The effectiveness of low- dosed outpatient biopsychosocial interventions compared to active physical interventions on pain and disability in adults with nonspecific chronic low back pain: A systematic review with meta- analysis

Objective: To evaluate the effectiveness of low- dosed outpatient biopsychosocial interventions versus active physical interventions on pain intensity and disability in adults with nonspecific chronic low back pain. Introduction: Research has shown that primary care biopsychosocial interventions (PCBI) can reduce pain intensity and disability. While scattered studies support low- dosed (≤ 15 treatment hours) PCBI, no systematic review exists comparing the effectiveness of low- dosed PCBI treatment with traditional physical activity interventions in adults with nonspecific chronic low back pain (CLBP). Inclusion Criteria: Randomized controlled trials that evaluate low- dosed PCBI compared to physical treatment with an active component such as exercise, physical activity or usual physiotherapy treatment for adult participants (18 years or older), who suffer from CLBP were included. Not recommended interventions that feature only passive therapies, spinal surgery or pharmacological treatment, and studies with inpatient multidisciplinary- based rehabilitation (MBR) were excluded. Methods: Databases were searched from inception to December 31, 2021. Language was restricted to English or German. Keywords and derivatives of “chronic back pain”


I N T RODUC T ION
Nonspecific chronic low back pain (CLBP) is a major health problem 1 with a great personal, social and economic burden. 2,3 It is defined as pain below the costal margin and the gluteal region, with or without radiation to one or both legs that persists for 12 weeks or more. 4 As CLPB has no clear pathology, it is often conceptualized as a biopsychosocial condition with complex physical, psychological, social and/or work-related interactions. 5 Treatment options for CLBP range from active (advice, exercise, cognitive behavioral therapy, etc.) to passive (massage, pharmacological, surgery, etc.) interventions that are as diverse as the condition itself. 6 Most clinical guidelines for the management of ongoing CLBP recommend inpatient multidisciplinary-based rehabilitation, which focusses on physical and psychosocial factors. 7 However, treatment can last up to 160 h and involves different types of providers, turning it into a time-consuming approach for participants and suppliers. 8 Further restrictions to access to multidisciplinarybased rehabilitation care are the cost-intensive approach and the scarce number of therapy slots since it is only offered in specialized centers. 5 An alternative are primary care biopsychosocial interventions (PCBI), which combine physical training (bio) and psychosocial cognitive treatment in forms of cognitive behavioral therapy, pain-neuroscience education or other psychosocial concepts in a primary setting. In doing so, they take into account the complex nature of CLBP and focus on physical capacity and cognitive elements such as kinesiophobia, pain-related behavior or perceptions of illness. 9 Primary care biopsychosocial interventions can be run by trained physical therapists, so they offer a combined intervention with low access barriers. Many patients can undergo the intervention simultaneously, as each session usually lasts between 30 and 120 min. 10,11 The best evidence for outpatient interventions is that intensive physical training of at least 20 h with the addition of psychosocial elements deliver the best results. 12,13 As noted by Kamper et al., a combination of different treatment elements can lead to higher costs and more time needed to complete the intervention. 5 Therefore, the question is, whether outpatient biopsychosocial interventions, focusing on all different aspects of the condition, with a low dose of a maximum of 15 h of physical treatment time achieve better results than physical activity alone? Or is it better to focus on physical activity alone in a setting where health budgets and time availability are limited. It is considered certain that physical activity is more effective than usual care. 14 Currently, there are a few systematic reviews available that focus on the effectiveness of primary care back pain interventions delivered by physiotherapists 11,15 compared to physical intervention. Furthermore, O'Keeffe et al. 16 assessed the comparative effectiveness of physical, behavioral/psychologically informed and combined interventions on pain and disability in patients with CLBP. Outpatient interventions were analyzed as a subgroup. The results are mixed and range from no effect 11 to significant long-term pain reduction compared to active treatments. 15 However, so far, no current or underway systematic review has explicitly focused on the effectiveness of low-dosed PCBI interventions compared to exercise therapy. Therefore, our research question is: Are low-dosed primary care biopsychosocial interventions, consisting of an active physical component and at least one psychological, social or occupational component, with a maximum of 15 treatment hours, are more effective in reducing pain intensity and improving physical function than other active outpatient physical treatment approaches for adult patients with nonspecific CLBP? short-term (SMD = −0.20, 95% CI = −0.36 to −0.04, I 2 = 0%, p = 0.01) and at longterm follow-up (SMD = −1.17, 95% CI = −2.06 to −0.28, I 2 = 98%, p = 0.01). The results were characterized by high heterogeneity due to different types (cognitive behavioral therapy, pain-neuroscience education, mindfulness, and motivation), delivery modes (individual and/or group), durations (3-12 weeks) and contact times (2-15 h) of PCBI. In sensitivity analysis outliers were removed to reduce heterogeneity. The results remained significant for pain intensity at short-term (SMD = −0.23, 95% CI = −0.39 to −0.08, I 2 = 0%, p = 0.004) and long-term follow-up (SMD = −0.22, 95% CI = −0.41 to −0.03, I 2 = 39%, p = 0.02). Conclusions: This meta-analysis suggests that low-dosed PCBI has favorable effects in terms of disability and pain intensity compared to active physical treatments alone. All conducted meta-analyses indicate that biopsychosocial interventions produce better outcomes than active physical treatment alone. Therefore, we strongly recommend decision makers and clinical practitioners to analyze how psychosocial elements can be introduced into outpatient (low-dosed) CLBP interventions.

