Study design and ethics

The study design was a longitudinal cohort study with two embedded RCT.

All employees (2480) in a forestry company were invited to respond to a postal questionnaire on LBP and LB-related physical impairment. Based on the responses (rate 71%), employees were divided into three main categories: no or mild LB symptoms or moderate LB symptoms that potential hamper work. Two separate RCT were performed among the subjects with mild LB symptoms and moderate LB symptoms. There is no gold standard that would distinguish mild LBP from moderate symptoms. Although the cut-off limit was chosen arbitrarily in our study, the categorization follows some previous recommendations, eg, according to Hawker et al (23), mild pain can be categorized between 5–44 mm in the visual analogue scale (VAS) and moderate pain between 45–74 mm. This paper focuses on the RCT targeted at mild LB symptoms. The results of the other RCT have been published elsewhere (24). The South Karelian Central Hospital Research Ethics Board approved the study, which was performed according to the Declaration of Helsinki.

Participants

Employees younger than 57 years were eligible for the study if they reported LBP intensity between 10–34 mm on a 100-mm VAS during the past week and fulfilled at least one of the following criteria: (i) LBP duration of ≥2 weeks in the past 12 months, (ii) LBP that radiates below the knee level at the time of responding to the questionnaire, (iii) recurrent LBP (≥2 episodes during the past 12 months irrespective of their duration), and (iv) self-reported work absence because of LBP during the past 12 months.

Exclusion criteria were retirement within the time span of the follow-up, pregnancy, presence of acute nerve root compression symptoms, suspicion of malignant tumors, recent fracture, severe osteoporosis or other specific disease preventing participation in the follow-up.

Included employees (N=312) were defined as a mild LBP cohort. Prior to the invitation to participate in the intervention study, a random sample of employees was extracted from the cohort in order to form a natural course (N=83) control group. The data of the natural course group is not shown in this paper.

The remaining employees (N=229) were invited to visit an OH nurse, who was specially trained in LBP and the main study procedures.

Randomization and blinding

An independent biostatistician prepared the stratification into two groups by using a computer-generated randomization table with a block size of 8. Based on the randomization scheme, a research assistant prepared the sealed envelopes before the start of the study containing either a referral to one of two groups: Back Book (BB) or Back Book with Advice (BB+A). The study design, implications of the trial, and alternative treatment options were explained to the study subjects personally at the randomization visit and also in the written informed consent form. After signing the informed consent form, the employee opened a sealed envelope containing their group assignment. The research assistant, OH nurse and the researchers were not able to identify the participants or the group assignments before the randomization. Due to the nature of the interventions, the participants and OH professionals were not blinded to the group assignment after randomization.

All employees would still get the best treatment and full attention of the OH service even if they did not want to participate. Information about the study was also provided in the company magazine and intranet. During the study period, the OH unit of the company operated as usual. Collaboration and communication between the study personnel, employers, trade union representatives, and the workers compensation board were assured prior to the study.

The study data were entered into the data file by persons who were independent of the research personnel in order to ensure blinded analysis.

After the randomization but well before the first 3-month follow-up visit, one person in the BB+A group retired and was excluded from the study. Thus, from a total of 182 randomized subjects, 181 were finally included in the study (table 1).

Table 1

Baseline characteristics of study subjects. [BB+A=Back Book with advice intervention group; BB=Back Book control group; FABQ=Fear Avoidance Back Questionnaire; HRQoL=health-related quality of life; LBP=low-back pain; RM-18=Roland Morris Disability Questionnaire (18-items); SA=sickness absence; SD=standard deviation; VAS=visual analogue scale.]

Interventions

At the beginning of the first visit, the OH nurse performed a clinical examination and explained the findings to the employee. If no exclusion criteria were noted and the subject signed the informed consent, the OH nurse performed the randomization according to the scheme. The visit lasted about 60 minutes in the BB group. In the BB+A group, the face-to-face information required an additional 20 minutes.

