Study design, settings, and participants
The study-design was a blinded four-arm randomized controlled trial of 6-month interventions with effectiveness and cost-effectiveness evaluations at 12 months. The Ethics Committee of Pirkanmaa Hospital District (ETL code R08157) approved the study protocol (ETL code R08157). The aim of the study, as well as risks and benefits, were clarified in a written information letter to those recruited to the study. Participants were encouraged to continue their usual physical activity and seek any medical or other treatments when needed. All participants provided their written consent to a research secretary at the beginning of the baseline measurements. The study protocol of NURSE-RCT is available at: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5117067/pdf/bmjsem-2015-000098.pdf [19].
Contrary to our original study plan to conduct a single RCT (ClinicalTrials.gov NCT01465698), we conducted three sub-studies to reach an adequate sample size. The sub-studies started consecutively in 2011, 2012, and 2013 at different locations in Tampere, Finland. Details of enrollment, settings, and time-points for screening, randomization, measurements, and interventions for each consecutive sub-study are shown in Fig. 1 of the trial protocol [19].
The target population was female nursing personnel from wards that required lifting and transferring patients, and direct healthcare workers from settings where the work was otherwise awkward for the lower back [6]. In the present paper, ‘healthcare workers’ refers to participants of the present study who were nurses, nurses’ aides, specialist nurses, assistant physiotherapists, physiotherapists, and midwives. The participants were individuals who voluntarily participated in the study on their own time outside working hours.
Study inclusion and exclusion criteria
The first author (JHS) was responsible for decisions regarding study inclusion or exclusion. The inclusion criteria were [19]: women aged 30–55 years; worked at current job for at least 12 months; intensity of LBP of at least 2 on the Numeric Rating Scale (scale 0–10) during the past 4 weeks [20]. The exclusion criteria were: serious former back injury (fracture, surgery, disc protrusion); chronic LBP defined by a physician or self-report of continuous LBP for 7 months or more [20]; disease or symptoms that limit participation in moderate intensity neuromuscular exercise; regular engagement in neuromuscular-type exercise more than once a week; pregnant or recently delivered. Altogether, 439 women responded to the screening questionnaire, 56% (n = 245) of whom met the inclusion criteria and 11% (n = 26) of whom refused to participate in the baseline measurements. The main back-related reasons for exclusion were intensity of LBP of less than 2 on the Numeric Rating Scale (22%) and having had continuous LBP for more than 7 months (12%) [17].
Randomization and masking
Participants were randomly assigned into one of the four study groups in equal proportions within each of the three consecutive sub-studies [19]: Combined neuromuscular exercise and back care counseling (Combined), Exercise alone (Exercise), Counseling alone (Counseling), and a non-treated group (Control). The statistician (KT) generated the random numbers utilizing the RAND function in Excel (Microsoft, Redmond, WA; ver. 2010). At the first appointment, the research secretary obtained a signed informed consent from all participants, after which she opened an envelope (next in order) to allocate the participant to a study group and provide information for participation. Research nurses conducting the study measurements were blind to the group allocation at the time of data collection, and the statisticians (KT, JR) until completion of the statistical analyses.
Interventions
Educated professionals provided the group exercise and back care counseling sessions near the workplaces of the participating healthcare personnel on weekdays, starting 15 min after the end of a typical day shift [19]. The instructors monitored adverse events related to exercise and adherence to both interventions during the group sessions. Participants in the Combined- and Exercise-arm received instructions to keep a diary of their exercise performed at home.
Researchers AT and JHS were responsible for the aims and training principles of the neuromuscular exercise. AT designed the exercise program and educated the exercise leaders, all of whom had a basic education in physiotherapy, a master’s degree in health sciences, or both. Researchers JHS and MR were responsible for designing the key issues and their realization of the back care counseling sessions. They also guided the counselors, all with a master’s degree in health sciences, to the content and materials of each counseling session at the beginning of each consecutive sub-study.
Neuromuscular exercise
The aim of the progressive neuromuscular exercise program (Supplementary appendix 1 of the study protocol: bmjsem-2015-000098supp_appendix1.pdf) was to enhance spinal stability by improving the movement control of the lumbar region of the back [19]. The training principles included maintenance of a neutral spine posture by co-contraction of the trunk muscles in all exercises [21,22,23,24,25]. Exercises demanding high muscular activity and inducing a low lumbar load [24], such as the side-bridge [23] and four-point kneeling [25], were preferred. In addition, exercises to increase the endurance and strength of the gluteal and lower extremity muscles [26] were included to meet the demands of the strenuous lifting tasks required of nursing [18]. The target dose for exercise was 48 sessions (60 min) twice per week for 24 weeks, and the expected minimum efficient dose was 24 based on a previous study by Suni et al. [27]. During the first 8 weeks, the goal was to participate in instructed exercise sessions twice a week, and during the next 16 weeks, in one instructed session and one home session with the help of a digital videodisc or booklet produced for the study [19].
Back care counseling
Cognitive behavioral learning theory was the framework for the back care counseling [28], and problem-based learning was the method used for implementation [19]. The main issues introduced and discussed in the group counseling sessions were: explaining LBP; how to avoid harmful loading of the lumbar spine in all daily activities; active strategies to cope with LBP; the role of physical activity in LBP, and overall health and well-being (Supplementary appendix 2 of the study protocol: bmjsem-2015-000098supp_appendix2.pdf). Researchers (JHS, MR) designed the specific learning targets, materials, and “take home tasks” for each session. Safe methods of squatting, emphasizing a neutral spine posture for the lower back [19, 27, 29], were practiced for 5 min during counseling sessions 2–10. The target dose for counseling was 10 sessions (45 min) once a week for the first month and then every third week for 24 weeks.