K E Y W O R D S
adult, exercise, human, low back pain, motivation, pain, outpatient To answer this research question, we systematically reviewed the literature until December 31, 2021 and underwent several meta-analyses to determine the effectiveness in terms of pain intensity, disability and healthrelated quality of life of these low-dose PCBIs compared to physical treatment alone.

M ET HOD S
This systematic review and meta-analysis included randomized controlled trials (RCTs) published in full text in peer reviewed journals. We included trials from the journals' inception date to December 31, 2021 that were published in English or German language and enrolled adults with CLBP in a primary care biopsychosocial intervention. The selection of studies was conducted in accordance with the JBI methodology for systematic reviews of effectiveness evidence. 17 According to the guidelines, our systematic review was registered with the International Prospective Register of Systematic Reviews (PROSPERO 2022) on February 27, 2022 (registration number CRD42022302771). This review was conducted in accordance with an a priori published protocol. 18 We followed the Preferred Reporting Items for Systematic review and Meta-Analyses (PRISMA) guideline 19 and the available checklist (see Appendix S7).

Participants
Studies with adult participants (18 years or older) who suffer from nonspecific CLBP were included. CLBP was defined as pain below the costal margin and the gluteal region, with or without radiation to one or both legs that persists for 12 weeks or more. 4 Trials were excluded when the sample included participants with acute or subacute LBP (unless subacute LBP subjects comprised 15% or less of the total study population (≥ 85% should be CLBP), or results of patients with CLBP were presented separately). Further, studies with participants with specific low back pain due to a known, specific pathology (ie, stenosis, spondyloarthritis, ankylosing, fractures, infection or spinal cord compression) or women suffering from pregnancy-related back pain were also excluded. Besides, study samples including pre-or postoperative patients as well as minors (younger than 18 years of age) were excluded.

Intervention(s)
Studies that evaluated low-dosed outpatient biopsychosocial interventions were considered. Following previous systematic reviews, biopsychosocial was defined as a multicomponent intervention that includes an active component (exercise, physical activity or physiotherapy) and at least one psychological, social or occupational component. 5,11,20 A low dosage was defined as a maximum of 15 face-to-face treatment hours. Unsupervised home exercises did not count toward treatment hours. Included were individually (one-toone) and group-supervised (two or more participants) therapy sessions. The active component had to be delivered face-to-face. Interventions had to be delivered in primary care (referral by general practitioner, physiotherapist in local facilities, primary care practice in hospitals, other outpatient healthcare professions). Mono-and multidisciplinary delivered interventions were included. The psychological, social or occupational component(s) could have other types of delivery (eg, telephone, web-based).
First-line interventions that are not recommended according to CLBP treatment guidelines and feature only passive therapies (eg, acupuncture, electrotherapy, traction, massage), spinal surgery or sole pharmacological treatment were excluded. 6,21 Studies including interventions that were performed completely online or via telephone (all components not equal to face-to-face) or interventions delivered inpatient (eg, in pain clinics or rehabilitation centers) were excluded. Studies conducted in an occupational setting were excluded.

Comparator(s)
Studies that compare low-dosed (≤ 15 h) outpatient biopsychosocial interventions with physical treatment with an active component such as exercise, physical activity or usual physiotherapy treatment were included. The comparison could be delivered in any type of delivery mode (face-to-face, as home training, webbased, app, etc.). Excluded were control groups that were formed from a waiting list, underwent a complete biopsychosocial intervention (as defined above), got invasive therapy or received a sole pharmacological intervention.

