Different outcome measures
The results of the present randomized comparative trial depended partly on the outcome measure chosen. At the one-year follow-up, the number of weeks on sick leave was statistically lower in the brief intervention group than in the multidisciplinary group which indicated that this intervention was the more effective. The other two outcome measures showed the same tendency, but the differences were not statistically significantly different. Even if the conclusion thus depends on the outcome measure chosen it remains clear that all three outcome measures pointed in the same direction.
At the two-year follow-up, the relative effects of the two interventions were similar to those obtained at the one-year follow-up. The “survival analyses” showed that about 5% more patients had achieved a four week RTW period during the second year, but this was the case for both intervention groups. The number of sick leave weeks was much lower in the second year than in the first year; again, this was the case for both intervention groups. The percentages of former patients having resumed work were not statistically different between the intervention groups at any of the follow-up points. However, it should be noted that the percentage at work in the 104th week was slightly lower (approximately 60%) than the percentage at work in the 52nd week (approximately 64%). The lower prevalence of sick leave in the second year thus testifies to the sustainability of RTW, but the slightly lower fraction of employees with regular work in the second year may be ascribed to the fact that more employees were engaged in modified work or were no longer part of the labour market due to early retirement or for other reasons. The percentage of patients with RTW at the two-year follow-up was higher than the percentage at the one-year follow-up because patients with RTW could not enter the study again if they had new sick leave spells. The number of RTW events will therefore either remain constant if no new patients return to work or increase if at least one new RTW event is registered. It is therefore crucial to any comparison of RTW rates between studies that the outcome measures are exactly the same.
Our “survival curves” showed that 74% returned to work during the first year, which may be compared with the results of Dutch studies that also used a four-week period without sick leave in their definition of RTW. Anema et al. reported that 91% of their intervention group accomplished RTW during the one-year follow-up period, and Heymans et al. reported that approximately 80% of their intervention groups achieved RTW during their six-month follow-up period. The RTW rates in our study were also lower than those in the control groups of the two Dutch studies. This indicates that the Dutch context may facilitate RTW better than the Danish context. A more elaborate dismissal protection legislation in Holland than in Denmark may lay at the root of this difference. However, “usual care” in Holland is also different from that in Denmark, as Holland operates a system where an occupational practitioner deals with sick leave problems, whereas in Denmark, the general practitioner is the primary health professional involved. Furthermore, one third of our patients had radiculopathy and 10% in each intervention group had surgery, most often due to radiculopathy, whereas the Dutch study only included patients with non-specific LBP. A comparison with Norwegian studies is also difficult as outcome measures were defined differently[10, 15].
It is possible that our interventions were less effective than usual care in Denmark. However, we consider this unlikely as an early intervention including a thorough clinical examination and reassuring advice is considered beneficial and has proven effective in other countries [6;10;12]. The only Danish study that could be used for comparison reported that at the 12-month follow-up, 78% were at work and 22% were on sick leave in the intervention group, and 62% were at work and 38% on sick leave in the primary care control group. These figures may be compared with our status at the one-year follow-up, where 25% and 28% were on sick leave in our two intervention groups. Our percentages at work in the 52nd week, i.e. 66% and 61%, were lower than the percentage of 78% reported in the previous Danish study. However the definition of RTW may have differed between our study and the previous Danish study; moreover, we separately measured other possible outcomes, like for instance “modified job or training”, which we do not know if was the case in the previous Danish study.
A major weakness of conclusions based on a comparison of the brief and multidisciplinary intervention groups was the existence of subgroups in which the interventions seemed to affect return to work rates in opposite directions. The stratified analyses made clear that the brief intervention was statistically significantly more effective than the multidisciplinary intervention in “Subgroup 1” in which the patients reported to have influence on work planning and were not at risk of being dismissed. We previously reported that effect modification was present at the one-year follow-up; that is, the multidisciplinary intervention was more effective than the brief intervention in the other subgroup without job control or where the patients felt at risk of losing their job. The differences between the intervention groups were not statistically significantly different if analysis was confined to “Subgroup 2”, even if the average differences were similar to those found in the other subgroup. The reason for this lay in the difference in statistical power due to the lower number of patients in “Subgroup 2”. The higher number of patients in “Subgroup 1” was also the most important reason for the tendency towards a better effect of the brief intervention in the total sample of subjects, i.e. the brief intervention was more effective in about two-thirds of the patients, which pushed the average result in favour of the brief intervention. For the other one-third of the patients, the relative benefits of the multidisciplinary intervention seemed just as large. The correct conclusion of the study would probably be that the brief intervention worked better for about two-thirds of the patients, and the multidisciplinary intervention was more effective for the remaining one-third of the patients. This result was based on post-hoc subgroup analyses in the randomized trial, and it therefore should be verified in a randomized trial stratifying patients into appropriate subgroups before randomization.