Follow-up visits were scheduled with the OH nurse at 3, 6, 12, and 24 months after the baseline visit. At every study visit, the OH nurse checked the returned questionnaire information, performed a simple clinical examination, and answered any patient questions. Employees completed the baseline and follow-up questionnaires within one week prior to the visit date. The intervention groups were comparable concerning follow-up intervals, visit activity, or the time spent (30 minutes) at the follow-up visits.

All subjects had unlimited access to the usual OH care during the entire study period and were free to obtain additional healthcare services if needed.

Outcome measures

The results of some previous studies in LBP show that there may be changes in, for instance, pain (25) or SA (26, 27), without any correlation between these variables. In this study, we have used several LB specific variables that were assessed separately in order to evaluate the effect of our intervention (25).

Power calculations

In our population, the standard deviation of RM-18 score was estimated to be 4 units. A difference of 2 units between treatment arms will be detectable with 85% power in two-tailed tests with the alpha of 0.05 for a sample of 73 employees in each group; the standardized effect size will be 0.50.

Statistical analysis

All statistical analyses were performed at employee level, according to the intention-to-treat principle. Missing values in the questionnaire data were imputed using the last observation carried forward (LOCF) principle. Baseline characteristics were compared using descriptive statistics. The effectiveness of the intervention was estimated primarily by the difference of questionnaire variables between two study groups (BB+A versus BB) at 3, 6, 12 and 24 month time points. Respective baseline values were used as covariates. The 95% confidence intervals (95% CI) for the mean differences between groups were computed using the generalized linear model (GLM). We used the statistical package of SPSS, version 17 (SPSS Inc, Chicago, IL, USA).

Regarding SA variables, the data were gathered at 12, 24, 36, and 48 months from the individual baseline date of each study subject. For all the SA variables, there was a great over-dispersion and excess of zeros in relation to the Poisson model. We used a hurdle model that consisted of a logistic part modeling the excess of zeros and a zero truncated negative binomial part modeling the count data.

In order to model nonlinear covariate effects and interactions between covariates, we used a hierarchical Gaussian process model with a neural network covariance function (34). The constructed hierarchical model included a common effect, an effect for the baseline, effects for each intervention group, and effects for each person (also called “random effects”). For the logistic model, we reported the probability of SA (days and periods) and odds ratios (OR) for the group differences, respectively. For the zero-truncated negative binomial model, we reported mean SA days or periods and mean ratios of the group difference. For all the reported values, we included 95% CI. The method is described in more detail elsewhere (24).

Employee flow

Figure 1 presents the study flow chart. From company personnel of 2480, 1754 individuals (71%) answered the questionnaire. Based on the inclusion and exclusion criteria, 312 employees were eligible for the study. From this cohort, 83 employees were randomly selected into the natural course group (data not shown), 47 refused participation, and one was excluded. Finally, 181 persons were randomized into two intervention groups (BB=92, BB+A=89).

Figure 1

Flow diagram about the course of the study showing the number of subjects at different phases of the trial.

Loss to follow-up

Four subjects from the BB+A and five from the BB group quit the study due to personal reasons within the first three months but gave permission to use their data. Some participants from both groups failed to return the questionnaires at all time-points because of changes at work or reasons unknown to the researchers resulting in missing data. For the missing data, the LOCF principle was followed. At the final check-up visit, the activity rates in the BB+A and BB groups were 75% and 73%, respectively.

Comprehensive SA data were available for all study subjects who were enlisted at the company for the entire follow-up period. Subjects who were no longer employed at the end of the follow-up years were excluded annually from the SA data analysis.

Primary outcomes

Physical impairment, low back pain (VAS) and HRQoL

Compared to baseline (table 1), there were no differences between study groups concerning physical impairment, intensity of LBP, or HRQoL at any time point during the 24-month follow-up (table 2). However, there were some differences during the follow-up time within the groups. Compared to baseline, physical impairment was lower in both groups up to 12 months but the trend remained only in the BB+A group until 24 months. Compared to baseline, VAS was lower up to 6-months but only in the BB+A group. Compared to baseline, HRQL was higher in the BB group up to 12 months (table 2).