Outcomes
Primary outcomes were physical functioning/disability and pain intensity. According to the recommendation for core outcomes in clinical trials with participants with nonspecific low back pain, 22 the outcomes of interest were physical functioning (measured with the Oswestry Disability Index (ODI) version 2.1a 23,24 or the Roland-Morris Disability Questionnaire (RMDQ) of 24-items 25 ) and pain intensity (measured with a numerical rating scale (NRS) or visual analogue scales (VAS) 26 ). Outcome of secondary interest was healthrelated quality of life measured with SF-12, 27 SF-36 28 (dimensions: physical function, general health, mental and emotional health) or 10-item PROMIS Global Health short form. 29 We included outcome measurements at any time point, that is, from directly after the intervention was completed (post-treatment) to longterm follow-up.

Search strategy
The search strategy aimed to locate published randomized controlled trials. In this review, a three-step search strategy was used. First, an initial limited search of MEDLINE (Ovid) and CINAHL (EBSCO) was undertaken to identify articles on the topic. The text words contained in the titles and abstracts of relevant articles, and the index terms used to describe the articles, were used to develop a full search strategy for: CINAHL, Cochrane Central Register of Controlled Trials (CENTRAL), Ovid Medline, Physiotherapy Evidence Database (PEDro), PubMed and Web of Science. The databases were additionally searched for relevant systematic reviews and meta-analyses. Titles, abstracts, key words and reference lists were scanned to refine search terms.
The search strategy, including all identified keywords and index terms, was adapted for each included database and information source (see Appendix S1). The search was conducted on 07 and 08 March 2022. The reference list of all included sources of evidence was screened manually for additional studies. To identify studies potentially missed by text-based database search, a systematic forward reference search of all included studies was performed using the R package citationchaser. 30

Study selection
Following the search, all identified citations were collated and uploaded to Mendeley Desktop 1.19.8/2020 (Mendeley Ltd.) and duplicates were removed. Titles and abstracts were screened by two independent reviewers (MH and PR) for assessment against the inclusion and exclusion criteria of the review in PICO Portal. 32 Potentially relevant studies were retrieved in full and their citation details were imported into the PICO Portal. The full text of the selected citations was assessed in detail against the inclusion and exclusion criteria by two independent reviewers (MH and PR). Reasons for exclusion of full text articles that did not meet the inclusion criteria were recorded and reported (Appendix S4). Any disagreements between the reviewers at each stage of the review process were resolved through discussion. The results of the search and the study inclusion process are reported in the following Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) flow diagram ( Figure 1). 19 Assessment of methodological quality Eligible studies were critically appraised by two independent reviewers (MH and PR) at the study level using standardized critical appraisal instruments from JBI for experimental studies. The results of critical appraisal are reported in narrative form and in Figure 2.
Following critical appraisal, studies that did not meet a certain quality threshold were excluded for the main meta-analysis. This decision was based on the JBI 13item checklist for randomized controlled trials. 33 Studies were excluded if one or more of the following checklist questions were answered with no: • Was true randomization used for assignment of participants to treatment groups? (Question 1/selection bias) • Were treatment groups similar at baseline? (Question 3/selection bias) • Were treatment groups treated identically other than the intervention of interest? (Question 7/performance bias) • Were outcomes measured in the same way for treatment groups? (Question 10/measurement bias) The methodological quality was assessed into three categories: Low risk of bias (≥ 9 questions answered with yes), some concerns (≥ 6 questions answered with yes) or high risk of bias (< 6 questions answered with yes). Review authors were not blinded for authors, journals or institutions.

Data extraction
Data was extracted from studies included in the review by an independent reviewer (MH) using the data extraction module of the PICO Portal. 32 The data extraction form can be found in the Appendix (S2). The data extracted was checked by a second review author (PR). The data extracted included specific details on (1) study characteristics (number of participants, age, sex, length of follow-up), (2) type of interventions (delivery, dosage, content), (3) baseline and follow-up patient-reported outcome measures of significance to assess pain intensity, disability and health-related quality of life and (4) summary of findings. In case there were multiple publications regarding one RCT, all available publications were checked and relevant data extracted. To avoid double counting that would bias the meta-analysis, the outcome measure data for eligible interventions was selected based on the follow-up time. In case of reporting the same outcomes of the same RCT in different reports, the report with the higher methodological quality was selected. Study authors were contacted to request missing or additional data. In cases where authors were uncontactable, a web-based tool 34 was used to extract raw data from the graphs.

