Multidisciplinary integrative care versus chiropractic care for low back pain: a randomized clinical trial

Background Low back pain (LBP) is influenced by interrelated biological, psychological, and social factors, however current back pain management is largely dominated by one-size fits all unimodal treatments. Team based models with multiple provider types from complementary professional disciplines is one way of integrating therapies to address patients’ needs more comprehensively. Methods This parallel group randomized clinical trial conducted from May 2007 to August 2010 aimed to evaluate the relative clinical effectiveness of 12 weeks of monodisciplinary chiropractic care (CC), versus multidisciplinary integrative care (IC), for adults with sub-acute and chronic LBP. The primary outcome was pain intensity and secondary outcomes were disability, improvement, medication use, quality of life, satisfaction, frequency of symptoms, missed work or reduced activities days, fear avoidance beliefs, self-efficacy, pain coping strategies and kinesiophobia measured at baseline and 4, 12, 26 and 52 weeks. Linear mixed models were used to analyze outcomes. Results 201 participants were enrolled. The largest reductions in pain intensity occurred at the end of treatment and were 43% for CC and 47% for IC. The primary analysis found IC to be significantly superior to CC over the 1-year period (P = 0.02). The long-term profile for pain intensity which included data from weeks 4 through 52, showed a significant advantage of 0.5 for IC over CC (95% CI 0.1 to 0.9; P = 0.02; 0 to 10 scale). The short-term profile (weeks 4 to 12) favored IC by 0.4, but was not statistically significant (95% CI − 0.02 to 0.9; P = 0.06). There was also a significant advantage over the long term for IC in some secondary measures (disability, improvement, satisfaction and low back symptom frequency), but not for others (medication use, quality of life, leg symptom frequency, fear avoidance beliefs, self-efficacy, active pain coping, and kinesiophobia). Importantly, no serious adverse events resulted from either of the interventions. Conclusions Participants in the IC group tended to have better outcomes than the CC group, however the magnitude of the group differences was relatively small. Given the resources required to successfully implement multidisciplinary integrative care teams, they may not be worthwhile, compared to monodisciplinary approaches like chiropractic care, for treating LBP. Trial registration NCT00567333.


Background
Low back pain (LBP) is one of the most prevalent and disabling chronic health conditions. An estimated 40-80% of adults experience low back pain (LBP) at some point in their lives [1,2]. Further, LBP related disability continues to increase, making it a leading cause of disability globally [3]. Approximately 20% of acute cases become chronic [4], and it is these individuals that bear a disproportionate share of LBP associated burden [5]. Importantly, LBP is one of the stronger risk factors for chronic opioid use [6].
While the 'biopsychosocial model' for LBP has been promoted for decades, it is still incompletely and inadequately applied in both research and clinical practice [7][8][9][10][11]. Indeed, the majority of back pain cases remain poorly treated with a heavy emphasis on symptom management [11] using a 'one size fits all' approach that fails to address sufferers' unique needs [7,[11][12][13]. This has resulted in the persistent use of marginally effective and potentially harmful unimodal therapies (injections, drug therapies, etc.) with a primarily physical focus. Further, current back pain management practices often contradict clinical guideline recommendations by failing to offer treatment options with scientific support, including complementary approaches [13][14][15][16]. This includes spinal manipulation, exercise, acupuncture, cognitive behavioral therapy, selfcare strategies, and others [15,17,18].
Integrating complementary modalities with conventional approaches has shown promise for LBP in previous studies [19]. Team based models of care with multiple provider types from complementary professional disciplines has been one way of integrating different therapies to more comprehensively address individual patients' needs [11,20]. Such approaches combine a range of viable treatment options to synergistically address multidimensional causes of pain, with the goal of exceeding the therapeutic effect of any one therapy alone [21,22]. A previous manuscript by our group described one approach for a team based model of care including acupuncturists, chiropractors, psychologists, exercise therapists, massage therapists, primary care physicians with case managers coordinating care [23].
The purpose of this manuscript is to report the primary and secondary clinical outcomes of a randomized trial of monodisciplinary chiropractic care (CC), versus multidisciplinary integrative care (IC) for sub-acute and chronic LBP.

