Work Loss Duration and Predictors Following Rectal Cancer Treatment among Patients with and without Prediagnostic Work Loss

Rectal cancer contributes to approximately 40,000 new cancer cases annually in the United States (1) and 120,000 in the European Union (2). In recent years, due to utilization of total mesorectal excision combined with neoadjuvant radiotherapy, the overall survival of rectal cancer patients has improved to about 60% at 5 years and 40% at 10 years following diagnosis (3).

Rectal cancer survivors may experience long-term complications of treatment and a reduced self-reported health compared with the general population (4–9). Among working-age patients, studies have also shown decreased work ability (10–18). Higher risk of work loss among colorectal cancer survivors (and cancer survivors in general) has been associated with demographic factors such as lower education (10, 15, 17), unemployment at diagnosis (15), and previous sick leave (SL) episodes (10, 15, 17). Although some clinical predictors of work loss have also been identified, including advanced stage (15, 17), it is unclear to what extent these associations could be explained by cancer relapse during follow-up. Also, most previous studies have been based on small numbers of rectal cancer patients (10, 11, 15) or self-reported data (18) which could introduce bias.

We have shown in a previous register-based study that a 2-fold increased risk of disability pension (DP) remained among relapse-free rectal cancer patients compared with the population (14). However, if rates of SL are also increased among relapse-free patients, and how they vary by clinical and demographic factors, is unclear.

The aim of this study was to evaluate the duration and predictors of work loss (SL and DP) within 5 years of rectal cancer diagnosis, with a focus on relapse-free patients, to identify patient groups that may benefit from early intervention and rehabilitation. Given the importance of work loss history as a predictor of continued work loss in other settings (10, 11, 14, 15, 18), the patients were stratified by the presence or absence of work loss 1 to 2 years before diagnosis.

In this cohort study, Swedish health care registers were used to compare prospectively recorded work loss days, defined as days of SL or DP, among subgroups of working-age rectal cancer patients, with a matched general population cohort as benchmark. Compensated net days with partial or complete SL or DP were provided by the Swedish Social Insurance Agency (Försäkringskassan; ref. 19). The Swedish health care system is tax-funded and offers universal access.

Patients diagnosed with rectal cancer in stage I–III 18–61 years of age 1996–2009, and treated curatively (n = 3438) were identified in the Swedish Colorectal Cancer Register. The register contains comprehensive clinical information and covers >98% of all national invasive rectal adenocarcinomas (20, 21).

Clinical data used included stage at diagnosis (I–III), operation type [anterior resection (AR), abdominoperineal resection (APR), Hartmann procedure], treatment strategy [abdominal surgery only, neoadjuvant (chemo)radiotherapy or adjuvant chemotherapy], surgical complications within 30 days (yes/no), and recurrence (yes/no) and dates of local recurrence and/or distant metastasis during follow-up. Surgical complications were defined as any kind of the following conditions occurring within 30 days of primary rectal cancer surgery—wound infection, intra-abdominal infection, dehiscence (rupture), bleeding, anastomosis leakage, or ostomy complication.

Prediagnostic work loss was defined as receiving compensation for SL and/or DP during 1 to 2 years before rectal cancer diagnosis (yes/no). Unemployment (yes/no) within 1 year prior to diagnosis (or match date for comparators), and highest attained level of education (>9/10–12/12+ years) was obtained from the Longitudinal Integrated Database for Health Insurance and Labour Market Studies (LISA) maintained by Statistics Sweden (22).

To quantify excess levels of work loss for the patient cohort, five population comparators per patient were randomly selected (with replacement) from a subset of the Total Population Register that encompasses all Swedish residents and is held by Statistics Sweden (23). They were alive and rectal cancer-free at the diagnosis date of the patient, and were matched individually to the patient by sex, age, geographic region of residence, calendar period, and education level.

Net days of SL and DP were used as the primary outcome to evaluate the work loss burden. Net days were calculated by multiplying the degree of compensation with the number of compensated days. Dates and extent of SL and DP are recorded in the Micro Data for Analysis of the Social Insurance database (MiDAS) maintained by the Swedish Social Insurance Agency since 1994 (24). Individuals who are registered as unemployed are also entitled to sickness absence compensations.

SL episodes longer than 14 days were recorded from the first day. SL episodes shorter than 14 days are paid by the employer (exception was January 1997–March 1998 when the upper bound was 28 days) and are therefore not recorded in MiDAS. Consequently, only SL episodes longer than 2 weeks contributed to work loss days, although also shorter episodes (regardless of length) have been captured if they occurred within 5 days of a previous episode. DP is registered and provided by the social insurance system from the first day. Compensation can be partial (25%–75%) or full (100%).

The study was approved by the regional ethics board in Stockholm, Sweden.

The median, 25th and 75th percentiles of work loss days were calculated for patient subgroups. Also, the mean annual work loss days during follow-up were calculated and plotted for patients and matched comparators. To compare the incidence rates of work loss between patient subgroups, a negative binomial regression model was used to estimate incidence rate ratio (IRR) with 95% confidence intervals (CI). By incorporating an offset term consisting of each individual's person-time at risk for work loss, the negative binomial model was adjusted for varying lengths of follow-up.