Table 2

Results of primary and secondary outcomes (questionnaire variables) within the groups BB+A (Back Book with advice, N=89) and BB (Back Book, N=92) along with the group comparisons (BB+A versus BB) at the 2-year follow-up. [FABQ=Fear-Avoidance Beliefs Questionnaire (range 13–78); RM-18=Roland Morris Disability Questionnaire (0–18); SD=standard deviation; VAS=visual analogue scale, (0–100 mm); 15D=15-dimensional health-related quality of life (range 0–1, higher values indicate better quality of life), 95% CI=95% confidence interval]

Sickness absence days

In comparison with BB, BB+A did not reduce the probability or amount of total or LB-specific SA days (table 3) at any follow-up point during four years. At the end of the 4-year follow-up, the mean differences of accumulated SA days between BB+A and BB were -3 [95% CI -28–21] (total) and 1 [95% CI -3–5] (LBP specific).

Between BB+A and BB, there was a difference of -124 (total) and 115 (LB-specific) days in the accumulated SA days in four years. The difference in total SA days favored the BB+A group and the difference in the LB-specific days favored the BB group. However, there were some differences during the follow up-time within the groups. During the first three years, the amount of LB-specific SA days per year decreased or remained at the same level, but increased during the fourth year in both groups (table 3). The increase of LB-specific SA days per year was significant between the 3^rd and 4^th year in both groups (BB+A: P=0.001; BB: P=0.001). A similar increase occurred also in the BB group concerning the amount of total SA days per year between the 2^nd and 4^th year (P=0.030). Thus, SA days seem to remain at the same level or even decrease during the first three years, but increased thereafter, in the fourth year.

Table 3

Sickness absence (SA) days [total and low-back (LB)-specific] in the BB+A (Back Book with advice, N=89) intervention and BB (Back Book, N=92) control groups and the change in time within the groups. Columns show the probability (P) for the occurrence of SA days (>0), odds ratio (OR) of the group comparison BB+A versus BB, the number of SA days in the intervention groups and the Ratio of the group comparison between BB+A versus BB. [95%CI=95% confidence interval. Bold denotes significance]

Secondary outcomes

There was no consistent or clinically important difference in pain-related fear (FABQ) between the BB+A and BB groups during the first two years (table 2). Concerning the differences within the groups during the follow up time, FABQ decreased in both groups up to 6 months but remained low until 24 months only in the BB group.

Sickness absence (SA) periods

In comparison to BB, BB+A did not reduce the probability or amount of total or LB-specific SA periods during the four-year follow-up (table 4). The mean differences of accumulated SA periods during four years between BB+A and BB were -1 [95% CI -5–2] (total) and -0.1 [95% CI -0.6–0.4] (LBP specific).

Table 4

Total and low-back (LB)-specific sickness absence (SA) periods in the BB+A (Back Book with advice) intervention and BB (Back Book) control groups. Probability (P) of SA periods that are >0 and odds ratio (OR) of the group comparison BB+A versus BB. The number of SA periods in the intervention groups and the Ratio of the group comparison BB+A versus BB. [95%CI=95% confidence interval]

At the end of the 4-year follow-up, the difference in the sums of accumulated total and LB-specific SA periods between BB+A and BB were -184 and -20 periods, respectively. The group differences in both SA (total and LB-specific) periods favored the BB+A group.

Strengths and weaknesses of the study

Main strengths of this study lie at the pragmatic approach and the recruiting strategy of the subjects. All the employees at the company personnel were invited to participate in the questionnaire survey (N=2480). The response rate was particularly high (71%). Patient information was delivered in a practical way by an occupational health (OH) nurse for employees with mild LB symptoms. In Finland, every OH unit has a nurse but not all have physiotherapists. Therefore, we assumed that LB-specific patient information should be delivered by an OH nurse, especially when the symptoms are minor. In the majority of previous studies on LBP, however, personal patient information has been provided by a physician. In some other fields of medicine, self-care has also been promoted by a nurse or other healthcare professional without losing the effectiveness of the intervention (35, 36).