Methods
This was a parallel group randomized clinical trial funded by the US Department of Health and Human Services. It was conducted from May 2007 to October 2009 with follow-up data collection through August 2010.
Institutional Review Boards at participating institutions (Northwestern Health Sciences University Study #1-32-10-06 and Minneapolis Medical Research Foundation Study # 07-2785) approved the study protocol which has been described elsewhere [24]. The study was monitored by a Data Safety and Monitoring Board. Written consent was provided by all participants.

Settings and participants
The study was conducted at Northwestern Health Sciences University (Minneapolis, Minnesota). Participants were recruited from the surrounding Minneapolis/ St. Paul metropolitan area primarily through targeted postcard mailings, brochures at community events, and advertisements in online local newspapers.

Inclusion criteria
Inclusion criteria were 18 years of age or older, LBP categories 1, 2, 3, or 4 according to the Quebec Task Force classification system (individuals with back pain, stiffness, or tenderness with or without leg pain or neurological signs) [25]; current episode of LBP 6 weeks or longer in duration; and LBP rating of ≥ 3 on a 0-10 scale during the previous week.

Exclusion criteria
Individuals were excluded if they had contraindications to study treatments (i.e. active inflammatory disease of the spine, blood clotting disorders, or severe osteoporosis) or who were pregnant or nursing, had current or pending spine-related litigation, a history of multiple lumbar surgeries, or progressive neurological deficits.

Randomization
The study statistician utilized a computer-generated randomization scheme and applied a 1:1 allocation ratio with randomly permuted block sizes that was concealed from the study team. As individuals became eligible, sequentially numbered sealed, opaque envelopes were drawn in consecutive order and opened in the presence of the study participant by the study team.

Interventions
All participants in the study received 12 weeks of either monodisciplinary chiropractic care (CC) or multidisciplinary team-based integrative care (IC). CC was delivered by a team of chiropractors allowed to utilize any non-proprietary treatment under their scope of practice not shown to be ineffective or harmful including manual spinal manipulation (i.e., high velocity, low amplitude thrust techniques, with or without the assistance of a drop table) and mobilization (i.e., low velocity, low amplitude thrust techniques, with or without the assistance of a flexion-distraction table). Chiropractors also used hot and cold packs, soft tissue massage, teach and supervise exercise, and administer exercise and self-care education materials at their discretion. IC was delivered by a team of six different provider types: acupuncturists, chiropractors, psychologists, exercise therapists, massage therapists, and primary care physicians, with case managers coordinating care delivery. Interventions included acupuncture and Oriental medicine (AOM), spinal manipulation or mobilization (SMT), cognitive behavioral therapy (CBT), exercise therapy (ET), massage therapy (MT), medication (Med), and self-care education (SCE), provided either alone or in combination and delivered by their respective profession. Participants were asked not to seek any additional treatment for their back pain during the intervention period. Standardized forms were used to document the details of treatment, as well as adverse events. It was not possible to blind patients or providers to treatment due to the nature of the study interventions. Patients in both groups received individualized care developed by clinical care teams unique to each intervention arm. Care team training was conducted to develop and support group dynamics and shared clinical decision making. A clinical care pathway, designed to standardize the process of developing recommendations, guided team-based practitioner in both intervention arms. Evidence based treatment plans were based on patient biopsychosocial profiles derived from the history and clinical examination, as well as baseline patient rated outcomes. The pathway has been fully described elsewhere [23]. Case managers facilitated patient care team meetings, held weekly for each intervention group, to discuss enrolled participants and achieve treatment plan recommendation consensus. Participants in both intervention groups were presented individualized treatment plan options generated by the patient care teams, from which they could choose based on their preferences.
To assess response to treatment during the intervention phase, patients completed a Patient Self-Assessment Form (PSAF) at each of their visits which was adapted from the Measure Yourself Medical Outcome Profile [26]. Patients chose a symptom and an activity most affected by their LBP, and then rated it on a 0-10 scale. Treating providers monitored patient progress by assessing patients' PSAF in relation to benchmarks for improvement generated from previous studies [23]. When benchmarks for improvement were not met, providers brought the case back to their respective care team for review and potential alteration of the treatment plan. Table 1 describes the specific details of the treatments using the Template for Intervention Description and Replication (TIDieR) [27].

