Strengths and limitations of this study
-
This longitudinal observational study was the largest cohort to date investigating beliefs about low back pain (LBP) (n=2734).
-
The cohort provided an opportunity to investigate associations in acute episodes of LBP as well as in long-lasting LBP.
-
The Back Belief Questionnaire (BBQ) is a widely used questionnaire that has previously shown good reliability but has not been tested in the Danish version, so we assessed its construct validity and scale reliability before conducting the primary analysis.
-
The BBQ mainly measures beliefs regarding negative consequences of LBP, thus neglecting other potentially relevant aspects of beliefs.
-
The cohort only consisted of chiropractic patients with generally positive beliefs and is thus not generalisable to all patients with LBP.
Introduction
Low back pain (LBP) is a common condition that is mostly non-specific, which means no single structure can be identified as the cause of the pain.1 Both biophysical, psychological and social factors are recognised to contribute to pain perception and disability.2 Among these biopsychosocial factors, one aspect that is considered important in relation to both disability and recovery is what people think and believe about their back and LBP.3 This could involve beliefs that LBP is a sign of structural damage and, consequently, the back is fragile and needs protection. Such beliefs can affect the behaviour of a person with LBP, and thereby influence recovery if a person adopts unhelpful behaviour such as fear-avoidance behaviour or overprotective behaviour.3–7
Multiple questionnaires have been developed to measure beliefs about pain and investigate the association between beliefs and LBP. A systematic review of back beliefs in the general population from 2018 found that negative beliefs, measured using the Back Belief Questionnaire (BBQ), were cross-sectionally associated with higher levels of pain and disability.8 Similarly, a systematic review from 2018 found a moderate level of evidence for a cross-sectional association between maladaptive illness perceptions, measured by the Illness Perception Questionnaire (IPQ), and pain intensity and disability in patients with musculoskeletal pain.9 The evidence regarding the prognostic value of illness beliefs was inconclusive due to lack of longitudinal studies.9 However, a recent longitudinal study from 2021 found that the IPQ only added a small and non-substantial predictive value for poor recovery at 3 months in people with musculoskeletal pain.10 For recovery expectations as a prognostic factor for LBP, a Cochrane review from 2019 concluded that having positive expectations towards recovery might be associated with a reduction in pain and disability, although the evidence was of low quality.11 In general, there is evidence supporting a cross-sectional association between negative beliefs regarding LBP and higher levels of pain and disability. However, as longitudinal studies are few and of low quality and mostly investigate recovery expectations, the relationship between other aspects of beliefs and clinical outcomes over time is uncertain.8 9 11–15
Longitudinal studies can help to determine if specific beliefs are associated with clinical outcomes, which is relevant as beliefs are potentially modifiable and could therefore be targets for clinical interventions. It has been proposed that the association between psychological factors, such as beliefs, and long-term disability might be more relevant for those with persistent pain compared with those with subacute pain.16 A verification of this theory would be clinically relevant as it could help clinicians prioritise when to address beliefs.
The objectives of this study were therefore to investigate if back beliefs at baseline, measured by the BBQ, were associated with pain intensity and disability at the 2-week, 13-week and 52-week follow-ups in patients with LBP who consulted a chiropractor, and whether the association differed according to pain duration. Also, we assessed if any items of the BBQ had a stronger association with pain intensity and disability at the 13-week follow-up compared with the other items.
Methods
Study design
This study was an observational cohort study based on data from the Danish Chiropractic Low Back Pain Cohort (ChiCo).17 The study was reported according to the STROBE (Strengthening the Reporting of Observational studies in Epidemiology) cohort reporting guidelines, and a STROBE checklist has been completed.18
Patient and public involvement
Patients were not involved in designing the study or interpreting the results.
