Breast cancer and long-term employment: A retrospective cohort study from Norway.

Introduction
Breast cancer is a leading cause of death among female cancer patients
1
and comprised 24% of all female cancers in Norway in 2022.
2
The incidence of breast cancer has increased over the past decades,3,4 and the relative survival has increased continuously due to earlier diagnosis and better treatment.5,6 Unfortunately, surgery, chemotherapy, radiotherapy, hormone treatment and other treatments alone or in combination may have detrimental effects to the daily life and well-being of many breast cancer survivors (BCS).
7
Moreover, fatigue, memory complaints and pain may result in reduced ability to work, occupational concerns, struggle in returning to work, and risk of permanently leaving the labor force.8
–10 Compared with other cancers affecting women, breast cancer is more frequently diagnosed in the working age group (<65 years),
9
and the negative consequence of reduced employment for the individual BCS may be substantial. Not only does discontinuation of employment result in financial instability,
11
work and the workplace are also important for fostering social relationships, contributing to people’s quality of life, and they give meaning to life.
12

Unfortunately, few longitudinal studies have examined employment outcomes among women diagnosed with cancer,13
–15 and few studies have included a matched control group.9,16,17 A Finnish study found that employment among BCS was reduced during the first 5 years following diagnosis,
18
however earnings reductions were most severe during the first 2 years. A Dutch study found that BCS more often received disability benefits and lost their paid employment compared to the general population 5–10 years after diagnosis.
15
A German study comparing patients whose primary diagnosis was breast cancer with matched controls from the general population found that nearly three times as many cancer patients had left their job compared to the reference group, and the probability of returning to work 6 years after surgery was only half of the reference group’s.
19
However, they included patients up to 70 years of age, resulting in a substantially lower number of patients at follow-up since many reached the retirement age of 65 years during the study. Thus, inclusion of more patients below retirement age is warranted in future studies.
Both health-related and socioeconomic factors negatively influence employment in BCS.20,21 To our knowledge, only a few studies have assessed the effects of a wider variety of sociodemographic risk factors in a retrospective cohort study. Factors such as chemotherapy, cancer stage, older age and less accommodating employers are associated with reduced working hours, long-term disability, and early retirement in the short-term.
22
However, a Dutch study showed that work related factors (e.g. type of contract, physically demanding jobs, and current work ability) were associated with adverse work outcomes 5–10 years after diagnosis, whereas clinical factors (e.g. surgery, chemotherapy, radiotherapy) were not.
23
Importantly, recent studies have shown that the time needed to adapt and manage late effects of cancer treatment during the return-to-work period is a process that may last for several years.24
–27 Since many breast cancer patients are diagnosed relatively early in their life
1
and are expected to live for many years,
28
there is a need for studies with longer follow-up periods. In fact, only a few studies have addressed the influence of clinical and socioeconomic factors on long-term employment after diagnosis.22,29 Thus, the influence of these factors after the first years remains largely unknown in many countries. Long-term studies should preferably apply a matched control group, reducing the influence of bias because of country specific labor market structures and inherent changes in employment rates within the country’s overall population.
People with chronic diseases have reduced employment prospects compared to people without health challenges, in part because they have more difficulty re-entering the labor market, or they leave employment earlier.
30
However, most studies of employment among BCS have focused on those working at the time of diagnosis. An important but little studied component of breast cancer is its role in reducing entry into employment among individuals not-working before cancer diagnosis.31,32 Thus, it is important to include information regarding employment status at diagnosis because BCS in employment and BCS outside employment may not have the same likelihood of being in employment after cancer.
The primary aim of this study was to investigate the likelihood of employment during the first 6 years after diagnosis among BCS and matched controls who were in employment at the time of diagnosis. The secondary aim was to investigate the likelihood of entry into employment within 6 years after diagnosis among BCS and matched controls who were not employed at the time of diagnosis.

Design and methods

Registries
In this retrospective register-based cohort study, BCS were identified by use of the Cancer Registry of Norway, which is a comprehensive population-based registry with mandatory reporting. Norway has a universal, and largely free for all, health care system, thus the Cancer Registry encompasses virtually all Norwegian cancer patients.33,34 Demographic data (e.g. age, gender, education) and work-related data (work contract, industry code) for BCS and controls were received from Statistics Norway. Data from the registries were linked through a personal identification number.
35