Data synthesis
Studies were pooled in a statistical meta-analysis using RevMan Web from the Cochrane Collaboration. 35 Effect sizes were expressed as standardized mean differences (SMD) and their 95% confidence intervals were calculated for analysis for all outcome measures. As SMD did not correct for differences in the direction of the scale, an opposing direction was corrected by subtracting the mean from the maximum possible value for the scale. 36 In the case where the mean was not available, it was imputed using the method by Wan et al. 37 If correction or imputation were used, it was marked. Effect sizes were interpreted according to Cohen: A SMD of 0.2 or less was interpreted as a small effect, an effect size between 0.2 and 0.8 as a medium effect and a SMD of 0.8 or more as a large effect. 38 Statistical analyses were performed with a random effects model using the DerSimonian and Laird method. 39 This model did not only account for the statistical heterogeneity of the included studies, but allowed for generalization beyond these studies. 40,41 Subgroup analyses were performed to investigate different delivery modes (individual versus group setting), length of follow-up (short, medium, long-term, very long term), types (cognitive behavioral therapy vs. other psychological component), doses (< 10 h vs. ≥ 10 h) and delivery (individual vs. group setting) of biopsychosocial interventions.
Heterogeneity between the studies was evaluated graphically with forest plots and statistically with the chi-squared and I squared tests. As the effect sizes  are expressed in SMDs, statistical tests for funnel plot asymmetry were not performed. 42,43 The level of heterogeneity was interpreted according to the latest version of the Cochrane Handbook for Systematic Reviews of Interventions, which provides the following guidance for the I 2 statistic: 0%-40%: might not be important; 30%-60%: may represent moderate heterogeneity; 50%-90%: may represent substantial heterogeneity; and 75%-100%: considerable heterogeneity. 41 Sensitivity analyses were performed to investigate whether removal of outliers deemed responsible for heterogeneity had an influence on effect estimates.

F I G U R E 2 Risk of bias of the included studies.
Assessing certainty in the findings The Grading of Recommendations, Assessment, Development and Evaluation (GRADE) approach for grading the certainty of evidence was followed 44 and a GRADE evidence profile was created using GRADEpro GDT (McMaster University and Evidence Prime, ON, Canada). 45 This was carried out by two independent reviewers (MH and PR) at the outcome level. Any disagreements that arose between the reviewers were resolved through discussion. The evidence profile presents the following information: Number of studies and participants, a ranking of the quality of the evidence based on the risk of bias, directness, consistency, heterogeneity, precision and risk of publication bias of the review results. The outcomes reported in the Summary of Findings (Table 4) are pain intensity, disability and health-related quality of life (physical).

Study inclusion
The study identification process is summarized in Figure 1. The computer-aided database search identified 2921 study records. After removal of 792 duplicates, 2129 records were screened on title and abstract. After abstract screening 150 reports were assessed as full text for eligibility. Of those, 132 reports were excluded because they did not meet the eligibility criteria. The main reasons for exclusion were type (not biopsychosocial) or duration (> 15 h) of intervention and population (not CLBP). Excluded articles are listed in Appendix S4, including reasons for exclusion. In total 29 studies were retrieved that were neither written in English nor German. As all abstracts were translated into English, an assessment based on abstract and title was possible. The 29 studies were excluded according to the inclusion criteria -mainly type of study (n = 12), population, intervention or combination of those. Accessibility due to language was not a reason for exclusion.

Characteristics of included studies
All included studies were RCTs written in English and published between May 1998 and October 2021 (Table 1). Five RCTs were reported to be pilot studies. 46,53,60,61,65 Three studies came from Italy, 48,49,57 two from Australia 54,58 and two from the UK. 50,52 The remaining studies were conducted in Austria, 47,51 Brazil, 55,56 Egypt, 66 Hong Kong, 62 Nigeria, 53 Norway, 63,64 Portugal, 59 Singapore, 46 Switzerland, 65 the Netherlands 61 and the USA. 60 Population Participants were recruited via physiotherapists, primary care practitioners, medical specialist referral or advertisement. Treatment was conducted in outpatient departments of hospitals, healthcare and rehabilitation centers as well as primary care physical therapy and chiropractic clinics. Baseline sample size ranged from 20 to 248 participants, resulting in an overall included sample size of 1859 participants ( Table 1). The age of participants ranged between 28 and 60 years. The female sex distribution ranged from 10% to 75%.