Setting and procedures
Participants were recruited from 10 chiropractic clinics in Denmark between November 2016 and December 2018. At the initial visit to the chiropractor, the patient filled out a baseline questionnaire, divided into two parts. The first part included items that might be influenced by consulting the chiropractor and was therefore filled out before the initial consultation (baseline 1). The second part was filled out after the initial consultation and included demographic and background data less likely to be influenced by the consultation (baseline 2). Follow-up questionnaires were obtained at 2, 13 and 52 weeks after inclusion. Participants who did not respond to the follow-up questionnaires at 13 and 52 weeks received a phone call for a structured interview on a limited number of questions from the survey. Data were collected electronically and stored using the online system REDCap (Research Electronic Data Capture) hosted and supported by the Odense Patient data Explorative Network. Further details on the data collection procedure have been described elsewhere,17 as have cross-sectional data from the BBQ in some of the study sample.19
Participants
To be enrolled in the study, the patient needed to be 18 years of age or older, be seeking a consultation with the chiropractor with a new onset of LBP with or without leg pain, and be able to complete electronic questionnaires in Danish. A new onset of LBP was defined as a new or recurring LBP problem for which the patient was not currently receiving treatment or long-term management. Patients referred for acute surgical assessment or patients with suspicion of pathology leading to referral for further diagnostic assessment were not enrolled in the study.17
Variables
Primary measures
Beliefs about LBP were measured at baseline 1, before consulting the chiropractor, using a Danish version of the BBQ. The BBQ consists of 14 statements regarding inevitable negative consequences of LBP that are scored on a 5-point Likert scale. Five statements are not included in the final score, and thus the score ranges from 9 to 45. The scores are reversed so that higher scores indicate positive beliefs.20 The translation process has been described in a previous paper.19 The questionnaire has been widely used in research and has previously been validated and translated into multiple languages, showing good test–retest reliability and demonstrating good construct validity (measuring only one construct).21–25
Disability was measured by the 23-item Danish version of the Roland-Morris Disability Questionnaire (RMDQ) (0–100, higher scores indicating higher levels of disability),26 and LBP intensity on a Numerical Rating Scale (NRS) examining typical LBP intensity during the previous week (0=no pain to 10=worst imaginable pain).24 25 Both disability and LBP intensity were measured at baseline 1 (before the consultation), and at the 2-week, 13-week and 52-week follow-ups. Only LBP intensity was part of the telephone interview with non-respondents.
Additional baseline variables
Baseline 1: age and sex (derived from the patient’s personal identification (social security) number); duration of current pain episode (1–2 days, 3–7 days, 1–2 weeks, 2–4 weeks, 1–3 months, 3–12 months, more than a year).
Baseline 2: previous treatment for LBP (yes/no); previous episodes of LBP (none, 1, 2–3, more than 3); number of days with LBP last year (≤30 days, >30 days).
Statistical methods
Missing responses on the BBQ and previous treatment for LBP were imputed using chained multiple imputations. For BBQ, we excluded participants who answered six or fewer items at baseline, and then used imputation for the remaining incomplete questionnaires. For both BBQ and previous treatment for LBP, the imputations were informed by age, sex, RMDQ scores, LBP intensity at baseline, duration of current pain episode, previous treatment and number of days with pain last year. Multiple imputations of missing RMDQ sum scores were performed as part of the standard preparation of ChiCo data.17
Construct validity and scale reliability
Before conducting the analyses, we tested the construct validity and scale reliability of the Danish version of the BBQ. The scale showed acceptable reliability (Cronbach’s alpha=0.77), but our findings did not support a unidimensional structure of the scale. However, as we were unable to detect a better factor structure of the scale, we decided to use the scale as originally intended and as it had been applied in previous studies. The process is described in online supplemental file 1.
Supplemental material
Data analysis
Baseline characteristics were presented as means with SDs or proportions.
To estimate associations between BBQ and outcomes, we used a linear mixed model with random intercept (taking repeated measures into account) to conduct longitudinal regression analysis with baseline BBQ score, follow-up time point (categorical), and the interaction between the BBQ score and follow-up time point as independent variables. This model was used for both LBP intensity and RMDQ score as the dependent outcome variable. We performed unadjusted analyses and adjusted analyses controlling for age, sex, baseline LBP intensity, baseline RMDQ score and previous treatment. Variables controlled for were chosen as they have been shown in a previous study on the same population to be associated with baseline BBQ scores.19 Results were presented as coefficients with p values and 95% CIs. P values for the interaction term were based on analysis of variance tests using the ‘contrast’ command in STATA. A linear relationship was assumed between baseline BBQ and both LBP intensity and RMDQ score at follow-up, based on inspection of a Locally Weighted Scatterplot Smoothing plot. For visualisation of the findings, the adjusted analyses were repeated with BBQ scores divided into quartiles, which were used to create a marginsplot for the association. The quartiles of BBQ scores had the following division: scores 9–29 (n=846), 30–32 (n=525), 33–37 (n=823) and 38–45 (n=540). Results were presented as regression coefficients with 95% CIs and p values.
To investigate if the association differed according to pain duration, the analyses with the outcomes on the original scales were repeated, stratified on the following four groups categorised by duration of the current episode and number of previous LBP episodes: group 1 (acute new): onset within 2 weeks and no previous LBP episodes; group 2 (acute episodic): onset within 2 weeks but with one or more previous LBP episodes; group 3 (subacute): pain for more than 2 weeks but less than 3 months; and group 4 (long-lasting): pain for more than 3 months.
