Background
Low back pain (LBP) is a highly prevalent [
1,
2] and costly [
3,
4] condition considered to cause more disability than any other disorder in the world [
5]. The vast majority of people will experience LBP at some point in their life and about two-thirds will experience recurrences [
4]. The individual course of LBP may follow a number of different trajectories [
6‐
8] and rather than the common categories of acute, sub-acute or chronic [
1] LBP could be described as a long-term recurrent condition. In about 90 % of patients suffering from LBP, no underlying spinal pathology or red flags can be identified, and their LBP is classified as non-specific [
4].
Due to the limited knowledge concerning the etiology of LBP, a multitude of treatments has been developed. At best, interventions have shown moderate effects in treatment outcome [
9]. One possible reason for not finding highly effective interventions for LBP is the heterogeneity of the condition [
4]. A recent literature review suggests that subgrouping patients according to genetic predisposition, psychological and activity related factors holds much promise and may be a suitable way to tailor treatments to yield better treatment outcomes [
10], identified in previous research as a priority [
11‐
13].
Psychological [
14,
15], behavioral [
16] and social factors [
17] have been associated with the transition from sub-acute to chronic pain [
18‐
21] and the bio-psycho-social model has become the leading theory in the management of LBP [
17,
22‐
24]. The West Haven-Yale Multidimensional Pain Inventory (MPI) is an instrument that has been widely used to measure and capture the chronic pain experience from the cognitive-behavioral perspective [
25]. The MPI instrument has been used to derive clinically meaningful clusters/subgroups [
26] named Adaptive Copers (AC), Interpersonally Distressed (ID) and Dysfunctional (DYS) which have been shown to be reliable, valid and sensitive to changes in treatment outcomes [
27,
28]. These subgroups have been used to investigate a number of chronic pain conditions including neck pain and LBP [
29‐
31], temporomandibular disorders [
32], headaches [
33], fibromyalgia [
34] and cancer pain [
35].
Attempts have been made to define psychological and behavioral factors that predict treatment outcome in chiropractic patients, but the results have been inconclusive [
36‐
40]. In a series of articles titled “The Nordic Subpopulation Program” a number of prognostic factors and their relation to treatment outcome have been investigated as a means of subgrouping patients. One of the most distinctive findings was that outcome at the 4
th visit was strongly predictive of the long-term outcome at 3 and 12 months [
41]. These results have been replicated in chiropractic LBP populations in Norway [
41], Finland [
42] and Sweden [
43]. Why some patients appear to respond better to treatment is not known, and further investigations are warranted. Psychological and behavioral factors may be what differentiate these patients.
The natural course of LBP has been studied and the results suggest that for a majority of patients a rapid reduction in pain occurs during the first few weeks following an acute episode [
44‐
47]. However many patients do not recover completely and show little further improvement past 3 months [
47‐
49]. The MPI-S subgroups have been shown to have different natural courses of LBP where the ID and DYS subgroups are more likely to have more persistent pain than the AC group [
30]. In the chiropractic setting it is possible to predict long-term improvement and non-improvement by the 4
th visit [
41]. If such a prediction was possible even earlier in the course of treatment (at the first visit) for those with a poor prognosis, extra resources might be allocated or a different approach altogether might be chosen. If the MPI-S subgroups were predictive of short-term (at 4
th visit) improvement they might also show similar properties long-term (at 3 and 12 months).
The aim of this study was to investigate if subgroup assignment at the 1st visit using the MPI-S instrument predicts the short-term clinical course among patients with recurrent and persistent LBP receiving chiropractic care.
Discussion
This is the first study to investigate if psychological/ behavioral profiles classified according to the MPI instrument can predict the short-term clinical course of persistent and recurrent LBP in a chiropractic primary care population. The main strength is the large sample and the use of valid instruments, thus the data are considered robust and reliable.
Our results suggest that the MPI-S subgroups could not predict short-term clinical course at the 4
th visit, despite previous research showing MPI-S to have predictive properties concerning long-term sick leave in chronic LBP populations [
30]. In the study from Bergström et.al [
30] they used absolute scores for sick leave as well as change scores for health related quality of life and found significant differences for the absolute scores but not the change scores. The DYS and ID groups reported significantly higher absolute values in pain intensity both at the 1
st and the 4
th visit which is in line with the previous study [
30]. The DYS and ID groups also reported the largest difference in pain levels between the 1
st and 4
th visit as well as the largest proportion of subjects with a clinically relevant difference. This indicates that the MPI-S instrument has worked well to classify the subjects into subgroups with significant differences of mean absolute values at the 4
th visit; however the change scores of subjective improvement and clinically relevant pain intensity failed to reflect a difference between the subgroups. Although the difference of absolute scores in pain intensity was an important finding the main focus of this article has been to investigate self-reported global improvement, thus change scores were the most appropriate choice of outcome.
