This study is based on the data of a four-arm randomised controlled trial (RCT) among female healthcare personnel (NURSE-RCT, clinical trial registration NCT01465698) , in which healthcare personnel with sub-acute or recurrent LBP were randomised to participate in neuromuscular exercise/non-exercise and to receive/not receive back care counselling for 6 months . In the secondary analysis, those receiving exercise (combined exercise + counselling, and exercise only) were merged to be exercisers, and non-exercisers (counselling only and controls) were merged to be the controls .
The study was conducted in the form of three identical consecutive sub-studies. The participants were female healthcare workers in physically demanding duties: in old people’s homes and geriatric wards (in the first sub-study in 2011, n=56); in home service, public healthcare units, and community hospital wards (in the second sub-study in 2012; n=80); and in university hospital wards (in the third sub-study in 2013; n=83) in the city of Tampere, Finland. The protocol and time frame of each sub-study are presented in the study protocol . The recruitment of participants, eligibility criteria, and reasons for exclusion have been previously described in detail . Briefly, 30–55-year-old female healthcare workers were eligible if they had worked in their current job for at least 12 months and had experienced LBP of an intensity of 2 or above on a numeric rating scale (NRS; 0–10)  during the previous four weeks. The exclusion criteria were a specific or serious earlier back condition (disc protrusion, fracture, surgery), chronic LBP (pain duration ≥7 months), pregnancy or recent delivery (<12 months), and engagement in a neuromuscular-type exercise (NME) more than once a week.
The power calculations, recruitment process, randomisation, and ethical issues of the NURSE-RCT have been presented previously , as have the contents of the exercise intervention .
The study design and flow of the participants are shown in Figure 1.
The participants were female healthcare workers who engaged in physically demanding work (including lifting, patient transfer and working in awkward positions) and suffered from sub-acute or recurrent LBP. The mean age of the participants was 46 years, and they had worked in their current job on average for 11 years . Some 87% were nurses or nursing assistants, and 70% did shift work . In the pre-study screening, most of the study subjects (82%) experienced LBP on a few or most days of the week, but not daily, while 18% had LBP daily . At the baseline, the mean of the pain intensity measured on a visual analogue scale (VAS; 0–100)  during the previous 4 weeks was 36.2 (SD 22.6) . The majority (77 %) of the study sample can be described as having sub-acute, mild-to-moderate, recurrent or fluctuating non-specific LBP (4). Among those with daily pain (18%), the pain intensity was higher the mean in VAS being 55.7 (25.3).
A wide range of measurements was taken at the baseline. In addition to background factors (age, education level, marital status, occupation, number of working years in the current job, working hours, smoking, perceived health and perceived fitness in comparison to persons of the same age and gender, current use of medication, high blood pressure (yes/no), and hormonal status); LBP intensity (VAS; 0–100) during the previous 4 weeks ; the frequency of LBP; the number of musculoskeletal pain sites ; quality of life (RAND 36) with eight sub-scales [39, 40]; depression (using the modified Finnish version of the Patient Health Questionnaire; PHQ-9); the short form of the workability index ; physical functioning in nursing tasks ; tiredness, sleepiness, and difficulties in recovering from work ; work-induced exertion in different body parts ; and psychosocial factors at work (Finnish work satisfaction questionnaire)  were investigated by questionnaires. FABs were measured with a questionnaire assessing FABs related to work (FAB-W) and physical activity (FAB-PA) . Three questions considering long-term sick leave were removed from the original FABs questionnaire, because the participants were still in work .
Physical measurements included body mass index (BMI), movement control of the low back [47, 48], and performance tests for physical fitness, namely aerobic fitness by the 6-minute walk test , muscular strength (modified push-up , one-legged squat with progressively increasing external load (10% of body weight after each performance up to 40%) , vertical jump , modified sit-ups ), agility by running a figure-of-eight , flexibility by trunk lateral side bending , and rhythm coordination . More precise information on the measurements is given in the study protocol article , the article on the repeatability of the physical measurements , and the baseline analysis of the study sample .
Exercise interventions in the NURSE-RCT
The contents of the 6-month exercise intervention have been described previously . The modified 6-month Pilates-type exercise intervention programme, which focused on controlling the neutral spine posture, started with light and easier exercises, and it was progressive in terms of demands for coordination, balance, and muscular strength over three stages. The goal was to exercise twice a week; during the first 2 months (stage I) in supervised neuromuscular exercise (NME) classes (lasting 60 minutes) and during the next 4 months (stages II and III) in one supervised class and one home session with the help of a DVD (lasting 50 minutes) or booklet produced for the study . During stages II and III, the participants were also allowed to exercise in supervised group sessions more than once a week if exercise at home was inconvenient, and also only at home if the group sessions were difficult to attend. During the progression (stages II and III), the participants were allowed and/or advised to do easier exercises from the previous stage if the more challenging exercises proved too demanding.