To explore the association between single items of the BBQ and LBP intensity and RMDQ, we performed a linear regression analysis with LBP intensity or RMDQ at the 13-week follow-up as the dependent variable and each BBQ item at baseline as independent variables, controlling for age, sex, baseline LBP intensity, baseline RMDQ score and previous treatment. The 13-week follow-up was chosen based on inspection of the overall change in LBP intensity and disability at follow-up, as most of the change had occurred by 13 weeks. All items were included in one model for each outcome and results were presented as regression coefficients with 95% CIs and p values. Variance inflation factors (VIFs) were calculated to check the influence on estimates from multicollinearity. With LBP intensity as the dependent variable, the mean VIF was 1.28 (range 1.06–1.57) and with RMDQ score as the dependent variable, the mean VIF was 1.27 (range 1.06–1.56). Thus, both models indicated no sign of multicollinearity.
The impact of single items on the amount of variance explained was explored by noting the reduction in the R2 value obtained from the linear regression model with a single item removed from the model at a time compared with a model with all items.
All analyses were performed using Stata/MP V.16 (StataCorp, Texas, USA).
Results
A total of 3165 participants were included in the ChiCo, and of those, 2734 were included in the current study (figure 1). Mean age was 44 years and 41% were female. The mean baseline score of LBP intensity was 6.7 and the mean RMDQ score was 55 (table 1). Follow-up data on LBP intensity were available for 72%, 69% and 65% (at 2, 13, 52 weeks, respectively) of the participants, and data on disability were available for 71%, 61% and 52% (at 2, 13, 52 weeks, respectively) (figure 1). Baseline characteristics were similar regarding pain intensity, RMDQ scores and BBQ scores between participants who completed the 52-week follow-up and those who were lost to follow-up, but those not completing the follow-up were younger than those who did (table 1).
Characteristics of study population
Flow chart of the study population. Explanatory text: partly completed data on BBQ or RMDQ were filled out using chained multiple imputations. BBQ, Back Belief Questionnaire; LBP, low back pain; RMDQ, Roland-Morris Disability Questionnaire.
The association between BBQ scores at baseline and LBP intensity and disability after 2, 13 and 52 weeks
Higher BBQ scores at baseline, indicating positive back beliefs, were weakly associated with lower LBP intensity at follow-up in both unadjusted and adjusted analyses (table 2).
Association between back beliefs at baseline and LBP intensity and disability at follow-up
The coefficient of the interaction between BBQ and LBP intensity over time denotes the additional reduction in LBP intensity for each additional point on the BBQ scale. This means that if two participants are alike on all parameters except that one scores 10 points higher on the BBQ at baseline, then that patient would be expected to have an additional reduction in LBP intensity at 13 weeks of −0.3 points (10×−0.03 (13-week coefficient)) compared with the other participant.
The association between quartiles of BBQ at baseline and LBP intensity at follow-up indicated higher reduction of LBP intensity for patients with the most positive beliefs compared with those with more negative beliefs (figure 2).
Marginsplot of the associations between quartiles of BBQ scores at baseline and LBP intensity at follow-up. Explanatory text: the association between quartiles of BBQ scores at baseline and LBP intensity at follow-up had a p value of 0.0030. BBQ, Back Belief Questionnaire; LBP, low back pain; NRS, Numerical Rating Scale.
Associations between BBQ at baseline and disability at follow-up were weak and had large p values (table 2). The association is visualised in a marginsplot in figure 3.
Marginsplot of the associations between baseline quartiles of BBQ scores at baseline and disability at follow-up. Explanatory text: the association between quartiles of BBQ scores at baseline and RMDQ scores at follow-up had a p value of 0.1071. BBQ, Back Belief Questionnaire; RMDQ, Roland-Morris Disability Questionnaire.
The association between BBQ scores and LBP intensity and disability stratified by LBP history
Dividing the populations into groups based on episode duration and number of previous episodes (‘acute new’ n=209, ‘acute episodic’ n=932, ‘subacute’ n=615 and ‘long-lasting’ n=473) did not show any substantial difference between the groups in the associations between BBQ at baseline and LBP intensity or disability at follow-up. The results are shown in online supplemental file 2.
Supplemental material
The association between single items on the BBQ and LBP intensity and disability at 13 weeks
Higher scores on an item (more positive beliefs on a scale from 1 to 5) were generally associated with slightly lower LBP intensity and disability scores at 13 weeks (table 3).