Methodological quality
Risk of Bias was assessed using the JBI 13-item checklist for randomized controlled trials 33 and visualized with robvis 67 (Figure 2). Overall, 16 out of the 21 study reports scored a low risk of bias. The quality of three studies was rated as having some concerns (6-8 categories low). Two studies were excluded for the main meta-analysis due to an overall high risk of bias. 46,66 The main reason for exclusion was that treatment groups were not similar at baseline. Even though Elabd et al. 66 stated that there was no significant difference between groups, we reran a twotailed t-test with pain intensity at baseline. The p-value equaled 0.0148. By conventional criteria, this difference was considered to be statistically significant. Therefore, we also marked this study as having inappropriate statistical analysis. Devasahayam et al. 46 reported a significant difference in pain intensity at baseline between both groups. The effect of excluding studies with a high risk of bias is assessed in a sensitivity analysis.
Due to the content and type of the interventions participants and providers could not be blinded in most studies. Only the study of Petrozzi et al. 58 was able to blind therapists as the psychosocial element of the intervention was delivered online and the active part was the same for the intervention and control group. Overall, there were four studies, which had the lowest risk of bias with 12 of the 13 categories ranked as low risk. 57,58,61,62 In the case where two reports reported outcomes of the same study 47  all follow-up points, the report with lower risk of bias score was selected for the meta-analysis. 56,64 Intervention The interventions (intervention and control group), mode, provider, duration and summary of the main results as well as an indication of inclusion in the metaanalysis of the included studies are presented in All interventions were of low intensity (≤ 15 h). Ten studies had a contact time of 10 or more hours . [48][49][50][53][54][55]57,59,61,66 Total duration of included PCBI ranged between 3 and 12 weeks. One study provided two booster sessions after completion of initial intervention (at 4 and 10 months 54 ). Nine PCBI interventions were delivered in an individual, 47,51,52,54-58,62-65 seven in a group 46,[48][49][50]59,60,66 and two in a mixed setting. 53,61 All PCBI interventions except two were delivered monodisciplinary by physiotherapists or chiropractors. The two remaining interventions 48,57 were delivered multidisciplinary by physiotherapists, psychologists and physiatrists.

Comparator
Characteristics of the comparison group are listed in Table 2. All studies had an active physical treatment comparison element. Eight studies used standard physiotherapy as comparator. 49,50,52,55,56,58,61,62,65 Other forms of exercise were generic mat exercises, 46 motor control exercise, 53,54 aquatic exercise, 59 stretching 60 or backspecific exercise programs. 47,48,51,57,63,64,66 The reported contact time was lower in the control group than intervention group in eight studies. 48,50,53,[57][58][59]63,64,66 In the remaining RCTs contact time was equal 46,47,49,51,[54][55][56]60,62,65 or slightly higher in control group. 52,61 Mode and provider were the same between intervention group and control group in all but three studies. Cecchi   This study had an instructive phase for 5 weeks and a reinforcement phase with monthly follow-up meetings. We only used the instructional phase and the PT results even though the authors published more follow-up data.

Quality of the evidence and effects of intervention
Fifteen of the 18 studies were included in the metaanalyses of pain intensity, disability and health-related quality of life. Three studies were excluded. 46,65,66 Two studies were excluded due to high risk of bias 46,66 and one study 65 because the outcome measurements were reported as within-group mean differences and original values could not be retrieved. Three main metaanalyses based on the outcome were conducted. For these meta-analyses' outcomes were analyzed at different time points (ie, post-treatment [PT], short-term [ST]; 1-3 months), medium-term (MT; 4-6 months), long-term (LT; 12 months) and very long-term (VLT; 18+ months). The evidence of the included RCTs was summarized using the GRADE approach and is presented for the main outcomes and subgroup/sensitivity analyses in the summary of findings table (Table 4). The grade evidence profiles for each outcome with explanation for assessment can be found in Appendix S5. The quality of evidence ranged between very low and high mainly due to attrition and measurement bias as well as heterogeneity between studies.

Subgroup analysis
Subgroup analyses were performed to investigate the effect of different delivery modes (individual versus group setting), doses (< 10 h vs. ≥ 10 h) and types (cognitive behavioral therapy vs. other psychological component) of PCBI on the main outcomes of pain intensity and disability (Appendix S6). When stratifying by dichotomized dosing, the effects on pain intensity were not significantly different. No difference was also found between individual and group setting on physical function. Positive effects on pain intensity were seen in subgroups of delivery (favoring individual) and type (favoring cognitive behavioral therapy). The type of psychological component had also an effect on physical functioning (favoring cognitive behavioral therapy). The dosage influenced the outcome of disability, favoring very low doses of 10 h or less. (2 studies, n = 271, SMD = −1.66, 95% CI = −2.17 to −1.15, I 2 = 70%, moderate certainty evidence) follow-up ( Figure 6). No significant difference was found at shortterm follow-up (2 studies, n = 91, SMD = −0.04, 95% CI = −0.45 to 0.38, I 2 = 0%, low certainty evidence).