Study measurements
The participants took part in study measurements at baseline, immediately after the interventions at 6 months and after follow-up at 12 months. Research nurses conducted the physiologic measurements at the research institute, and the participants responded to the study questionnaire during the measurement sessions or posted them later to the institute.
Outcomes of effectiveness
The main outcome measure of effectiveness [19] was intensity of LBP as measured with the Visual Analog Scale (VAS, 0–100 mm) [30]. Secondary outcomes were bodily pain interfering with work [31] and FABs related to work and physical activity [8]. Contrary to the original plan of the NURSE-RCT [19], the results of the test battery regarding movement control impairment [32] are not included in the present paper due to the poor reliability of several test items [33] assessed as part of the first sub-study.
Outcomes of cost-effectiveness
We evaluated cost-effectiveness of the three intervention-arms compared with the control-arm in terms of days of sickness absence due to LBP reduced and QALY gained. The QALY were calculated from the SF-6D score [34] derived from the original SF-36 data [31], which is a validated instrument for measuring the physical and mental components of quality of life. Cost assessment included direct healthcare costs (visits to a physician, nurse, physiotherapist, inpatient days, and medication) and days of sickness absence for each 6-month period, retrospectively collected via a questionnaire at baseline [17] and as continuous self-reporting with the same questionnaire during the intervention (0–6 months) and follow-up (6–12 months) periods. We calculated the costs of the delivery of the three intervention-arms: salaries of instructors with administrative costs, material costs, and opportunity cost for home exercise. Additional file 1 provides further information on the SF-6D score [34], assessment of cost-effectiveness and cost calculations, and reports the calculated costs.
Statistical analysis
Sample size calculation
Sample size was calculated [19] based on the intensity of LBP in terms of an absolute change [35] of at least 15 mm in VAS. We expected that there would be a minimal difference of 20% between the intervention groups with improved VAS, and 15% in the control group. Thus, to detect a difference in main effects (i.e., exercisers vs non-exercisers and counseling vs non-counseling) with a significance level of 0.05 and a power of 80%, the study required at least 160 participants (40 in each study-arm). For compensation of probable loss of participants to follow-up, the aim was to recruit 240 participants, see Fig. 1 of the study protocol [19] and the CONSORT flow-chart (Fig. 1) of the present paper.
Analysis of effectiveness
The present paper introduces the results of the primary analyses of effectiveness based on a superiority design of any single intervention-arm compared with the Control-arm. Statistician (KT) performed all analyses according to the intention-to-treat principle. The change in the primary outcome of intensity of LBP in VAS and the other outcomes of effectiveness were analyzed as percentage of change [35] over time at three time-points (baseline, 6 months, and 12 months) using a generalized linear mixed model (GLMM) with gamma or log-normal distributions using SPSS statistics software, version 22 (IBM, Chicago, IL).
Statistician KT first conducted the GLMM analyses without any adjustments (crude analysis). Next, the GLMM analyses were first adjusted [36] as follows: Background variables, including age, civil status, level of education, and smoking. Work-related factors covering shift work, perceived physical exertion at work, perceived work-induced lumbar exertion [37], support from superiors [38], and work stress as effort-reward imbalance [39]. Health-related factors included perceived health, perceived fitness, body mass index, meeting the aerobic part of the physical activity recommendation [40], and fitness in a modified push-up test [41]. Only covariates that improved the model at both follow-up stages in the sense of Bayesian information criteria were included in the final models.
Second, the sub-study was included as a random effect in all the GLLM analysis models to indicate possible heterogeneity between the study sites and study time in the three consecutive sub-studies (see Fig. 1 of the study protocol) [19]. KT calculated the continuity-corrected confidence intervals for proportions with the statistical software R function prob.test [42].
We used Cohen’s d to calculate the effect size for the primary outcome measure. The proportion of participants with an improvement of at least 15 mm in the absolute VAS score [30, 35] at 6 and 12 months in each study group is also reported: the between-group differences at the two time-points (baseline and 6 months; baseline and 12 months) were analyzed using the chi-square test.
Analysis of cost-effectiveness
We evaluated cost-effectiveness ratio for each intervention-arm in comparison with non-treatment control-arm [43]. Cost-effectiveness is expressed as incremental cost-effectiveness ratios (ICERs), calculated as the ratio of the difference in mean total costs (including LBP related healthcare costs, medication, costs of sickness absence, and intervention costs) and mean effects (i.e., change in number of days of sickness absence or QALY) at the level of the study-arms. Regarding the cost-effectiveness analysis for sickness absence, the costs of sickness absence days of regular workers (i.e. study participants) were not included in order avoid double counting. The ICER indicates the amount of money required to decrease a day of sickness absence or gain QALY.
We estimated the uncertainty regarding the ratios in mean total costs and mean effects using bootstrapping with 5000 iterations to generate 95% confidence ellipses for the joint distribution of cost and effectiveness outcomes, and graphically represented them on a cost-effectiveness plane. Cost-effectiveness acceptability curves indicate the probability of any of the alternative interventions being cost-effective. JR conducted the cost-effectiveness analyses using Stata statistics software, version 12.1 (StataCorp LP, College Station, TX).
The costs of implementing the interventions were higher because we conducted three consecutive sub-studies instead of a single study [19]. To evaluate the robustness of the findings, we performed sensitivity analysis assuming a single intervention for all participants. Thus, the intervention costs related to group-sessions would be one-third of the actual costs.