PROGNOSTIC FACTORS FOR PHYSICAL FUNCTIONING—NARRATIVE AND QUANTITATIVE ANALYSES
The association between baseline pain intensity and physical functioning after MDR was assessed in 16 studies,41–43,48,51–55,57,59,62,63,65,68,69 including a total of 8191 participants.
The narrative analyses indicated inconclusive results. Eight studies42,43,52,53,55,63,68,69 reported no association between pain intensity at baseline and outcome. Four studies54,57,59,65 showed that lower levels predicted positive outcomes while 2 studies41,51 showed that high pain levels at baseline predicted positive results at follow-up. Two studies had conflicting results, depending on pain location48 or type of analysis (uni/multivariate)62 (Table 4).
Five studies (4 low, 1 high RoB) provided continuous data for inclusion in a meta-analysis (n=2676). Results of the meta-analysis showed that initial pain intensity was not associated with improvement in physical function at follow-up, OR=0.84; 95% CI, 0.65-1.07; P=0.16 (Fig. 3A).
The association between pain duration before MDR and physical functioning was assessed in 8 studies43,54,55,62,63,65,68,69 including a total of 3800 participants.
Four of 8 studies54,55,62,69 reported no association with outcome, 2 studies showed a negative association,63,65 and 2 studies43,68 reported conflicting results on multiple outcome measures, showing either no association or a negative association in favor of short duration (Table 4).
Five studies (3 low, 1 moderate, 1 high RoB) were included in a meta-analysis (n=2978). The pooled OR (95% CI) showed no association with physical functioning; that is, the results indicate pain duration at baseline is not a prognostic indicator for outcome, OR=0.97; 95% CI, 0.93-1.00; P=0.08 (Fig. 3B).
Sensitivity Analyses and LoE (GRADE)
The sensitivity analyses for both pain intensity and pain duration showed that our results remained robust when examining the influence of study quality, follow-up time, measurement instruments, uni/multivariate analyses, and when compared with a fixed-effects model. The GRADE analyses of pain intensity as well as pain duration showed that, due to downgrading as a result of “inconsistency of the results,” there is evidence of moderate quality that baseline pain level and pain duration cannot predict physical functioning at ≥6-month follow-up of MDR (Table 7).
Physical Function–related Factors
The association between baseline and follow-up physical functioning was assessed in 15 studies (n=4868).41,43,48,51–53,55–57,60–62,65,68,69 Physical function was assessed either by patients’ actual performance of physical tests (and evaluated by therapists)—or by patients’ own reporting of their function, activities, or disability, that is completing questionnaires (PROMs). The factors were divided into 2 groups and analyzed separately due to the qualitative differences of the assessment methods (Table 5).
Performance-based Physical Factors
Two studies48,60 investigated 6 performance-based physical factors (n=783). The tests evaluated isometric endurance, mobility, and aerobic capacity as prognostic factors. The narrative analyses indicated no prognostic value for outcomes related to physical function, both studies reported no significant association. Both studies were rated as having high RoB. Because of limited data, a meta-analysis was not appropriate.
Self-reported Function, Activities/Disability
Fourteen studies examined the association between self-reported physical functioning and outcome (n=4706).41,43,48,51–53,55–57,61,62,65,68,69
The narrative analyses of self-assessed physical function revealed inconclusive results. Higher levels of function at baseline were significantly associated with a positive outcome in 6 studies, while low levels of function associated with a positive outcome were reported in one study and no significant association was reported in another one study. However, 6 studies presented inconclusive results depending on measures used, either showing an inconsistency between a positive association and no association (3 studies) or between a negative association and no association (3 studies).
Eight studies (5 low, 2 moderate, 1 high RoB) were included in a meta-analysis (n=3444). The pooled OR (95% CI) showed that high baseline function was associated with positive outcome, OR=1.07; 95% CI, 1.02-1.13; P=0.01 (Fig. 4).
Sensitivity Analyses and LoE (GRADE)
The results of self-reported physical function remained robust when excluding high RoB studies, and were independent of a fixed or random model. However, when analyzing the 3 studies43,52,61 with shorter follow-up times, there was no longer any significant association between physical function at baseline and outcome. Moreover, in studies with univariate analysis only,51,56,61 the associations disappeared as well.
The Grade synthesis showed there was no evidence (−) of prognostic value of performance-based physical function and that there was low evidence (++) of a small effect of self-rated initial high physical functioning as prognostic for good physical functioning at follow-up after MDR (Table 7). Downgrading was due to “study limitations” and “inconsistency of the results.” For performance-based physical function, the initial GRADE LoE was set at +++, due to unclear study phases.