Heymans et al. reported higher RTW rates and less sick leave days in a group receiving low-intensity back-school intervention than in a group receiving high-intensity back-school intervention or usual care. High-intensity back-schools were not superior to usual care. However, only some of the analyses showed significant differences between groups, and subgroup analyses were not performed. Others have compared low-intensity interventions with usual care[7, 10] or more intensive interventions with usual care[6, 9, 16], and positive effects have been reported for both types of interventions. In a recent review of randomized controlled trials, it was suggested that brief interventions (<12 hours spent on the intervention) were more effective than interventions where the efforts were more extensive, at least as compared with efforts lasting more than 32 hours. Our subgroup analyses and those of others[8, 18] indicated that it is very likely that both types of interventions may be effective, but that the effectiveness depends on other risk factors than those that were used to include sick-listed employees in the RCTs. The identification of such factors is important to the provision of the right kind of “treatment”. However, these factors may differ between different countries, occupational groups and so forth. The tendency for the brief interventions to be more effective than more intensive interventions may be explained in two ways. Like in our study, the subgroup who benefited more from the intensive intervention counted fewer members than the other subgroup, i.e. those who lacked job control and who were at risk of losing their jobs, which was reported by one-third. In Dutch studies, the risk of being dismissed during sick leave was probably much lower than in our study as the labour market legislations differ between Denmark and Holland. Like in Holland, dismissal protection is high in Norway, but low job control and other adverse factors that may require more intensive RTW efforts may be equally prevalent in all three countries. The other explanation lies in the duration of the intervention. The duration of a multidisciplinary intervention is longer than the duration of a brief one and may postpone RTW even if patients are, indeed, told to RTW as soon as possible during the course of the intervention.
We ascertained the same effect of intervention, whether brief or multidisciplinary, in the one-year follow-up and the two-year follow-up. In both subgroups, the outcome that measured work status at the 104th week featured the largest differences between the interventions, as the percentages of patients at work were significantly different between intervention groups in”Subgroup 1”. In”Subgroup 2”, the same tendencies were seen at the two-year and the one-year follow-up and the percentage with sick leave in the 104th week was considerably lower in the multidisciplinary intervention group than in the brief intervention group. Thus, the reported subgroup differences appeared sustainable.
Long-term sickness relapse was not common within the first two years after the intervention was initiated. More than half of the employees who accomplished RTW did not experience new episodes of long-term sick leave, neither in “Subgroup 1” nor in “Subgroup 2” irrespective of intervention. However, it should be noted that we only measured new spells with durations of more than two weeks’ absence. Shorter sick leave spells may be more common and previous reports of frequent sick leave relapse after RTW cannot be contradicted.
The duration of the period without sick leave used to define RTW was crucial when estimating the percentage of patients returning to work during the follow-up period based on “survival analyses”. When choosing a longer duration than one week (or one day for that matter), such as four weeks which has been the “tradition” in Dutch studies, one is more certain that RTW is sustainable, i.e. that the employee is capable of staying at work without new sick leave spells. However, the percentage of patients with RTW decreased gradually when the required duration of the period without sick leave was increased from one week up till 26 weeks as indicated in Table1. No threshold was observed that could have served to define the temporal boundaries of the concept of sustainability. More importantly, the relative effects of the two interventions on the RTW rates did not change when the duration of the period with sick leave was changed. Thus, to compare effects of RCTs, the duration of the period required for an individual to have fully returned to work to define RTW is probably of minor importance.
The most important strength of the present study was that the outcome measures were based on registers that ensured follow-up for all participants. The DREAM register was established by the Ministry of Employment to be able to monitor all social transfer income, that is, tax-paid benefits due to social or health-related events with economic consequences such as sick leave. This also explains why we could not measure sick leave spells shorter than two weeks as this period is paid by the employer or the employee without compensation from the tax-paid public insurance system. Another shortcoming of the DREAM register is that the cause of sick leave is not registered. Thus, the cause was registered at the first visit at the Spine Center only for the initial sick leave which was required for being enrolled in the study. The cause of sick leave relapse was not known.