Adverse outcomes of chronic widespread pain and common mental disorders in individuals with sickness absence – a prospective study of Swedish twins

This prospective twin cohort study was based on data from the Swedish Twin project of Disability pension and Sickness absence (STODS). Data from the following sources were included, which were merged using the unique personal identification number of participants:

Data from two surveys conducted by the STR were used: Screening Across the Lifespan Twin Study (SALT) [21], performed 1998–2003, and the Study of Twin Adults - Genes and Environment (STAGE) [22], conducted 2005–2006. SALT included interview data collected from all twins in the STR born before 1958 while STAGE was an extensive web-based questionnaire sent to all twins in the STR born between 1959 and 1985. The date that each participant answered the questionnaire/interview was set as the baseline.

Individuals who had at least one sick-leave spell due to any cause during the 2 years preceding the baseline were included in the study (n = 9827). Individuals who had a sick-leave spell longer than 1 year, had emigrated, were on DP or old-age pension at baseline were excluded (n = 1943) (Fig. 1). The final cohort contained 7884 twins (aged between 20 and 66 years, 65% women), including 1178 complete twin pairs, 466 monozygotic (MZ) pairs, 400 same-sex dizygotic (DZ), and 298 opposite sex DZ. In the sample, 162 pairs were discordant for CWP and CMDs.

Information on CWP and CMDs was collected in the SALT and STAGE surveys. The questions were based on a screening algorithm for CWP according to the classification criteria for fibromyalgia proposed by the American College of Rheumatology [3]. However, the difference was that our definition of CWP was based on self-reports without clinical evaluations by physicians or inspection of medical records, and that a count of tender points was not included. In the questionnaire and interview data, individuals who responded “yes” on all four of the following questions were defined as having CWP [15]: 1) “Have you suffered from general pain during the last three months?” 2) “Did you have continuous pain during all three months?” 3) “Do you suffer from pain in both the upper and lower body?” and 4) “Do you suffer from pain in both the right and left sides?”. CMDs was measured by the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV) [23] major depression and anxiety. For a more detailed description of criteria for major depression and anxiety, see Mather et al. [24]. A categorical variable which combined CWP and CMDs was created and used as exposure variable: 0 = No CWP or CMDs (reference group), 1 = No CWP but having CMDs, 2 = Having CWP but no CMDs, 3 = Having both CWP and CMDs.

The participants were followed up for all-cause DP, long-term unemployment (> 90 days) and all-cause mortality from baseline until the 31st of December 2012.

Information on sex (women/men), age (continuous variable), and zygosity (MZ, DZ) were obtained from STR. Data on education (elementary, secondary or higher education), having children under 18 living at home (yes/no), type of living area (urban or semi-urban/rural) at the end of the year of baseline was obtained from LISA. Self-rated health (excellent, good, moderate, fairly poor/poor), pain (pain in neck, shoulder or low back) (yes/no), migraine (yes/no), headache (yes/no), cancer (yes/no) at baseline was collected from the survey data. The cohort effect was binary i.e. belonging to either SALT or STAGE.

In general, all individuals aged 16 or older with an income from work or unemployment benefits, who have a reduced work capacity due to disease or injury, can be granted sickness benefits [25]. For employees this is paid by the employer during the first 14 days, and by the Social Insurance Agency thereafter. Hence, national register data covers SA lasting more than 14 days. There is one qualifying day (more for self-employed) without benefits. All residents with a permanently impaired work capacity due to disease or injury can be granted disability pension. SA benefits covers up to 80% of lost income, DP to approximately 65%.

Cox proportional hazards regression was performed to estimate the associations between CWP and CMDs and DP, long-term unemployment and mortality separately in the whole sample. The proportional hazards assumption was explored via graphs and was met in all the models (See supplementary materials). The follow-up started after the baseline until the end of follow-up 31st of December 2012. Follow-up (time in days) was censored for emigration and the respective outcomes. In the analysis of unemployment as outcome, DP and old-age pension were censored while in the analysis of DP, old-age pension was censored. Besides the crude model, the multivariate model was adjusted for sex, age, cohort effect, education, having children under 18 living at home, type of living area, pain in neck, shoulder or low back, migraine, headache, and cancer. As pain in neck, shoulder or low back, migraine and cancer were highly correlated with CWP, sensitivity analyses were conducted where individuals with those disorders were excluded from the analyses.

Co-twin analyses were performed separately for each outcome on exposure discordant twin pairs (i.e. one twin in a pair had CWP and/or CMDs and the other twin did not) using conditional Cox proportional hazards models. As twins are matched on genetics (100% for MZ twins and on average 50% for DZ twins) and shared environment (100% for both MZ and DZ) while growing up (if raised together), the co-twin analysis adjusts for these factors. If the association found in the age and sex adjusted analyses of the whole sample disappears or becomes weaker in the analyses of discordant twin pairs there is a suggestion that familial factors are influencing the association. If an association is still observed after controlling for familial factors this would instead suggest a direct link between the exposure and the outcome [26]. The analyses were based on discordant twin pairs (n = 162) for the exposure where one twin in a pair had CWP and/or CMDs and the twin sibling had neither CWP nor CMDs. MZ and DZ twin pairs were pooled for the analyses as there was not enough power to stratify by zygosity.