All regression models were adjusted for sex, age at diagnosis, region, diagnosis year, education level, and unemployment. Potential effect modification was assessed with interaction terms and formally tested by likelihood ratio tests (significance level: 0.05).

Missing values on were modeled as a separate category in the regression analyses. Patients were followed from the rectal cancer diagnosis date to emigration, retirement (65th birthday), death, or December 31st 2013 (whichever occurred first). In analyses of relapse-free time (most analyses), we additionally censored at first relapse date (local recurrence or distant metastasis). Eighty-six patients (2.5%) had missing relapse dates and were then excluded. Analyses were conducted separately for the 1st year of diagnosis and treatment, and the 2nd–5th post-treatment years to distinguish work loss attributable to acute and late treatment effects.

All statistical analyses were performed on SAS (version 9.3, SAS institute Inc.) and STATA (StataCorp, release 12.0.).

The study population consisted of 3,438 rectal cancer patients with a median age of 56 years at diagnosis, of whom 42% were women. Forty-one percent of the patients were diagnosed with stage III disease, 55% were treated with neoadjuvant (chemo)radiotherapy, and 63% were operated with AR. Within 30 days of surgery, 27% experienced surgical complications. Relapse occurred in 24% of the patients during the 5-year follow-up and local recurrences occurred in 5%. Nine hundred and nine patients (26%) had prediagnostic work loss, whereas the majority (n = 2,529; 74%) had none. Among patients with prediagnostic work loss, 500 (55%) had experienced SL only and 409 (45%) patients had DP. Patients with prediagnostic work loss were more often women (50% vs. 40%; P < 0.001), had lower education level (39% vs. 26% educated < 9 years; P < 0.001), and were slightly older when diagnosed (62% vs. 49% diagnosed at 51–61 years of age; P < 0.001) compared with patients without prediagnostic work loss. Overall, 12% of the patients were unemployed in the year prior to diagnosis (Table 1).

Patients were followed for a median of 4 (maximum 5) years. For all patients, the mean days of work loss peaked during the 1st year after diagnosis, and then decreased from the 2nd year (Fig. 1). Patients with prediagnostic work loss had consistently higher mean levels of work loss days than patients without such history. The mean number among the patients was slightly lower when restricted to relapse-free time than among all. However, relapse-free patients without prediagnostic work loss still had significantly more work loss days than comparators up to the 5th year of follow-up (P < 0.001). Among these patients, up to the 5th year, the majority (74%) did not have any recorded work loss the 5th year, whereas 6% were absent all year (Fig. 2). In contrast, for the same year among patients with prediagnostic work loss, 30% had no recorded work loss and 34% were completely absent.

Among patients without prediagnostic work loss, the median duration of work loss days during this period was 147 days (55 and 281 for the 25th and 75th percentile, respectively; Table 2). Clinically, advanced stage (compared with stage I), neoadjuvant (chemo)radiotherapy and adjuvant chemotherapy (compared with surgery alone) all rendered an elevated rate. Being operated with APR also increased rate of work loss in comparison with AR. Patients' work loss rate was further elevated following surgical complications within 30 days of primary surgery. For patients with prediagnostic work loss, however, stage and treatment did not affect work loss risk. Prediagnostic work loss status significantly modified the association between work loss rate and stage (P = 0.000), treatment (P = 0.000).

Among patients without prediagnostic work loss, the median work loss days were 7 (0 and 172 for the 25th and 75th percentile, respectively) during the 2nd–5th years. Advanced stage, combined treatments, and having surgical complications were associated with increased rates of sustained work loss up to the 5th year. For the 5th year, the patients operated with APR had significantly higher rate of work loss compared with those who were operated with AR, and to the matched population (P < 0.001; Fig. 3; Supplementary Table S1). However, among patients with prediagnostic work loss, there were no associations with clinical characteristics and work loss (Table 2).

During the 1st year, women were at an increased rate of work loss compared with men (Supplementary Table S2). Among patients without prediagnostic work loss, both unemployment and lower education were associated with higher work loss risk.

Among patients with prediagnostic SL only, 25% were absent all year the 5th year whereas this was true for 45% among those with prediagnostic DP (Supplementary Fig. S1). During the entire follow-up, in patient subgroups with prediagnostic SL only, or with prediagnostic DP, clinical determinants were not associated with work loss (Supplementary Table S3).

In this Swedish population-based cohort study of work loss predictors following rectal cancer diagnosis and treatment, we observed large differences by prediagnostic work loss in the number of work loss days up to 5 years after diagnosis. Patients without prediagnostic work loss were absent median 5 months from work during the 1st year. Up to 5 years, advanced stage, APR, combined treatment [neoadjuvant (chemo)radiotherapy and/or adjuvant chemotherapy] and surgical complications were associated with increased work loss. These increased rates among relapse-free patients indicate long-standing treatment side effects with consequences for work ability. One-fourth of the patients had prediagnostic work loss and they continued to be absent to a high extent with little variation by clinical factors. Our results emphasize that work loss prediction following rectal cancer diagnosis should be based first on work loss history and second on disease- and treatment-related factors.