We included employees who were generally healthy and able to continue working. They were not primarily seeking care but were expected to benefit from self-care information for their LB symptoms. Employees had a history of LBP or ancillary symptoms for about 11–13 years and an average of 12 total SA days during one year prior to study inclusion, of which 1–2 days were LB specific. Most of them had a history of LB treatment, eg, a self-care program. According to study group characteristics and the pragmatic approach in this study, our results are most applicable in the OH setting. We believe that the study group was especially suitable for the trial on self-care information. We are not aware of other comparable studies in an OH setting.

All subjects received the intervention as intended and 74% of them continued to the final visit. Adherence to follow-up visits and response to the questionnaires was quite high throughout the two-year follow-up though some workers expected more intense treatment or rehabilitation for their LB trouble.

Personal, face-to-face information was performed with the help of a slide show, which was a complete review of the BB’s content; no other information was included. The outcome variables have previously been validated and are widely used. Our SA data has good coverage, accuracy, and consistency (37) as salaries and other employee benefits are based on the same information. Still, our study may have been somewhat underpowered as regards SA variables, which can, for example, be seen from the broad CI in the differences between the treatment arms.

We chose well-described and validated LB-specific outcome variables in our study although they have been shown to function best in their mid-range (38). For example, RM-18 is rather insensitive to change when the impairment levels are low (39).

One reason for the lack of effectiveness might be that face-to-face information is rather a tenuous complement to the booklet information. Furthermore, it is not possible to exclude the possibility of group contamination since some of the employees in the BB group requested additional information on LBP during their follow-up visits. We have not systematically recorded or responded to these requests, but are aware that contamination to some extent is possible. All scheduled follow-up visits in this study should be considered as part of the intervention, but there were no differences between the groups in follow-up visit activity, intervals, or frequency. On the other hand, we made no systematic attempts to find out if the study subjects in the BB group actually read and understood the booklet. However, these facts originate from the pragmatic approach of the study, ie, these concerns cannot be ruled out in common practice either.

Comparison with previous studies

Educational booklets have been used in general practice and research purposes for many years. Results from RCT on their effectiveness in LBP are mixed (3, 8, 11, 35, 40–44) and these prior studies have included patients with existing LB symptoms, either acute or chronic LBP. We are not aware of comparable RCT about LB-specific self-care information for non-sick-listed employees in an OH setting. There is an obvious need of evidence in this field.

The few prior RCT of non-sick-listed populations in an OH setting (45–48) have dealt with more intensive patient education and focused on general symptoms or risk of work disability. A population-based study in Denmark showed that a psychosocially oriented educational booklet without personal contact was not successful in reducing work absence due to general musculoskeletal pain (48). Interventions in two other studies (46, 47) were more intensive than provision of simple patient information and the study subjects were already at risk of work disability.

According to the recent Cochrane review, strong evidence supports that an individual 2.5-hour oral educational session is more effective on short- and long-term return to work than no intervention for patients that were absent from work due to acute or sub-acute LBP (9). Less-intensive interventions or written information alone were not effective, and there was no evidence of pain relief in these subjects. Another systematic review (13) concluded that simple patient information for subjects with chronic LBP increases patient knowledge about LBP and reduces pain, disability, and fear but not employee absenteeism. A positive result was strongly related to the personal contact between the information provider and the patient.

Obviously, the concept of self-care is quite different from the traditional care-giving concept in healthcare (49). Patients may gather patient information from various sources of their own choice and for their individual purposes. The quality of such information may range from non-factual to proper evidence-based information.

When OH professionals promote the self-care of LB patients, written information looks superior compared to oral information alone (3, 12, 50). On the other hand, we have found no effectiveness of the additional face-to-face information in our study. Like in some other studies (51–54), previous SA predicted future work loss also in this study. In the OH setting, there is an urgent need for simple and reliable LB-specific patient information that can be delivered to employees during their health surveillance visits. In the view of our findings, further research is needed to analyze the effectiveness of patient information compared to usual care in the secondary prevention of mild LB symptoms in the OH setting.

Concluding remarks

Face-to-face patient information, based on a LB-specific booklet, was not more effective in combination with the booklet than the booklet information alone in reducing mild LB symptoms among non-sick-listed employees.