Does a patient’s physical activity predict recovery from an episode of acute low back pain? A prospective cohort study

A prospective cohort study, conducted from March 2008 to April 2009, measured observed change in physical activity in patients with acute LBP. The main outcome measure used for calculation of sample size was change in the Roland Morris Disability Questionnaire (RMDQ) score from baseline to 3 months, and calculated for 80% power with a two-sided test and significance set at p = 0.05. Based upon a clinically meaningful change of 4 points (and SD = 5.4) in the RMDQ [17], and assuming unequal group sizes between those whose PA levels change by 4 points or greater, versus those whose PA levels change minimally over a 3 month period, required a sample size of 65 participants. An estimated attrition rate of 40% over the two time points increased the required sample size to 100 participants.

Following ethical approval from the New Zealand Lower South Regional Ethics Committee (LRS/07/11/043), 101 participants were recruited by public advertising within physiotherapy clinics (n = 77) and general practice (GP) clinics (n = 16), as well as by newspaper advertisement (n = 8). Eligible participants had an episode of LBP of 6 weeks or less, preceded by a minimum period of 3 months during which participants had not sought treatment for LBP. Participants were aged between 18 and 65 years, and had no other pre-existing conditions which limited their mobility. All participants were required to provide written informed consent to participate in this study and were given a $10 voucher as a measure of thanks and recompense for their participation. All participants were required to be receiving physiotherapy treatment for the current episode of acute LBP to be eligible to participate in the study. Physiotherapy treatment was administered by their treating physiotherapist and they were free to take a pragmatic approach with regards to the types and duration of treatment the patients received. The principal investigator (PI) screened each participant for the following exclusion criteria: serious or systemic spinal pathologies including persistent or progressive neurological deficit, intractable pain, spinal surgery, or inflammatory disorders.

The RT3 accelerometer (Stayhealthy, Inc., Monrovia, California) is a small, pocket sized, portable triaxial activity monitor. The RT3 monitor is a triaxial accelerometer used to measure physical activity in free living and is approximately the size of a personal pager (71 × 56 × 28mm, weight 65g, with one AAA sized battery). When firmly attached to clothing or a belt at the waist, it measures the acceleration of bodily movement in three directions, vertical (X), anterioposterior (Y) and mediolateral (Z), converting this movement into raw counts. Vector magnitude (VM) is the square root of the sum of the squared counts in each direction and is utilized as a measure of accelerometry output representing PA. The RT3 monitor can store data for up to seven days while simultaneously recording from all three axes at one-minute intervals. Proprietary interface software allows the download of data to an appropriate computer database. The software converts activity counts into kcals/minute from a physiological regression equation developed by the manufacturer; helping to calculate both activity related energy expenditure (EE) and total EE.

The RT3 is capable of collecting and storing data in 1-minute epochs for 21 days, and has no external controls that allow the person being tested to manipulate or change recorded data. The RT3 is therefore potentially suitable for long term tracking of PA data in the home environment [18]. The RT3 demonstrates high levels of intra-monitor reliability [19], although levels of inter-monitor variability are more variable [20]. To evaluate the technical performance of the nine RT3 accelerometers in this study for field use, each monitor was subjected to specific vibration testing along each sensitive axis in isolation. Inter-instrument CV across all monitors on each of the 3 axes ranged from 10.8 – 35.7, and the intra-class correlation (ICC) for intra-axis reliability ranged from 0.98-0.99 on the 3 axes. Due to the variable inter-monitor reliability, each participant was given the same monitor on the repeat testing measurement. The RT3 tri-axial accelerometer accumulated vector magnitude (VM) activity counts for each one-minute epoch over the 7 days of monitoring.

