Cohort profile: why do people keep hurting their back?

Low back pain (LBP) is one of the most disabling and costly conditions worldwide. It remains unclear why many individuals experience persistent and recurrent symptoms after an acute episode whereas others do not. A longitudinal cohort study was established to address this problem. We aimed to; (1) evaluate whether promising and potentially modifiable biological, psychological, social and behavioural factors, along with their possible interactions, predict LBP outcome after an acute episode; (2) compare these factors between individuals with and without acute LBP; and (3) evaluate the time-course of changes in these factors from LBP onset. This paper outlines the methodology and compares baseline characteristics between acute LBP and control, and LBP participants with and without follow-up. 133 individuals with acute LBP and 74 pain-free individuals participated. Bio-psycho-social and behavioural measures were collected at baseline and 3-monthly for 12 months (LBP) or 3 months (control). Pain and disability were recorded fortnightly. Baseline characteristics were mostly similar between those who did and did not return for follow-up. Initial analyses of this cohort have revealed important insights into the pathways involved in acute-to-chronic LBP. These and future findings will provide new targets for treatment and prevention of persistent and recurrent LBP.


Introduction
Low back pain (LBP) is the world's leading cause of disability [1] and is associated with enormous and escalating costs to society [2]. Most of this burden is attributed to the condition when LBP becomes persistent or recurrent. Why some individuals with acute LBP recover, whereas others do not [3,4], is largely unknown.
Although psychosocial factors have generally been considered stronger predictors of long-term outcome than diagnostic or injury-related factors [5,6], they only explain a small proportion of the variance in outcome [6,7]. Biological factors have largely been dismissed and the few that have been comprehensively addressed (e.g., muscle strength/endurance [8]) have little relation to outcome [9]. With this incomplete understanding of factors related to LBP outcome, it is not surprising that most treatments have modest effects at best [10] and are generally unable to prevent recurrence/persistence of pain [11].
We argue that three issues underlie a fresh approach to this problem. First, novel biological factors that could online questionnaires) were repeated at 3, 6 and 9 months for LBP participants, and at 3 months for control participants. At 12 months, questionnaires were completed by all participants in the LBP group in addition to a separate 12-month recall questionnaire relating to the trajectory of their LBP since initial assessment for the study. Participants were also instructed (and reminded) via email to report their pain and disability level every fortnight for 3 (controls) or 12 months (LBP) via an online survey. For some analyses that have been conducted to date, these pain and disability data have been used to classify LBP participants as either "unrecovered", "partially recovered" or "recovered" at follow-up (for details see Additional file 1: Table S2).

Participant recruitment
A total of 1849 individuals from Brisbane (and surrounds), Australia, were screened between April 2012 and September 2017 (Fig. 1). Participants were recruited through advertisements around the University campus and local community, social media, three nearby hospitals and via a professional recruitment agency (Trialfacts).
Screening was conducted using two different methods. Initially, eligibility was determined via email and/ or phone, and when this method was used the reason(s) for exclusion at this initial screening were not recorded. This was replaced with an automated online screening questionnaire from April 2014, with reasons for exclusion recorded. The inclusion and exclusion criteria for LBP participants are outlined in Additional file 1: Table S3. Participants did not need to be experiencing their first ever LBP episode. Previous LBP was recorded for inclusion as a covariate. Control participants were included if they had not experienced LBP within the last month in addition to meeting the exclusion criteria in Additional file 1: Table S3.
At the first laboratory testing session (after initial screening), eligibility for inclusion were confirmed using data from the baseline questionnaire (completed within 24 h of the first laboratory-based session), to ensure that the participant's average level of pain and LBPrelated disability in the past week exceeded the inclusion threshold (≥ 1/10 for pain; ≥ 1/24 for disability). Potential control participants who reported a score > 0 on a 0 ("no pain") to 10 ("worst pain imaginable") numerical rating scale (NRS) and/or the Roland Morris Disability Questionnaire (RMDQ [35], [for definition of measures see Table 1]), or provided no scores were excluded from the study (N = 14). Potential LBP participants who reported < 1 on the pain NRS and/or the RMDQ, or provided no scores in the past week were excluded from the study (N = 8). After data collection, criteria for exclusion Responses to questions are quantified on a five-point Likert scale ranging from 0 ("not at all") to 4 ("all the time") with respect to how often the respondent experiences certain thoughts and feelings when in pain. Yields a total score as well as three subscale scores of magnification ("I become afraid that the pain will get worse": 3 items), rumination ("I worry all the time whether the pain will end": 4 items) and helplessness ("I feel I can't go on": 6 items) 0-52: ↑score = ↑pain catastrophizing Subscales: magnification (0-12), rumination (0-16) helplessness (0-24) Fear-Avoidance Beliefs Questionnaire (FABQ [20], measured in LBP only) Questionnaire: evaluates fearful and avoidant behaviours.
Consists of 16 items in which participants' rate their agreement with each statement on a seven-point Likert scale ranging from 0 ("completely disagree") to 7 ("completely agree"). Two subscales measure the agreement of statements related to physical activity (FABQ-PA: 4 items) and work (FABQ-W: 7 items) Zero was allocated for values below the reported sensitivity of the test [14,16] TNF (pg/ml), IL-6 (pg/ml), IL-1β (pg/ml), CRP (ng/ml)