Dosage
The dose of the intervention was less than 10 h in eight studies that compared the effect of PCBI versus physical training interventions on physical functioning. Significant differences were found post-treatment (5 studies, n = 348, SMD = −2.75, 95% CI = −5.30 to −0.19, I 2 = 99%, moderate certainty evidence), at short-term (3 studies, n = 354, SMD = −0.26, 95% CI = −0.46 to −0.05, I 2 = 0%, moderate certainty evidence) and longterm (5 studies, n = 570, SMD = −1.30, 95% CI = −2.52 to −0.09, I 2 = 98%, moderate certainty evidence) followup ( Figure 9). As there was no statistical difference in favor of higher doses (equaling an intervention time of 10-15 h) for the outcome of physical function and only very slight differences for pain intensity, there is no forest plot depicted in this section, but they can be found in Appendix S6.

Sensitivity analysis
Sensitivity analyses were performed to assess the robustness of our findings based on the decision to include studies in the meta-analysis. Therefore, we assessed the impact of including studies marked as having a high risk of bias and the effect without outliers (removal by visual inspection) on the results.

Analysis without outlier
Outliers were removed by visual inspection to assess their influence on the overall effect. Three articles 48,57,64 included in the main analyses showed extreme outliers and raised the heterogeneity to very high values. All three used cognitive behavioral therapy with task-oriented exercises with a focus on reducing fear-avoidance behavior and integrating training in their lifestyle to continue exercising after the end of treatment. After exclusion no significant difference was found on physical functioning at any term. The I 2 statistic ranged between 0% and 17% so that heterogeneity was not deemed important anymore (Appendix S6).

Analysis with high-risk studies included
In the main analysis, two studies were excluded due to an overall high risk of bias. 46,66 For the study of Devasahayam et al. 46 we imputed the means based on the summary statistics of lower quartile, median and upper quartile. 37 Exclusion of both studies had an impact on post-treatment for pain intensity and physical function. If we decided differently and included them, the  Figure 11). However, due to the high risk of bias of both studies, the overall GRADE certainty of evidence assessment would be downgraded to a low grade.

Summary of key findings
To our knowledge, this systematic review is the first to examine the effects of low-dosed outpatient biopsychosocial interventions versus active physical interventions on pain intensity and disability in adults with CLBP. Eighteen RCTs were found eligible in this systematic review and 15 trials comprising a total of 1531 participants suffering from CLBP were entered into the metaanalyses. Slightly more women (55.4%) than men were included in the examined trials. The participants were of middle age (main age group 45-60). Significant effects in favor of PCBI were found concerning the primary outcome pain intensity post-treatment (large effect size), as well as at short-term (medium effect size), long-term (medium effect size) and very long-term (large effect size) follow-up. Short to moderate significant effects in favor of PCBI for physical function were found post-treatment (large effect size), at short-term (small effect size) and at long-term follow-up (large effect size). The results were characterized by high heterogeneity due to different types (cognitive behavioral therapy, pain-neuroscience education, mindfulness, and motivation), delivery modes (individual and/or group), durations (3-12 weeks) and F I G U R E 9 Subgroup analyses of low dosage (< 10h) for physical function immediately after intervention and during the followup period.
contact times (2-15 h) of biopsychosocial interventions. Subgroup analyses suggest moderate evidence with large effect sizes in favor of cognitive behavioral therapy as the psychosocial component at post-treatment, longterm and very long-term follow-up for the outcomes of pain intensity and physical function. Subgroup analysis suggests that individual settings are more effective than group settings in reducing pain intensity and disability. Furthermore, interventions with a dose of less than 10 h might be more effective than interventions with 10-15 h treatment time. Contradicting results were achieved in the sensitivity analyses without outliers. After their removal, a small significant effect with low evidence on physical function was only found at short-term followup. Regarding pain intensity, the results remained significant with moderate evidence of a medium effect in favor of PCBI at short-term and long-term follow-up. The secondary outcome health-related quality of life was