Material
In the Cancer Registry of Norway, we identified 10,454 new breast cancer cases (ICD-10 code C50 Malignant neoplasm of breast) diagnosed from 2004 through 2008 for all ages. We excluded patients with multiple cancers (n = 1589), those who died during follow-up (n = 1646), a small group of stage IV patients (n = 32) and two male patients. Only BCS aged 30 through 55 years were included in the analyses; patients outside this age range (n = 3451) were excluded for reasons described below. Self-employed individuals were not included in this study because it was impossible to ascertain whether self-employed individuals in fact were working at all or at different times during the year, for example, at the time of the cancer diagnosis. The age range of 30–55 years was chosen based on labor-force participation in general, and that breast cancer is most often diagnosed among people aged 40 and above. Among younger people, many may still be in education in their 20s. Early retirement increases from about 60 years of age
36
and participants close to this age at the end of our follow-up may more likely choose to retire instead of making efforts to return to employment. By excluding BCS’ early retirement availability, the employment outcome should not be biased by this alternative.
Each BCS was matched with two unique cancer free controls randomly extracted from the remaining general population (i.e. after the aforementioned exclusions). Controls had to be cancer free during follow-up (i.e. through 2014). Matching was based on age, education level and employment status at time of diagnosis. Our data allowed us to use exact matching, that is, BCS and controls were identical regarding the matching variables. Exact matching is the preferred matching method
37
but can result in larger bias if there is a failure to match a large share of cases.
38
We were able to match all selected BCS to identical controls, thus there was no selection bias on the matching variables. Controls were given a pseudo-diagnosis date according to their corresponding BCS’ diagnosis time. When controls were matched to a cancer case, they were blocked from being selected again.
Some BCS (N = 131) were excluded because of missing information on employment industry and/or sector, and a few controls (N = 19) did not have valid information on industry, thus 19 BCS have only one control.
The dataset ultimately included 3550 BCS and 7081 controls, who were followed from 2004/2008 through 2010/2014. The year of diagnosis (2004/2008; baseline) is labeled T0, the first year after diagnosis is labeled T1 (2005/2009) and so on to T6 (2010/2014). Thus, we followed up five pooled cohorts of BCS for a total of 6 years.

Variables
Employment was the dependent variable and for t0 (baseline), it was defined as being registered with an employer within the 90 days prior to- and including the date of cancer diagnosis (e.g. if a BCS was diagnosed on December 31st 2004, then she had to have a valid work contract for October–December 2004). All employment contracts were included; thus employment includes both part-time and full-time employees as well as employees on temporary contracts. For the follow-up years (t1–t6), employment was similarly defined as having a valid contract within 90 days prior to the start of the given follow-up year (i.e. for t2: a valid employment contract in October–December 2006). BCS not employed in a given year were regarded as out of work in that year but were re-included as employed if a new work contract was subsequently registered.
In the regression analyses, we adjusted for cancer stage (I–III) among BCS (cancer diagnosis; C50), whereas controls had no cancer (=ref.). The cancer stages are defined as follows: Early stage (I)—all cases where the tumor is confined to the primary organ; Localized stage (II)—all cases with metastasis to nearby lymph nodes; Regional stage (III)—all cases where the tumor has invaded neighboring tissue outside of the primary organ or metastasized to regional lymph nodes.
2

Independent variables at t0 (baseline) included: Age (in years) and age-squared; Educational level coded as basic/unknown education (=ref.), secondary education, short tertiary education (BA-level) and long tertiary education (MA-level or higher).
Family type was coded as single individuals without children (=ref.), singles with children, couples with children, couples without children, and unknown family type; Industry (eight groups as shown in Table 1, e.g. transportation, public administration, human health and social work (=ref); and Sector coded as private sector (=ref.) and central/local government.

Statistics
The analyses included descriptive characteristics of the study population (BCS and controls), and overall share of BCS and controls in employment during follow-up according to employment status at time of diagnosis (Figure 1). Differences in the share employed between BCS and controls were tested by means of t-tests. If our matching removed all observed and unobserved imbalances between BCS and controls, the differences between BCS and controls in Figure 1 are correct and unbiased. However, this is both uncertain and probably unlikely. To investigate imbalances between BCS and controls we included the variables described above in our subsequent analyses. Due to our initial exact matching, BCS and controls had very similar family, industry and sector characteristics at the time of diagnosis. Thus, on a group level, BCS and controls were also closely matched on family type, industry and sector at t0. Separate multiple logistic regression models for each year during follow-up (t1 through t6) were estimated for two BCS groups: (i) employed at diagnosis or (ii) not employed at diagnosis, and their respective controls. The advantage of this approach is that BCS or controls temporarily not in employment at a given time during follow-up were retained if they returned to employment at a later stage. In the multiple logistic regression results: we presented the average marginal effects for each covariate, which indicates the average change in the predicted probability of the outcome measure (e.g. employment at t6), while keeping the values of all the other covariates constant across all observations in the sample. Stata’s “margins” command was used to estimate the standard errors and confidence intervals (CIs) for the marginal effects shown in the Tables 2 and 3. Logistic regression tables results are presented in Supplemental Files 1 and 2.
The study included treatment variables: surgery, chemotherapy, radiation, and hormone therapy. However, in a model with cancer stage, these treatment variables were not significant due to high correlation with cancer stage (multicollinearity). This is because cancer stage (to a certain extent) determines treatment. The treatment variables were thus excluded from our final analysis.
Matching variables were included in our analyses, thus adjusting for group residual covariate imbalances38,39 and enabling us to check whether the likelihood of employment post-diagnosis between BCS compared to controls varied by age and education level; we found no significant interaction effects (not shown). All analyses were performed in Stata/SE 16.1 for Windows. This manuscript was prepared following the STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) statement
40
(Supplemental File 3).