Outcomes measures
Participant demographic and clinical characteristics were collected during the baseline visits through selfreport questionnaires and a health history and physical examination. Self-reported outcomes were collected at 2 baseline visits (7-14 days apart) and at 4,8,12,26, and 52 weeks after enrollment using questionnaires administered independent of staff or clinician influence.

Power calculation and sample size
In a previous chronic LBP trial of exercise and spinal manipulation conducted by our team, group differences in pain of up to 8 percentage points after 3 months of treatment were observed. Informed by these results, the scientific literature at the time, and consensus of study investigators and clinicians regarding clinical importance, we were interested in detecting an 8-percentage point between group difference in pain after 12 and 52 weeks of treatment. Based on an α of 0.05 and 80% power, 85 participants per group were required. An allowance of 15% drop-out rate resulted in a total sample size of 200 participants.

Statistical analysis
We used an intention-to-treat approach, analyzing all observed data from participants according to their allocated treatment assignment. Data were prepared for analysis by a data manager masked to group status; analyses were performed in SAS, version 9.1. The primary and most secondary outcomes were analyzed using linear mixed effect models including fixed effects for time, treatment, and a time-by-treatment interaction in addition to a random intercept to account for within-subject correlation. Hierarchical linear models are a robust method for analyzing ordinal outcomes using a Likert scale [47]. Secondary outcomes collected only at week 12 (i.e. pain management inventory) were analyzed using linear regression. Generalized estimating equations were used to analyze missed work and reduced activity days. The Binomial family was used to analyze the proportion of subjects with any missed work or reduced activity and the Poisson family was used to analyze the number of missed or reduced days. All models included the baseline outcome as a covariate except for analyses of improvement and satisfaction where baseline measures are not applicable.

Primary outcome measure analysis
The primary outcomes were short-term (4 to 12 week) and long-term (4 to 52 weeks) group differences in pain intensity derived from the linear mixed effect model. We used Fisher's protected least significant difference approach to control for multiplicity [48]. An area under the curve minus baseline summary measure was used as the omnibus test to determine if the long-term pain profile (including 4, 12, 26, and 52 weeks) was different between groups [49,50]. The omnibus test needed to be significant (p-value ≤ 0.05) for group differences in the short term (weeks 4 through 12) to be determined. Clinical and demographic variables were included as covariates if they were at least moderately correlated (|r|≥ 0.5) with change in outcomes [51]. Linear mixed effect model analyses provide unbiased estimates when data are missing at random [52]. The pattern and reasons for missing data were assessed to determine whether sensitivity analyses to address data missing not at random were required. In addition, a sensitivity analysis including patient expectations as a covariate was conducted to assess impact on study results. Secondary analyses of the primary outcome measure included group differences at weeks 4, 12, 26, and 52. Additionally, responder analyses for no pain reduction, or pain reductions of 30% (minimal improvement), 50% (moderate improvement), 75%, and 100% (substantial improvement) were performed at weeks 12, 26, and 52 [53]. Differences in proportions of responders between groups were calculated and 95% confidence intervals were analyzed using the Wilson method for risk differences [54]. Cumulative responder analysis graphs were created to display the proportion of responders for all possible levels of pain reduction [55]. Differences in cumulative response curves were assessed by determining the area under the response curve using the trapezoidal rule and 95% confidence intervals were calculated using bias-corrected bootstrapping with 5000 iterations [56].

Secondary outcome measure analysis
Analyses included group differences at the relevant individual time points for all measures, in addition to shortterm (including 4 and 12 week outcomes) and long-term (including all time points) profiles when possible.

Baseline characteristics
A summary of study participants is provided in Fig. 1. A total of 384 participants were assessed for eligibility and 201 were enrolled. Demographic and baseline clinical characteristics are provided in Table 2. On average, participants were just over 50 years of age and had pain that was chronic in nature (8 to 9 years); low back pain intensity was moderate (~ 5 on a 0-10 scale) as was disability level (~ 40 percent, 0-100). The two groups were comparable at baseline in terms of demographic and clinical characteristics. Pain intensity was the only baseline clinical or demographic characteristic that was moderately correlated with changes in pain intensity and was included as a covariate in the primary analysis. Expectation of improvement was very weakly correlated with change in pain intensity (r = − 0.13 or weaker).