Interpretation of review findings
Comparison with other reviews As this is the first systematic review focusing on lowdosed PCBIs there are no direct comparable papers available; however, some meta-analyses with similar study questions 5,10,11,15,16 exist. Those focus either on biopsychosocial physiotherapist-led interventions in primary care 10,11,15 or combined interventions. 5,16 However, none of these reviews explicitly focused on low-dosed, primary care biopsychosocial interventions compared to physical activity, but act as best possible and available comparable options for our systematic review with metaanalysis. The identified significant effects in the reduction of pain intensity after treatment, in the short-term, long-term and very long-term, have also been reported in the meta-analyses by Hall et al. 10 and O'Keeffe et al. 16 Zhang et al. 15 only identified significant reductions in pain intensity in favor of group-based BP interventions provided by physiotherapists compared to other active treatments in the long-term. However, the authors included only studies delivered by physiotherapists and in a group format. In the subgroup analysis we revealed that especially individual delivered interventions are more effective at post-treatment and short-term follow-up, but not at long-term follow-up. Therefore, these results are also in line with our findings. Van Erp et al. 11 identified low quality evidence that there is no difference in improving functional disability or pain intensity between PCBI and other physical active treatments. However, their conclusion was based on a descriptive systematic review that focused on physiotherapist-delivered biopsychosocial interventions without meta-analysis. The additional fact that they only included four studies in their descriptive analysis can explain the differences in findings. Similarly to the findings of O'Keeffe et al., 16 Hall et al. 10 and Kamper et al. 5 significant effects in favor of PCBI for physical function were found at long-term follow-up. Regarding physical function, we also identified significant effects at post-treatment. O'Keeffe et al., 16 Hall et al. 10 and Kamper et al. 5 did not report any outcomes for the post-treatment consideration. Zhang et al. 15 did not analyze the outcome physical function. The effects in secondary outcome health-related quality of life only included few studies and, therefore, yielded an imprecise estimate with nonsignificant findings as previously reported by Hall et al. 10 and Kamper et al. 5 Other comparable studies did not analyze health-related quality of life.
In general, our findings indicate that biopsychosocial interventions -even when administered at a low dosage -are more effective in reducing pain intensity and disability than physical active interventions alone. This is in line with the findings of Hayden et al., 12 who concluded after a large meta-network analysis, that adding psychosocial elements to most types of exercises, improved the effectiveness for outcomes concerning pain and functional limitation.

Limitations of included studies
The reported results in the meta-analyses were characterized by high heterogeneity, which is why they have to be interpreted with caution. One the one hand, reasons for high heterogeneity were different types (cognitive behavioral therapy, pain-neuroscience education, mindfulness, motivation), delivery modes (individual and/ or group), durations (3-12 weeks) and contact times (2-15 h) of biopsychosocial interventions. This was partly induced by our proposed definition. We wanted to include all true biopsychosocial studies, to make sure that we have an all-encompassing picture. As we already set strict inclusion criteria on dose, setting and comparator, we did not want to reduce our potential studies even further by specifying the type of biopsychosocial intervention in more detail.
On the other hand, heterogeneity was induced due to the three outlier articles 48,57,64 and accounted for in sensitivity analyses and with the use of SMD and random effects model in the meta-analyses. Significant differences prevailed for pain intensity at short-term and long-term follow-up in favor of PCBI, which is why these findings are believed to be robust. In addition, the total sample size with 1859 participants is not large. Only seven RCTs had more than 100 participants and five were pilot studies with very few participants. One reason for the high number of pilot studies could be that primary biopsychosocial interventions are gaining research interest only recently. 11 Most identified studies were published between 2015 and 2020 and due to the unknown feasibility and additional effects of PCBI versus active only therapies, many authors chose to start with a pilot study. It must be noted that we divided one study 57 into two parts and only used the first part. The study had an instructive phase for 5 weeks and a reinforcement phase with monthly follow-up meetings.
We only used the instructional phase and the posttreatment results even though the authors published more follow-up data. But if the instructive and the reinforcement phase were taken together, they would have resulted in an intervention dose of more than 15 h.
In general, most included studies had a low risk of bias. However, due to the trial design, blinding of personnel and participants was not possible, which is a potential source of bias. As with this type of intervention, blinding could not be performed; we did not rate down the grade of evidence for nonblinding. Another possible source of bias is that all the outcomes chosen were self-reported. However, since pain intensity and physical function are almost always measured by patient reports, we did not downgrade the evidence for this risk of bias.
We restricted the language to German and English. In the screening process we identified 29 studies that were reported in another language. However, we were able to assess them based on a translated abstract. All studies were excluded according to our inclusion criteria. Accessibility due to language was not a reason for exclusion, which is why we believe that we did not introduce a language bias.

Limitations of review process
We also identified limitations introduced by the review process. In the associated study protocol we described focusing on the main outcomes, pain intensity measured with NRS or VAS, and physical function measured with ODI or RMDQ according to the recommendation for core outcomes in clinical trials with participants with nonspecific low back pain. 22 During the revision process, it was suggested not to limit the choice of measurement instruments -especially as we used a random effects model and standardized -not total -mean difference in the meta-analyses. This led to the inclusion of one more study on the outcome of pain intensity and three more studies on the outcome of physical function. Another limitation in the review process was the missing of a filter of our literature search on human subjects and, therefore, included unnecessary animal studies in screening. This in one of the reasons why we had to exclude approx. 93% of all records by abstract screening only. Further, we limited our search to published studies and, therefore, introduced a risk of publication bias. As a consequence, we did not assess this meta-analysis for risk of reporting bias. But still with the much focused search strategy and the usage of backward and forward citation screening we believe we were able to present a very accurate overview of the current knowledge on this topic. As studies from every continent were included, we depicted the global importance of accessible interventions for CLBP. Further, we extensively used sensitivity and subgroup analyses to rule out sources of uncertainty.
In total, the included studies had a high methodological quality and a low risk of bias, therefore, we believe that the presented findings are robust.