Results

Sample characteristics
Among the BCS included, 44% were diagnosed with Stage I, 46% with Stage II, whereas Stage III and unknown stage were less common (5.5% and 4.5%, respectively; Table 1). The age of the BCS were on average 47.6 years when diagnosed, more than half (53%) of them were living with a spouse/partner and children; and the most common educational level was secondary education (42%) followed by lower-level university education (30.7%), basic education (20.2%) and higher-level university education (7.2%). Seventy-eight percent of BCS were employed at the time of diagnosis, and most of them were working in the human health and social work sector (27.9%). BCS and controls were matched by age, educational level and employment status, thus the corresponding distributions among controls were identical on these variables (t-test/Chi-squared tests: p > 0.05). BCS and controls were not matched according to family type, industry, or sector, but their distributions in these groups were largely comparable. The most notable difference was the larger share of controls in the “other/unknown” category of family type compared to BCS.
Among individuals employed at the time of diagnosis, the percentage employed during follow-up decreased, with a slightly more rapid decrease among BCS compared to controls (Figure 1). The difference between BCS and controls is at its largest at t2. Among individuals not in employment at the time of diagnosis, the percentage employed during follow-up increased persistently during follow-up, although mostly over the first 2–3 years, but less so among the BCS compared to controls. All differences between BCS and controls were statistically significant from t1 through t6.

BCS and controls in employment at diagnosis
Holding all other variables constant, BCS in employment at diagnosis with Stage I, II, III and other unknown cancer stage had a 2.9 (95% CI: −0.044; −0.013), 3.3 (95% CI: −0.048; −0.018), 3.7 (95% CI: −0.081; 0.007), 6.8 (95% CI: −0.124; −0.012) percentage point lower probability of being employed compared to matched controls at 1-year post-diagnosis. BCS consistently had lower probability for employment during follow-up compared to controls, regardless of cancer stage at baseline; and there were generally no differences in the probability of being employed between cancer stages (Table 2).
Increased age was associated with increased probability of being employed among BSC and controls at t1 and t3. At t6 age reduced the probability for employment, implying that the probability of employment decrease as individuals grow older.
Among BCS employed at diagnosis and controls, individuals with university education were consistently more likely to be in employment compared to those with basic education. In addition, couples with children were more likely to be in employment compared to single individuals without children. Industry and sector at the time of diagnosis had no impact on employment 1-year post-diagnosis. Over subsequent years, the probability of being in employment was higher among people working in the public sector (central and local government) compared to the private sector; and among those in the education sector compared to the human health and social work sector. Employees in transportation, administrative and support service activities, public administration and defense and “other industries” were more likely to be employed at t2 and t3, compared to the human health and social work sector.

BCS and controls not employed at diagnosis
Holding all other variables constant, BCS not employed at diagnosis with Stage I, II, III and other unknown cancer stage had a 5.6 (95% CI: −0.089; −0.022), 6.5 (95% CI: −0.096; −0.034), 10.3 (95% CI: −0.159; −0.046, 9.8 (95% CI −0.159; −0.038) percentage point lower probability of being employed compared to matched controls at 1-year post-diagnosis (Table 3). BCS not employed at diagnosis had lower probability for employment during follow-up, and the pattern was most persistent for Stage I patients. Five and six years after diagnosis Stage II BCS were not significantly different from matched controls. Stage III BCS not employed at diagnosis had a lower probability for employment at 5 and 6 years after diagnosis. There were no differences between BSC with different cancer stages in the probability of being employed during follow (t1–t6; Table 3).
Among BCS not employed at diagnosis and controls: increased age was associated with reduced probability of being employed throughout follow-up. Additionally, individuals with university and secondary education were consistently more likely to be in employment compared to those with basic education. Couples with children and single individuals with children were also more likely to be in employment compared to single individuals without children