All participants were recruited through Physiotherapy clinics in the Dunedin region and all were currently receiving Physiotherapy treatment for their LBP. For the purposes of this study, the form of treatment was not specified and Physiotherapists were advised to educate their patients around PA according to standard accepted guidelines [1]. The RT3 was attached over the right hip at the level of the iliac crest for seven days in a harness. Participants wore the RT3 during waking hours (except when performing activities which might cause it to become wet, such as bathing or swimming) for 7 consecutive days. Participants also completed an activity diary [21] which detailed their predominant daily activities during each hour, their sleep times, and also noted removal of RT3 and reasons for such removal. Participants were contacted twice during the week to improve compliance in wearing the RT3, recording activity in the activity dairy, and to address any problems the participants were having with either the RT3 or in using the activity diary. After seven days the participant met with the PI at either the School of Physiotherapy or the participant’s home/work. RT3 data was downloaded to a portable computer by the PI and the daily activity diary collected.

A 7-day recall questionnaire (7D-PAR) [22] was completed by interview at the end of each week of activity assessment. This questionnaire is accepted as providing reasonable validity as a measure of free living energy expenditure (EE) within various populations [23, 24]. Participants were asked to recall the amount of time spent in sleep, moderate, hard, and very hard activities during weekdays and weekend days of the past 7 days. This allowed an estimate of the average total daily energy expenditure (TDEE) and physical activity energy expenditure (PAEE) for each participant, as well as an estimate of the time spent in each of the activities indicated. Participants also completed an activity diary [25] which detailed their predominant daily activities during each hour, their sleep times, and also noted removal of RT3 and reasons for such removal.

Habitual activity levels prior to the onset of LBP were measured with the Baecke Physical Activity Questionnaire (BPAQ), which gathers information on “usual” activity levels over the previous year. The BPAQ demonstrates good repeatability and relative validity in free living populations [26], and high test-re-test reliability in low back pain populations [27]. The questionnaire is divided into three sections (work, sports, and leisure), which are individually scored [28]. The questionnaire was worded such that participants were asked about activity levels over the previous year prior to the onset of LBP.

Participants completed a number of validated LBP outcome measures at baseline and at 3 months. The primary outcome measure was a change in RMDQ. The 24 item RMDQ has been shown to be a valid measure of LBP disability and sensitive measure of change in functional disability in LBP populations [28]. This self-administered questionnaire consists of 24 items which refer to limitations of daily activities as a result of LBP. Total score ranges from 0 to 24 (higher scores indicate more disability). Secondary outcomes were pain intensity, measured with the Visual Analogue Scale (VAS) which is a scaled measurement of pain (0 – 100) in which the participant was asked to rate the average level of pain over the past 7 days. This tool has been shown to be a valid, reliable and appropriate tool for use in clinical practice [29], and sensitive to change [30]. A specific activity question, developed for this study, asked participants Do you consider that you have returned to full “normal” activities since this current episode of low back painn (Yes/No).

Participants also completed the Fear-Avoidance Beliefs Questionnaire (FABQ), and the 12-item General Health Questionnaire (GHQ12). The FABQ has been shown to be a reliable measure of pain-related fear in acute LBP populations [31] and demonstrates strong predictive validity for functional disability in both acute [32] and chronic LBP populations [33]. The GHQ12 is a validated measure of psychological distress in the general population [34] as well as LBP populations [35]. Prospective research has shown that GHQ12 scores can predict future episodes of LBP [36]. Main occupation was also recorded and dichotomised into manual (involving manual lifting, heavy labour, or regular bouts of physical activity or exertion) or non-manual occupation. Each participant’s height and weight were also recorded to calculate their BMI.

Full details of the methodology are available in Hendrick et al. [37]. In brief, participants were recruited from physiotherapy practices within the Dunedin and Otago region (approx. population of 250,000). The principal investigator (PI) met with the participant at the physiotherapy practice or at the participant’s home or work. Each participant completed the baseline LBP outcome measures and the FABQ, GHQ12, and BPAQ. The participant’s age, height, weight, age, occupation, and ethnicity were recorded, and the participant instructed to wear the RT3 over the right pelvis during waking hours, and only to remove it for sleep and water-based activities. Participants were also asked to record their main activity for each hour in an activity diary, and to note the time and reason for removing the RT3. In order to improve compliance, participants were contacted twice during the week that they wore the RT3. Participants were instructed to continue their “normal” activity levels; they did not receive specific physical training during this week nor given any specific instructions regarding activities. At the completion of the week of monitoring, participants met the PI, the RT3 was removed and the data downloaded to a computer and checked to ensure that measurement protocols were met. To be included in the analyses, each participant was required to have recorded a minimum of 10 hours of RT3 data on five or more days, including one weekend day [38]. Sleep times were also ascertained, the data scanned and possible RT3 malfunctions identified. Data were then scanned for non-worn periods (> 20 minutes of zero VM activity) and such data were set to an electronic file labelled as ‘missing’ [39]. Participants returned the activity diary and completed the 7D-PAR under the guidance of the PI. Each participant repeated the activity monitoring procedure and the LBP outcome measures as per baseline at 12 weeks following the first measurement point.