Recommendations for practice or policy
This research project had a strong focus on practical use. We explicitly focused on low-dosed primary care interventions to identify whether combined approaches are more effective than physical treatments alone. As the results showed a mostly large effect size with moderate evidence in favor of PCBI, it is highly recommended to introduce psychosocial concepts in existing or new lowdosed physical interventions for patients with nonspecific CLBP. The introduction of psychosocial elements helps the patients to better cope, understand and manage their pain after the end of the intervention. As CLPB is a chronic condition, it is very important that the affected understand the influence of their behavior on their pain intensity as well as perception and act appropriately. With low-dosed PCBI, this knowledge and guidance is provided. This may be one reason for better results, especially in long-term follow-up. As long-term effective PCBI interventions have been shown to be possible and create long-term improvements, we suggest that decision makers strongly support the integration of psychosocial components in CLBP interventions to improve the outcome. Compared to multidisciplinary interventions in secondary care, this low-dose primary care approach has the potential to reduce direct healthcare costs and improve access options to effective treatment for nonspecific CLBP globally. We strongly suggest continuing this road of treatment in the practice landscape.

Recommendations for research
The findings of this systematic review suggest the need for more high-quality studies with larger study samples. We focused our analysis on primary biopsychosocial interventions with a low-dose compared to physical activities. With this focused research question, we reduced potential selectable studies and, therefore, also the size of the total population. As active interventions are generally more effective than the usual care/waiting list, this seemed reasonable. 14 We believed that due to the strict inclusion criteria, we would have less heterogeneity. As described above, this was not the case. There were different active parts, psychosocial parts and also settings, providers and doses. For further research we recommend an even more focused approach. As seen in the subgroups, an exclusive analysis of individual PCBIs or PCBIs with cognitive behavioral therapy could yield clearer results with less heterogeneity. To reduce the risk of bias, further systematic reviews could also focus on objective measurable outcomes (eg, sick days or doctoral visits). Another recommendation would be to conduct cost-effectiveness analyses of PCBIs and to elaborate if they are as cost-effective as physical interventions alone. And lastly, since we saw three studies with very large effect sizes, we strongly suggest a reproduction of these studies to verify the results.

CONC LUSION
The presented meta-analysis suggests that low-dosed PCBI has favorable effects in terms of physical function and pain intensity compared to physical active treatments alone at post-treatment, short-term and long-term follow-up. Subgroup analyses indicate that cognitive behavioral therapy and individual treatment approaches may produce better outcomes than group settings or other psychosocial components. Due to the identified high heterogeneity, further high-quality RCTs with large sample size are needed to validate the effect. However, as no meta-analysis indicated that physical activity treatment alone produces better outcomes than biopsychosocial interventions, we strongly recommend decision makers and clinical practitioners to analyze how psychosocial elements can be introduced into outpatient CLBP interventions. Based on our results we believe that lowdosed PCBI can lead to better outcomes than physical activity treatments alone.

AU T HOR CON T R I BU T ION S
MH, PR, and VA involved in conceptualization review and editing. MH and PR involved in data curation. MH involved in formal analysis and prepared the original draft. MH and PR involved in investigation and visualization. MH involved in methodology and project administration. VA involved in supervision.

AC K NOW L E DGM E N T S
This systematic review is to count toward the dissertation of MH through Hannover Medical School (MHH). Open Access funding enabled and organized by Projekt DEAL.

CON F L IC T OF I N T E R E ST
There was no conflict of interest in this project.

DATA AVA I L A BI L I T Y STAT E M E N T
The critical appraisal form is publicly available in JBI SUMARI and in the JBI Manual for Evidence Synthesis. The data extraction module is available in PICO Portal. A modified form can be found in the Appendix. Extracted data and data used for all analyses is available upon request.

R EGI ST R AT ION A N D PROTOCOL
This systematic review and meta-analysis was registered a priori in the International Prospective Register of Systematic Reviews (PROSPERO) (CRD42022302771, 27.02.2022). The associated study protocol has been published elsewhere. 18 Minor amendments to that protocol were made: In the protocol, we stated that studies that do not meet a certain quality threshold in critical appraisal will be excluded. However, we decided to keep those in the narrative synthesis and only excluded the studies for the main meta-analysis and included them in a sensitivity analysis. Further, in contrast to the statement in the protocol, we did not report a GRADE SoF table to assess the certainty of findings. Instead, we chose to include GRADE evidence profiles for each outcome assessed.