Treatment frequency and adherence
Overall, 96% of study participants attended treatment visits as recommended; 93% for the CC group and 98% in the IC group. The mean number of visits in the CC group was 18.1 and for the IC group, was 23.8. Participants in the IC group received the following types of care: exercise therapy (ET, n = 96); self-care education (SCE, n = 59); traditional Chinese medicine (TCM, n = 51); massage therapy (MT, n = 37); chiropractic care (CC, n = 19); cognitive behavioral therapy (CBT, n = 35) and medication (MED, n = 5). The most

Effectiveness assessments Primary analysis of primary outcome measure
The longitudinal omnibus test for pain showed IC to be significantly superior to CC over the 1-year period (P = 0.02). The long-term profile for pain intensity (0-10) which included data from weeks 4 through 52, showed a significant advantage of 0.5 for IC over CC (95% CI 0.1 to 0.9; P = 0.02). The short-term profile (weeks 4 to 12) favored IC by 0.4, but was not statistically significant (95% CI − 0.02 to 0.9; P = 0.06). Primary results are shown in Table 3 and Fig. 2.

Secondary analysis of primary outcome measure
Group differences for pain intensity at individual time points favored IC and ranged from 0.4 (weeks 4 and 12) to 0.6 (week 52), with the only significant finding

Analysis of secondary outcome measures
Long term longitudinal profiles significantly favored IC over CC for disability (Fig. 6), improvement (Fig. 7), satisfaction, and low back symptom frequency (Table 5). Medication use, quality of life, leg symptom frequency, fear avoidance beliefs, and self-efficacy did not significantly differ between groups over the 1-year period. Cross-sectional group differences for secondary outcomes mainly favored IC over CC, but most differences were not significant. The exceptions were improvement, satisfaction with care, and frequency of low back symptoms with IC consistently demonstrating a significant advantage over CC. IC also demonstrated a significant advantage over CC for passive coping strategies at week 12 in addition to disability and self-efficacy at week 52.

Missing data and sensitivity analyses
Among the 201 participants, 182 (91%) provided data on back pain intensity at every time point, and 166 (83%) provided the secondary outcomes at every time point. A total of 14 participants in the CC group and 5 in the IC group did not provide primary outcome data at all time points and the pattern of missingness seemed to be nonrandom. Participants with missing data at any time point reported higher pain intensity (when data was available) than participants with no missing data and this pattern was similar between groups. Sensitivity analyses for data missing not at random were performed using pattern mixture methods [57]. Missing pain intensity outcomes were imputed separately for each treatment group using multiple imputation (Procedure MI in STATA). Five imputed data sets were created using a multivariate normal model including baseline covariates associated with missing data. The sensitivity analysis for data missing not at random assumed the imputed pain intensity outcomes were worse by 50%. The estimated group differences from the missing data sensitivity analyses were similar in magnitude and in the same direction as the primary analysis, and all statistically significant between-group differences remained. The sensitivity analysis adjusting for expectations led to very small increases of group differences

Adverse events
There were five serious adverse events (SAEs) that occurred during the course of the trial (CC = 4, IC = 1); all were classified as unrelated to study interventions. Four SAEs were reported by patients assigned to CC  in which three were hospitalized (pneumonia, surgical intervention for fractured foot, and syncope), and one diagnosed with a brain tumor. One patient assigned to IC was hospitalized for nephrolithiasis evaluation and management.
Minor self-limiting adverse events during the 12 weeks of intervention were reported with about equal frequency in both groups ( Table 6). The most commonly reported participants were unusual or increased soreness (51-54%) and a different type of pain (31-34%).

Summary of findings
While there have long been calls to address LBP from a more comprehensive biopsychosocial perspective, there is still little research to date that has done so in a rigorous fashion. This study examines one approach for remedying this gap by comparing an integrative care (IC) intervention applying a multidisciplinary team-based approach to monodisciplinary chiropractic care. Overall, analyses demonstrated a consistent trend in favor of the IC group.