Discussion
The aim of this study was to examine the prevalence and probability of employment, and to investigate the factors affecting employment up to 6 years after diagnosis among BCS of working age (30–55 years) in Norway. The results of this nationwide study have shown sustained lower employment in BCS employed and not employed at baseline, compared to controls over a 6-year period after diagnosis. Furthermore, our study suggests that cancer stage (among BCS compared to controls) and sociodemographic factors (among BCS and controls) such as caring for children and being educated are significant determinants of employment.
Reduced ability to work has been shown to be higher among BCS than among non-cancer groups and can be a consequence of treatment-related late effects and the psychological burden of cancer diagnosis.41,42 A reduction in the share of employed BCS found in our study confirms previous research.15,41,42 We found that between 72 and 74 percent of BCS were employed 1–6 years after diagnosis. This is in line with a meta-analysis reporting a 28% higher non-employment rate among BCS compared to a control population.
43
In comparison, a prospective cohort study of Danish BCS reported 81% employment 2 years after diagnosis; however, they included older cancer survivors who received early retirement- or anticipatory pension.
44
Future studies should also consider whether patients received social benefits or pensions during the study period given that a possible mechanism behind high unemployment rate among cancer survivors is a higher disability rate.
43

Numerous barriers and facilitators that could have a negative impact negatively on employment following BC diagnosis have been identified. They include factors such as age, lack of social support, comorbidities, treatment modalities, education and income.
45
In our study, severe cancer stage was not associated with lower likelihood of employment. This finding is in contrast with previous studies showing that women with advanced stages have lower employment rates after BC diagnosis compared to those with other stages of cancer.11,46,47 Normally, BCS with advanced stage diagnosis may suffer from side effects that may require continued treatment and can limit employment or lead to disability. Oncological surgery has been shown to be associated with long-term adverse effects, and one study found that BCS who had undergone a mastectomy were more likely to have stopped working Seven months post- diagnosis.
46
A French study showed that ongoing chemotherapy-induced peripheral neuropathy was associated with lower employment 5 years after diagnosis.
48
However, the present study only included BCS who were alive 6 years after diagnosis, thus the BCS included may have been in better health compared to patients in previous studies.
Consistent with prior studies,49,50 the present study showed that being educated increased the likelihood of employment. Education level has been shown to be stratified by the type of job, whereby low educated patients were more likely to work as manual laborers and thus engage in heavy lifting jobs resulting in experiences of fatigue.
51
This study did not examine the conditional effects involving level of education and employment sector; but we found significant group differences in the effect of education on employment net of employment sector.
In the current study BCS and controls with children showed a greater likelihood of employment compared to BCS and controls without children. This is in line with numerous studies that have found that women with children return to employment more often compared to women without children, which could be due to financial necessity.52,53 Our study did not differentiate between for example, having young/dependent versus adult/independent children. This is an area that should be examined in future studies; for example, financial and other forms of support from adult children might reduce the financial strain experienced by BCS and the attendant pressure to prematurely return to work.
BCS in this study could be out of employment for different reasons, with some of them potentially being out of the labor force due to cancer. However, it should be noted that the current study included a relatively healthy group of BCS by excluding those who had more than one cancer diagnosis. In addition, BCS who died during follow-up were not included in the analyses. It is likely that breast cancer patients with the most severe illness died during the study period. Thus, our study addressed BCS with a long life-expectancy following diagnosis. Reports from Norway have shown that in the general population, employment in Norway peaks around the age of 40–50 years Norway.
36
Since BCS up to 55 years were included, only a few may have chosen early retirement due to reassessment of their priorities following a cancer disease. Therefore, our participants were also rather young compared to other comparable studies, with a mean age of 47.6 years at baseline. However, it might be that the physical and psychological impact of the disease and treatment affected the patients’ employment seeking behavior. At later ages, the employment rate for both genders in the general population start decreasing and the rate of sickness absences and disability pensions start increasing.
36
These factors may also explain the decrease in employment over time among the controls in the present study.
It should also be noted that differences in results between studies may be explained by differences in the definition of employment. They may also be due to studies being conducted in countries with a diversity of social security systems. By international comparison, the employment rate is very high in Norway, both for men and women.
54
One important factor behind the well-functioning Norwegian labor market is the so-called “tripartism.” This involves cooperation amongst unions, employers and the government for a more inclusive working life through the inclusive workplace agreement, which is aimed at helping achieve high employment and mobilizing the workforce by preventing and reducing sick leave and withdrawal from working life.54,55 It should also be noted that employment rates and education levels are high for Norwegian women compared to those in other countries; and the generosity of the welfare state in Norway, which makes it possible to combine family life and work, has contributed to the high participation rate of women.54,56 In countries with lower female employment rates and less supportive welfare systems, the differences in employment between BCS and controls may be more substantial compared to our findings. However, our findings corroborate overall findings from various welfare systems and labor markets,
22
but it should be noted that results may vary substantially according to welfare regime.
13