Single-item association with LBP intensity or disability at 13 weeks
Item 3 ‘Back trouble means periods of pain for the rest of one’s life’ had the strongest association with a reduction in both LBP intensity and disability at 13 weeks. For LBP intensity, the coefficient was −0.29 (95% CI −0.4 to −0.19, p<0.001) and for disability, −2.42 (95% CI −3.52 to −1.33, p<0.001).
For LBP intensity, the second strongest association was with item 11 ‘Medication is the only way of relieving back trouble’ (coefficient −0.16, 95% CI −0.28 to −0.04, p<0.007). For disability, the second strongest association was with item 9 ‘Alternative treatments are the answer to back trouble’ (−1.31, 95% CI −2.36 to −0.26, p=0.015) (table 3).
When removing one item at a time from the model, the reduction in R2 was low for all items. Item 3 showed the greatest reduction in R2 accounting for 1.5% of the explained variance in the association with LBP intensity and 1% in the association with disability. Among the other items, the variance explained ranged from 0% to 0.35%.
Discussion
Main findings
To our knowledge, this is the first study using longitudinal data to investigate if back beliefs, measured by the BBQ, are associated with LBP intensity and disability at follow-up in patients with LBP who consult a chiropractor. Overall, we found that more positive beliefs at baseline were associated with decreasing LBP intensity at follow-up. However, the coefficients were small, and thus might not be of clinical relevance. There was no certain association between back beliefs and disability outcomes. The associations were not substantially different between groups with different LBP history. Assessment of the individual BBQ items showed that the item ‘Back trouble means periods of pain for the rest of one’s life’ had the strongest association with a reduction in both disability and LBP intensity at 13 weeks.
Interpretation
The BBQ is designed to measure beliefs regarding negative consequences of LBP.20 Based on the Common-Sense Model (CSM), beliefs regarding consequences represent one particular type of health beliefs. The CSM depicts how beliefs about LBP potentially affect disability as it explains how individuals respond to and manage health threats based on the way pain or stimuli related to illness is understood. The representation of health threats is described in five different domains: identity (what is this pain?), cause (what caused this pain?), consequence (what consequences will this pain have?), control (how can I control this pain?) and timeline (how long will this pain last?).3 4 6 27 It is our interpretation that the questions in BBQ primarily reflect the consequence domain, yet not entirely. Our findings indicated that perceptions related to consequences are not strongly related to outcomes in this population, whereas one item related to timeline (‘Back trouble means periods of pain for the rest of one’s life’) had a noticeably stronger association with LBP intensity and disability at the 13-week follow-up compared with the other items. This might imply that recovery expectations are an important subdomain in the BBQ, which is in line with the finding from other studies reporting that recovery expectations are a predictor of prognosis for LBP.11
The consequence domain was reported in a systematic review to be a prognostic factor for pain outcomes in people with musculoskeletal pain.9 The review investigated relationships of illness perceptions, pain intensity and disability in people with musculoskeletal pain. However, only two of the included studies focused on LBP in a longitudinal design and both these studies only investigated outcomes of disability.9 Nevertheless, both studies found maladaptive illness perceptions to be associated with worse outcomes regarding pain-related disability at follow-up, whereas our findings did not provide such evidence.28 29 Similar to our findings, a prospective cohort study (2020) of people with acute LBP found that maladaptive illness perceptions measured by IPQ were predictive of pain but not disability at 12 weeks although the predictive value was low.30 The same trend was seen for musculoskeletal pain, where IPQ did not add substantially to the prediction of recovery.10 Similarly, a secondary analysis of a randomised controlled trial published in 2018 showed that high levels of fear-avoidance beliefs measured by the Fear-Avoidance Beliefs Questionnaire in patients with LBP were only weakly associated with worse outcomes in LBP and disability at 12 months, yet the association was much stronger for sick leave.31 However, both the IPQ and Fear-Avoidance Beliefs Questionnaire cover more domains than the BBQ and the results are therefore not directly comparable.
It is questionable as to whether the observed association between positive back beliefs and the reduction in LBP intensity is clinically relevant. There is not a generic meaningful minimal clinically important change for pain scores, as it is always content specific,32 33 but a change of 2 points on the NRS has been proposed to be clinically significant in people with long-lasting LBP.34 In our study, a 10-point higher score on the BBQ translated into an expected additional reduction in LBP intensity of 0.3 points which we doubt to be clinically relevant. However, as LBP is complex and many factors are considered important contributors, it is unlikely that the BBQ score would be able to independently predict future LBP intensity with high precision in a one-size-fits-all model. BBQ scores were generally high (mean BBQ sum score=32) indicating overall positive beliefs, and more important associations can perhaps be demonstrated in populations with a larger variation in back beliefs.