Surgery, particularly extensive surgery, as the essential treatment for rectal cancer, could render high morbidity (9). APR often causes more complications compared with AR due to ensuing permanent stomas and more complications in perineal wounds (25). Neoadjuvant (chemo)radiotherapy, which controls local recurrence and improves survival for locally advanced rectal cancer (26), may additionally increase late morbidity (27–29). Clinical covariates, including advanced stage based on the TNM system, and having complications as registered through hospital admission codes have previously been reported as work loss predictors among colorectal cancer patients (15, 17). However, previously observed associations could have been driven by relapse during follow-up, as advanced stage is associated with an increased risk of relapse that in turn is associated with work disability. For the first time, we have shown that rectal cancer stage, surgical technique (APR, in particular) and combined treatment modalities affect work loss rates up to 5 years even among relapse-free patients. These findings suggest long-standing treatment side effects with consequences for the ability to return to work. For patients that remain in complete remission (especially young patients), characterization and management of side effects (due to treatment or not), with the aim to restore function and work ability is important. So far, according to our experience, these issues have not been addressed enough in clinical cancer survivorship care.

Independent of the underlying disease, work loss can be seen as a proxy for comorbidity or poorer health condition, and is therefore expected to be an important predictor for future SL and/or DP (30, 31). In a previous Swedish study of cancer patients overall, SL before cancer diagnosis conferred an approximately doubled risk of having SL 5 years after diagnosis (11). In Danish colorectal cancer patients, a significantly lower probability of returning to work was found among those who had previous SL (17).We previously showed that relapse-free rectal cancer patients with prior SL had an 8-fold risk of getting DP compared with general comparators without SL (14). For individuals with a history of work loss, an additional burden of a rectal cancer diagnosis plausibly lowers the likelihood of returning to work, and we show in this study that clinical characteristics and given treatment is subordinate in this situation. Still, patients with prediagnostic work loss constitute a heterogeneous group and the potential benefit of rehabilitation programs should be evaluated individually.

Low education level (<9 years) was a risk factor for work loss both among patients with and without prediagnostic work loss, in line with previous studies (10, 15, 17). People with low education level are more likely to have physically demanding jobs and less influence over work situations (32), which may contribute to difficulties to cope with previous tasks and to adjust back to the work environment following cancer treatment.

People who were voluntarily away from the labor market, including those who were not registered as unemployed (e.g., stay-home parents), were not entitled to sickness absence compensation in case of disease. However, this group is estimated to be small (approximately 3% of the population 20–64 years of age), and so any resulting misclassification of needs should also be small. Sweden has a generous family policy which allows compensation during registered parental leave (normally about 1 year; ref. 33).

In this study, we used prospectively recorded work loss data in population-based registers, minimizing the risk for selection bias and misclassification (34). Other strengths were the detailed clinical data with high validity and coverage (35, 36) and use of a comparison cohort. However, whereas register records of neoadjuvant (chemo)radiotherapy reflects given treatment, adjuvant chemotherapy reflects planned care only and might deviate somewhat from given treatment. As relapse information was missing or erroneous among 10% to 20% patients in the clinical register (37, 38), the impact of relapse on work loss could have been slightly underestimated. Furthermore, we have only captured SL episodes longer than 14 days, leading to some degree of underestimation of the total work loss days. However, the potential underestimation is not likely to be related to clinical characteristics and treatment.

In conclusion, among relapse-free rectal cancer survivors without prediagnostic work loss, clinicians should be aware of a potential need of early vocational rehabilitation or adjustments, specifically among patients with advanced stage, operated with APR, treated with neoadjuvant (chemo)radiotherapy, or with surgical complications. Prediagnostic work loss should be used to identify rectal cancer patients that are likely to remain absent from work and where rehabilitation initiatives should be individualized.

M. Neovius reports receiving commercial research support from Astra Zeneca and commercial research grant from Pfizer, and is a consultant/advisory board member for Pfizer and Itrim. No potential conflicts of interest were disclosed by the other authors.

Conception and design: L. Chen, I. Glimelius, M. Neovius, S. Ekberg, A. Martling, S. Eloranta, K.E. Smedby

Development of methodology: L. Chen, M. Neovius, A. Martling, S. Eloranta, K.E. Smedby

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A. Martling, K.E. Smedby

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): L. Chen, I. Glimelius, M. Neovius, S. Ekberg, A. Martling, S. Eloranta, K.E. Smedby

Writing, review, and/or revision of the manuscript: L. Chen, I. Glimelius, M. Neovius, S. Ekberg, A. Martling, S. Eloranta, K.E. Smedby

Study supervision: I. Glimelius, M. Neovius, K.E. Smedby

This study was funded by the Strategic Research Program in Epidemiology at Karolinska Institutet.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.