Seventeen studies41–43,49,51,52,54,56–58,62,64–69 investigated baseline psychological factors. Of these, most were categorized as either emotional factors or cognitive behavioral factors. For the purpose of analyses, cognitive-behavioral factors were divided into protective factors or risk factors. A few remaining factors, mostly relating to personality traits,51,64,67 were considered too compound or dissimilar and were therefore not synthesized in this context.
Fifteen studies (n=4358)41–43,51–53,55–57,62,64–66,68,69 investigated emotional factors relating to mood/distress, for example, depression and anxiety and their association to physical functioning at follow-up.
The narrative analyses showed inconclusive results concerning their prognostic value. Six studies43,51,53,55,56,69 did not demonstrate any significant associations, 6 studies41,52,62,64,66,68 showed differing results between anxiety and depression, 2 studies57,65 showed that low levels of depression/anxiety at baseline could predict positive results at follow-up, while 1 study42 showed some degree of initial anxiety/depression was associated with a positive outcome. Anxiety and depression were analyzed both separately and in combination with each other (Table 6).
Eight studies (5 low, 3 high RoB) with continuous data were included in a meta-analysis (n=3483). The pooled OR (95% CI) showed that there was a small, statistically significant, association between low baseline emotional distress and a positive outcome, OR=0.77; 95% CI, 0.65-0.92; P=0.003 (Fig. 5A).
Cognitive and Behavioral Factors—Protective Factors
Nine studies (n=2288)43,52–56,62,65,69 examined various cognitive and behavioral factors relating to self-efficacy, control beliefs, and health optimism; factors commonly attributed to strengthening a person’s resilience, that is with a protective effect.
The narrative analyses showed diverse results. Three studies43,56,69 found no association from 6 examined protective factors, while 3 studies showed a positive association favoring high levels of 3 identified protective factors56,62,65 and 1 study43 showed a negative association, indicating low levels of 1 factor was associated with a positive outcome.
Four studies (3 low, 1 moderate RoB) were included in a meta-analysis (n=1392). The pooled OR (95% CI) showed, contrary to the narrative analysis, an association between high levels of protective cognitive behavioral factors and a positive outcome, OR=1.49; 95% CI, 1.17-1.90; P=0.001 (Fig. 5B).
Cognitive and Behavioral Factors—Risk Factors
Eleven studies (n=4068)41,43,51,52,54–56,58,65,66,68 examined the association between various “negative” cognitive and behavioral factors and outcome, that is potential risk factors. These were related to illness and self-efficacy beliefs, fear-avoidance beliefs and behavior, catastrophizing, and dimensions of somatic discomfort/somatization.
The narrative analyses of cognitive and behavioral risk factors indicated a majority of nonsignificant associations. Results identified 20 items with no association and 9 in favor of low levels for a positive outcome.
Six studies (2 low, 3 moderate, and 1 high RoB) were included in a meta-analysis (n=1173). The pooled OR (95% CI) showed, contrary to the narrative analysis, an association between low levels of cognitive and behavioral risk factors and a positive outcome, OR 0.85; 95% CI, 0.77-0.93; P=0.0008 (Fig. 5C).
Sensitivity Analyses and LoE (GRADE)
Sensitivity analyses of emotional factors showed that the significant associations disappeared when including only studies with low RoB and the OR increased from 0.77 (95% CI, 0.65-0.92) to 0.89 (95% CI, 0.75-1.04) and to 0.90 (95% CI, 0.78-1.03) when only including studies with multivariate analyses. In addition, when only the 2 studies with a 6-month follow-up were included, the association disappeared (OR, 0.86; 95% CI, 0.69-1.08). However, the results remained robust when comparing anxiety/depression separately and when compared with a fixed effects model.
The results remained robust through all sensitivity analyses of protective factors; study quality, follow-up time, univariate/multivariate data and when compared with a fixed effects model. The OR increased from 1.49 (95% CI, 1.17-1.90) to 1.67 (95% CI, 1.12-2.49), when including studies with follow-up periods of longer than 6 months.
Sensitivity analyses of risk factors showed that the significant associations disappeared when including only studies with low RoB or studies with short follow-up time (6 mo). However, the OR changed by <0.06 and the results remained robust when comparing univariate/multivariate data and when compared with a fixed effects model. All in all, sensitivity analyses of the psychological factors clearly showed that the results were robust.
In summary, based on a GRADE analysis of these results including sensitivity analyses, the results showed that (a) there is moderate quality evidence that low initial emotional distress predicts a positive outcome on physical functioning at follow-up after MDR, (b) there is moderate quality evidence that high levels of protective cognitive behavioral factors predict a positive outcome of physical functioning at follow-up after MDR, and (c) there is moderate quality evidence that low levels of cognitive behavioral risk factors predict a positive outcome (Table 7). Downgrading was due to “study limitations” (a, c) and suspected “publication bias” (b).