Clinical importance
Discussions of clinical importance for group differences requires consideration of the broader context regarding the condition being studied, what treatments are available, and the overall risk-benefit ratio of available options, which goes beyond criteria for establishing a clinically important change at the individual patient level [58]. Several factors should be considered when assessing the clinical importance of study results including magnitude of group differences in primary and secondary outcomes, proportion of responders, consistency of findings across outcomes, durability of effects, adverse events, treatment adherence, and costs [58]. While there was an advantage for the IC group in terms of the primary outcome, pain intensity over 1 year, the magnitude of the group difference fell below the threshold of a moderate effect size that was used to power the study. Further, despite the IC group consistently reporting greater percentage reductions of pain intensity than the CC group, they were generally small (< 10%) with the exception of the 50% threshold at Week 26 and the 30% threshold at Week 52. There was a significant advantage over the long term for IC in terms of some secondary measures (disability, improvement, satisfaction and low back symptom frequency), but not for others (medication use, quality of life, leg symptom frequency, fear avoidance beliefs, selfefficacy, active pain coping, and kinesiophobia). Importantly, no serious adverse events resulted from either of the interventions, and less serious events were approximately equal in both groups. Functional change in objective measures of torso strength and endurance, as well as lumbar dynamic motion characteristics, may provide further clinical implications and it is our intent to analyze and report these outcomes in future manuscripts. Similarly, qualitative analysis of interviews collected pre and post study invention will give insight to the patients' perspective. Finally, while a formal cost-effectiveness analysis has yet to be performed, the IC group did require substantially more resources to implement, and patients in that group had more visits raising the question of cost-benefit. When considered together, these factors suggest that while the IC group tended to do better in the long term on some important outcomes, the overall clinical importance of these findings remains debatable for chronic LBP populations similar to the one studied.

Comparison with other studies
The most recent Cochrane systematic review by Kamper et al. [20] examining multidisciplinary biopsychosocial rehabilitation programs for chronic LBP found a modest advantage compared to usual care or physical treatments for reducing pain and disability and increasing the likelihood of return to work. We have identified 6 additional trials published after the Cochrane review which had similar magnitudes of improvement in pain intensity and disability relative to physical treatments, generally confirming our findings [59][60][61][62][63][64][65]. Two low risk of bias trials conducted by Monticone et al. are notable exceptions [59,60]. These trials reported much larger treatment effects for a multidisciplinary program incorporating CBT with manual treatment and task-oriented exercise compared to manual therapy and exercise, with average reductions in pain intensity and disability around 50-75%.
The Kamper et al. review reported on subgroup analyses assessing the impact of baseline symptom severity on the effectiveness of the multidisciplinary programs and found the results were inconclusive; however, given the modest benefits and relatively high resource and cost      Mean at baseline (SD) 6.0 (4.9) 6.6 (6.8) Mean at week 4 (95% CI) 6 burden, the authors suggest these programs be reserved for the most severe cases where marked psychosocial distress is present. This approach is consistent with emerging risk-stratification and stepped care models aimed towards improving the efficiency and quality of care for musculoskeletal conditions [66].

Strengths and limitations
This study had several strengths, including the longterm follow up and high intervention adherence and follow up data collection rates. Expectations, a potentially important contextual factor that can influence treatment outcomes [67,68] were measured, and were found to be similar between groups and had little impact on the primary outcome. Importantly, side effects and adverse events were systematically collected and reported (see Table 6). Another strength was the development and application of a clinical care pathway that integrated the best available evidence with patient-rated outcome instruments and patient preferences to create biopsychosocial patient profiles to inform team based shared decision making [23]. As with any study there are also limitations that must be considered when interpreting the results. An important one which is common to many studies of complementary therapies, is the lack of representation of individuals from more diverse backgrounds, especially in terms of race and income. Also, despite careful attention to applying a clinical care pathway process [23], it was not validity and reliability tested, limiting its replication to other research and clinical settings, and potentially impacting the results of this trial. For example, while an important aspect of the pain pathway included assessing patients' psychosocial needs with established patient rated outcome instruments, treatment plans were still very much oriented towards managing pain, versus the whole person with pain, which is advocated for a truly biopsychosocial approach [11]. This was evidenced by providers recommending and patients mainly choosing treatments that were focused on physical symptoms and function in the IC group [23]. Of note is that social factors were not thoroughly addressed which is a common limitation of current pain research and should be remedied in future trials [11]. Admittedly, assessing psychosocial risk is still an area in the LBP field that remains relatively underdeveloped [69,70]. Patients in this study had relatively low baseline measures of fear-avoidance beliefs, and high self-efficacy and active pain coping which all would suggest these patients to be only modestly psychosocially impacted, and which potentially explains the lack of use of CBT in this study. Because of the pragmatic nature of the study with intervention delivery designed to approximate how it could be implemented in practice, it is difficult to discern between the active elements of treatment and potentially important contextual and structural intervention qualities (e.g. influence of the practitioner, therapeutic relationship, number of visits, time spent, etc.). Additionally, fidelity assessments of the interventions were limited to documentation by providers and patient-self report. Finally, another limitation of the study is the time it has taken from completion of the trial to publishing of the results, which came about from several members of the investigative team changing institutions. The findings however remain very relevant especially in light of continued and pervasive use of biophysically focused mono-disciplinary treatments (e.g. medications, surgery, etc.) that still plague the LBP field, despite their limited effectiveness.