The leaders of the neuromuscular exercise groups were all certified Pilates instructors with a background in physiotherapy, a master’s degree in health sciences, or both . Supervised exercise groups were organised in facilities near the workplaces of the healthcare personnel. Group sessions were provided on weekdays starting 15 minutes after the typical work shifts ended. The exercise classes, videos, and booklets were free for the participants, but they exercised in their own time .
Adherence to exercise
The instructors monitored the participation of supervised group exercise, and study subjects kept an exercise diary of their home practice. The structured exercise diaries were returned at the end of stage II (week 16) and stage III (week 24). Attendance of the supervised exercise sessions and the number of home exercise sessions were added together to determine the total exercise attendance rate.
All participants in the exercise group received an information letter at the beginning of the exercise intervention about the goals and principles of the exercise programme. During the 4th week of the first-stage exercise period, those who had not participated in any group-based exercise sessions received a telephone call from a research nurse (not involved in the exercise intervention or measurements), who encouraged them to start to exercise. All participants received two material packages (between stages I and II, and between stages II and III), which included an exercise DVD, exercise booklet, exercise diary for home practice, and a letter including information about the study and the importance of regular exercise. They also received two e-mails during stage II in order to encourage exercise, and a letter before the 6-month follow-up measurements from the principal investigator (JS).
To avoid any contamination to the back care counselling intervention (in the original four-arm setting), and to ensure exactly the same information to all who were allocated to the exercise group, the exercise instructors focused on instructing the standardised exercise programme (individual modifications due to musculoskeletal problems other than LBP were allowed). All other kinds of counselling (e.g. lifestyle, pain management, and ergonomics) were avoided in the exercise classes.
Power calculations (at least 160 subjects needed) for the original NURSE-RCT have been reported previously , as has the randomisation of the participants .
Partial correlation analysis was conducted between all background and baseline variables and the adherence rate to determine which of the 60 different variables could have an association with the exercise adherence. Those variables showing a statistically significant association with the exercise adherence rate were selected for bivariate analysis with the adherence rate; the analytical methods were Spearman’s correlation for continuous variables and the Kruskall–Wallis test for the categorical variables.
The median (24 times) was used to split the exercise group into the compliers (those who exercised once a week or more; ≥24 times during the 24 weeks) and non-compliers (those who exercised 0–23 times) We examined the baseline characteristics of the participants randomised to the exercisers by the adherence status for those variables showing statistically significant associations with the adherence rate in the bivariate analysis. The analytical methods were the independent samples t-test, the χ2 test, or the Mann–Whitney U test as applicable.
To analyse the effects of the exercise programme on FABs, the mean differences in time (at three measurement points: baseline, 6 months, and 12 months) between the two groups (exercisers vs non-exercisers) were tested using a generalised linear mixed model (GLMM) (Fig. 1). To take the interaction between back counselling and exercise into consideration, all analyses were first adjusted for counselling. Second, the sub-study was included as a random effect in all the GLLM analysis models to indicate the possible heterogeneity between the study sites and study time in the three consecutive sub-studies. Other confounding factors were background variables (age, civil status, education), work-related factors (shift work/regular work, psycho-social factors at work , perceived work-induced lumbar exertion ), and health-related factors (BMI, hormonal status, perceived health, perceived fitness, blood pressure, current medication, self-reported physical activity and fitness components). Only those confounding factors that improved the model in the second stage in the sense of Bayesian information criteria were included in the final adjusted model.
After analysis according to the intention-to-treat (ITT) principle (Fig. 1), the study sample was assigned into two groups in order to investigate the effectiveness of the exercise on FABs, based on a per-protocol (PP) analysis. The mean difference in time (0, 6, 12 months) of exercise compliers (≥24 exercise sessions) were estimated and compared to the results of a combined group of non-compliers and non-exercisers (0–23 exercise sessions + controls).
The correlation between the change in LBP intensity from the baseline to 6 months  and the change in the results of the FAB measurements after the intervention period were calculated by Spearman’s correlation coefficient (rs). Associations between professional status and fear avoidance at the baseline were analysed by analysis of variance (ANOVA).
All the analyses were conducted using SPSS (IBM Corp. Released 2017. IBM SPSS Statistics for Windows, Version 25.0 Armonk, NY: IBM Corp.).