The main outcome variables were RMDQ at 3 months and change in RMDQ from baseline to 3 months. The change score was calculated by subtracting each participant’s 3 month RMDQ score from their baseline RMDQ score. The sum of RT3 activity counts for each day was calculated as well as the total number of hours of activity data collected on each day. The RT3 activity levels were calculated from the estimated wear periods and total weekly activity counts were then divided by the total number of hours worn. The RT3 score was expressed as VM counts/hour/week. Change in RT3 activity was calculated as VM counts/hour/week (VM/hr/wk) at 3 months – VM/hr/wk at baseline. The SD of RT3 VM/hr/wk was calculated at the two time points as a measure of variability in PA [40]. High RT3 VM/hr/wk change was determined as greater than the 75th percentile change in RT3 VM/hr/wk. Daily physical activity energy expenditure (PAEE) from the 7D-PAR (kcal/kg/day) was calculated as the average number of hours in each activity multiplied by the metabolic equivalent (MET) value assigned to the activity category (light = 1.5, moderate = 4, hard = 6, very hard = 10) [41]. Change in PAEE was calculated as the average daily PAEE at 3 months – average daily PAEE at baseline. Paired t-tests and Pearson correlations analysed the difference and the correlation between the groups of parametric scale variables respectively. Comparisons of the binary variable “returned to ‘normal’ activities” at the two time points were made using McNemar’s test. Kendall tau correlations assessed the correlation between the dichotomous variables (return to ‘normal’ activities) and RT3 VM/hr/wk at baseline and 3 months

Simple linear regression was used to assess the unadjusted relationship between PA and PA change and the two main outcome measures ΔRMDQ and RMDQ at 3 months. The explanatory variables included activity levels at baseline (measured with the RT3 and 7d-PAR), change in PA (baseline to 3 months with the RT3 and 7d-PAR), age, sex, occupation, BMI, as well as baseline levels of pain, depression, anxiety, emotional distress and fear avoidance (GHQ12 and FABQ) and activity levels prior to the onset of LBP (Baecke work, sport and leisure scores). A separate analysis was performed to investigate whether change in activity in participants with a low activity at baseline as recorded by the RT3 and 7D-PAR predicted RMDQ change. For the purposes of this research a low activity was defined as below the mean value for both RT3 VM/hr/week and PAEE (kcals/kg). Two types of adjusted analyses were carried out. The first adjusted for all variables whose p-value was < 0.1 in the unadjusted analyses. The second began with these variables and performed a backwards selection multiple linear regression model, which forced RT3 activity level at baseline and change in PAEE as recorded by the RT3 to be included in the model. Assumptions for the regression model were tested by investigating the normalized distribution of residuals and also testing for collinearity and homoscedasticity of the data. All analyses used SPSS software version 14.0 (SPSS Inc., Chicago, Ill).

Measures of PA at baseline and 3 months are shown in Table 4 for the 83 data sets with complete RT3 data and for the 90 participants with complete data from the 7D-PAR. There were no significant differences in (1) the activity levels recorded with either the RT3 or the 7D-PAR from baseline to 3 months; (2) the amount of hours recorded as moderate, hard, and very hard, from the 7D-PAR from baseline to 3 months. The only difference in baseline measures noted between those lost to follow-up and those with complete data was a greater percentage of those with complete RT3 data having not returned to full ‘normal’ activities at baseline (p < 0.05) (data not shown).