Implication for clinical practice
Team based models of care with multiple provider types from complementary disciplines has been one way of integrating different therapies to more comprehensively address patients' needs [11,20]. However, these approaches can have significant challenges including inconvenience and inaccessibility due to multiple appointments with different providers along with substantial system resources needed to coordinate care across provider types, all which can contribute to disjointed and unsatisfactory care [11]. While our study was able to successfully coordinate multidisciplinary treatment plans for participants as evidenced by high adherence and satisfaction rates, achieving this required significant time and resources which is a major barrier in most clinical settings [11,23]. Further, the findings of this trial, along with other research evidence, suggest only modest advantages in terms of clinical outcomes of multidisciplinary team based interventions [20]. Importantly, Table 6 Adverse events during the 12-week treatment* *Analysis included 98 participants in Chiropractic care group and 101 in Integrated care group; † Participants reporting at least one event during treatment, participants could report more than one event Bothersomeness on 0-10 scale; bothersomeness was averaged for participants with more than one of the same event during the 12 weeks of treatment the IC group in this study had a mean of nearly 24 visits compared to 18 in the CC group, with many of the visit lengths substantially longer than a typical CC visit. This has potential cost implications to both systems and patients, posing additional hurdles to effective pain care, especially for those with challenging socioeconomic circumstances. It is likely that offering these resource intensive approaches are not going to be cost-effective. As one example, cost-effectiveness comparisons of chiropractic care have been shown to be more advantageous to more structurally intensive interventions for older neck pain patients [71].

Implication for future research
One potential alternative to integrative multidisciplinary interventions is to train individual front-line providers to address patients' biopsychosocial needs within the scope of their professional discipline, using multiple modalities to support guideline recommendations. Indeed, there have been shifts in both the chiropractic and physical therapy fields to take such an approach, integrating more psychosocial elements to these professional care models [72][73][74][75]. To facilitate, future attention is required to developing translational support tools to comprehensively address the full range of patients' biopsychosocial needs in a manner that facilitates shared decision making and monitoring in an accurate, systematic and practical manner [66,76]. Additionally, future studies should consider 'intervention mapping' in the trial design phase, where the mechanisms of action of each element of multimodal care packages are more carefully aligned with patient needs and perhaps a greater range of more relevant psychosocial outcomes. This can facilitate results interpretations, as well as reporting and replication of interventions, and ultimately broader dissemination to those who could most benefit [77,78]. Greater attention should also be given to fidelity assessment (i.e. video recordings) to ensure interventions are being delivered as intended. Finally, given the resource intensive nature of the IC group, cost-effectiveness analyses comparing CC to IC from societal and healthcare perspectives are warranted; these are forthcoming in a future publication.

Conclusion
Low back pain patients who received integrative care by a multidisciplinary integrative care team tended to have better outcomes than those who received chiropractic care. However, given the relatively small magnitude of between group differences and the extensive resources required to successfully manage and implement, the team based integrative care might not be worthwhile. More efficient models for addressing biopyschosocial care for low back pain should be explored with greater emphasis on addressing the full spectrum of related psychosocial mechanisms and ensuring equitable access for all.