Summary of the Results
To synthesize the evidence on prognostic factors for long-term (≥6 mo) physical functioning in patients with chronic musculoskeletal pain after MDR treatment, we examined 25 studies (n=9436) that included a total of 87 potential prognostic factors relating to initial pain and physical and psychological functioning.
The key finding of this review confirmed that pretreatment psychological factors as well as physical function/disability are important prognostic indicators of functional outcome after MDR while common pain variables did not appear to provide evidence on prognosis.
Regarding psychological factors, results showed a moderate LoE that low levels of emotional distress, high levels of cognitive and behavioral protective factors, and low levels of cognitive and behavioral risk factors predicted a better physical functioning in long-term follow-up. Moreover, results showed a low LoE that high levels of self-reported physical function predicted better physical functioning. Our results also indicated, with moderate levels of evidence, that pain severity and pain duration did not predict physical functioning after MDR in patients with chronic musculoskeletal pain at least 6 mo after treatment.
Comparison With Previous Reviews
Our study found that pain severity and pain duration did not have any prognostic value (moderate LoE), indicating that pretreatment information on pain per se is not informative for the further clinical course, at least not where physical function is concerned. The review of van der Hulst et al20 also reported that pain duration lacked prognostic value. But contrary to our study, they found evidence that higher pain intensity was associated with worse outcome. However, this conclusion was based on only 2 articles, one of which is included in our study,48 while the other study included findings on a dissimilar subgroup of population, intervention, and outcome. On the other hand, the review of de Rooij et al19 reported the opposite, that is high pain intensity being associated with a better outcome, though this conclusion was based on only 1 study. In previous reviews22,24 that have investigated prognostic ability in earlier phases of pain chronicity (acute and subacute), pain variables presented with evidence of a negative impact on outcome. In our results, however, pain ratings were not significantly related to the outcome, in this case physical functioning, although the direction of the association was in accordance to these previous results, maybe indicating a less prognostic value over time.
In the synthesis we differentiated between objectively measured performance-based and self-assessed physical functioning. The assessment of performance-based function was only investigated in 2 studies, and showed no association, which is in line with van der Hulst et al.20 Moreover, the study of Wessels et al,21 which investigated the association of changes in physical performance factors with improvement in disability, also reported that there was no association with outcome. Further research is needed to elucidate the topic, to investigate whether more objectively measured dimensions of physical functioning could have a prognostic value for outcome. On the other hand, self-assessed physical functioning emerged as a major outcome topic, and proved valuable in predicting outcome. We found, with low levels of evidence, that self-assessed physical function predicts physical functioning 6 mo after MDR. Our meta-analysis strengthened the results from the qualitative analyses of van der Hulst et al,20 where it was found that self-assessed physical functioning could predict physical functioning. Also, as the findings were reproduced in a mixed-diagnosis chronic pain population—instead of a more homogenous chronic low back pain population—the generalizability of the findings increased. However, the reasons for the inconsistency in reported direction of the association (either favoring higher or lower baseline status), which were also noted by van der Hulst and colleagues, need to be further examined.
We found high levels of emotional distress predicted poor outcome, which is in line with previous assumptions and reports19,70,71; however, there is a lack of consistent evidence.20 This is the first time it has been shown in a meta-analysis based on >3000 participants, and our results confirm the importance of patients’ emotional functioning for treatment outcome.
Cognitive and behavioral factors are implied to have an impact on treatment outcome19,20,70 and this was also confirmed by our results. These essential factors of the pain experience may both strengthen the ability to deal with chronic pain as well as hinder patients’ adaptation. The narrative analyses of cognitive behavioral risk factors indicated a majority of nonsignificant associations but the meta-analysis revealed them to be significant prognostic factors for a negative outcome. While addressing these factors is at the core of pain management in MDR, our results show that high levels on cognitive and behavioral risk factors are related to poorer functional outcome. This implies that our current best evidence practice may not be addressing the coping problems of these patients satisfactorily. Indeed, Morley et al72 pointed out that results of cognitive-behavioral therapy pain management programs are modest at best, and these results have led to calls for improvements in treatment models.73,74
High levels of cognitive and behavioral protective factors predicted a better level of physical functioning in long-term follow-up. The results confirm the importance of factors attributed to a person’s resilience in determining outcome. Indeed, in a recent publication, the importance of factors related to a positive affect has been lifted forward as one way to improve treatments for chronic pain.75 As psychological risk and protective cognitive and behavioral factors are not mutually exclusive, MDR treatment should focus on both lowering the psychological risk factors and enhancing the protective psychological factors.
