Several issues should be considered when interpreting our findings. The main strength of this study is the use of routinely collected registry data, with high validity and completeness [
24]. The clinical data registered in the DBCG have high validity [
41,
53], and the expected direction and association of prognostic characteristics, e.g. stage, were evident (data not shown). Nonetheless, based on a previous study validating DBCG data against medical records [
41], our observed incidence of recurrence may be underestimated.
We used registry-based individual-level measures of SEP, with high validity and completeness [
24,
30‐
32]. As such, our classification of marital status had high validity for marriage and registered relationships [
24], but does not capture unregistered cohabitation, which is a common way of living in Denmark [
54]. We therefore included cohabitation, which, besides marriage and registered partnerships, also considers two persons of different genders at the same address as cohabiting partners if the age difference is less than 15 years, or if they have children. As such, women with a female roommate or partner would be registered as living alone, leading to misclassification. We identified a group of cohabiting singles; this may include women in unregistered partnerships or divorced women living with their old or a new partner. However, the suggested mediation of cohabitation on marital status was imprecise and presumably a chance finding.
Our findings may be prone to unmeasured confounding. Lundqvist et al. [
12] found that reproductive and lifestyle factors contributed to the association between SEP and mortality in breast cancer patients. Naturally, reproductive factors in singles may differ to those in married women. Although we incorporated CCI as a proxy for patient health, we had no information on lifestyle factors, e.g., smoking and diet, which may have influenced our observed associations. However, reproductive and behavioral factors in relation to breast cancer are more relevant in postmenopausal women [
55]. The CCI does not include mental disorders, which account for 50% of all long-term sick leave in Denmark [
56]. Mental disorders before breast cancer are associated with early retirement [
57], suggesting exacerbation after diagnosis. This may have contributed to our findings in women with health-related absenteeism.
The average time from recurrence to death varies by breast cancer subtypes, being longer in women with ER+ disease (~ 2–3 years) and shorter in women with ER− (~ 1–2 years) [
58]. Thus, we may have underestimated the 5-year recurrence incidence in our sensitivity analysis where we considered BCSMs as recurrences.
Last, docetaxel was the taxane-compound used during the study period, but this has gradually been replaced by paclitaxel due fewer adverse effects [
26]. Docetaxel and paclitaxel exert similar structural and microtubule-stabilizing effects [
59] and have similar effectiveness [
60]. Therefore, our findings are relevant to current clinical practice.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.