In the whole sample with a history of SA, having CWP and/or CMDs were associated with a higher risk for DP compared to those without CWP and CMDs in the crude model. In Model 1, we adjusted for socio-demographic factors including sex, age, education, children living at home, type of living area and cohort effect and the HRs for DP showed somewhat of changes (range of HRs: 1.54–2.84), indicating some influence from socio-demographic factors on the associations. The increased risk of DP remained even after additional control for health-related factors (i.e. self-rated health, pain, headache, migraine, and cancer), particularly for those having only CWP (HR: 2.43, 95% CI: 2.03–2.93) and both CWP and CMDs (HR: 2.65, 95% CI: 1.89–3.71). In the analyses of discordant twin pairs, the HR was lower and became non-significant for those who only had CMDs (HR: 0.96, 95% CI: 0.52–1.81) compared to the estimate of the whole sample. This result suggests that familial factors influence the association between CMDs and all-cause DP (Table 2).

An association between CMDs and subsequent unemployment > 90 days was found in the crude model as well as (HR: 1.38, 95% CI: 1.19–1.60) after adjustment for all the covariates (HR: 1.36, 95% CI: 1.17–1.58). The HRs changed slightly after controlling for the covariates, meaning a minor influence from those factors on the associations between CMDs and unemployment > 90 days. The HR was attenuated in the co-twin model (HR: 1.14, 95% CI: 0.67–1.92), which suggest that the association between CMDs and unemployment was influenced by familial factors (Table 3).

Table 4 shows HRs for all-cause mortality in the whole sample with a history of SA and among twins discordant for the exposures. Having either CWP or CMDs was associated with a higher risk for all-cause mortality after controlling for all the covariates (Having only CWP: HR: 1.95, 95% CI: 1.39–2.74; Having only CMDs: HR: 1.60, 95% CI: 1.19–2.16). The HRs increased more after controlling for socio-demographic factors and cohort effects compared to the model controlled with additional health-related factors. This indicates that socio-demographic factors and cohort effects had more impact on the associations between CWP and/or CMDs and mortality than health-related factors. In the analyses of discordant twins, the HR increased, especially for those only with CMDs. This may indicate a direct (i.e. not affected by familial confounding) link between CMDs and all-cause mortality.

Sensitivity analyses were conducted for DP, long-term unemployment and mortality among all twins and twins discordant for the exposure after excluding neck-, back-, and shoulder pain, migraine, and cancer at baseline (data not shown). The results of the sensitivity analyses were of the same magnitude and direction as the main analyses which included pain sites, migraine, and cancer.

Strengths of this study include the large population-based sample of twins and the use of Swedish nationwide register data of high quality [41], which reduced the risk of loss to follow-up and recall bias. Moreover, we could include discordant twin pairs in the analyses, which enable to control for the influence from familial factors. However, some analyses were still underpowered with broad CIs. In this study, we were also able to include several relevant covariates in relation to the exposure and outcomes. Still, there might be other factors than those included here that are associated with the outcomes, such as unhealthy lifestyle (i.e. smoking behaviour, alcohol consumption, lack of physical activity) etc.

The limitations of the study include that data on sick-leave spells < 14 days were not available for employed individuals. Thus, the findings from the study may not be generalizable to individuals with short-term SA spells (< 14 days). In addition, since the survey was extensive and included many questions, there are some internal missing data. This might lead to underestimation of the results. Also, it was not possible to perform analyses with only same-sex discordant twins due to low numbers, or analyses stratified on zygosity, making further investigation of familial factors impossible. Moreover, the co-twin models lacked power and findings from these analyses needs to be interpreted with caution. Moreover, because our measures of CWP and CMDs relied on self-reports and the diagnostic criteria for CWP has changed over time [42], it is likely to have some misclassification. However, misclassification is possible to be non-differential in MZ and DZ twins and it is therefore not likely that our estimation of the influence of familial factors is biased. Our measure of CWP was not based on clinical examination and thus inferences should be with caution when our findings are compared with studies based on clinical examination. On the other hand, questionnaire- and interview-based screening have shown relatively high positive predictive value and test-retest reliability [43] and these are the only realistic means to gather data over ten of thousands individuals in population-based samples. The questionnaire and interview data for measuring CWP and CMDs were collected during 1998–2006. Hence, a new collection of data in this field is recommended for facilitating future studies with updated findings. In addition, the Swedish government launched stricter rules for granting DP since 2008, which might result in that individuals with DP had a higher medical severity than before these regulations were implemented [11]. Our measures of CWP and CMDs were self-reported, hence they may or may not require any medical attention which could lead to underestimated risk of DP as the severity and chronicity of CWP and CMDs is unknown. Still, the findings in this study might be generalised to working-aged individuals living in countries with comparable social insurance systems.