Overall, beliefs about LBP seem to be associated with pain intensity and disability at a cross-sectional level, but the longitudinal relationship remains unclear.8 9 11–15 Due to only weak associations between beliefs and reduction of LBP intensity, and uncertainty regarding the domains of beliefs measured, the BBQ does not seem suitable for predicting or explaining the course of LBP in our setting. However, based on the cross-sectional association, and as other domains of beliefs could be relevant to patients with LBP, we still encourage clinicians to address beliefs with their patients preferably using an individual approach.
In our sample, the associations between back beliefs and LBP intensity and disability were not influenced by the number of previous pain episodes and the duration of pain. This finding contradicts the theory that the association between beliefs and disability is most relevant for those with persistent pain.16 This is important as it implies that the decision to discuss beliefs with a patient should not be based on the duration of pain or number of previous LBP episodes.
Limitations
As discussed previously, BBQ focuses on the consequence domain of beliefs. For a more thorough investigation of the association between beliefs and clinical outcomes, the use of different questionnaires could add information on beliefs from other domains, and thereby give a broader perspective on potential associations.
Before conducting the primary analyses, the construct validity and scale reliability of the BBQ were evaluated. The internal consistency and scale reliability were considered acceptable, and in line with other studies.20 21 24 However, other studies have found the BBQ to be unidimensional, although the fit of item 1 has been questioned, which we could not confirm.22–24 When interpreting the results, it should therefore be kept in mind that it is unclear what constructs the BBQ sum score represents in this sample. Another consideration is that the BBQ might be outdated as it was created in 1996 and a lot has happened in the field of LBP since then and perhaps in the public perception of LBP. This may explain why a questionnaire from 2014, the Back Pain Attitudes Questionnaire, which was developed based on in-depth interviews with people experiencing LBP, asks questions very different from those of BBQ.35 For future studies investigating beliefs about LBP, we recommend researchers carefully consider the suitability of the different instruments.
This study did not account for the treatments the patients received from the chiropractor (eg, advice, education, exercise, manual therapy), and it is unknown to what extent beliefs were discussed and addressed as part of treatment in a way that potentially affected outcomes. This could have blurred an otherwise stronger association than observed. However, BBQ sum scores were previously observed to be relatively constant over time in this sample, suggesting that negative beliefs were not effectively changed after initiating care.19
In this study, we explored the prognostic effect of baseline beliefs. In addition, it would be relevant investigating if changes in beliefs as a result of a healthcare consultation mediate treatment effects. However, our sample would not be very suitable for this purpose as beliefs were generally positive at baseline, and optimally it would require a randomised design.
Generalisability
Data were collected from a limited number of chiropractic clinics in Denmark, yet we have no reason to believe that data were not representative of Danish chiropractic clinics in general. Demographic baseline data were similar to a previous Danish chiropractic cohort based on a national sample.36 However, a population of patients consulting a chiropractor cannot be fairly compared with other patients in primary care.36 Further, the study sample’s overall positive beliefs with a mean BBQ sum score of 32 differ from the findings from a systematic review that found the majority of mean BBQ sum scores in the general population were below 27.8 Also, a recent study from 2021 exploring back beliefs in the general population reported a mean BBQ sum score of 27.37
Conclusion
Positive beliefs regarding LBP at baseline, measured by the BBQ, were weakly associated with a reduction in LBP intensity but not disability at the 2-week, 13-week and 52-week follow-ups in people with LBP seeking chiropractic care. Whether the association with LBP intensity was clinically relevant is questionable. The BBQ is therefore not promising for predicting or explaining the course of LBP in this setting. Future research should focus on exploring the associations between beliefs and clinical outcomes in different patient populations and with instruments covering all pain belief domains or more unambiguously covering a single domain.
Data availability statement
Data are available upon reasonable request. Data used and analysed in the current study are available from the corresponding author on reasonable request. If interested in using data from the ChiCo for other research projects, the Chiropractic Knowledge Hub should be contacted directly.
Ethics statements
Patient consent for publication
Ethics approval
This study involves human participants but the Health Research Ethics Committee for Southern Denmark determined (S-20,162,000-109) that the Danish Chiropractic Low Back Pain Cohort did not require ethical approval according to Danish regulations. Participants gave informed consent to participate in the study before taking part.
This post was originally published on https://bmjopen.bmj.com