The results show that this population is clearly affected by psychological and behavioral distress (MPI-S), however similar to other studies on chiropractic patients the clinical course in this population does not differ significantly between the MPI-S subgroups [
36‐
40]. The study population is a self-selected sample which may have resulted in a selection bias for individuals with a more favorable psychological profile, possibly with a higher degree of self-efficacy. As self-efficacy was not measured this will have to remain a speculation. Another possibility may be that the clinical encounter itself affects psychological factors and although the treatment is mainly physical, it inevitably includes strategies aimed at affecting non-physical factors [
40]. Field et.al [
40] found that patients’ mean scores of self-efficacy, fear avoidance beliefs and catastrophizing improved within a few days after the initial consultation with a chiropractor. A systematic review found evidence of the long-term effectiveness of SMT for chronic LBP [
62] and non-physical effects may very well be important mediators.
The STarT Back screening tool (SBT) is an instrument designed to stratify LBP patients according to modifiable risk factors for poor outcomes [
63] and has been shown to improve effect and cost-effectiveness of physiotherapy interventions by using stratified care [
64]. A previous study by Field and Newell found that the SBT could not predict the prognosis for LBP patients receiving chiropractic care at 30 and 90 days [
36], in line with the results in this study. In Field and Newell’s study one of the dependent variables were pain intensity (NRS-11) and they analyzed both the absolute values and change scores associated with a 2 point reduction, and their results were similar to the results reported here. Similarly Kongsted et al found in a prospective cohort study that the prediction of chiropractic patients’ individual treatment outcome (at 2 weeks, 3 months and 12 months follow-up) was not improved by using the SBT in a stratified treatment approach compared to treatment as usual [
65]. Thus, both the MPI-S and SBT seem unable to predict short-term outcome in the chiropractic population. It is important to note that the MPI-S instrument was not designed as a screening tool for fear avoidance, catastrophizing and self-efficacy as the SBT instrument, therefore a direct comparison may be misleading. However it is interesting to see that studies with instruments measuring psychological variables not captured with the MPI-S instrument have arrived at similar conclusions in that they seem to not be associated with treatment outcomes among chiropractic patients. An instrument such as the Örebro Musculoskeletal Pain Screening Questionnaire (ÖMPSQ) that includes a wider range of psycho-social factors may be more sensitive as a prediction tool for LBP in this population [
66].
Further, the MPI-S instrument was originally designed to measure the chronic pain experience and is perhaps not suitable for patients experiencing an acute flare-up. Possibly the acute pain disease-pathway affects the subjects differently compared to the chronic pain experience. Previous research has shown that patients who seek chiropractic care in Sweden mostly do so due to an acute episode/exacerbation of LBP (typically with a duration of less than 1 month) [
67]; their symptoms are likely to subside given the natural course with a regression towards the mean. Therefore our data may simply illustrate the natural course of the acute flare-up. Nevertheless, research has shown the long-term predictive value of the improvement measure at the 4
th visit to have strong correlations with outcomes at 3 and 12 months [
41‐
43].
All subjects in the study sample had recurrent pain and a total of more than 30 days with pain in the previous year. A large proportion (62 %) also had comorbidity with neck and thoracic pain. Therefore chronicity may also have confounded the results as a chronic condition may need more time to respond to an intervention. A long term follow up might reveal different predictive results. Even though the MPI instrument could not predict the short term clinical course, the instrument successfully classified subjects according to psychological and behavioral subgroup profiles. This information may be clinically relevant for other aspects of the patient encounter such as the indication for co-management with other health professionals (e.g. psychologists) for the DYS individuals or involving significant others (spouse, relative or friends) in the treatment plan for the ID individuals. Longitudinal research with a stratified care model in this way could answer these questions.
The study has some limitations. Most important is the possible selection bias from the dropout at the 4th visit (47 % of the study population). Reasons for not completing follow-up may include a fast recovery warranting no further treatment or negative reactions to treatment – both of which may have resulted in discontinuation of the treatment plan. Also the administrative procedures at the clinic may have failed, so that the administration of the 4th visit questionnaire may have been forgotten, contributing to attrition. However, when comparing the descriptive data and the cluster distribution of the dropouts they were similar to the subjects who completed follow-up (results not shown) and the risk of attrition bias is considered to be low. The “perceived chance of improvement” which was slightly higher (and significantly different) in the subjects who completed follow-up may have overestimated the results in a non-differential manner. As in many other outcome studies without a control group, unknown confounders may have biased the results. Therefore randomized controlled longitudinal studies which take into account the long- and short-term effects should be conducted to confirm or reject the results from this study.
As the follow-up data was not recorded at a fixed time point, but during the 4
th visit, the data may be confounded by the variability in the time of the follow-up period. As this information was not collected, it is difficult to say to what extent this may have affected the results. However, as previous research have shown the predictive value (regarding long term outcomes at 3 and 12 months [
41]) of improvement at the 4
th visit (irrespective of when this occurs) it is unlikely that this issue have affected the conclusions to any greater degree.
As some of the descriptive data (age and activity limitation) was recorded at the start of the RCT, they were available only for approximately half of the subjects; which is another limitation of the study.
Competing interests
The authors declare that they have no competing interests.
Authors’ contribution
AE has been the main author and drafted the manuscript. IA has supervised and taken part in most aspects of the study. GB has been consulted as an expert in advising on matter of study design, epidemiology and psychology. With regards to epidemiology and statistics LB has served as an expert consultant. All authors have been involved with the design of the study and critical revision of the manuscript. All authors have read and approved the final manuscript.