In univariate analyses none of the measures of activity at baseline or change in activity from baseline to 3 months predicted RMDQ score at 3 months (Table 5). Significant variables in univariate analyses included in the multiple linear regression model are presented in Table 6. Increasing age was found to be the only variable which predicted RMDQ at 3 months; all other variables were not significant predictors of RMDQ score at 3 months.

Multiple linear regression analyses including all variables with a p value < 0.1 from the unadjusted analyses and measures of PA change from the RT3 and 7D-PAR are presented in Table 7. The variable associated with a change in RMDQ was the patient’s report of a return to full ‘normal’ activities at the 3 month point. All measures of PA measurement were not predictive of RMDQ change in either univariate or multivariate analyses.

Figure 1 shows the RMDQ change score for those with a dichotomized (high/low) RT3 VM/hr/wk change score from baseline to 3 months. The figure demonstrates that there was no difference in the RMDQ change score between the two groups (high and low RT3 change score)

Post-hoc analyses (presented below) were carried out in order to examine for potential correlations between baseline activity measures and disability and also measured activity levels and self-report in an attempt to explain and better understand the observed results.

There was no significant correlation between the observed measures of activity (RT3 VM/hr/wk) at baseline and RMDQ and VAS scores at baseline (Table 8).

There was no significant correlation between the observed measures of activity (RT3 VM/hr/wk) at baseline and 3 months, and the participant’s report of a return to full ‘normal’ activities at these two time points (Table 9).

The setting and choice of the MCID for the RMDQ is open to some debate. Previous research has demonstrated that baseline RMDQ scores to be relatively high in an ALBP population [58], and taking into account that a 30% change in score has recently been proposed as a meaningful change score for the RMDQ [42] we therefore conservatively set the MCID as a score of 4. Although there was a non-significant change in activity levels over this timeframe it is not known whether the choice of a smaller effect size (MCID < 4) and therefore a larger sample size requirement would have altered the observed results.

The units of activity employed in this research (RT3 VM/hr/wk), although extensively employed in field research as a measure of activity, have not been fully investigated for validity and responsiveness to change in free living research, an essential component of functional activity measures [59]. However, there is no acknowledged and standardised field research protocol for accelerometry use, and few instruments have been evaluated for their ability to reliability evaluate change in activity over time within LBP populations, which remains a weakness of PA measures within populations with disability [60]. Also, the current research employed a number of statistical models to analyse potential predictive relationships within the data sets (n = 83) and acknowledge that although the study was powered to assess such predictive relationships and suitable adjusted analyses performed there is a potential that over fitting of the statistical models may have occurred.

The report of a return to ‘normal’ activities was a dichotomous questionnaire variable and there is scope for a degree of subjectivity as to what patients perceived as their normal level of activity. Further exploration of the perception and return to ‘normal’ activities is warranted based on the findings from the current study.

The current study did not measure the type of physiotherapy intervention that the patient received, and how this might have influenced the patient’s activity and recovery; however, it was presumed that all therapists were adhering to physical activity guidelines in respect to the advice given to patients. Although recommendation exist for evidence based physiotherapy management of acute LBP [61], research demonstrates that there is much variability in the primary care management of LBP [62] in that not all primary care health practitioners adhere to current best practice guidelines. Therefore, further research is required to investigate whether specific therapeutic interventions have a greater effect on a patient’s activities during an episode of acute LBP

There is mixed evidence for activity behavior to change when being monitored [63, 64], and higher levels of activity recorded at baseline may be a result of the novelty of activity measurement. The two measurement points chosen mean that activity changes may have occurred outside this temporal window, perhaps as a result of treatment, and thus activity levels may have regressed to a mean level of activity by 3 months. A greater number of measurement points, coupled with the use of other activity measurement tools offer further opportunities to study the relationship between activity and LBP outcomes.

The strength and validity of the findings from this study are supported by powered sample size to detect a MCID in the RMDQ, coupled with the relatively high participant response rate at 3 months, and low levels of RT3 data loss. RT3 data loss was predominately due to RT3 factors such as malfunction and RT3 loss rather than participant factors, which means that data are likely to be completely missing at random and therefore unlikely to have caused any systematic bias in the results.