Strength and limitations
A strength of the study was the use of registry-based individual level measures of employment, with high validity and completeness, allowing the selection of data according to inclusion criteria. The longitudinal design and inclusion of a control group of age-matched women was another strength of the study. Our use of a control group allowed us to account for sociodemographic factors (age, education, family type, employment industry) that may have influenced employment participation during the study period.
Findings from our study should be interpreted in light of its limitations. Although we matched breast cancer survivors (BCS) and controls on key sociodemographic variables, including educational level and employment status, unobserved differences may still exist—particularly with regard to lifestyle-related factors such as smoking, alcohol consumption, physical activity, obesity, and hormone use. These factors are known to influence both cancer risk and employment outcomes, but were not available in our dataset. While education is a strong proxy for many health-related behaviors,
57
residual confounding cannot be ruled out. Nonetheless, the use of a matched control group without a cancer diagnosis during the study period strengthens the internal validity of our comparisons, as it allows us to isolate the potential impact of cancer and its treatment on employment outcomes.
The present study did not account for working hours, but included all employment, both part-time and full-time. BCS may work reduced hours to a larger extent than controls, but this was not addressed here. Future studies may want to investigate whether there are differences in working hours between BCS and controls. Previous studies have shown that BCS may experience both reduced
18
and increased earnings,
16
suggesting that working hours may both increase and decrease following a breast cancer diagnosis. These variations may reflect individual coping strategies, workplace accommodations, or economic necessity, and could have important implications for understanding the full impact of breast cancer on labor market participation.
Additionally, our study did not include information regarding employment before diagnosis. A study from Denmark showed that the duration of unemployment before diagnosis was the most important determinant of unemployment following cancer diagnosis.
44
This result is in accordance with studies from the economic field, suggesting that unemployment is an important risk factor for future unemployment.
58
It might be that the worsening of health may be accelerated by the stress and psychological and socio-economic disadvantages of unemployment, and that marginalization in the labor market may be accelerated by health-related factors. Thus, unemployment before diagnosis should be considered in future studies of BCS.

Conclusion
The current study showed reduced employment in BCS compared to controls from the general population over a 6-year period after diagnosis. The findings indicated that disease burden and socioeconomic status is important for employment in BCS of working age. These factors in combination with other socioeconomic factors allowed for an identification of BCS with an increased risk of reduced long-term employment.

Supplemental Material

sj-docx-1-phj-10.1177_22799036251410249 – Supplemental material for Breast cancer and long-term employment: A retrospective cohort study from Norway

Supplemental material, sj-docx-1-phj-10.1177_22799036251410249 for Breast cancer and long-term employment: A retrospective cohort study from Norway by Roy A. Nielsen, Eliva Atieno Ambugo, Steffen Torp, Torgrim Tandstad, Guro Birgitte Stene, Line Oldervoll, Alain Paraponaris and Harald K. Engan in Journal of Public Health Research

sj-docx-2-phj-10.1177_22799036251410249 – Supplemental material for Breast cancer and long-term employment: A retrospective cohort study from Norway

Supplemental material, sj-docx-2-phj-10.1177_22799036251410249 for Breast cancer and long-term employment: A retrospective cohort study from Norway by Roy A. Nielsen, Eliva Atieno Ambugo, Steffen Torp, Torgrim Tandstad, Guro Birgitte Stene, Line Oldervoll, Alain Paraponaris and Harald K. Engan in Journal of Public Health Research

sj-docx-3-phj-10.1177_22799036251410249 – Supplemental material for Breast cancer and long-term employment: A retrospective cohort study from Norway

Supplemental material, sj-docx-3-phj-10.1177_22799036251410249 for Breast cancer and long-term employment: A retrospective cohort study from Norway by Roy A. Nielsen, Eliva Atieno Ambugo, Steffen Torp, Torgrim Tandstad, Guro Birgitte Stene, Line Oldervoll, Alain Paraponaris and Harald K. Engan in Journal of Public Health Research