Units/range
Pain processing Laboratory measure: Pain thresholds to pressure (PPT), heat (HPT) and cold (CPT) were assessed at the back (LBP -site of most pain on palpation; control -fixed site ~ 5 cm rostral [toward the head] and lateral to the center of the lumbo-sacral junction divided randomly between the left and right side) and either the thumb nail bed (PPT) or proximal volar aspect of the forearm (HPT and CPT) [12,15]. CPM was assessed based on our previous work [23] that validated the use of PPT as a test stimulus (TS) and noxious contact heat as the conditioning stimulus (CS). TS and CS were applied to the lower back or forearm. CPM was measured on three occasions (separated by a 15-min break) using different anatomical locations and stimuli (TS/CS) arrangements as reported previously [12,15].
The CPM response was calculated as the difference between the TS scores obtained before and during the CS. A higher TS score during the CS than baseline indicated pain inhibition (expressed as a positive value). A lower TS score during the CS Multifidus muscle morphology Laboratory measure: Multifidus muscle cross sectional area was measured at the level of each spinous process between the first lumbar (L1) and first sacral (S1) vertebra on both sides of the body (totalling 12 images) using a high resolution ultrasound system (LOGIC 9; GE Company, Milwaukee, WI), with a linear array 10 MHz transducer [24] Cross-sectional area (cm 2 ) Trunk muscle coordination Laboratory measure: Latency of response of superficial trunk muscle activity to unloading was assessed using an established paradigm [25]. Participants sat in a semi-sitting position with their pelvis fixed and a cable attached either behind or in front of their trunk via a harness. On instruction, participants pulled against the cable with a force of 12.5% of their body weight using visual feedback (target on a computer screen). The cable was released at an unpredictable time for 10 repetitions in each direction and the response of 12 abdominal and back muscles were recorded with surface EMG electrodes using placement described previously Muscle activity (EMG)

Units/range
Trunk mechanical properties Laboratory measure: Effective trunk stiffness, mass and damping was estimated following trunk perturbation with the trunk modeled as a linear second-order system [26]. Participants sat in a frame with equal weights (7.5% body weight) attached to the front and back of the trunk via pulleys such that the masses were balanced and the trunk could move freely with minimal muscle activity. The trunk was perturbed by the unexpected release of one of the weights. The task was repeated 10 times in each direction with the order of direc-

Measurements
Details of the measures and at what time-point(s) they were implemented are presented in Table 1 and Additional file 1: Table S1. All variables were measured in a standardised order for each participant. Biological measures were those we considered to be the most promising candidate factors for predicting LBP recurrence/persistence based on previous research and plausible rationales founded on clinical, epidemiological and fundamental research. For psychological measures, we considered three key domains of relevance in LBP: cognitive (expectations, beliefs, and perceptions concerning pain) [6,[36][37][38], emotional (distress, anxiety, and depression) [5], and behavioural (coping, pain behaviour, and activity/activity avoidance) [6,36,37]. Social measures were selected based on the Multinational Musculoskeletal Inception Cohort Study (MMICS) guidelines [39]. These guidelines were developed by an international expert team with review of the best available evidence from systematic/ narrative reviews and expert consensus. As it was not our intention to withhold treatment over the study period, we collected information regarding health care and medication use so that treatment variables can be included as covariates. The total number of variables was restricted to limit the required participant sample size, minimise the potential for over-fitting, and for cost-benefit.