The prognostic ability of the psychological factors with a negative bearing, emotional distress (OR=0.77), and cognitive and behavioral risk factors (OR=0.85), respectively, was somewhat lower compared with the prognostic ability of the psychological protective factors (OR=1.49). This could be due to treatment effects, as in most MDR treatment programs the negative psychological factors are often targeted, while protective psychological factors may not be as commonly addressed. As previously put forward by de Rooij et al,52 prognostic factors that are targeted and altered during treatment can lose their prognostic ability, which may also be reflected in the present results. This could point to a more active clinical use of these positive, psychological protective factors for prognosis.
On the whole, as today’s management of chronic pain still gains only moderate effects, and the evidence to guide optimal treatment tailoring is limited, the importance of identifying prognostic indicators is of major clinical relevance. A prerequisite is that we are able to identify who is at risk of poor outcomes and who is most likely to benefit. Until now, no previous meta-analysis review studies have been conducted on this topic and, to our knowledge, this study is the first well-powered systematic review to summarize the available literature on prognostic factors specifically for this major patient group.
The strength of this systematic review is that it synthesizes factors of importance for physical functioning, one of the main targeted outcomes of MDR, rather than exploring a single prognostic factor impact or a selected part of the chronic pain-population, for example based on diagnosis. The study takes its standing point from a pragmatic perspective, hypothesizing that some factors probably exist that are common for the chronic pain population in general, irrespective of initial pain diagnosis, that is generic factors of importance for treatment outcome. From a methodological point of view, a body of evidence derived from longitudinal and pragmatic cohort studies enables high confidence in the field of prognosis, in comparison to more selected experimental randomized controlled trial studies.47 On the other hand, attrition and confounding can limit the internal validity of observational studies. The way of creating high-level evidence by unifying these observational studies with systematic synthesis methods is therefore a strength of this study.
The interdisciplinary review team with expertise in all fields relating to the aim of the study enabled a precise study selection, which led to great confidence in the identification of both the population of interest and the intervention of interest. The team was generally in agreement during the study selection process, despite the heterogeneity of retrieved studies. Good inter-rater agreement was strived for in all selection steps and RoB ratings, by introducing every phase with a pilot.
Omitting gray literature is likely to have introduced some information bias; however, it would be too time consuming to also collect and deal with this type of spread-out information, which is often not reported in enough detail. Including only articles in English is a potential source for information bias as well; however, it was a necessity for maintaining the strictness and specificity during the scrutiny of the study selection process. In addition, some reporting biases, for example publication bias or selective reporting of outcomes or analyses, cannot be ruled out. Significant results have a greater chance of being made available. Still, we found many studies presenting nonsignificant results. We believe this was partly due to our broad review scope and an exploratory search strategy, which permitted a vast amount of material, independent of primarily targeted prognostic factors in the original research publications. We put great effort into using these, often nonsignificant, variables in our syntheses, either narratively or quantitatively if data were provided. This has hopefully led to adding power and reducing possible asymmetry. As the relatively small number of studies reporting on each comparison precluded a detailed and meaningful analysis of funnel plots for publication bias, we attempted to visually analyze the narrative tables for symmetry of significant versus nonsignificant reporting. For some results, for example, the synthesis of protective psychological factors, the effect emerged stronger in the meta-analysis, which could likely be a result of missing nonsignificant data.
The risk of selection bias may have been introduced in the initial screening of titles, which was performed by one reviewer instead of 2. However, it was necessary to reduce the recall volume resulting from the broad and sensitive search strategy—and this stage therefore dealt only with identifying titles that precluded inclusion. The following screening process had a robust arrangement with randomization of studies and independent teams constituted by a senior and junior researcher.
Other sources for limitations of the study results may arise if narrative and quantitative syntheses are based on incompatible study heterogeneity or low study quality. We aimed to provide a well-powered overview of potential prognostic indicators of various MDR outcomes—as a result heterogenous studies were included with regard to types of pain conditions/regions and clinical settings. This was based on the premise that common prognostic factors for “the chronic pain disease itself” probably exist. Although unique in its kind, some loss of specificity is therefore a consequence and limit to this review. To the best of our ability, great effort was put into a sensible study selection and a coherent collating of our found predictors and outcomes, in the sense of minimizing incompatible (noncomparable) factors. We are thus confident that the study populations and study interventions constituted a sample in accordance with the pragmatic, wide selection of individuals with chronic pain that would normally participate in MDR. The same applies for the grouping of factors and outcomes, which were measured with various instruments; however, all with the intention of capturing dimensions of the same construct. Incompatible measures or measures with measurement properties considered to be too vague were not included in analyses. In the present study, the OR was used as the common index in the meta-analysis, although the OR has sometimes been criticized for its difficulty in interpretation. We stated in our study protocol that we will present associations between prognostic factors and outcome by means of OR, and this could enhance comparisons with future MAs.25 A random effects model was chosen for the statistical analyses, as it assumes and deals better with the anticipated heterogeneity.