Sample size
A sample size of 217 was calculated based on power to detect predictor variables using complex multiple regression models (i.e., 28 predictor variables, 5 a priori selected interactions) and growth curve modelling methods, while allowing for loss to follow-up. The planned sample size was not achieved due to feasibility issues as outlined in the "limitations" section. Although the achieved sample size (N = 133) limits the ability to examine numerous interactions simultaneously, reduced model sizes and alternative methods (e.g., cluster analysis) have been applied successfully on data from this cohort [12][13][14][15][16].

Data analysis
All questionnaire-based measures (that could be quantitatively analysed) at baseline were compared between: (1)

Fig. 1 Cohort flow diagram
LBP and control participants, and (2) LBP participants who did and did not follow-up at 3, 6, 9 and 12 months.

Participant characteristics at baseline
The characteristics of the study participants are described in Additional file 1: Table S4. Compared to controls, LBP participants were/had: taller and heavier, a higher BMI, a higher prevalence of comorbidities, a higher incidence of previous LBP, higher depressive and pain catastrophizing symptoms, higher self-reported job demands, more sick days (over the last 12 months), poorer sleep quality, more likely to have a history of cigarette smoking, and more likely to have performed vigorous physical activity on less days in the previous week.

Participant attrition
Of the 133 eligible acute LBP participants who were enrolled in the study and provided baseline data, 35 (26%) were lost to follow-up for their laboratory-based measures (i.e., did not attempt/complete any biological measures) at 3 months, a further 9 at 6 months (total lost to follow-up = 44, 33%), and a further 5 at 9 months (total lost to follow-up = 49, 37%). As biological measures were not performed at 12 months, all participants were invited to complete the standard 3-montly questionnaire in addition to a separate recall questionnaire at 12 months, irrespective of whether or not they had continued or dropped out earlier. One or both of these were fully/partly completed by all but 41 of the 133 LBP participants who started the study (follow-up at 12 months: N = 92). For participants that did follow-up, Additional file 1: Table S5 shows the number of those that provided valid data for each of the 3-monthly questionnaire-based measures at each respective time-point. Ten control participants did not return for follow-up at 3 months. With respect to the completion rate of fortnightly pain (NRS) and disability (RMDQ) questionnaires, 85% (1505 of 1770) were completed by LBP participants who were retained for follow-up (i.e., up to 3, 6, 9 or 12 months) within 7 days of each questionnaire being issued, and 91% (282 of 310) were completed by control participants (i.e., up to 3 months).

Comparison of follow-up and non-follow-up participants
Comparison of baseline characteristics between LBP participants who did and did not follow-up for laboratorybased measures at 3, 6 and 9 months, and questionnaire measures at 12 months, revealed some differences, as shown in Table 2.

Discussion
This paper profiles the only acute LBP cohort in which detailed biological, psychological, social and behavioural factors have been longitudinally and frequently collected, to date. The cohort has great potential to provide unique insight into the features that may predict and/or mediate long-term outcome [40]. The findings of baseline (acute LBP) characteristics presented here provide a foundation for future longitudinal analyses. Whether the findings can be generalised to a larger or clinical sample of individuals with early-acute LBP requires further and detailed studies of the condition during the early-acute phase. Despite the rate of loss to follow-up, most occurred after the first session, and baseline characteristics were generally similar between those who did and not return for follow-up.
Initial analyses of this cohort have revealed specific immune and nervous system features associated with the transition to persistent/recurrent LBP, and that various psychological and behavioural factors shape these relationships [12][13][14][15][16]. Ongoing analyses focus on elucidating the role of trunk neuromuscular, kinematic, mechanical and morphological properties, along with their possible interactions with psychosocial/behavioural features, in predicting LBP outcome.
∆ Participants who answered "other" to question 25 of the Job Content Questionnaire were removed prior to analysing the "job insecurity" scale.