Heterogeneity, measured by I 2, was generally high for almost all comparisons (range: 48% to 94%). Although I 2 indicated high heterogeneity, our attempts to investigate the source for these differences did not reveal any systematic reasons for the variance. Sensitivity analyses proved our results were in general robust. The direction of the associations remained stable and did not result in any major change of variation in the effect, except for the factor physical functioning. The effect estimates remained stable when comparing studies based on statistical analyses (univariate vs. multivariate) and study quality (low vs. high), and follow-up time (shorter vs. longer), although the statistical significance level occasionally decreased to nonsignificant for the emotional distress and cognitive and behavioral risk factors. Sometimes the effects of the prognostic factors seemed to be strengthened over time, when comparing shorter versus longer follow-up time (eg protective cognitive and behavioral factors), but the limited number of included studies in each meta-analysis did not permit further detailed moderator analyses of follow-up time or further aspects of clinical diversity.
Although our sensitivity analysis of potential factors influencing the stability of our results was generally stable, we cannot exclude true heterogeneity. With more unexplained variance across studies, some caution in the interpretation of the results was required and we therefore downgraded all pain and physical function domains in the GRADE, due to “inconsistency.”
Study quality, that is poor methodological quality may also impose limitations to the validity of study results, for example, low power, low attrition rates, or inadequate analyses are likely to affect the estimates and widen the 95% CIs in smaller studies. Our included studies were to a large extent of good methodological quality, with at least two thirds having low or moderate RoB. Still, “study limitations” was the most common reason for downgrading the GRADE. The measures for both outcomes and prognostic factors were mainly of “good” quality and statistical analyses were relevant but attrition and dealing with confounding were the weakest domains—which can seriously impact the results in prognostic factor studies. The assessment of study quality relies to a great extent to the level of relevant reporting. Often study quality was downgraded due to unclear detailing on, for example, study participation and attrition, which might not have been actual sources for bias. Moreover, for some RoB domains, the PABAK-OS was found to be unacceptably low. However, it was easy to obtain consensus on the overall RoB scores during the consensus discussions. On a general note, it was apparent that reporting has improved over the past decades, possibly as a result of the devise of reporting guidelines, for example the STROBE checklist. All in all, we believe our results have external validity and can be generalized within the context of the population and intervention of interest—still keeping in mind that our findings may apply to this specific outcome “Physical functioning” and possibly not to the other dependent variables that will be analyzed in subsequent reviews.
Physical functioning at 6 months or longer after MDR was not predicted by initial pain level or pain duration (chronicity), contrary to previous indications, and therefore should not be used for assumptions of treatment prognosis. Better physical functioning was predicted by high levels of initial self-assessed physical functioning. Furthermore, a better outcome was predicted by low levels of emotional distress and low levels of cognitive and behavioral risk factors, indicating that treatment should further target and optimize these modifiable factors. Finally, high levels of protective cognitive and behavioral factors were strong prognostic indicators of better physical functioning at 6 months or more after MDR, and an increased focus on positive, psychological protective factors may perhaps provide an opening for yet untapped clinical gains. The prognostic ability of the investigated factors may have been confirmed, but substantial heterogeneity between the studies was present and the effect sizes were in general fairly low, explaining only a limited part of the variance of outcome. Further research is naturally warranted to identify more important prognostic factors. Ultimately, this body of evidence can contribute to the development of clinical prediction models, which, in turn, will generate a basis for the future optimization of multidisciplinary biopsychosocial rehabilitation in chronic pain.
The authors would like to thank the Swedish Research Council, the Doctoral School in Health Care Sciences, Karolinska Institutet, Sweden, AFA-Insurance Sweden and Research-ALF, County Council of Östergötland, Linköping and the Swedish Research Council for Health, Working Life and Welfare (FORTE) for financial support. None of the funders have had any influence in the systematic review. The authors would also like to thank the Karolinska Institutet University Library for valuable support in developing and reviewing the electronic searches. The authors would also like to thank the research network within The Swedish Quality Registry for Pain Rehabilitation, in particular the Predictor group, for valuable contributions during the review process.
1. Taylor AM, Phillips K, Taylor JO, et al. Is chronic pain a disease in its own right? Discussions from a Pre-OMERACT 2014 Workshop on Chronic Pain. J Rheumatol. 2015;42:1947–1953.
2. Breivik H, Collett B, Ventafridda V, et al. Survey of chronic pain in Europe: prevalence, impact on daily life, and treatment. Eur J Pain. 2006;10:287–333.
3. Leadley RM, Armstrong N, Lee YC, et al. Chronic diseases in the European Union: the prevalence and health cost implications of chronic pain. J Pain Palliat Care Pharmacother. 2012;26:310–325.
4. Turk DC, Dworkin RH, Revicki D, et al. Identifying important outcome domains for chronic pain clinical trials: an IMMPACT survey of people with pain. Pain. 2008;137:276–285.
6. Gatchel RJ, McGeary DD, McGeary CA, et al. Interdisciplinary chronic pain management: past, present, and future. Am Psychol. 2014;69:119–130.
7. Hoffman BM, Papas RK, Chatkoff DK, et al. Meta-analysis
of psychological interventions for chronic low back pain. Health Psychol. 2007;26:1–9.
8. Kamper SJ, Apeldoorn AT, Chiarotto A, et al. Multidisciplinary biopsychosocial rehabilitation for chronic low back pain: Cochrane systematic review and meta-analysis
. BMJ. 2015;350:h444.
9. SBU. Methods for Treatment of Chronic Pain a Systematic Review of the literature (In Swedish: Metoder för behandling av långvarig smärta: en systematisk litteraturöversikt). Stockholm: SBU—Swedish Council on Health Technology Assessment; 2006.
10. SBU. Rehabilitation of Chronic Pain [In Swedish: Rehabilitering vid långvarig smärta En systematisk litteraturöversikt] SBU-rapport
. Stockholm: SBU—Swedish Council on Health Technology Assessment; 2010.
11. Scascighini L, Toma V, Dober-Spielmann S, et al. Multidisciplinary treatment for chronic pain: a systematic review of interventions and outcomes. Rheumatology (Oxford). 2008;47:670–678.
12. van Geen JW, Edelaar MJ, Janssen M, et al. The long-term effect of multidisciplinary back training: a systematic review. Spine
13. van Middelkoop M, Rubinstein SM, Kuijpers T, et al. A systematic review on the effectiveness of physical and rehabilitation interventions for chronic non-specific low back pain. Eur Spine J. 2011;20:19–39.
14. Craig P, Dieppe P, Macintyre S, et al. Developing and evaluating complex interventions: the new Medical Research Council guidance. BMJ. 2008;337:a1655.
15. Kaiser U, Treede R-D, Sabatowski R. Multimodal pain therapy in chronic noncancer pain—gold standard or need for further clarification? Pain. 2017;158:1853–1859.
16. Riley RD, Hayden JA, Steyerberg EW, et al. and for the PG. Prognosis Research Strategy (PROGRESS) 2: Prognostic Factor Research. PLoS Med. 2013;10:e1001380.
17. Clark GM. Prognostic factors
versus predictive factors: examples from a clinical trial of erlotinib. Mol Oncol
18. Croft P, Altman DG, Deeks JJ, et al. The science of clinical practice: disease diagnosis or patient prognosis? Evidence about “what is likely to happen” should shape clinical practice. BMC Med. 2015;13:20.
19. de Rooij A, Roorda LD, Otten RH, et al. Predictors of multidisciplinary treatment outcome
in fibromyalgia: a systematic review. Disabil Rehabil. 2013;35:437–449.
20. van der Hulst M, Vollenbroek-Hutten MM, Ijzerman MJ. A systematic review of sociodemographic, physical, and psychological predictors of multidisciplinary rehabilitation—or, back school treatment outcome
in patients with chronic low back pain. Spine (Phila Pa 1976). 2005;30:813–825.
21. Wessels T, van Tulder M, Sigl T, et al. What predicts outcome in non-operative treatments of chronic low back pain? A systematic review. Eur Spine J. 2006;15:1633–1644.
22. Artus M, Campbell P, Mallen CD, et al. Generic prognostic factors
for musculoskeletal pain in primary care: a systematic review. BMJ Open. 2017;7:e012901.
23. Laisne F, Lecomte C, Corbiere M. Biopsychosocial predictors of prognosis in musculoskeletal disorders: a systematic review of the literature (corrected and republished)*. Disabil Rehabil. 2012;34:1912–1941.
24. Valentin GH, Pilegaard MS, Vaegter HB, et al. Prognostic factors
for disability and sick leave in patients with subacute non-malignant pain: a systematic review of cohort studies. BMJ Open. 2016;6:e007616.
25. Tseli E, Grooten WJA, Stålnacke B-M, et al. Predictors of multidisciplinary rehabilitation outcomes in patients with chronic musculoskeletal pain
: protocol for a systematic review and meta-analysis
. Syst Rev. 2017;6:199.
26. Taylor AM, Phillips K, Patel KV, et al. Assessment of physical function and participation in chronic pain clinical trials: IMMPACT/OMERACT recommendations. Pain. 2016;157:1836–1850.
27. Main CJ, Keefe FJ, Jensen MP, et al. Fordyce’s Behavioral Methods for Chronic Pain and Illness: Republished With Invited Commentaries. Philadelphia, PA: Wolters Kluwer; 2014.
28. Moher D, Liberati A, Tetzlaff J, et al. Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement. J Clin Epidemiol. 2009;62:1006–1012.
29. Stroup DF, Berlin JA, Morton SC, et al. Meta-analysis
of observational studies in epidemiology: a proposal for reporting. JAMA. 2000;283:2008–2012.
30. Turk DC, Dworkin RH, Allen RR, et al. Core outcome domains for chronic pain clinical trials: IMMPACT recommendations. Pain. 2003;106:337–345.
31. Vertitas Health Innovation Ltd. Covidence systematic review software
. Vertitas Health Innovation Ltd: Melbourne, Australia. Available at: https://www.covidence.org/home
32. Shamseer L, Moher D, Clarke M, et al. Preferred reporting items for systematic review and meta-analysis
protocols (PRISMA-P) 2015: elaboration and explanation. BMJ. 2015;349:g7647.
33. Hayden JA, van Der Windt DA, Cartwright JL, et al. Assessing bias in studies of prognostic factors
. Ann Intern Med. 2013;158:280–287.
34. Cooper HM. Research Synthesis and Meta-Analysis
: A Step-by-Step Approach. Los Angeles, CA, London: SAGE; 2010.
35. Cooper HM, Hedges LV, Valentine JC. The Handbook of Research Synthesis and Meta-Analysis
. New York, NY: Russell Sage Foundation; 2009.
36. Wilson DB. Practical meta-analysis
effect size calculator (Online calculator). The Campbell Collaboration; 2015.
37. Lenhard W, Lenhard A. Calculation of effect sizes. Dettelbach, Germany: Psychometrica; 2016.
38. Lipsey MW, Wilson D. Practical Meta-Analysis
(Applied Social Research Methods). Thousand Oaks, California: SAGE Publications; 2000.
39. The Cochrane Collaboration. Review Manager (RevMan)
(Computer Program). Version 5.3. Copenhagen: The Nordic Cochrane Centre; 2014.
40. Egger M. Systematic Reviews in Health Care : Meta-Analysis
in Context. Chichester, GB: BMJ Books; 2008.
41. Angst F, Gantenbein AR, Lehmann S, et al. Multidimensional associative factors for improvement in pain, function, and working capacity after rehabilitation of whiplash associated disorder: a prognostic, prospective outcome study. BMC Musculoskelet Disord. 2014;15:130.
42. Bremander AB, Holmstrom G, Bergman S. Depression and age as predictors of patient-reported outcome in a multidisciplinary rehabilitation programme for chronic musculoskeletal pain
. Musculoskeletal Care. 2011;9:41–48.
43. Glattacker M, Heyduck K, Meffert C. Illness beliefs and treatment beliefs as predictors of short-term and medium-term outcome in chronic back pain. J Rehabil Med. 2013;45:268–276.
44. Riley RD, Elia EG, Malin G, et al. Multivariate meta-analysis
of prognostic factor studies with multiple cut-points and/or methods of measurement. Stat Med. 2015;34:2481–2496.
45. Atkins DBD, Briss PA, Eccles M, et al. GRADE
Working Group. Grading quality of evidence and strength of recommendations. BMJ. 2004;328:1490.
46. Huguet A, Hayden JA, Stinson J, et al. Judging the quality of evidence in reviews of prognostic factor research: adapting the GRADE
framework. Syst Rev. 2013;2:71.
47. Iorio A, Spencer FA, Falavigna M, et al. Use of GRADE
for assessment of evidence about prognosis: rating confidence in estimates of event rates in broad categories of patients. BMJ. 2015;350:h870.
48. Bendix AF, Bendix T, Haestrup C. Can it be predicted which patients with chronic low back pain should be offered tertiary rehabilitation in a functional restoration program? A search for demographic, socioeconomic, and physical predictors. Spine (Phila Pa 1976). 1998;23:1775–1783.
49. Bergstrom G, Jensen IB, Bodin L, et al. The impact of psychologically different patient groups on outcome after a vocational rehabilitation program for long-term spinal pain patients. Pain. 2001;93:229–237.
50. Bergström M, Ejelöv M, Mattsson M, et al. One-year follow-up of body awareness and perceived health after participating in a multimodal pain rehabilitation programme—a pilot study. Europ J Physiother. 2014;16:246–254.
51. Ciechanowski P, Sullivan M, Jensen M, et al. The relationship of attachment style to depression, catastrophizing and health care utilization in patients with chronic pain. Pain. 2003;104:627–637.
52. de Rooij A, van der Leeden M, Roorda LD, et al. Predictors of outcome of multidisciplinary treatment in chronic widespread pain: an observational study. BMC Musculoskelet Disord. 2013;14:133.
53. Dobkin PL, Liu A, Abrahamowicz M, et al. Predictors of disability and pain six months after the end of treatment for fibromyalgia. Clin J Pain. 2010;26:23–29.
54. Farin E, Gramm L, Schmidt E. The patient-physician relationship in patients with chronic low back pain as a predictor of outcomes after rehabilitation. J Behav Med. 2013;36:246–258.
55. Gerdle B, Molander P, Stenberg G, et al. Weak outcome predictors of multimodal rehabilitation at one-year follow-up in patients with chronic pain-a practice based evidence study from two SQRP centres. BMC Musculoskelet Disord. 2016;17:490.
56. Harkapaa K, Jarvikoski A, Estlander A. Health optimism and control beliefs as predictors for treatment outcome
of a multimodal back treatment program. Psychol Health. 1996;12:123–134.
57. Lillefjell M, Krokstad S, Espnes GA. Prediction of function in daily life following multidisciplinary rehabilitation for individuals with chronic musculoskeletal pain
; a prospective study. BMC Musculoskelet Disord. 2007;8:65.
58. Lüning-Bergsten C, Lundberg M, Lindberg P, et al. Change in kinesiophobia and its relation to activity limitation after multidisciplinary rehabilitation in patients with chronic back pain. Disabil Rehabil. 2012;34:852–858.
59. McGeary DD, Mayer TG, Gatchel RJ. High pain ratings predict treatment failure in chronic occupational musculoskeletal disorders. J Bone Joint Surg A. 2006;88:317–325.
60. Moradi B, Benedetti J, Zahlten-Hinguranage A, et al. The value of physical performance tests for predicting therapy outcome in patients with subacute low back pain: a prospective cohort study. Eur Spine J. 2009;18:1041–1049.
61. Moss-Morris R, Humphrey K, Johnson MH, et al. Patients’ perceptions of their pain condition across a multidisciplinary pain management program: do they change and if so does it matter? Clin J Pain. 2007;23:558–564.
62. Persson E, Lexell J, Rivano-Fischer M, et al. Occupational performance and factors associated with outcomes in patients participating in a musculoskeletal pain rehabilitation programme. J Rehabil Med. 2014;46:546–552.
63. Ruscheweyh R, Dany K, Marziniak M, et al. Basal pain sensitivity does not predict the outcome of multidisciplinary chronic pain treatment. Pain Med (Malden, Mass). 2015;16:1635–1642.
64. Trief PM, Yuan HA. The use of the MMPI in a chronic back pain rehabilitation program. J Clin Psychol. 1983;39:46–53.
65. van Hooff ML, Spruit M, O’Dowd JK, et al. Predictive factors for successful clinical outcome 1 year after an intensive combined physical and psychological programme for chronic low back pain. Eur Spine J. 2014;23:102–112.
66. Vendrig AA, Derksen JJ, de Mey HR. Utility of selected MMPI-2 scales in the outcome prediction for patients with chronic back pain. Psychol Assess. 1999;11:381–385.
67. Vendrig AA, Derksen JJ, de Mey HR. MMPI-2 Personality Psychopathology Five (PSY-5) and prediction of treatment outcome
for patients with chronic back pain. J Pers Assess. 2000;74:423–438.
68. Verkerk K, Luijsterburg PA, Heymans MW, et al. Prognosis and course of disability in patients with chronic nonspecific low back pain: a 5- and 12-month follow-up cohort study. Phys Ther. 2013;93:1603–1614.
69. Lemstra M, Olszynski WP. The effectiveness of multidisciplinary rehabilitation in the treatment of fibromyalgia—-a randomized controlled trial. Clin J Pain. 2005;21:166–174.
70. Gatchel RJ, Peng YB, Peters ML, et al. The biopsychosocial approach to chronic pain: scientific advances and future directions. Psychol Bull. 2007;133:581–624.
71. McCracken LM, Turk DC. Behavioral and cognitive-behavioral treatment for chronic pain: outcome, predictors of outcome, and treatment process. Spine
72. Morley S, Eccleston C, Williams A. Systematic review and meta-analysis
of randomized controlled trials of cognitive behaviour therapy and behaviour therapy for chronic pain in adults, excluding headache. Pain
73. McCracken LM, Morley S. The psychological flexibility model: a basis for integration and progress in psychological approaches to chronic pain management. J Pain. 2014;15:221–234.
74. Morley S, Williams A. New developments in the psychological management of chronic pain. Can J Psychiatry. 2015;60:168–175.
75. Hanssen MM, Peters ML, Boselie JJ, et al. Can positive affect attenuate (persistent) pain? State of the art and clinical implications. Curr Rheumatol Rep. 2017;19:80.
chronic musculoskeletal pain; GRADE; interdisciplinary rehabilitation; meta-analysis; prognostic factors; treatment outcome
Supplemental Digital Content
Copyright © 2019 Wolters Kluwer Health, Inc. All rights reserved.